Coacervates of Lactotransferrin and β- or κ‑Casein: StructureDetermined Using SAXSC. G. (Kees) de Kruif,*,†,∥ JanSkov Pedersen,‡ Thom Huppertz,§ and Skelte G. Anema∥

†Van’t Hoff laboratory for Physical and Colloid Chemistry, Padualaan 8, Utrecht University, The Netherlands‡Department of Chemistry and Interdisciplinary Nanoscience Center (iNANO), University of Aarhus, Langelandsgade 140,DK-8000 Århus, Denmark§NIZO food research, PO Box 20, 6710 BA, Ede, The Netherlands∥Fonterra Research Centre, Private Bag 11029, Dairy Farm Road, Palmerston North, New Zealand

ABSTRACT: Lactotransferrin (LF) is a large globular protein in milk with immune-regulatory and bactericidal properties. At pH 6.5, LF (M = 78 kDa) carries a net(calculated) charge of +21. β-Casein (BCN) and κ-casein (KCN) are part of thecasein micelle complex in milk. Both BCN and KCN are amphiphillic proteins witha molar mass of 24 and 19 kDa and carry net charges of −14 and −4, respectively.Both BCN and KCN form soap-like micelles, with 40 and 65 monomers, respec-tively. The net negative charges are located in the corona of the micelles. On mixingLF with the caseins, coacervates are formed. We analyzed the structure of thesecoarcervates using SAXS. It was found that LF binds to the corona of the micellarstructures, at the charge neutrality point. BCN/LF and KCN/LF ratios at thecharge neutrality point were found to be ∼1.2 and ∼5, respectively. We think thatthe findings are relevant for the protection mechanism of globular proteins in bodilyfluids where unstructured proteins are abundant (saliva). The complexes will pre-vent docking of enzymes on specific charged groups on the globular protein.

I. INTRODUCTION

The interaction and self-assembly of oppositely charged proteinsis a topic of recent research. Historically, it starts with Tiebackx1

who first reported on the coacervation of gum Arabic (GA) andgelatin. However, it was Bungenberg de Jong and Kruyt2 whomade the first systematic investigation into the phenomenon,using a positively charged protein (gelatin) and the negativelycharged polysaccharide GA. The interaction of oppositely charged(bio)polymers has an important application in encapsulation.Again, Bungenberg de Jong and Kruyt2 were first to showencapsulation of carbon particles and dyes in the GA/gelatinsystem. The National Cash Register company developedcarbonless copy paper in the early 1950′s based on this system.Nowadays, for encapsulation, the so-called layer by layertechnology is often used.3 Recent reviews on the interaction ofoppositely charged (bio)polymers are presented by Izmudov4

and in a special issue of Sof t Matter entitled “Polyelectrolytes inbiology and soft matter”.5

In Figure 1, we present a schematic illustration of the maintypes of coacervates. Systems containing a polyelectrolyte/polysaccharide and an oppositely charged protein or colloid havebeen extensively researched and were reviewed by De Kruifet al.,6 Schmitt and Turgeon,7 and, more recently, Kayitzmayer.8

In general, complexes and coacervate phases are formed whenthe two colloids are on the opposite side of the isoelectric point(IEP) and at low(er) ionic strength. Proteins are patchy colloidsand they may even form coacervates at the “wrong” side of the

Received: June 17, 2013Revised: July 16, 2013Published: July 16, 2013

Figure 1. Schematic illustration of the morphology of variouscoacervates. Red color: positively charged; blue color: negativelycharged (or vice versa). A: chromatin, the DNA-histone complex(from Arcesi et al.15,17). B: gum Arabic + β-lactoglobulin (fromWeinbreck et al.10). C: gum Arabic + gelatin (from Bungenberg deJong and Kruyt2). D: BCN + lactotransferrin (from Pan et al.,14

Anema and De Kruif,12 and this paper). E: lysozyme in a poly-electrolyte brush (from Witteman and Ballauf18). F: complexcoacervate core micelles (from Voets et al.19) or surfactant micelles+ polyeletrolyte (from Chiappisi et al.20). G: ionic crystals madeof PMMA spheres (from Leunissen et al.21). H: (fractal) aggregatesof lysozyme and β-lactoglobulin (from Howell et al.22) or silicaparticles (from Kim and Berg,23 Rasa et al.24). Note: drawings are notto scale.

