the role of decorated sds micelles in sub-cmc protein denaturation and association

doi:10.1016/j.jmb.2009.06.019 J. Mol. Biol. (2009) 391, 207–226

Available online at www.sciencedirect.com

The Role of Decorated SDS Micelles in Sub-CMC ProteinDenaturation and Association

Kell K. Andersen1,2, Cristiano L. Oliveira3, Kim L. Larsen2,Flemming M. Poulsen4, Thomas H. Callisen5, Peter Westh6,Jan S. Pedersen6 and Daniel Otzen1⁎
1Interdisciplinary NanoscienceCenter, University of Aarhus,Gustav Wieds Vej 10C,DK-8000 Aarhus C, Denmark2Department of Life Sciences,Aalborg University,Sohngaardsholmsvej 49,DK-9000 Aalborg, Denmark3Department of Chemistry,iNANO InterdisciplinaryNanoscience Center and Centerfor mRNP Biogenesis andMetabolism, University ofAarhus, Langelandsgade 140,DK-8000 Aarhus C, Denmark4Structural Biology and NMRLaboratory, Department ofMolecular Biology, University ofCopenhagen, DK-2200Copenhagen, Denmark5Novozymes A/S, DK-2880Bagsværd, Denmark6NSM Functional Biomaterials,Roskilde University, PO Box260, DK-4000 Roskilde,Denmark
Received 28 January 2009;received in revised form28 May 2009;accepted 4 June 2009Available online10 June 2009

*Corresponding author. E-mail addAbbreviations used: SAXS, small-

albumin; CMC, critical micelle concITC, isothermal titration calorimetryFourier transformation.

0022-2836/$ - see front matter © 2009 E

We have combined spectroscopy, chromatography, calorimetry, and small-angle X-ray scattering (SAXS) to provide a comprehensive structural andstoichiometric description of the sodium dodecyl sulfate (SDS)-induceddenaturation of the 86-residue α-helical bovine acyl-coenzyme-A-bindingprotein (ACBP). Denaturation is a multistep process. Initial weak binding of1–3 SDS molecules per protein molecule below 1.3 mM does not perturb thetertiary structure. Subsequent binding of ∼13 SDS molecules per ACBPmolecule leads to the formation of SDS aggregates on the protein and changesin both tertiary and secondary structures. SAXS data show that, at this stage, adecorated micelle links two ACBP molecules together, leaving about half ofthe polypeptide chain as a disordered region protruding into the solvent.Further titrationwith SDS leads to the additional uptake of 26 SDSmolecules,which, according to SAXS, forms a larger decorated micelle bound to a singleACBP molecule. At the critical micelle concentration, we conclude fromreduced mobility and increased fluorescence anisotropy that each ACBPmolecule becomes associatedwithmore than onemicelle. At this point, 56–60SDS molecules are bound per ACBP molecule. Our data provide keystructural insights into decoratedmicelle complexeswith proteins, revealing aremarkable diversity in the different conformations they can stabilize. Thedata highlight that a minimum decorated micelle size, which may be a keydriving force for intermolecular protein association, exists. This may alsoprovide a structural basis for the known ability of submicellar surfactantconcentrations to induce protein aggregation and fibrillation.

© 2009 Elsevier Ltd. All rights reserved.

Keywords: ACBP; surfactant; isothermal titration calorimetry; small-angleX-ray scattering; dimerization
Edited by K. Kuwajima
ress: [email protected] X-ray scattering; ACBP, acyl-coenzyme-A-binding protein; BSA, bovine serumentration; ANS, 1,8-anilino-nathphalene-sulfonic acid; CE, capillary electrophoresis;; EGPC, eluent gel permeation chromatography; RI, refractive index; IFT, indirect

lsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.jmb.2009.06.019

mailto:[email protected]

208 SDS Aggregates in Protein Denaturation-Association

Introduction

Protein–surfactant interactions have both funda-mental and applied interests. They play a significantrole in the food industry, pharmaceutical industry,and, not the least, fabric and homecare detergentindustry. Enzymes such as proteases, amylases,lipases, and cellulases are typically present in manydetergency formulations.1,2 Furthermore, protein–surfactant interactions reveal new facets about thetypes of conformational changes that proteins mayundergo in response to a changing environment.While both nonionic and ionic surfactants canincrease the activity of lipases,3,4 ionic surfactantsgenerally denature proteins at low concentrationsthrough a combination of ionic and hydrophobicinteractions.5,6 The classical model of protein dena-turation by sodium dodecyl sulfate (SDS) is based onthe pioneering work of Tanford7 and has beenconfirmed by more recent reports.8 In this model,individual SDS molecules at low surfactant concen-trations first bind to a number of high-affinity sites,accompanied by a limiteddegree of structural change,followed by a plateau in the binding isothermbefore amassive uptake of SDS occurs in a cooperative step.However, many details on binding and accompany-ing structural changes remain to be elucidated. Thisparticularly relates to coupling between proteinstructure and susceptibility to SDS. Generally, SDSresistance cannot be linked to primary structure9 orthermodynamic stability,10 but appears to be morestrongly correlated with rigidity and lack of “breath-ing.” This is particularly found in proteins that arerich in β-sheet structure, often in combination witholigomeric assemblies, possibly due to the elaboratenetwork of global hydrogen bonding between β-strands. In contrast, the predominance of localinteractions stabilizing α-helical structures allowsSDS micelles to solubilize individual α-helices andthus unravel the tertiary structure of an α-helicalprotein while keeping the secondary structure moreor less intact. In support of this, our previous studiessuggest that proteins containing mixed α/β struc-tures such as S611 and α-lactalbumin12 are moresusceptible to denaturation than proteins consistingof all-β secondary structures,13 which also unfoldsignificantly more slowly.9,14 All α-helix proteinsunfold readily in SDS, but mechanistic details canvary significantly. Our study of the interactionsbetween SDS and the seven-helix heme proteinmyoglobin revealed a multistep denaturation15 thatis considerablymore complex than the behavior of thearchetypal surfactant-binding protein bovine serumalbumin (BSA).16,17

In addition to providing new insights into proteinconformational changes at the individual-moleculelevel, surfactants may also provide more insights intothe driving forces behind protein aggregation. Whilesurfactant micelles show strong dispersive or solubi-lizing abilities, submicellar concentrations of SDSshow a remarkable propensity to stimulate proteinaggregation. This is particularly pronounced at lowpH,where aggregation appears to lead to amorphous

precipitates,18 but also occurs at neutral pH, leadingto orderly or fibrillar aggregate structures for proteinsas diverse as Aβ,19 β2-microglobulin,20 S6,18 andlysozyme.21 The increased aggregation propensity isundoubtedly related to the ability of SDS to stabilizeaggregation-prone structures that represent species atthe very early stage of the aggregation process. Assuch, the mechanism by which surfactants stimulateaggregation may provide important informationabout aggregation pathways in biological contexts.However, the details of how surfactant and proteincooperate to facilitate this process remain obscure.The present study is devoted to the interactions of

SDS with a simple α-helix protein containing nocofactor, namely, the 86-residue four-helix bundlebovine acyl-coenzyme-A-binding protein (ACBP).ACBP has several properties that make it interestingin protein–surfactant studies: its unfolding behaviorin denaturants under equilibrium and kinetic condi-tions has been well characterized,22–24 and it has abinding site for amphiphilic ligands with Kd valuesfor C14–C22 acyl-coenzyme A around 2–10 nM,25

which might provide sites for initial SDS binding. Inthe present study, we combine a number of comple-mentary techniques that uncover molecular, meso-scopic, thermodynamic, and stoichiometricinformation to piece together the details of thedifferent binding steps involved in SDS denaturationof ACBP. Our work also provides low-resolutionstructural information on SDS–protein complexes,which we term “decorated micelles,” including aspecies inwhich twoproteinmolecules are bridgedbya cluster of SDSmolecules. This sheds new light on therole that SDS may play in promoting protein–proteininteractions.

Results

ACBP unfolding in SDS proceeds throughseveral transitions according to spectroscopyand capillary electrophoresis

Our aim is to provide as complete a description ofmultiple structural transitions that ACBP undergoesin the presence of SDS as possible. We start withconventional fluorescence spectroscopy, since theintensity I and the peak position of the emissionspectrum λmax can be very informative about the Trpside-chain environment. Two closely spaced trypto-phans (Trp55 and Trp58) are present in the primarysequence of ACBP. Upon titration of ACBP with SDSat pH 8, profound changes are observed in both I andλmax (Fig. 1a). No apparent change in fluorescenceintensity or λmax is observed in the concentrationrange 0–1 mM SDS. Between 1 and 3 mM, there is anincrease in I and λmax blueshifts, while furthertitration to 5 mM SDS resulted in a decrease influorescence with only a little change in λmax. Nosignificant change in fluorescence is observed abovethe critical micelle concentration (CMC) (∼5 mM; seethe text below). The baseline region (0–1mM) prior to

Fig. 1. ACBP denaturation followed by tryptophanfluorescence and far-UV CD. (a) Tryptophan fluorescenceintensity and λmax are unaffected in the SDS concentrationrange 0–1 mM SDS, followed by increased fluorescenceintensity and a blueshift in λmax (transition 1). Transition 2is observed between 3 and 5 mM SDS characterized by afluorescence decrease with only little change in λmax. (b)Far-UV CD data, plotted as the ratio between ellipticity at207 and 220 nm, report on secondary change in transition1, while transition 2 cannot be detected. Inset: Far-UV CDspectra of ACBP in 0 and 10 mM SDS.

Fig. 2. (a) Rotational mobility of the Trp side chains ofACBP in SDS as shown by fluorescence anisotropy (righty-axis) and ANS fluorescence versus SDS concentration(left y-axis). Trp rotational mobility is described in threestates. Native structure in the interval 0–1 mM SDS,followed by a decrease in anisotropy, indicating a lessordered and denatured structure in the range 2–4.5 mMSDS. Finally, associations with SDS micelles, at whichpoint anisotropy increases. ANS binds in the absence ofSDS, indicating a hydrophobic area present on the surfaceof ACBP. ANS–SDS cobinding is observed up to between 0and 1.75 mM SDS, followed by decreased ANS fluores-cence indicating denaturation of ACBP. (b) Formation ofSDS aggregates on the protein surface is revealed by theincrease in the I3/I1 ratio of the fluorescent probe pyrenearound ∼1.5 mM SDS.