Article

pubs.acs.org/Langmuir

© 2013 American Chemical Society 10483 dx.doi.org/10.1021/la402236f | Langmuir 2013, 29, 10483−10490

IEP.9 Small proteins (colloids) may decorate the polysaccharideas depicted by Weinbreck et al.10

Croguennec and co-workers investigated the formation ofhierarchical self-assembly of α-lactalbumin with lysozyme.11

Supramolecular structures are formed from tetrahedral proteinbuilding blocks containing two lysozyme and two α-lactalbuminmolecules interacting through their oppositely charged patches.Interestingly any further supramolecular assembly seems to bedriven by hydrophobic interactions.Anema and De Kruif12,13 studied the interaction between

folded proteins (lactotransferrin and lysozyme) with unstruc-tured proteins (e.g., αS-casein). These proteins form coacervatesand lactotransferrin forms stable (overcharged) associates oneither side of the charge neutrality point. αS-Casein can beconsidered as poly(ampho)eletrolyte and then there is a greatsimilarity with the coacervation of polysaccharides andoppositely charged colloids or proteins. Pan et al.14 interactedlysozyme with β-casein (BCN) and subsequently denatured thelysozyme. If the proteins (colloids) are large enough, the poly-electrolyte may wind around the protein. For instance, in thechromatin complex, DNA winds around the histones,15

therewith effectively folding the DNA into the cell. The foldingcomes with an energetic penalty, i.e., folding of the stiff poly-electrolyte chain. Park et al.,16 Arcesi et al.,17 and Anema and DeKruif13 showed that one can add small amounts of lysozyme to BCN(at low salt content) without flocculation; however, adding BCN tolysozyme flocculated the system and it was suggested that this wasdue to the folding energy penalty of BCN around the very smalllysozyme. If the polyelectrolyte forms a brush, e.g., by grafting to thesurface of a colloid or because it has a hydrophobic tail, then theproteins are attached to the charged polymer corona, as shown byWitteman and Ballauff,18 and in this work. Voets et al.19 reviewed thework on complex coacervate coremicelles inwhich diblock polymershave a neutral tail, but an oppositely charged head. These systemsform coacervates but cannot grow because of the neutral tail.Biesheuvel et al.25 studied the interaction between two folded

proteins, i.e., lysozyme and succinylated lysozyme which can becompared with the interaction of oppositely charged colloidparticles, e.g., by Kim and Berg23 and Rasa et al.24 The work onoppositely charged colloids shows rich phase behavior, fromfractal aggregates to ordered crystalline structures. Both Rasaet al.24 and Leunissen et al.21 used low dielectric solvents, inwhich the product of the Debye−Huckel screening parameter κ,and particle radius a, is large; i.e., κa ≫ 1. By manipulating theconditions, Leunissen et al.21 were able to mimic the crystallinestructures of simple salts. In addition to the long-range attractionthe PPMA particles have a short-range polymeric repulsion. Inthis paper we studied the adsorption of lactotransferrin to thepolyelectrolyte brush of BCN micelles.Lactotransferrin (LF) is a large (78 kDa) protein with pI ≈

8.32 and has immune regulatory and bacteriostatic proper-ties.26,27 BCN is an unstructured protein of 209 amino acids frommilk. It is amphiphilic with a molecular mass of 24 kDa and a pI≈5.6. The net number of charges on both proteins can be cal-culated using the primary sequence of the Swiss ProteinDatabase.28

Early investigations on the temperature driven hydrophobicself-association of BCN were made by Payens29 during the 1960susing light scattering, analytical ultracentrifugation, viscositymeasurements, and high pressure measurements. Kegeles30

proposed a so-called shell model for the self-association of BCN.Later, the work on BCN was extended often under varying anddifferent conditions.31,32 Interestingly, Leclerc and Calmettes33

measured second virial coefficients of the micelles and foundslightly negative values. This is consistent with the idea that theremust be an overall attractive (cooperative) interaction in order toform micelles. Mikheeva et al.34 made an extensive investigationof the thermodynamics of the micellization of BCN confirmingthe shell model proposed by Kegeles.30,35 The micellizationenthalpy is however endothermic, indicating that micellization isstrongly entropy driven.36

Thus, BCN is monomeric at temperatures below 5 °C andforms polymeric micelles containing about 40 monomers attemperatures above 30 °C. Details will depend on the buffersused and the addition of “chaotropes”. For example, guanidinehydrochloride or urea will dissociate the BCNmicelles. However,cross-linking the micelles using transglutaminase prevents thedissociation and leaves the micelle greatly intact.37,38 From smallangle scattering experiments, the size of the micelles in a 25 mMphosphate buffer is about 15 nm (hydrodynamic radius andinteraction radius) and the radius of gyration is 6 nm.32 Hence, itwas concluded that the micelles had a relatively dense core ofabout 7 nm radius and a fluffy outside, being the hydrophilicN-terminus of BCN. In a recent paper, the association of BCN(and KCN) was studied using SANS and SLS.39 The (truncated)scattering spectra were fitted to the sum of an oblate ellipsoid anda Debye function, and are generally consistent with previousstudies. Interestingly, Ossowski et al.39 studied the fibril forma-tion on standing and applying shear.On increasing the concentration, the BCNdispersions become

very viscous, but viscosity decreases exponentially with temper-ature and increases with ionic strength.40

BCN micelles have been claimed to have chaperone-likeproperties.41,42 It remains to be seen whether this is a realchaperone property, i.e., preventing proteins from aggregationduring unfolding/folding, or just the generic surfactant propertyof BCN in stabilizing hydrophobic particles like emulsion drop-lets. Actually, experiments on heating lysozyme in the presence ofBCN showed that refolding was inhibited in the presence ofBCN.43 We have used caseins to stabilize gold sols and wheyprotein particles (unpublished results), and BCN is known toabsorb small molecules such as vitamins and drugs.44

κ-Casein (KCN) is famous for its stabilization of the caseinmicelle. Often it is claimed to be of electrostatic origin but that isimprobable, because KCN caries only few net charges and, inaddition, the Debye−Huckel screening parameter in milk is only1 nm. A quantitative calculation of the steric, electrostatic, andvan der Waals potential was given by Tunier and De Kruif45

showing that electrostatic contributions are small. Numerouspublications discuss the interaction of KCN with the (heated)whey proteins or with added polysaccharides. There are fewpapers on the micellization of KCN. KCN forms micellarstructures, and the size and number of monomers (∼30) islargely independent of temperature and ionic strength, whichcontrasts with BCN. Much of the early work on KCN was doneby Vreeman and co-workers.46−49 Thurn et al.31 measured SANSspectra and concluded that KCN formed micellar structures justlike BCN, but these probably associated through (disulfide)bonding at the hydrophilic N-terminus. De Kruif and May50

measured SANS spectra of KCN micelles in which the disulfidebonds were reduced with dithiothreitol. They found structureswith a dense core and a fluffy corona with a radius of 12−15 nmdepending on the measurement method. Later, De Kruif et al.37

did further measurements using SANS confirming the earlierdata and showed, using NMR, that BCN micelles are less denseand more dynamic than KCN micelles. Ossowski et al. report a

Langmuir Article

dx.doi.org/10.1021/la402236f | Langmuir 2013, 29, 10483−1049010484

temperature independent association number of 30−35.39Hence, KCN appears to be a micelle with a dense core and a(very) fluffy corona. From a structural point of view, bothmicelles appear to be very similar.Often, micelles, such as those from BCN, are used to adsorb

materials in the hydrophobic core. However, for positivelycharged solutes adsorption in the corona of the micelles ispresent as well. Here, we investigate the adsorption of lacto-transferrin, a basic protein, into the corona of the micelles butalso at temperatures where the BCNmicelles are dissociated. Wewill use SAXS measurements to unravel the structure of thecoacervates of lactotransferrin with BCN or KCN. We willanalyze the SAXS spectra using an appropriate model for thesystem at hand and measure the scattering of both lactotransferrinand BCN (the last one also as a function of temperature).