209SDS Aggregates in Protein Denaturation-Association

the first transition shrinks with decreasing pH downto pH 4 (data not shown). In part, this reflects anincreased affinity of SDS at lower pHdue to increasedprotonation of anionic side chains;11 a reduction inprotein stability is less likely, as ACBP remains stablyfolded until pH 3.5.26 Far-UV CD spectra at pH 8show that both native ACBP and SDS-denaturedACBP possess a high degree of α-helix secondarystructure, as indicated by the two minima at 207 and220 nm (Fig. 1b). Using the ratio of these twowavelengths, we only observe one transition, whichcorresponds to the change in fluorescence emissionintensity and λmax between 1 and 3 mM. Thus, the 1–3mMtransition shows joint changes in secondary andtertiary structures, whereas the 3–5 mM transition isrestricted to changes in the aromatic environment.We used fluorescence anisotropy to measure

changes in the Trp side chains' rotational diffusionupon titrationwith SDS (Fig. 2a). In the range 0–1mMSDS, no obvious change was observed; however,

increasing the SDS concentration to 2–3 mMdecreased anisotropy, most likely due to a fasterrotation of the Trp residues and increased structuralflexibility. This observation corresponds well to theincrease in fluorescence intensity and the change insecondary structure. Anisotropy levels stayed con-stant up to around 5mMSDS, afterwhich it increasedto the initial native-state level. This change inanisotropy coincides with the formation of bulk SDSmicelles (at theCMC). It is likely thatdenaturedACBPassociates with SDS micelles, forming a large aggre-gate and leading to a slower rotation of the Trp sidechain. Unlike our small-angle X-ray scattering (SAXS)

Fig. 3. (a) Electropherograms showing the mobility ofACBP at SDS concentrations from 0 to 20 mM. Linesindicate the mobility of native ACBP and ACBP saturatedwith SDS, respectively. Figures indicate SDS concentrationin micromolars. (b) Mobility of ACBP, defined as themidway between the baseline before the peak region andthe baseline after the peak region, plotted against SDSconcentration. The broken line indicates the onset of thefluorescence-based transitions seen in Fig. 1a.


data for the sub-CMC ACBP–SDS complexes (see thetext below), we do not have any direct structuralinformation on the complex between ACBP and bulkmicelles. However, it is worthmentioning that severalstructures have been suggested for the complexbetween proteins and SDS above the CMC. Onbalance, there is most support for the “necklace-and-bead” model, which involves an unfolded andelongated protein that wraps around severalmicelles.27

The fluorescent probe pyrene was used to investi-gate whether small SDS aggregates, which wedesignate as decorated micelles, are formed on thesurface of ACBP. Pyrene is a highly hydrophobicmolecule with low solubility in water (2–3 μM); theratio of the intensity of its emission at 372.5 nm to theintensity of its emission at 383.5 nm (I3/I1) can be usedto evaluate the polarity of its environment.28 In theabsence of protein, this ratio changes from ∼0.65 to∼0.93 (Fig. 2b) when pyrene partitions into SDSmicelles, indicating micelle formation in the concen-tration range 4–6 mM SDS. In the presence of 2 μMACBP, it is clear that pyrene experiences a morehydrophobic environment than water in the range1.25–6 mM SDS. We explain this by the formation ofSDS aggregates on ACBP, into which pyrene canpartition. The point at which these aggregates aredetected agrees very well with the concentrationwhere we observe an increase in tryptophan fluores-cence and a change in secondary structure by CD. Puttogether, these studies indicate that denaturation ofACBP is due to the formation of hydrophobic clustersof SDS on ACBP.A second extrinsic probe, 1,8-anilino-nathphalene-

sulfonic acid (ANS), was used to study the initialbinding of SDS and denaturation. ANS binds tohydrophobic regions on protein surfaces with amany-fold increase in fluorescence intensity.29 Thishas made it a very good indicator for partially foldedstates,30 exposing contiguous hydrophobic regionsthat are buried in the native state. However, ANSwillalso bind to natively formed hydrophobic patchessuch as active sites or ligand-binding pockets.31 Wealso see ANS binding to native ACBP (Fig. 2a). Thisagrees with independent observations that ANSbinds to the active site of ACBP, reflecting theprotein's high affinity for hydrophobic substratessuch as long-chain acyl-coenzyme A esters (B. B.Kragelund and F.M.P., unpublished observations).Upon titration of ACBP with SDS, ANS fluorescenceincreases in the interval 0–1.5 mM SDS due to eithercobinding of ANS and SDS or exposure of hydro-phobic patches upon denaturation (Fig. 2a). Furthertitration results in a decrease in fluorescence, whichcoincides with denaturation of the native structure.As opposed to Trp fluorescence, which did not showany change in the interval 0–1 mM SDS, a significantchange in ANS fluorescence was observed in thisrange. This suggests that initial SDS binding takesplace in the absence of significant changes in thearomatic environment.To obtain independent information on this early

binding event,we used capillary electrophoresis (CE),

where solute mobility is based on a combination ofhydrodynamicdrag and charge,making it sensitive tothe binding of even small amounts of chargedligands.32,33 Electropherograms show changes in themobility of ACBP at 0–1 mM SDS without aconcomitant broadening of the ACBP peak (Fig. 3a),indicating that ACBP binds SDS without denaturingthe protein. Above 1mMSDS, a shoulder on the rightside appears and increases with SDS concentration,merging with the main peak at around 2.5 mM SDSuntil only a large smear is observed, indicatingdenatured ACBP with a range of bound SDSmolecules. Finally, between 10 and 20 mM SDS,ACBP is fully saturated with SDS and thus retainsconstant mobility over this concentration range.When plotting the mobility of ACBP (defined as themidway between the baseline before the peak regionand the baseline after the peak region) against SDSconcentration,we observe an almost linear increase inmobility in the range 0–1 mM SDS, followed by asteep increase inmobility that eventually levels out at10–20 mM SDS (Fig. 3b). Broken lines indicate the


onset of the fluorescence-based transitions seen in Fig.1a. CE data generally very well support the spectro-scopic techniques described above, apart from thecontinued shift in peak position between 4 and 10mMSDS, which is not reflected in fluorescence data. Thismay reflect additional binding of SDS (e.g., asnecklace-and-bead structure) without concomitantstructural changes in ACBP (see Discussion).

Isothermal titration calorimetry resolves thestoichiometry of SDS binding

Fluorescence and CD focus predominantly onACBP's structural changes. To obtain informationabout the stoichiometry and thermodynamics ofbinding, we turned to isothermal titration calorime-

Fig. 4. (a) ITC enthalpograms for the titration of SDS (99 mMsixty 4-μl titrations; however, for clarity, only a few poinenthalpogram. The experimental resolution is illustrated in (ACBP. (b) Enthalpogram for 140 mM ACBP, with indication ofcalculated according to Eq. (1) at different stages of the process.angle scattering and spectroscopy is discussed in the main texvisual guide to highlight the second endothermic peak with adifferent transition points indicated in (a), plotted as a functioconcentrations are derived from Eq. (1) and summarized in Tabof bound SDS molecules. This plot corresponds to Tanford's cSDS complexes are indicated in the appropriate concentrationsample names described in Table 3) under which the SAXS d

try (ITC), which measures the heat flow associatedwith binding even in the absence of structuralchanges.15,34,35 Enthalpograms from titrations ofACBP with SDS revealed a number of characteristicand reproducible transitions (Fig. 4a). Each titrationcurve is generally described by a large endothermicprocess peaking at 1.5–2.5 mM, followed by a dip to aminimum at 3–5mM, a second peak around 4–7mM,and a dip at 6–11 mM SDS. At higher SDSconcentrations, the signal levels out, indicating nofurther interaction, as the binding process hasreached saturation at this stage; thus, added SDS(injected as micelles) remains in micellar form anddoes not lead to any additional detectable heat flow.Each of the observed transitions represents distinct

SDS binding transitions, which we have described for

) into four different ACBP solutions. Each trial comprisedts are marked along the broken line identifying eachb), which shows all data points for the trial with 140 μMthe binding stoichiometry and the free SDS concentrationThe molecular interpretation of points 1–7 based on small-t. The dotted line between points 4 and 6 is inserted as amaximum at point 5. (c) Total concentrations of SDS, at then of ACBP concentration. Binding numbers and free SDSle 1. (d) Free SDS concentration plotted versus the numberlassical SDS binding isotherm.7 Different types of protein–range. The two arrows refer to the conditions (including

ata were recorded.


other proteins such as cutinase,34 myoglobin,15 andthe all-β proteins TII27 and Tnfn3.35 The secondendothermic peak may be visualized more clearly inFig. 4b, where the broken line indicates the plateauthat is seen for the other studied proteins, typicallytaking place after an exothermic peak.15,35,36 Thisplateau was investigated in some detail for SDS-titrated into cutinase,37 and it was concluded that itreflected the hydrophobically driven association ofSDS and denatured protein, which is exothermic atroom temperature and becomes endothermic above30–40 °C. We interpret the second endothermic peakas a specific structural change in the protein–SDScomplex occurring on top of a more general accretionof SDS. Details of this change will be discussed belowin light of SAXS results.The stoichiometry of SDS binding (the binding

isotherm) can be derived from a series of titrationsinto solutionswithdifferent protein concentrations, asdescribed previously.34 Briefly, the underlying prin-ciple is that a given trait or anomaly in theenthalpogram reflects a certain stage or transition inthe titration (e.g., protein unfolding, binding satura-tion, and so on, as discussed above). As more SDS isrequired to satisfy the stoichiometry of a certaintransition the higher is the concentration of protein,the location of a transition will be shifted to the rightwith increasing [ACBP]. This tendency is clearly seenin Fig. 4a, and it may be quantified by a massconservation approach, which separates the total SDSconcentration [SDS]tot into a free population and aprotein-bound population. Thus, at a certain stage(defined by maximum, minimum, or inflexion in Fig.4a) of the titration, we may write:

½SDS�tot¼ ½SDS�free +N ACBP�½ ð1Þwhere N is the number of SDS molecules bound perprotein. In Fig. 4c, we have plotted [SDS]tot as afunction of [ACBP] for the seven transitional pointsdefined in Fig. 4b and indeed found the linear relationstipulated by Eq.(1). Hence, the slope and interceptlisted inTable 1 specify the free SDS concentration andbinding number at the seven transition points. ThreeSDS molecules bind at a free SDS concentration of upto 1.3 mM, during which time ACBP does notundergo any significant changes. Themajor endother-

Table 1. Stoichiometry of binding at different stages ofITC titration

Transition pointa Number of bound SDSb [SDS]free (mM)b

1 2.97±0.28 1.25±0.022 6.95±0.06 1.57±0.003 11.33±0.14 1.96±0.014 16.44±0.62 2.77±0.055 25.27±1.53 3.85±0.126 42.47±0.32 5.19±0.027 56±5.07 6.46±0.36Final point (EGPC) 60.0 (CMC)

All data were determined in 10 mM Tris (pH 8.0) at 22 °C.a See Fig. 4b for definitions.b Based on fits to Eq. (1).

mic process occurs over a free SDS range of 1.3–2.8 mM, which corresponds nicely to the maindenaturation profile and the formation of decoratedmicelles and is accompanied by the binding of ∼16SDS molecules. The second endothermic peak, span-ning 2.8–5.2mM free SDS and leading to 42moleculesbound to SDS, coincides with the second fluorescencetitration, with minor changes in λmax. Finally, accre-tion of large micelle-like clusters occurs up to theCMC completion at around 6.5 mM, with 56 SDSmolecules bound and with resumption of increasedanisotropy (Fig. 2a). The enthalpies scale essentiallylinearly with protein concentration (data not shown),aswould be expected in a processwhere each bindingreaction contributes to calorimetric signal.Plotting the number of bound SDSmolecules versus

free SDS concentration (Fig. 4d) reveals the character-istic multistep binding isotherm described in Tan-ford's classical treatment.7 Thus, an initial strong (or“high-energy”) binding at low SDS is followed by aslow-rising part of the binding isotherm at intermedi-ate concentrations with a binding level of about 0.4 gSDS/g protein. Eventually, as the free SDS concentra-tion approaches the CMC, massive uptake (tradition-ally referred to as cooperative binding) sets in, leadingto a final binding level of about 1.7 g SDS/g protein.Although the final amount of SDS bound is signifi-cantly above the average value of 1.2 g SDS/gprotein,7 this binding profile mimics the generic SDSbinding isotherm,8 and the spectroscopic and SAXSresults from the current work identify several of thestructural details involved in the associated protein–SDS interactions. Some of these are delineated in thegraph and discussed further below.

Independent verification of binding stoichiometryby eluent gel permeation chromatography

To verify the saturation binding number (56±5)obtained by ITC, we turned to eluent gel permeationchromatography (EGPC).38 In this approach, acolumn is equilibrated with a certain concentrationof a desired ligand (in our case 9.0 mMSDS), which isdistributed between ∼5 mM monomer (the CMC)and ∼4 mM micelles (the exact CMC value is notessential for the accuracy of the binding number). Themicelle concentration is monitored by the refractiveindex (RI). We then inject samples of 50 μM ACBP indifferent concentrations of SDS onto the column,which leads to two RI peaks. The large positive peakaround 23 min (also seen by absorption at 280 nm)corresponds to ACBP saturated with SDS, whereasthe smaller peak around 31 min (invisible by 280-nmabsorption) corresponds to micelles. As illustrated inFig. 5, if there is less than 9.0mM free (i.e., monomericor micellar) SDS in the ACBP sample injected into thecolumn, then the micelle concentration will be lessthan ∼4 mM (the concentration in the column buffer,which is the background concentration on thecolumn), and this will be reflected in a negativemicelle peak at 31min. Conversely,more than 9.0mMfree SDS in the ACBP sample leads to a positivemicelle peak. By plotting micelle peak area against

Fig. 5. ACBP–SDS binding stoichiometry measured byEGPC. An eluent containing 9 mM SDS (5 mM monomerand 4mMmicelles) passes through a gel-filtration column,and a change in micelle concentration is measured with anRI detector. Samples of 50 μM ACBP incubated with 9.5–13.5 mM SDS are passed onto the column. A micelleconcentration smaller or higher than 4 mM results in adecrease or an increase in the RI, respectively. Inset: Areaof the micelle peak as a function of SDS concentration. Theintersect at 12 mM indicates that 3 mM SDS is bound to50 μM ACBP, which gives a binding stoichiometry of 60SDS molecules to one ACBP molecule.


total SDS concentration in the ACBP sample, weobtain a clear linear dependence (Fig. 5, inset), whichintersects the x-axis at 12.0 mM SDS. Thus, at a totalSDS concentration of 12.0 mM in the ACBP sample,the free concentration of SDS is 9.0 mM, which meansthat 3.0 mM SDS is bound. Given that the ACBPconcentration is 50 μM, the stoichiometry of bindingmust be 60.0 SDS to oneACBPmolecule.Within error,this value (whichwe estimate to have 5%uncertainty)is identical with the value (56±5) obtained from ITC.

SAXS identifies both dimeric and monomericACBP–SDS complexes

In previous sections, we have obtained informationabout the conformational changes that ACBP under-goes at increasing SDS concentrations, as well as thenumber of SDS molecules bound at various transi-tions. Based on these data, we identify two centralpoints where it is particularly relevant to obtainstructural information about the ACBP–SDS com-plexes by SAXS. These are as follows: (1) after the firstendothermic peak, where ∼16 SDS molecules arebound (and where ACBP has undergone significantchanges in secondary and tertiary structures), and (2)after the second endothermic peak, where a total of∼42 SDS molecules are bound, accompanied by amore modest change in tertiary structure and nosecondary structure change. These are points 4 and 6in Fig. 4b, which correspond to distinct parts of theTanford binding isotherm depicted in Fig. 4d. Wereason that the additional uptake of SDS moleculesmust be reflected in a structural rearrangement of theACBP–SDS complex, which SAXS can shed light on.

In order to give a comprehensivepresentation of theinvestigations of the complexes by SAXS, we havealso performed SAXS on the nativeACBP protein andon pure SDS samples. The study of the native proteinis important with respect to addressing whether thesamples with ACBP and SDS consist of mixtures orwhether they are homogenous, with only one type ofcomplex. The samples of pure SDS serve to demon-strate that the new modeling method that we use foranalyzing the SAXS data from the complexes, whichis based on Monte Carlo simulation integration, isapplicable for describing SAXS from SDS micellarstructures.The analysis of SAXS data from the SDS micelles

and complexes progresses in two steps: First, theindirect Fourier transformation (IFT) method39,40 isused for obtaining model-independent informationon the structures of the objects. Second, structuralmodels are constructed, partly based on model-independent information from IFT and partly basedon information from spectroscopic, chromatographic,and calorimetric titration experiments and on knowl-edge of constituting molecules and their concentra-tions within the samples.The IFT method provides the pair distance dis-

tribution function p(r), which gives direct informationabout the structure in real space. The function is ahistogram of all distances between a pair of pointswithin the particles weighted by excess electrondensity relative to the buffer (which can be bothpositive and negative) at each point. The functiongoes to zero at r=Dmax, where Dmax is the maximumdistance within scattering objects. A micelle-likestructure of SDS has a negative excess scatteringlength density in the core, as the electron density ofhydrocarbon chains is lower than that of the buffer,and a positive excess scattering length density in theheadgroup region, as the electron densities of thesulfate and sodium ions are larger than that of thebuffer. Therefore, the pair distance distribution func-tion p(r) of micelle-like structures of SDS will havecharacteristic oscillationswith amainmaximum closeto the diameter of the headgroup+counterion shell.For proteins, the p(r) function gives themaximum sizeof the protein and, by comparison with p(r) functionsfor objects of known shape, also low-resolutioninformation on the structure of the protein.The new modeling method is very similar to the

traditional modeling of SAXS data in which ageometric structure is assumed and parametersdescribing the geometry are optimized by weightedleast-squares methods when the scattering intensityof the model is fitted to the experimental data.41Constraints in terms of partial specific volumes andexcess scattering length densities of various parts ofthe molecules and in terms of concentrations can beincluded in the model. This allows the scatteringintensity to be calculated on an absolute scale, andgood-quality fits ensure that models are consistentwith prior knowledge of the samples. The maindifference between the traditionalmodeling approachand the new one introduced here is that Monte Carlosimulation techniques are used for integrating over


the volume of the objects in connection with thecalculation of the scattering intensity.42,43 A finite setof points generated by Monte Carlo methods is usedfor representing the structure. Note that it isstraightforward to introduce smeared distributionsor interfaces in the model by adding random vectorswith a certain distribution to the points. Theimplementation used in the present work employssmeared distributions in terms of Gaussians, as wellas constraints, so that the SAXS intensitywas fitted onan absolute scale. The Monte Carlo method has thegreat advantage that one can apply very complexstructuralmodels since one is not limited to structuresfor which the intensity can be calculated analytically.SAXS data for pure nativeACBPprotein in solution

are shown in Fig. 6. The data decay monotonicallywith increasing q, and most of the decays are beyondq=0.1 Å− 1. The p(r) function (inset) obtained by IFTanalysis shows that the protein in solution forms acompact, slightly flattened structure. Comparingexperimental SAXS data with the theoretical intensityfor the protein in solution calculated for the atomiccoordinates of the protein (Protein Data Bank entry1HB6) using the program CRYSOL,44 we obtain agood agreement (Fig. 6). This shows that the nativeprotein in solution has a similar shape as in the crystalstructure.The experimental data for pure SDS samples are

displayed in Fig. 7 (left). The characteristic bump athigh scattering vectors q originates from a core–shellmicelle structure with a negative scattering lengthdensity in the core and a positive scattering lengthdensity in the shell.45 At low q, SAXS data at high SDSconcentrations display a decrease as q goes to zerodue to interparticle interference effects caused mainlyby electrostatic repulsion between micelles. The p(r)functions obtained by IFT are displayed in Fig. 7(right). The p(r) functions have a characteristic bumpat short distances and an oscillatory behavior that

originates from the core–shell structure of the micelle.The functions furthermore demonstrate that themaximum diameter of the micelles is about 60–70 Å.The Monte Carlo modeling method was applied to

SAXS data from pure SDS solutions. The micellestructure was described by a core–shell ellipsoidalmodel, with a constant thickness of the headgroupshell and core semiaxis of (Rm,Rm,ɛRm). A fitparameter was used for describing the concentrationof SDS in micelles. This value can be compared to theexpected value estimated from the overall concentra-tion and the estimated CMC of 4 mM (Table 2). It wasnecessary to fit the number of headgroup excesselectrons in terms of a scale factor Sehead to account forthe reduction in excess electrons due to dissociation ofsodium counterions. As illustrated in Table 2, theapproach gave good fits to the data and micelleaggregation numbers of 66±1 for oblate ellipsoidswith Rm=20.3±0.3 Å and ɛ=0.663±0.005, except atthe lowest concentration where the micelles appearmore flattened. The number of headgroup excesselectrons is reduced by a factor of about 0.86, and thisparameter was kept fixed during the fit to the datafrom the complexes described below. The obtainedparameters for the SDS micelles are in good agree-ment with those obtained in a previous work by Vasset al. using a more traditional modeling approach.45

The results demonstrate that the model can very welldescribe the scattering from SDSmicelles, andwe canproceed to applying it for modeling the complexes.The experimental SAXS data for ACBP–SDS com-

plexes are shown in Fig. 8a for all studied concentra-tions. A characteristic bump at high scattering vectorsq demonstrates that, in all cases, we have theformation of micelle-like structures. At low q, theSAXS data at high protein concentrations as for thepure micelles display a decrease as q goes to zero dueto interparticle interference effects. The p(r) functionsobtained by IFTare shown in Fig. 8b, and they have a

Fig. 6. SAXS results for nativeACBP in solution at 5 mg/mL.Experimental data (open circles),IFT fit (continuous line), and fitusing an atomic-resolution crystalstructure model (dotted line).Upper inset: Pair distance distribu-tion function for ACBP. Lowerinset: Atomic structure of ACBPprotein (Protein Data Bank entry1HB6).