II. EXPERIMENTAL METHODSExperimental Solutions. Lactotransferrin (97% purity) was

obtained from DMV International (Veghel, The Netherlands). KCN(>95% purity) was prepared according to the method of Leaver andLaw.51 BCN (>95% purity) was prepared according to the method ofPayens and Heremans.29 Proteins were dissolved in Milli-Q water(Millipore, MA, USA) containing 0.02% NaN3 and the pH was adjustedto 6.52−6.54. After filtering through a 0.2 μm filter, the concentrationwas measured by absorbance (A280 − 1.7 × A320). The A320 should be asmall number as it measures background scattering. Specific absorbancesat 280 nm:H2O, 0.001 cm

2.mg−1; lactotransferrin, 1.27 cm2.mg−1; KCN,0.95 cm2.mg−1; BCN, 0.46 cm2.mg−1. The concentrations were:lactotransferrin, 14.35 g/L; KCN, 14.93 g/L; BCN, 16.1 g/L.SAXS Measurements. A commercially available SAXS camera,

NanoSTAR (Bruker AXS, Karlruhe, Germany) with a pinholeconfiguration, was modified to optimize its use for weakly scatteringsolution samples and is described in detail by Pedersen.52

III. RESULTS AND DISCUSSIONLactotransferrin and BCN. First we will present the SAXS

results of the pure compounds. Figure 2 shows the scattering

curve of lactotransferrin. The protein structure as presented inthe Swiss Protein Data Bank28 shows a dumbbell structure.Therefore, data were fitted to a halter of two lobes and the resultis the drawn line in Figure 2. The radius of each lobe is 2.5 nmand the center to center distance is found to be 11 nm. The lobesize seems reasonable, but the center to center distance issomewhat large and probably indicates the presence of dimers orsmall aggregates, as suggested by the SLS data (Figure 2).

Figure 3 shows a plot of BCN scattering at temperaturesbetween 2 and 30 °C for BCN concentrations of 4 and 16.1 g/L.The spectra were the same for increasing or decreasing tempera-tures and completely reproducible. It is known that BCN formsmicelles at ambient temperatures34,53 but is predominantlymonomeric at 2 °C. For such behavior, one would expect anincreasing low Q scattering with increasing temperature as aresult of increasing particle number and size. However, thespectra show a very peculiar behavior, namely, that they arevirtually unchanged at both low and high Q. How the calculatedcurves were obtained will be explained below.It was our aim to analyze the data in a quantitative way, so as to be

able to point out major changes in the system, with an eye on what isphysically reasonable given other evidence. We experimented withusing a polydisperse form factor for themicelles. First, this introducesadjustable parameters, and second, we found that using a so-calledgeneralized expression (Beaucage function54) allowed a goodquantification of the data with good physical significance. Thegeneralized expression assumes one ormore structural elements (seeeq 1) in the scattering objects and is then fitted to the experimentalscattering spectrum. We aim to use the fitted parameters to indicatetrends and changes, rather than correlating them to exact physicalvalues. We intend to get a grasp on the formation of the complexesrather than analyzing them in all possible detail. For that, further andmore detailed data will be needed.In view of the pronounced correlation, we first determined

experimental “structure factors” by dividing the experimentaldata by the lowest temperature scattering. This gives a function(not shown) that is similar to a structure factor of an excludedparticle interaction. We determined the relative increase inamplitude of the scattering spectra at a few wave vector values, inparticular at low Q-values and at the maximum of the correlationpeak (Figure 4).The lowQ-values vary only slightly with temperature, whereas

the value of S(Q = Qmax) increases with temperature. The highconcentration data show a clear sigmoidal increase attributed toan increasing micellization, which is then fully consistent withSANS data of Leclerc and Calmettes,55 calorimetry data ofMikheeva and Grinberg,34 and dynamic light scattering results ofO’Connell, Grinberg, and De Kruif,32 O’Connell and De Kruif,38

and De Kruif and Grinberg.53 The value ofQmax = 310 Å−1 and is

independent of temperature and the same for the two con-centrations. It must be remarked that there are some quantitativedifferences with the references mentioned. We attribute this tothe fact that we did not use a 25 mM sodium phosphate buffer inthis investigation. Adding salt further promotes micellization(data not shown), but we intended to investigate the bindingof lactotransferrin and therefore did not add salt as this wouldreduce the interactions.The (excluded volume) correlation peak at Q = 310 Å−1

corresponds to a half-distance (radius) of 20 nm. This amplitudeincrease is to be expected when volume fraction, φ, of the micellesincreases at the expense of monomers. For noninteractingparticles, the ordinate would be proportional to φ. For interactingparticles, the ordinate would be proportional to φ × S(Q = 0, φ),where S(Q, φ) is the structure factor of the interacting particles,assuming that the form factor P(Q) does not change.For a hard sphere or repulsive particle system, S(Q = 0, φ)

decreases exponentially with increasing φ. For an attractiveinteraction, SAhs(Q = 0, φ) increases with increasing attractionand eventually diverges at phase separation.We therefore used anadhesive hard sphere (Ahs) structure factor with the condition thatφ × SAhs(Q = 0.01 Å−1, φ) is constant. This is realized for a second