Fig. 7. SAXS results and associated IFT results for pure SDSmicelles. (a) The experimental SAXS data are shown as opencircles, and the model fits using the methods described in the text are shown as continuous lines. The overall SDSconcentrations are given. The same structurewas used for fitting the samples bS2–bS6, but itwas only fully optimized againstthe bS5 and bS6 data sets. The structure had to be slightly modified to fit the data for bS1 for which the concentration is veryclose to CMC. The structures of themicelles are shown as spheres at the position of theMonte Carlo points. The hydrocarbontail region is represented by red spheres, whereas the headgroup and counterion region are represented by green spheres.Fivehundredpoints areused for each of the two contributions (SDSheadgroup+counterions andSDSC12 tails, respectively).(b) Pair distance distribution functions obtained from the IFT fit of each experimental data set.


Table 2. Parameters obtained by model fits to SAXS data for pure SDS micelles

Fit results

Sample

bS1 bS2 bS3 bS4 bS5 bS6

CSDS−CMC (mM) 1.8 (0.8) 2.8 (4.5) 12.4 (15.7) 3.7 (6.1) 16.1 (15.3) 52.0 (44.6)Nagg

a 67±5 69±1 66±1 66±1 66±1 66±1Rm (Å)b 22.9±2.8 20.7±0.9 20.3c 20.3c 20.3±0.3 20.3±0.2ɛb 0.476±0.03 0.663c 0.663c 0.663c 0.663±0.005 0.662±0.01Sehead

d 1.06±0.21 0.878c 0.878c 0.878c 0.878±0.07 0.855±0.003

Nominal concentrations assuming a CMC of 4 mM are given in parentheses.a Number of SDS molecules per micelle.b Parameters referring to the oblate ellipsoid formed by micelles with long axis length Rm and axis scaling factor ɛ.c Parameters were kept fixed.d Scale factor accounting for reduction in excess electrons due to dissociation of Na+ counterions.


characteristic bump at short distances and an oscilla-tory behavior. The functions are similar to thoseobtained for pure micelle solutions, however withoutnegative portions. The similarity shows that there is amicelle-like core–shell structural organization in thecomplexes. The functions furthermore demonstratethat the maximum diameter of the complexes is 60–80 Å, with a clear tendency for larger diameters forcomplexes with less SDS bound (samples S1–S3,corresponding to point 4 in Fig. 4b) as compared tocomplexes with more SDS bound (samples S5–S6,corresponding to point 6 in Fig. 4b). We note that theSAXS intensities and the p(r) functions are both verydifferent from those of the native protein and muchmore similar to those of pure micellar samples. If thesamples have a significant part of heterogeneity interms of complexes and native protein, which a prioricould be expected in particular for sampleswith a lowSDS/protein ratio (samples S1–S3), the bump at highq in the SAXS data and the oscillatory behavior of p(r)would be much less pronounced. From the data andthe model-independent analysis, we can thereforealready conclude that the samples are quite homo-geneous, and we will use this in the modeling.The Monte Carlo modeling approach was also

applied to the five SAXS data sets from the protein–SDS complexes. A large set of possible structures wastested for the ACBP+SDS complexes, and theirscattering intensities were compared with the experi-mental SAXS data. The models consisted of cylind-rical structures (both long and flat) for the protein andvarious structures for SDS contribution: micelles ofSDS associated with the cylinder, rims of SDS, andhemirims of SDS. The surfactant aggregate wasplaced at several positions around the proteincontribution. However, none of the structures couldreproduce the experimental SAXS intensities or the p(r). After further tests, it was realized that only a core–shell structure could give the features observed in p(r).For the data at high SDS concentration, a Janus-typeparticle consisting of an ellipsoidal SDS micelle withthe protein as shell covering part of the surface of themicelle and mixing with the SDS headgroups couldperfectly reproduce the data. The micelle corestructure is described as an ellipsoid of revolutionwith axes (Ri,Ri,ɛRi). The protein covers the fullangular range around the z-axis, except the rangebetween −α and α (a region between −wp and wp

along z), and has a thickness of Wp. The fits for theACBP–SDS complex at high SDS/protein ratio areshown in Fig. 8a (S5 and S6), which also shows theresulting structures of the complexes. The modelreproduces the experimental SAXS data very well.Table 3 displays the fitting results and aggregationnumbers derived from them. The model has oneACBP molecule per decorated micelle, and theaggregation number of the micelle is 35–38, in verygood agreement with the value obtained by ITC(42.5). The protein almost fully covers the surface ofthe micelle, and such a distribution is in goodagreement with the loss of tertiary structure and theconservation of part of the secondary structure, asobserved by spectroscopic techniques. Note that themicellar cores are nearly spherical, with a smallprolate tendency, and thus the micelles are quitedifferent from those formed in pure SDS solutions.The volume of the distribution of the protein can alsobe estimated from the fit results. The proteindistribution, as displayed in Fig. 8a, is partly mixedwith the SDS headgroups, but these only contributeabout 5000 Å3. As mentioned, the distribution coversalmost the full surface area on the decorated micelle,and it has a thickness of about 14 Å. It has a volume of70,000–80,000 Å3, which is six to seven times that ofthe dry protein (12,000 Å3), suggesting that a largefraction (∼80%) of this volume is water.For the data at low SDS/protein ratio, the same

structure could reproduce the data quite well;however, very importantly, the total amount of SDSrequired was much higher than what was present inthe samples. Such a model is clearly inconsistent andhas to be disregarded. Calculations showed thateither there has to be free protein in solution or twoproteins have to be associated with each decoratedmicelle. Further test showed that, in the former case,the free protein has to be associated in dimers and haseither a large dilute structure or a shell-like structure.As none of these possibilities seems likely, they werediscarded. For the second case with two proteinsassociated with each decorated micelle, a significantpart of the protein should only contribute to thescattering at low q, as it would otherwise mask thescattering at high q from the micelle-like structure.This is possible if parts of the protein are disordered.In order to describe this,wemasked amodelwith twoACBPmolecules permicelle, with half of each protein

Fig. 8. SAXS intensities and associated IFT results for the ACBP+SDS complexes. (a) Model fits and data for thedifferent ACBP–SDS complexes. The experimental data are shown as open circles, and the theoretical fits are shown ascontinuous lines. Concentrations of ACBP (in mg/mL; 1 mg/mL corresponds to 100 μM ACBP) and SDS (in mM) areindicated in brackets. The sample concentrations are indicated for each case. The sketch of the proposed model is shownas inset, and the resulting model is shown near each curve. Red points represent the hydrocarbon tails of the SDS, greenpoints represent the headgroup and counterions of SDS, and blue points represent protein distribution. (b) Pair distancedistribution functions obtained from the IFT fit of each experimental data set.


Table 3. Parameters obtained by model fits to SAXS data for ACBP protein+SDS micelles

Fit results

Sample

S1 S2 S3 S5 S6

Cp (mg/mL)a 1.0 2.5 5.0 2.5 5.0CSDS (mM)b 4.3 6.4 9.9 14.8 24.3Nagg

c 25±14 18.6±1.1 18.7±0.7 37.6±4.1 40.1±1.3Molecular mass per complex (kDa)d 20 20 20 10 10Wp (Å)e 5.2±1.2 8.6±0.8 8.1±0.2 14.2±0.2 14.2±0.4wp (Å)e 14.3±2.8 18.6±2.3 20.4±1.6 21.8±3.0 20.8±1.3α (rad)e 3.86±1.4 3.43±0.18 3.40±0.19 1.92±0.21 3.26±0.12Ri (Å)

f 14.2±0.8 13.6±0.3 14.2±0.2 11.8±0.4 12.1±0.2ɛf 1.42±0.84 1.26±0.07 1.12±0.04 1.92±0.21 1.90±0.06SRW (Å)g 12±4 16±2 16.2±0.7 — —rRW (Å)g 12±6 9.1±0.9 8.5±0.4 — —

a ACBP concentration.b SDS concentration.c Number of SDS molecules in the complex.d Protein mass per complex.e In the SAXSmodel, the protein molecule covers the full angular range around the z-axis, except the range between −α and α (a region

between −wp and wp along the z-axis), and has a thickness of Wp.f The micelle core structure is described as an ellipsoid of revolution with axes (Ri,Ri,ɛRi), where ɛ is a constant defined in the table.g These parameters refer to the random configurations of ACBP outside the SDS micelle. Random walks were constructed using step

lengths of SRW with spheres of radius rRW at each point of the walk.


located in a shell around the micelle and with theother two halves in random configurations protrud-ing into the solvent on opposite sides of the micelle.The random configurations were generated as ran-dom walks starting from a flat surface and onlyoccupying half of the space. The walks each had 10points and a step length of SRW. A small ensemble of10 different configurations for the complex wasgenerated. The walks were placed at two oppositesides of the decoratedmicelle along the z-axis. A finitevolume was associated with random walks byplacing spheres of radius rRW at each point of thewalk.The model with two proteins bound per decorated

micelle was fitted to the data at low SDS/proteinratio. The results are given in Table 3, and the fits andstructures are displayed in Fig. 8a (S1, S2, and S3). Themodel fits the SAXS data well. Importantly, thestructure is also in agreement with the larger size ofthe structures at low SDS concentrations, as observedin the p(r) functions, which underlines that this‘dimer’ model is correct. As already mentioned, themodel employs molecular constraints; therefore, wecan calculate the amount of SDS per ACBP, as well asthe overall SDS concentration. For themodelwith twoACBP proteins per complex, the protein and theoverall SDS concentrations are in agreement with theactual sample concentrations, and the model is thusconsistent. For the samples S1, S2, and S3 with lowSDS/protein ratios, the number of associated SDSmolecules perACBP, given asNagg in Table 3, is 15–17,which agrees very well with the 16 moleculesdetermined by ITC. The core of the SDS micellestructure is also for these samples that are close tospherical, however, with a small tendency towards aslightly oblate shape. The distribution of the protein inthe structures is in good agreement with the loss oftertiary structure and the conservation of part of thesecondary structure, as observed by spectroscopic

techniques.We note that the dimeric association of theACBP in the model for the ACBP–SDS complexes issimilar to the behavior of cutinase at low SDSconcentrations36 where dimers also are found.The volume of the protein distribution for the low

SDS/protein ratio samples (S1–S3) is 40,000–50,000Å3,which is about twice the volume of two proteins(2×12,000 Å3), meaning that the protein distribution issomewhat spread out on the surface of micelle. Theprotein distribution, as displayed in Fig. 8a, alsoincludes the SDS headgroups, but these only con-tribute about 2200 Å3. Thus, there is also a significantamount of water present in these distributions. Theprotein layer on the micelle is only about 9 Å thick,corresponding to the thickness of a single α-helix.Themodels derived from the SAXS data for the low

and high SDS/protein ratio samples for the formedcomplexes are displayed in two different views inFig. 9. Although the decorated micelle SAXS modelsare of low resolution, the figure provides a visualimpression of the characteristic features of thecomplexes formed between the ACBP protein andthe SDS micelles. It should be noted that the modelsagree both with SAXS data and with all theadditional information that we have available fromspectroscopic, chromatographic, and calorimetricmeasurements, and that we have not been able tofind any other model that could fit the data despiteextensive tests on alternative structures.