Figure 2. Scattering spectrum of lactotransferrin. The low-Q data arefrom light scattering. Data were fitted to a dumbbell structure.

Langmuir Article

dx.doi.org/10.1021/la402236f | Langmuir 2013, 29, 10483−1049010485

virial coefficientB2/Vhs =−2.36, whereVhs is the hard sphere volume4/3 π (20 nm)3. After some preliminary data fitting, it appeared thatB2/Vhs = 0 gave an excellent fit of the data. It is interesting to notethat the effective second viral coefficient is zero, which is to beexpected for particles in equilibrium with a monomeric system. Theosmotic compressibility (S(Q = 0, φ)) of the particles must be verysmall. Adding particleswill tend to dissociate them intomonomers orthe increased number density will be compensated by attractiveinteractions and as a result the osmotic pressure does not increase.Leclerc and Calmettes55 measured second virial coefficients as afunction of temperature and found slightly negative values.We fitted the data to the following expression:

= ×

+ ×

Q a a Q a a

a Q a

FF( , ) ln[( IAHS( , , )

PDebye( , ))]0 1 2

3 4 (1)

= ×Q r a Q r a S QIAHS( , , ) Ihspol( , , ) ( )0 2 0 2 (2)

= × − ·

+·

⎡⎣⎢⎢

⎤⎦⎥⎥

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

( )( )

Q R a a QR

aQ

Q

Ihspol( , , ) exp( )

3

erf

g

R

1 12 g

2

26

3 P1g

(3)

Equation 3 is the Beaucage result.54 Thus the fitting functionFF(Q, a) is the sum of an adhesive hard sphere contributionIAHS(Q, r0, a2) including a structure factor S(Q). The symbolsrepresent the following: a is a matrix containing elements a0 to a4,where a0, a1, and a3 are amplitudes; a2 is the radius of gyration ofthe particles, Rg; and a4 is the polymer radius of gyration, Rgp.In view of the presence of BCN polymers, we added a Debye

polymer scattering term PDebye(Q, Rgp) under the assumptionthat the two contributions are not correlated. For the structurefactor S(Q, φ), we used the expressions of Menon et al.56 for thecalculation of the structure factor of adhesive hard spheres.Volume fraction φ increased with temperature (2 to 30 °C)from 0.12 to 0.2 for the 16.1 g/L data, which corresponds to avoluminosity of 12 mL/g, a reasonable value in view of the lowsalt concentration.. We found a value of 3.5 mL/g in a phosphatebuffer.34 Panouille et al.40 observed a divergence of the viscosityat 90 g/L (in salt buffer), which would correspond to avoluminosity of 0.63/0.09 = 7 mL/g. Sood et al.57,58 found fromsedimentation experiments a value of 3 mL/g for the monomer.These voluminosity results indicate that the BCN micelles arequite fluffy objects. Here we used a voluminosity of 12 mL/gfor all the experiments. For the 4 g/L data, we divided volumefractions of the 16.1 g/L data by 4. With these values, we fittedthe experimental data to eq 1.Equation 3 contains 5 fit parameters: a0, the amplitude of the

Guinier function; a1, the radius of gyration; a2 the amplitude ofthe power law function, which for a given structure should beapproximately constant. In fitting the data, we found that thecontribution of the power law function was very small. Actually,small negative values were also found, which is not physicallyrealistic; therefore, we set a2 = 0. The amplitude of the Debyefunction is a3, and a4 is the radius of gyration of a polymer chain.P1 is the power law exponent and should be 4 in the Porod limit.We used P1 = 4, but results are rather insensitive to the precisevalue of P1. It appeared that a value of τ, the “adhesiveness”parameter, in the expressions of S(Q, φ) of 0.25 gave a very goodfit of the data, especially the low concentration data. Setting thepotential well width to 0.1 then gives a value for B2/Vhs = 0. Thisvalue was used for all calculations and the volume fraction wasdetermined by varying the temperature. The correlation lengthwas set to 20 nm because of the position of the correlation peakin Figure 3. Therefore, the fit parameters for the data in Figure 3were the size (Rg) of the micelles, the radius of gyration of the