Discussion

We have used a wide variety of techniques tocharacterize the changes in ACBP's structure and inthe extent of binding of SDS molecules over a rangespanning the whole bulk premicellar regime. Thisallows us to compare ACBP with a large number ofother proteins and to establish a general pattern of

Fig. 9. Model representation of the ACBP+SDS complex structures using the spheres at the position of the MCpoints used in the analysis of the SAXS data. Note that 500 points are used for each of the three contributions (SDSheadgroup+counterions, SDS C12 tails, and protein, respectively). Green points represent the headgroup and counterionsof SDS, red points represent the hydrocarbon tails of the SDS, and blue points represent protein distribution. Left:Configuration for low concentrations of SDS. Right: Configuration for high concentrations of SDS. For details, see the text.


protein–surfactant interactions. We will start bydescribing the individual ACBP–SDS binding stepsin more detail. An overview is provided by thetentative structures given in Fig. 10 and the summarygiven in Table 4. It should be noted that thedenaturing steps and structures given in the figureare based on an interpretation of all the experimentalevidence on paper, as direct structural investigations

Fig. 10. Schematic representation of the different stages of A3 SDS molecules without losing the native structure. Stage Bmolecules that binds 2 ACBP molecules. Further binding of SDa shell-like structure of SDS. The structure presented in stagemodel that has been proposed for protein interactions with SD

by SAXS have only been performed on structures atstages B and C:

Stage A (0–1.3 mMSDS): No change in tryptophanfluorescence, pyrene fluorescence, or secondarystructure is observed, indicating that ACBP retainsits native structure. ANS fluorescence and CE,however, clearly reveal the binding of SDS

CBP denaturation. In stage A, ACBP binds between 1 andinvolves the formation of a decorated micelle of 37 SDSS to a total of 40 in stage C leads to monomeric ACBP withD is speculative, but it represents the “beads-on-a-string”S micelles above the CMC.

Table 4. Summary of conformational changes that ACBP undergoes in SDS

Stage Event MethodsNumber ofSDS bounda [SDS]free (mM)a

A Binding of a few molecules (around the active site,possibly making the ANS site more hydrophobic);no significant structural or calorimetric impact

ANS, CE 3 1.25

B Major alterations in secondary and tertiary structures,decrease in anisotropy, and displacement of ANS;

Endothermic endothermic interactions leading to theformation of decorated micelles

ITC, Trp, CD, CE,ANS, anisotropy, pyrene

16 (ITC), 18–19 (SAXS) 2.8

C Second exothermic series of interactions, minor changesin tertiary structure, and possibly second phase of

ANS displacement

ITC, Trp (λmax), ANS 42 (ITC), 38–40 (SAXS) 5.8

D No changes in ACBP structure; onset of micellization ITC, CE, anisotropy 56±5 (ITC), 60±6 (EGPC) 6 (CMC)

All data were determined in 10 mM Tris (pH 8.0) at 25 °C.a Based on ITC data fitted to Eq. (1), unless otherwise stated.


molecules without accompanying denaturation. Itis very difficult to determine the exact number ofSDSmolecules bound to ACBP by ITC because thenumber is very low, leading to a very low heatflow. A reasonable approximation would bebetween 1 and 3 SDS molecules. Individual SDSmolecules bind to specific sites on the surface of thenative state, probably around the ligand-bindingsite where we have identified several positivelycharged patches. The binding behavior in thisregion is considerably less complex than that ofSDS towards myoglobin, where the predenatura-tion zone can be divided into two regions (0–0.15and 0.15–0.4 mM SDS). SDS–myoglobin interac-tions are detected by ITC already at 0.05–0.1 mMSDS,15 and decorated micelles start to form ataround 0.15–0.4 mM SDS, although it is notpossible to determine the stoichiometry of bindingon myoglobin until around 0.4–1.2 mM SDS. It ispossible that the presence of predeterminedhydrophobic ligand-binding sites on the surfacemakes this process stoichiometrically more well-defined for ACBP, while stoichiometric binding tomyoglobin only sets in above a critical SDSconcentration of about 0.15 mM.Stage B (1.3–2.8 mM SDS): In this region, properdenaturation occurs; there are changes in Trpfluorescence intensity, λmax, and secondary struc-ture, as well as formation of SDS-decoratedmicelles from an uptake of 3–16 SDS molecules.Anisotropic measurements show that the Trp sidechain tumbles faster, probably due to loss ofstructure in its surroundings. Also ANS fluores-cence decreases at this stage, showing that thestructural changes that take place in this region aresevere enough to disrupt the ANS-binding site(s).At the end of this transition, structural character-ization by SAXS revealed the formation of adecorated micelle, containing about 33 SDS mole-cules, that links twoACBPmolecules together. Partof the protein is bound to the decorated micelle,and part of it protrudes into the solvent. Clearly

ACBP has lost (most of) its tertiary structure at thispoint.Stage C (2.8–5.2 mM SDS) A massive uptake of anadditional 26 SDS molecules to a total of 42 isaccompanied by a decrease in Trp fluorescence.There are no additional changes in secondarystructure. An endothermic process, observed byITC, suggests an event involving bond breaking,which probably involves further unfolding. How-ever, the enthalpy of this process is not as large asobserved for the initial denaturation step, indicat-ing that structural loss is minor as compared toinitial denaturation in stage B. The changes in Trpfluorescence and the slow decline in ANS fluores-cence may reflect rearrangements of SDS micellesor tertiary changes. Both may occur according toSAXS, which reveals an ACBP–SDS complex withonly one protein per decorated micelle and all ofthe proteins located in a shell-like structure aroundthe micelle.Stage D (N5.2 mM SDS): This stage involves theformation of bulk SDS micelles and binding of 56–60 (according to ITC/EGPC) SDS molecules,leading to large micellar structures where Trp hasreduced rotation. CE also indicates that SDSmolecules continue to accumulate on ACBP alltheway up to 10mMSDS, possibly as large proteinmicelle complexes are formedwith ACBP bridgingbetweenmicelles. The fact that ITC does not reporton this interaction is probably due to the weakinteraction of this binding. The concentrations wehave sampled in our experiments at stage Dcorrespond to the lower to middle SDS concentra-tion range for species formed during SDS-PAGE.Proteins are typically dissolved in a loading bufferwith around 2% (70 mM) SDS. The SDS-PAGE gelsthemselves usually contain around 7mMSDS, andthe running buffer, which conducts current, con-tains around 3.5 mMSDS, although concentrationsof SDS can reach around 1M during stacking.46 Atthe buffer strengthused for SDS-PAGE (∼200mM),the CMC is well below 3.5 mM, so we can expect


micelles to be present throughout the electrophor-esis process. The stage D structure sketched out inFig. 10 represents a single protein moleculeinteracting with several micelles. This model isbased on a bead-on-a-string model in which thepolypeptide chain is flexible and micelle-likeclusters of SDS are scattered along the chain,47–49

as investigated by neutron scattering, bindingisotherms, and free-boundary electrophoresis.

Two-stage SDS denaturation is seen for several,but not all, α-helix-containing proteins andrequires binding of a critical amount of SDS

The enthalpogram profile for ACBP resembles theITC profiles not only for an all α-helix protein suchas BSA17 but also for the mixed α/β proteinlysozyme.50 Although studies on lysozyme andBSA have not used the same collection of techniquesas in the present study, the ITC titration profile is, inboth cases, described by two endothermic peaks, asis the case for ACBP. Endothermic processescorrespond to a loss of organized protein structure,while the formation of electrostatic protein–SDScontacts leads to an exothermic signal; in contrast,hydrophobic protein–SDS contacts are enthalpicallyessentially invisible.7,51

We propose that these proteins follow a commondenaturation mechanism in which initial denatura-tion at relatively low SDS concentrations, driven bythe formation of organized SDS aggregates on theprotein surface, leads to the first endothermic peak.The minor second peak may then be due tocondensation of the protein molecule on the micellarinterface (i.e., step 3 in Fig. 10). This secondrearrangement is accompanied by smaller structuralchanges but a larger number of SDS molecules perprotein molecule. It is difficult to say whether the firststep in all cases requires the formation of a proteindimer bridged by the decorated SDS micelle. Ifdimerization occurs over the entire protein rangeprobed by ITC scans, it will not show up as a proteinconcentration dependence phenomenon. The abilityto dimerize may depend on the details of the proteinsurface chemistry and the availability of otherdomains on the protein that could fold back ontothe decorated micelle to form an “internal arrange-ment.” Other α-helix-containing proteins such asmyoglobin15 and S618 do not show this doubleendothermic peak at pH 8.0, indicating that particularconditions need to be fulfilled to allow this process tohappen. SDS-induced dimerization detected bysmall-angle (neutron) scattering was also reportedfor cutinase, and this was hypothesized to rely ondecreased intermolecular repulsion, as positivelycharged cutinase molecules were gradually neutra-lized by the anionic surfactant.36 However, chargeestimates for ACBP based on the primary structure†suggest a negative charge of 2–3 at pH 8, and it

†www.expasy.ch

follows that the mechanism suggested for cutinasecannot account for the current results. Rather, transi-tion from stage B to stageC (Fig. 10) forACBPappearsto be driven byattractive interactionsof thedenaturedprotein and themicellar interface, and hence require acritical area of available interface to occur. Interest-ingly, this may be directly reflected in the ITC results.Thus, the thermograms for cutinase were similar tothose for ACBP in Fig. 4a in the early parts of thetitration, but did not show any signs of the secondendotherm found for ACBP. The binding of SDS tocutinase saturated at 0.5 g SDS/g protein, which isabout two to three times lower than typical values.52

This correlation between low binding numbers andlack of second transition supports the view that acritical amount of surfactant must be associated withthe protein to support the condensed structure (stageC in Fig. 10). This interpretation is corroborated by anumber of previously investigated proteins, includ-ing phytase,53 which has low saturation binding andno second endotherm (like cutinase), andlysozyme,54,55 BSA,17 α-lactalbumin,12 andmyoglobin,15 which all show normal saturationbinding (∼1 g/g) and a conspicuous secondendotherm (i.e., like ACBP). While a final conclusionawaits further experimental backup, the currentresults suggest (i) that the transition between thetwo forms in Fig. 9 requires a critical amount of SDS(in the 0.5–1 g/g protein range) onto which theprotein can bind, and (ii) that this transition may beidentified as the second endotherm in ITC trials.

α-Helix structure may be a prerequisite formultistep unfolding

All the proteins described above contain α-helixstructure and interact strongly with SDS under theCMC. Previously, we have studied proteins with ahigh content of β-structure, which only showedweakinteractionswithmonomeric SDS andwere only fullydenatured by micellar SDS.35 Thus, for Tnfn3, we didnot observe any endothermic processes below theCMC, showing that micelles were required todenature the proteins while monomeric SDS showedno denaturing effect. This highlights a fundamentaldifference between all β-proteins and proteins con-taining α-helical elements, namely, that α-helicalsegments are much more prone to taking uppremicellar SDS with dramatic conformational con-sequences. We have previously used protein engi-neering approaches to identify the α-helix as apreferred site of attack in the unfolding mechanismof the mixed α/β protein S6 in SDS.56 In addition, S6also forms premicellar SDS aggregates and undergoesa series of conformational changes according totryptophan fluorescence,18 which is remarkablysimilar to that of ACBP. The tryptophan fluorescenceprofile, which is a fluorescence intensity increasefollowed by an intensity decrease, is also observed forcutinase,34 α-lactalbumin,12 and myoglobin,15,57

albeit the fluorescence decrease in the latter case israther faint, probably due to the heme group whosedissociation leads to a subsequent rise in fluorescence.


In all cases where pyrene fluorescence experimentshave been conducted (for α-lactalbumin, S6, andACBP), the increase in tryptophan fluorescence isassociated with the formation of decorated micelles.Cutinase, myoglobin, and α-lactalbumin do not showany fluorescencebaselineprior tounfolding, probablydue to high susceptibility to SDS binding. As ACBP isdesigned to bind anionic substrateswith hydrophobictails, it may be able to bind SDS at very lowconcentrations, but without significant conforma-tional changes.For all-β proteins, the fluorescence and far-UV CD

signals change in parallel, indicating a simple bindingmechanism,35 in contrast to α or α/β proteins whereCD and fluorescence signals do not show an identicalnumber of steps. Since the β-proteins are onlydenatured by micelles, the denaturation processtakes place around the CMC. This means that theproteins bind a large amount of SDS in one steparound the CMC and, as a consequence, secondaryand tertiary structures change in parallel. In contrast,α and α/β structured proteins denature at SDSconcentrations well below the CMC. This leads to amore gradual denaturation process as the binding ofSDS takes place over a wider range of SDS concentra-tions. As a consequence, several stages in thedenaturation process can be observed for α and α/βproteins, as emphasized by the present study.

Decorated micelles linking protein moleculesmay provide a structural explanation forsubmicellar protein aggregation withphysiological parallels

SAXS analysis shows the formation of decoratedmicelles with an aggregation number of about 33,since two proteins are associated with each decoratedmicelle at the lower SDS concentrations investigated.It is likely that these micelles are the smallest onesthat can be formed and, thus, one can conclude thatthe structure of the formed complexes is dominatedby the properties of SDS. The initial binding of SDS atvery low concentrations predominantly occurs byelectrostatic interactions as single molecules. Thesebound SDS molecules have lost their mixing entropywith the solvent. Although they presumably bindprimarily to the most hydrophobic part of the proteinfacilitated by the loss of protein tertiary structure,they are likely to have their hydrophobic tails at leastpartially exposed to the solvent. They are thereforeprone to forming micelle-like structures at concentra-tions much below that of the bulk CMC value of SDS.For the formation of energetically favorable micelle-like structures with sufficient space and conforma-tional entropy for the C12 tails, the aggregationnumber has to exceed a minimum value. A way toachieve this is to link up with SDS molecules boundto another protein molecule. Thus, unsatisfied“hydrophobic bonds” drive the formation of inter-molecular contacts with another protein molecule.Furthermore, the disordered regions of ACBP outsidethe micelle-decorated region, which are exposed tothe solvent, may also, by themselves, provide anchor

points to other proteins, since conformational flex-ibility is a prerequisite for initial contacts in proteinaggregation.58 Note that these phenomena only occurwithin a narrow SDS concentration range. Once theSDS concentration of SDS is high enough to facilitatethe formation of full micellar structures on eachprotein surface or bulk micelles available for binding,the motivation for aggregation is removed, and SDScan revert to its role as an agent of dispersion andsolubilization.Ultimately, decoratedmicelles, in combinationwith

flexible protein regions in which protein–SDS inter-actions are mediated by electrostatic interactionswhile hydrophobic interactions mediate intermolecu-lar protein contacts, may be the force that drives theaccumulation of higher-order protein aggregates atcertain critical SDS concentrations. Although ACBPdoes not form higher-order aggregates with SDS atpH 8.0, this may only be a matter of altering, forexample, pH, since even small alterations in theionization of side chains can strongly affect the abilityof SDS to bind, unfold,11 and aggregate18 proteins.Suitable aggregation conditions with submicellarSDS, some at neutral pH, have been identified forother proteins such as S6,18 Aβ,19 β2-microglobulin,20

collagen,59 and lysozyme.21 Similar types of amphi-phile-driven aggregation may occur in vivo. Forexample, fatty acids such as oleic acid can formhigher-order complexes with α-lactalbumin andrelated proteins with potent anti-cancer properties.60

The large number of oleic acid molecules bound perprotein leaves plenty of room for the formation ofdecorated micelles.

Materials and Methods

Chemicals

Tris(hydroxymethyl)aminomethane (Tris) and SDS wereobtained from AppliChem (Darmstadt, Germany). ACBPwas purified as described previously.61 Pyrene and ANSwere obtained from Sigma-Aldrich (St. Louis, MO). Allchemicals were of the highest grade available. All experi-ments were performed in 10 mM Tris (pH 8).

Fluorescence measurements

All experiments were conducted in a 10-mm quartzcuvette (Hellma) on a LS-55 Luminescence spectrometer(Perkin-Elmer Instruments, UK) at 25 °C using 2 μMACBP(except for fluorescence anisotropy, for which 20 μMACBPwas used). Solutions were recorded with a scan speed of200 nm/min and three accumulations after at least 30min ofequilibration.

Tryptophan fluorescence

Excitation at 295 nm and emission at 345 nm weredetermined using 10-nm slit widths for both excitation andemission wavelengths. For anisotropy, 15 measurementsfor each samplewere averaged using an integration time of10 s.


Pyrene fluorescence

A stock solution of 200 μM pyrene in ethanol was madeand added to the samples up to a final concentration of1 μM. Excitation at 335 nmwas determined using excitationand emission slit widths of 3.5 nm. The ratio of the intensityof the emission at 372.5 nm to the intensity of the emission at383.5 nm was used for further analysis.

ANS fluorescence

ACBP (2 μM) was incubated with 0–10 mM SDS and40 μM ANS. Contributions from buffer and SDS weresubtracted at each SDS concentration. Excitation at 350 nmand emission at 500 nm were determined, with excitationand emission slit widths of 7.5 nm.

Circular dichroism

Far-UVCD spectra were recorded, using 10 μMACBP, ina 1.0-mm quartz cuvette on a JASCO J-715 spectropolari-meter (Jasco Spectroscopic Co. Ltd., Japan) equippedwith aJasco PTC-423S temperature control unit. Wavelength scanswere recorded in thewavelength range of 185–250 nm,witha bandwidth of 2 nm and a scanning speed of 50 nm/min.Five accumulations were averaged to yield the finalspectrum. Background contributions from the buffer weresubtracted.

Capillary electrophoresis

CE was carried out on a Beckman PACE-MDQ systemusing a capillary with an inner diameter of 25 μm. The totallength of the capillary was 60 cm, and the length to thedetector was 50 cm. The running buffer consisted of 10 mMTris (pH 8) and SDS in the range 0–20 mM. The samplecontained 1 mg/mL ACBP and 1 mM dimethylformamidein order to monitor electroosmotic flow. In each run, a plugcorresponding to 3% of the capillary length was injected,and electrophoresis was performed at 30 kV. Themobility ofACBP was followed using absorbance at 214 nm.

Isothermal titration calorimetry

Calorimetric measurements were conducted on a VP-ITC(MicroCal, Inc., Northampton, MA). The reference cell wasfilled with water and, in a typical experiment, the samplecell was loadedwith a solution of 18–141 μMACBP. The cellsolutionwas titrated with aliquots of 2.5–4 μl of 99 mMSDSin 10 mM Tris (pH 8). All experiments were performed at22 °C, where SDS demicellization is practically athermal.34

Therefore, enthalpic contribution from the demicellizationof SDS upon injection can be neglected in data analysis. Theobtainedheat signals from the ITCwere integratedusing theOrigin software supplied by MicroCal, Inc.

Eluent gel permeation chromatography

Measurements were carried out using an eluent buffercontaining 9 mM SDS in 10 mM Tris (pH 8). The eluent waspassed onto a 24-mL Superose 6 (GE Healthcare) gel-filtration columnwith a flow rate of 0.5mL/min. Followingthe column, a Hitachi L-4250 UV detector measuredabsorbance at 280 nm, and an Agilent 1100 detectormeasured the RI. A probe series of 50 μM ACBP was

incubated with varying SDS concentrations in the range of9.5–14 mM SDS for 1 h. After equilibration of the columnwith eluent, a probe sample was injected onto the column,and absorbance at 280 nm and the RI were followed overtime.

Small-angle X-ray scattering

SAXS experiments were performed on an in-houseinstrument at the Chemistry Department of the Universityof Aarhus.62 All experiments were performed at 25 °C in10 mM Tris (pH 8). For comparison, experiments wereperformed on the native protein at a concentration of 5 mg/mL, on pure SDS micellar solutions at six differentconcentrations (given in Fig. 6), and on samples withACBP–SDS complexes at five different concentrations (seelegend to Fig. 8a). The exposure time was 1800 s, and datatreatment, background subtraction of buffer, and normal-ization to absolute scale were performed using homemadesoftware. The resulting data on absolute scale are displayedas a function of themodulus of the scattering vector q=(4π/λ)sin(θ), where λ=0.154 nm is the X-ray wavelength and 2θis the angle between incident X-ray and scattered X-ray.The SAXSdatawere first analyzedby the IFTmethod,39,40

whichprovides the pair distance distribution function p(r). Itgives direct information about the structure in real space.The function is a histogramof all distances between a pair ofpoints within the particles weighted by the excess electrondensity (which can be both positive and negative) at thepoints. Structural information derived from the p(r) functioncan beused for identifying structural features that have to beincorporated into a model for the micelles and thecomplexes. The influence of interparticle interference effectswas eliminated by omitting the low-q part of the data.

A new modeling method was developed and implemen-ted in order to describe the SAXS data from the complexes.The approach is inspired by the work of Hansen42 andSpinozzi et al.43 The method is similar to the traditionalmodeling of SAXS data inwhich a structure is assumed andparameters describing the structure are optimized byweighted least-squares methods when fitting the scatteringintensity of the model to the experimental data.41 The maindifference of the newmethod introduced here is that MonteCarlo simulation techniques are used for integrating overthe volume of the objects in connection with the calculationof scattering intensity.42,43 This is accomplished by using afinite set of points generated by Monte Carlo methods forrepresenting the structure. TheMonte Carlomethod has thegreat advantage that one can apply very complex structuralmodels since one is not limited to structures for which theintensity can be calculated analytically. This allows us to testa large number of different structures against the experi-mental data. The newapproachwas used for analyzing boththe SAXS data for pure SDS solutions and the solutionswiththe SDS–protein complexes.It should be pointed out that the new approach is a

modeling method and not an ab initio method as thosedeveloped by Chacón et al.,63,64 Svergun,65Svergun et al.,66

Walther et al.,67 andVigil et al.68 The general feature of the abinitio methods is that very few assumptions aremade aboutthe structure of the particles and that random structures aregenerated by Monte Carlo simulations. The structures arethen optimized with respect to the measured data byvarious techniques such as genetic algorithms or simulatedannealing. Our approach is, as already mentioned, amodeling approach where a geometric model for thestructure, parameterized by only a few parameters, isassumed. The parameters are subsequently optimized by


traditional least-squaresmethods. For ab initiomethods, therandomprocess in the generation and selection of structurescan lead to a significant variation between the modelsderived in different runs. In our new approach,Monte Carlotechniques are only used for integration, and there aretherefore only negligible differences between modelsgenerated in different runs.In our new approach, a large set (108) of statistically

uniformly distributed points is first generated by MonteCarlo simulation within a cubic search volume of volume(2Rmax),

3 whereRmax is the expectedmaximum radius of thestructure. Rmax can be estimated from the p(r) function andshould fulfill RmaxNDmax/2, where Dmax is the distance atwhich p(r) goes to zero. Additional three-dimensionalrandomvectors with a Gaussian distribution are calculated,so that these can later be used for smearing the distributionsof components in the model. Note that when the vectorswith a Gaussian distribution are added to the points in thedistribution of a component, they correspond to a convolu-tion of the distribution of a component with the Gaussian.For each of the components of the structure, a subset of thepoints is selected by geometrical conditions for thecoordinates. The subset is defined by parameters that canbe varied during a least-squares optimization. The para-meters also define the various components of the model(hydrocarbon core of the micelle, shell describing thesurfactant headgroups and counterions, and protein,respectively), which each has a different scattering lengthdensity. For each component, 5000 points are selected, andthese points are divided into 10 subsets with 500 points.Excess electron density is associatedwith each point, so thatthe known number of excess electrons is obtained. Themolecular values used for calculating the values were asfollows: SDS headgroup+counterion, volume of 60.53 Å3

with 59 electrons; C12 tailgroup, volume of 355.1 Å3 with 97electrons.45 The protein volume was calculated from theprotein sequence and corresponds to a theoretical partialspecific volume of 0.7372 cm3/g. The excess electron densityof 2.0×1010 e/g was used for the estimation of molecularmass (J.S.P. and C.L.O., unpublished data). The bufferelectron density was set to 0.3334 e/Å3, so that the numberof excess electrons per headgroup, tail, and proteinwas 38.8,−21.4, and 1178, respectively. The core volume of themicelles/decoratedmicelles is expected to be dry; therefore,it was used for calculating an aggregation number bydividing the volume with that of a C12 tail.The pair distance distribution p(r), which is a histogram

over distances between the points weighted by the electrondensity of the points, was calculated for each subset andadded. The resolution of p(r) was chosen to be 0.1 Å. Notethat the distribution functions are calculated for each subsetand merged rather than calculated for the full set. Theprocedure using subsets is faster by a factor of 10. Theconvergence of p(r) is quite good, as the distances are moreindependent in the subsets (S. Hansen, private communica-tion). The function p(r) is calculated in a subroutine callbefore the q dependence is calculated by a simple Fouriertransformation of p(r). The resulting intensity on absolutescale is calculated by multiplying the Fourier transform bythe number concentration of particles and the square of theThomson radius of the electron (0.282×10−12 cm).For points representing the headgroup+counterions and

the micelle core, smearing of the distributions wasintroduced, as already mentioned, by adding randomvectors with a Gaussian distribution to the position of thepoints with scale factors for the width of the Gaussiandistribution. A width of σhead for the interfaces of theheadgroup+counterion distribution and awidth ofσcore forthe surface of the core were used. The width of the

distribution generated for the headgroup region was fixedat 3.0 Å, and σhead and σcore were fixed, respectively, at 4.0and 2.0 Å, in accordance with previous studies.45

Parameters for the structure and background wereoptimized using a standard Levenberg–Maquard weightedleast-squares algorithm.41 Despite the discretization ofstructures into limited sets of points, the gradients requiredfor the least-squares algorithm can be calculated for allparameters without any problems.The final intensity expression on an absolute scale that

was used for fitting the experimental data was:

I qð Þ = nP qð ÞS qð Þ ð2Þ

wheren is the number density of particles,P(q) is the Fouriertransform of the p(r) functionmultiplied by the square of theThomson radius, and S(q) is the average structure factor ofthe systemdescribing the effects of interparticle interactions.As the particles are charged, the structure factor for ascreened Coulomb potential69 is used.

Acknowledgements

K.K.A. was supported by a predoctoral grant fromthe innovation consortium BIOPRO (headed by Dr.Torben Madsen, DHI), financed by the DanishMinistry of Science, Technology, and Innovation. P.W. is grateful for support from the CarlsbergFoundation, the Danish Research Foundation, andthe Danish Research Agency (grants 26-02-0160 and21-04-0087 to P.W.). D.O. is grateful for long-termsupport from the Danish Research Foundation(inSPIN) and the Villum Kann Rasmussen Founda-tion (BioNET). We thank Dr. S. Vass for very helpfuldiscussions on SDS micellar structures.

References

1. Stoner, M. R., Dale, D. A., Gualfetti, P. J., Becker, T. &Randolph, T. W. (2006). Surfactant-induced unfoldingof cellulase: kinetic studies. Biotechnol. Prog. 22,225–232.

2. Kirk, O., Borchert, T. V. & Fuglsang, C. C. (2002).Industrial enzyme applications. Curr. Opin. Biotechnol.13, 345–351.

3. Mogensen, J. E., Sehgal, P. & Otzen, D. E. (2005).Activation, inhibition, and destabilization of Thermo-myces lanuginosus lipase by detergents. Biochemistry,44, 1719–1730.

4. Sonesson, A. W., Callisen, T. H., Elofsson, U. M. &Brismar, H. (2007). Imaging the detergency of singlecotton fibers with confocal microscopy: the effect ofsurfactants and lipases. J. Surfactants Deterg. 10,211–218.

5. Gitlin, I., Gudiksen, K. L. & Whitesides, G. M.(2006). Peracetylated bovine carbonic anhydrase(BCA-Ac18) is kinetically more stable than nativeBCA to sodium dodecyl sulfate. J. Phys. Chem., B,110, 2372–2377.

6. Otzen, D. E., Christiansen, L. & Schulein, M. (1999). Acomparative study of the unfolding of the endogluca-nase Cel45 from Humicola insolens in denaturant andsurfactant. Protein Sci. 8, 1878–1887.


7. Tanford, C. (1980). The Hydrophobic Effect. Formation ofMicelles and BiologicalMembranes, 2nd edit.Wiley&Sons,New York.

8. Turro, N. J., Lei, X.-G., Ananthapadmanabhan, K. P. &Aronson, M. (1995). Spectroscopic probe analysis ofprotein–surfactant interactions: the BSA/SDS system.Langmuir, 11, 2525–2533.

9. Manning, M. & Colon, W. (2004). Structural basis ofprotein kinetic stability: resistance to sodium dode-cyl sulfate suggests a central role for rigidity and abias toward beta-sheet structure. Biochemistry, 43,11248–11254.

10. Nelson, C. A. (1971). The binding of detergents toproteins. J. Biol. Chem. 246, 3895–3901.

11. Otzen, D. E. (2002). Protein unfolding in detergents:effect of micelle structure, ionic strength, pH, andtemperature. Biophys. J. 83, 2219–2230.

12. Otzen, D. E., Sehgal, P. & Westh, P. (2009). -αLactal-bumin is unfolded by all classes of detergents but withdifferent mechanisms. J. Colloid Interface Sci. 329,273–283.

13. Nielsen, M. M., Andersen, K. K., Westh, P. & Otzen,D. E. (2007). Unfolding of beta-sheet proteins in SDS.Biophys. J. 92, 3674–3685.

14. Cunningham, E. L., Jaswal, S. S., Sohl, J. L. & Agard,D. A. (1999). Kinetic stability as a mechanism forprotease longevity. Proc. Natl Acad. Sci. USA, 96,11008–11014.

15. Andersen, K., Westh, P. & Otzen, D. E. (2008). A globalstudy of myoglobin–surfactant interactions. Langmuir,15, 399–407.

16. Nielsen, A. D., Borch, K. & Westh, P. (2000). Thermo-chemistry of the specific binding of C12 surfactants tobovine serum albumin. Biochim. Biophys. Acta, 1479,321–331.

17. Kelley, D. & McClements, D. J. (2003). Interactions ofbovine serum albumin with ionic surfactants inaqueous solutions. Food Hydrocolloids, 17, 73–85.

18. Otzen, D. E., Nesgaard, L., Andersen, K. K., Hansen,J. H., Christiansen, G., Doe, H. & Sehgal, P. (2008).Aggregation of S6 in a quasi-native state by mono-meric SDS. Biochim. Biophys. Acta, 1784, 400–414.

19. Rangachari, V., Reed, D. K., Moore, B. D. &Rosenberry, T. L. (2006). Secondary structure andinterfacial aggregation of amyloid-β(1–40) on sodiumdodecyl sulfate micelles. Biochemistry, 45, 8639–8648.

20. Yamamoto, S., Kasegawa, K., Yamaguchi, I., Tsutsumi,S., Kardos, J., Goto, Y. et al. (2004). Low concentrationsof sodium dodecyl sulfate induces the extension ofbeta-2-microglobulin-related amyloid fibrils at neutralpH. Biochemistry, 43, 11075–11082.

21. Moosavi-Movahedi, A. A., Pirzadeh, P., Hashemnia,S., Ahmadian, S., Hemmateenejad, B., Amani, M. et al.(2007). Fibril formation of lysozyme upon interactionwith sodium dodecyl sulfate at pH 9.2. Colloids Surf. B,60, 55–61.

22. Kragelund, B. B., Poulsen, K., Andersen, K. V.,Baldursson, T., Kroll, J. B., Neergard, T. B. et al.(1999). Conserved residues and their role in thestructure, function, and stability of acyl-coenzyme Abinding protein. Biochemistry, 38, 2386–2394.

23. Kragelund, B. B., Osmark, P., Neergaard, T. B., Schiodt,J., Kristiansen, K., Knudsen, J. et al. (1999). Theformation of a native-like structure containing eightconserved hydrophobic residues is rate limiting in two-state protein folding of ACBP. Nat. Struct. Biol. 6,594–601.

24. Kragelund, B. B., Robinson, C. V., Knudsen, J.,Dobson, C. M. & Poulsen, F. M. (1995). Folding of a

four-helix bundle: studies of acyl-coenzyme A bind-ing protein. Biochemistry, 34, 7217–7224.

25. Knudsen, J. (1990). Acyl-CoA-binding protein (ACBP)and its relation to fatty acid-binding protein (FABP):an overview. Mol. Cell. Biochem. 98, 217–223.

26. Thomsen, J. K., Kragelund, B. B., Teilum, K., Knudsen,J. & Poulsen, F. M. (2002). Transient intermediarystates with high and low folding probabilities in theapparent two-state folding equilibrium of ACBP atlow pH. J. Mol. Biol. 318, 805–814.

27. Hazra, P., Chakrabarty, D., Chakraborty, A. & Sarkar,N. (2004). Probing protein–surfactant interaction bysteady state and time-resolved fluorescence spectro-scopy. Biochem. Biophys. Res. Commun. 314, 543–549.

28. Kalyanasundaram, K. & Thomas, J. K. (1976). Envir-onmental effects on vibronic band intensities inpyrene monomer fluorescence and their applicationin studies of micellar systems. J. Am. Chem. Soc. 99,2039–2044.

29. De, S., Girigoswami, A. & Das, S. (2005). Fluorescenceprobing of albumin–surfactant interaction. J. ColloidInterface Sci. 285, 562–573.

30. Semisotnov, G. V., Rodinova, N. A., Razgulyaev, O. I.,Uversky, V. N., Gripas, A. F. & Gilmanshin, R. I.(1991). Study of the “molten globule” intermediatestate in protein folding by a hydrophobic fluorescentprobe. Biopolymers, 31, 119–128.

31. Mogensen, J. E., Wimmer, R., Larsen, J. N., Spangfort,M. D. & Otzen, D. E. (2002). The major birch allergen,Bet v 1, shows affinity for a broad spectrum ofphysiological ligands. J. Biol. Chem. 277, 23684–23692.

32. Stutz, H., Wallner, M., Malissa, H., Jr., Bordin, G. &Rodriguez, A. R. (2005). Detection of coexistingprotein conformations in capillary zone electro-phoresis subsequent to transient contact withsodium dodecyl sulfate solutions. Electrophoresis,26, 1089–1105.

33. Gudiksen, K. L., Gitlin, I. & Whitesides, G. M. (2006).Differentiation of proteins based on characteristicpatterns of association and denaturation in solutionsof SDS. Proc. Natl Acad. Sci. USA, 103, 7968–7972.

34. Nielsen, A. D., Arleth, L. & Westh, P. (2005). Analysisof protein–surfactant interactions—a titration calori-metric and fluorescence spectroscopic investigation ofinteractions between Humicola insolens cutinase andan anionic surfactant. Biochim. Biophys. Acta, 1752,124–132.

35. Nielsen, M. M., Andersen, K. K., Westh, P. & Otzen,D. E. (2007). Unfolding of β-sheet proteins in SDS.Biophys. J. 92, 3674–3685.

36. Nielsen, A. D., Arleth, L. & Westh, P. (2005). Interac-tions of Humicola insolens cutinase with an anionicsurfactant studied by small-angle neutron scatteringand isothermal titration calorimetry. Langmuir, 21,4299–4307.

37. Nielsen, A. D., Borch, K. & Westh, P. (2007). Thermalstability of Humicola insolens cutinase in aqueous SDS.J. Phys. Chem., B, 111, 2941–2947.

38. Draper, M., Savage, M., Collett, J. H., Attwood, D.,Price, C., Booth, C. et al. (1995). Solubilisation of drugsin micellar systems studied by eluent gel permeationchromatography. Pharm. Res. 12, 1231–1237.

39. Pedersen, J. S., Hansen, S. & Bauer, R. (1994). Theaggregation behavior of zinc-free insulin studied bysmall-angle neutron scattering. Eur. Biophys. J. 23,379–389.

40. Glatter, O. (1977). A new method for the evaluation ofsmall-angle scattering data. J. Appl. Crystallogr. 10,415–421.


41. Pedersen, J. S. (1997). Analysis of small-angle scatteringdata from polymeric and colloidal systems: modellingand least-squares fitting. Adv. Colloid Interface Sci. 70,171–201.

42. Hansen, S. (1990). Calculation of small-angle scatter-ing profiles using Monte-Carlo simulation. J. Appl.Crystallogr. 23, 344–346.

43. Spinozzi, F., Carsughi, F., Mariani, P., Teixeira, C. V. &Amaral, L. Q. (2000). SAS from inhomogeneousparticles with more than one domain of scatteringdensity and arbitrary shape. J. Appl. Crystallogr. 33,556–559.

44. Svergun, D. I., Barberato, C. & Koch, M. H. (1995).CRYSOL—a program to evaluate X-ray solutionscattering of biological macromolecules from atomiccoordinates. J. Appl. Crystallogr. 28, 768–773.

45. Vass, S., Plestil, J., Laggner, P., Gilanyi, T., Borbely, S.,Kriechbaum, M. et al. (2003). Models of micellarstructure tested by SANS and SAXS (from a Kratkycamera) in cesium dodecyl sulfate solution. J. Phys.Chem., B, 107, 12752–12761.

46. Walkenhorst, W. F., Marzlyakov, M., Hristova, K. &Wimley, W. C. (2009). Polar residues in transmem-brane helices can decrease electrophoretic mobility inpolyacrylamide gels without causing helix dimeriza-tion. Biochim. Biophys. Acta, 1788, 1321–1331.

47. Ibel, K., May, R. P., Kirschner, K., Szadkowski, H.,Mascher, E. & Lundahl, P. (1990). Protein-decoratedmicelle structure of sodium-dodecyl-sulfate–proteincomplexes as determined by neutron scattering. Eur. J.Biochem. 190, 311–318.

48. Takagi, T., Tsujii, K. & Shirahama, K. (1975). Bindingisotherms of sodium dodecyl sulfate to proteinpolypeptides with special reference to SDS–polya-mide gel electrophoresis. J. Biochem. 77, 939–947.

49. Shirahama, K., Tsujii, K. & Takagi, T. (1974). Free-boundary electrophoresis of sodium dodecyl sulfate–protein polypeptide complexes with special referenceto SDS–polyacrylamide gel electrophoresis. J. Biochem.75, 309–319.

50. Chatterjee, A., Moulik, S. P., Majhi, P. R. & Sanyal, S. K.(2002). Studies on surfactant–biopolymer interaction: I.Microcalorimetric investigation on the interaction ofcetyltrimethylammoniumbromide (CTAB) and sodiumdodecylsulfate (SDS) with gelatin (Gn), lysozyme (Lz)and deoxyribonucleic acid (DNA). Biophys. Chem. 98,313–327.

51. Kauzmann, W. (1959). Some factors in the interpreta-tion of protein denaturation. Adv. Protein Chem. 14,1–63.

52. Pitt-Rivers, R. & Impiombato, F. (1968). Binding ofsodium dodecyl sulphate to various proteins. Biochem.J. 109, 825–830.

53. Bagger, H. L., Hoffmann, S. V., Fuglsang, C. C. &Westh, P. (2007). Glycoprotein–surfactant interactions:a calorimetric and spectroscopic investigation of thephytase–SDS system. Biophys. Chem. 129, 251–258.

54. Chatterjee, A., Moulik, S. P., Majhi, P. R. & Sanyal, S. K.(2002). Studies on surfactant–biopolymer interaction: I.Microcalorimetric investigation on the interaction of

cetyltrimethylammoniumbromide (CTAB) and sodiumdodecylsulfate (SDS) with gelatine (Gn), lysozyme (Lz)and deoxyribonucleic acid (DNA). Biophys. Chem. 98,313–327.

55. Lad, M. D., Ledger, V. M., Briggs, B., Green, R. J. &Frazier, R. A. (2003). Analysis of the SDS–lysozymebinding isotherm. Langmuir, 19, 5098–5103.

56. Otzen, D. E. & Oliveberg, M. (2002). Burst-phaseexpansion of native protein prior to global unfoldingin SDS. J. Mol. Biol. 315, 1231–1240.

57. Tofani, L., Feis, A., Snoke, R. E., Berti, D., Baglioni, P. &Smulevich, G. (2004). Spectroscopic and interfacialproperties ofmyoglobin/surfactant complexes.Biophys.J. 87, 1186–1195.

58. Pedersen, J. S., Christiansen, G. & Otzen, D. E. (2004).Modulation of S6 fibrillation by unfolding rates andgatekeeper residues. J. Mol. Biol. 341, 575–588.

59. Dombi, G. W. & Halsall, H. B. (1985). Collagen fibrilformation in the presence of sodium dodecyl sulphate.Biochem. J. 228, 551–556.

60. Svanborg, C., Ågerstam, H., Düringer, C., Fischer, W.,Gustafsson, L., Hallgren, O. et al. (2003). HAMLETkills tumor cells by apoptosis—cellular, molecular andtherapeutic aspects. Adv. Cancer Res. 88, 1–29.

61. Mandrup, S., Højrup, P., Kristiansen, K. & Knudsen, J.(1991). Gene synthesis, expression in Escherichia coli,purification and characterization of the recombinantbovine acyl-CoA-binding protein. Biochem. J. 276,817–823.

62. Pedersen, J. S. (2004). A flux- and background-opti-mized version of the NanoSTAR small-angle X-rayscattering camera for solution scattering. J. Appl.Crystallogr. 37, 369–380.

63. Chacón, P., Morán, F., Díaz, J. F., Pantos, E. & Andreu,J. M. (1998). Low-resolution structures of proteins insolution retrieved from X-ray scattering with a geneticalgorithm. Biophys. J. 74, 2760–2775.

64. Chacón, P., Díaz, J. F., Morán, F. & Andreu, J. M.(2000). Reconstruction of protein form with X-raysolution scattering and a genetic algorithm. J. Mol.Biol. 299, 1289–1302.

65. Svergun, D. I. (1999). Restoring low resolutionstructure of biological macromolecules from solutionscattering using simulated annealing. Biophys. J. 76,2879–2886.

66. Svergun, D. I., Petoukhov, M. V. & Koch, M. H. J.(2001). Determination of domain structure of proteinsfrom X-ray solution scattering. Biophys. J. 80,2946–2953.

67. Walther, D., Cohen, F. E. & Doniach, S. (2000).Reconstruction of low-resolution three-dimensionaldensity maps from one-dimensional small-angle X-ray solution scattering data for biomolecules. J. Appl.Crystallogr. 33, 350–363.

68. Vigil, D., Gallagher, S. C., Trewhella, J. & Garcia, A. E.(2001). Functional dynamics of the hydrophobic cleftin the N-domain of calmodulin. Biophys. J. 80,2082–2092.

69. Hayter, J. B. & Penfold, J. (1981). An analytic structurefactor for macroion solutions. Mol. Phys. 42, 109–118.

the role of decorated sds micelles in sub-cmc protein denaturation and association

Documents