Figure 3. Scattering spectra of BCN at temperatures between 2 and 30 °C. Correlation peak at intermediate wave vectors increases with temperature.(A) BCN 4 g/L. (B) BCN 16.1 g/L. The fitted spectra (drawn lines, see text) were offset by a factor of 4, for clarity.

Figure 4. Relative amplitudes of the scattering spectra at low Q-valuesand at the maximum of the correlation peak (S(Q =Qmax)) as a functionof temperature.○, S(Q = Qmax) 16.1 g/L; □, S(Q = Qmax) 4 g/L; closedsymbols experimental S(Q)-values at the three lowest Q-values.

Langmuir Article

dx.doi.org/10.1021/la402236f | Langmuir 2013, 29, 10483−1049010486

polymer, and the scattering amplitudes. Equation 1 gave a verygood fit to the data.We evaluated the fit parameters of both samples by plotting

them as a function of temperature (Figure 5). The amplitude a0

increases with temperature indicating the increased number ofmicelles, but the amplitude will also depend on the precise natureof the interparticle (20 nm radius) interaction. We observed thatthe Rg of the particle is very small (about 4 and 2 nm), but it isquestionable whether this has any physical significance, as thelow Q-data are multiplied by the S(Q) values. It is noted that thecalculated spectra show a much more detailed variation due tothe use of a monodisperse structure factor S(Q, φ). Poly-dispersity in size and the distribution in wavelengths of the SAXSwill effectively smooth the data. From SANS measurements (in25 mM sodium phosphate buffer) we found a Rg of 8 nm and aninterparticle radius of 15 nm.37 Fitting the data with the Beaucagefunction only gives a Rg of 12 nm. In any case, the BCN micellesare rather fluffy and mass is concentrated in their core, as onewould expect.Summarizing the results of Figures 3−5, we can say that the

BCN (in low salt conditions) forms fluffy micelles with aninteraction radius of 20 nm. In phosphate buffer the interaction

radius was 15 nm.34 The volume fraction (number) of themicelles increases up to 30 °C. Themicelles have a small radius ofgyration and a large specific volume with a compact core and afluffy “hairy” outside. In order to resolve the precise massdistribution in the BCN micelles, it would be required to docontrast variation using a technique such as SANS.

BCN into Lactotransferrin. We titrated BCN intolactotransferrin and measured the scattering spectra at eachcomposition. The results are presented in Figure 6 and the drawnlines are fits to the data using the Beaucage function with two“levels”. For comparison we have plotted the spectra of BCN andlactotransferrin at 25 °C as well.In Figure 7 we plot the coefficients of the fit. The data were fit

to a Beaucage function with two Guinier and two power law

regimes. The fitting function FF(Q, a) is given by eq 4, followingPignon et al.59

= × − × +×

+ × − × +×

⎡

⎣

⎢⎢⎢⎢⎢

⎡

⎣

⎢⎢⎢⎢⎢

⎡⎣⎢

⎤⎦⎥

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥⎡

⎣

⎢⎢⎢⎢

⎡⎣⎢

⎤⎦⎥

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

⎤

⎦

⎥⎥⎥⎥

⎤

⎦

⎥⎥⎥⎥⎥

⎤

⎦

⎥⎥⎥⎥⎥

( )

( )

( )

( )

Q a a Qa

aQ

Q

a Qa

aQ

Q

FF( , ) ln exp( )

3

erf

exp( )

3

erf

a

a

02 1

2

26

3 P1

32 4

2

56

3 P2

1

4

(4)

In total there are 6 fit parameters. a1 and a4 are the radii ofgyration. The other parameters are amplitudes. We plot the fitparameters in Figure 7 as function of the mixing mole fraction.Only the first amplitude changes with the molar fraction of

BCN, which means that the number of particles increases up to amaximum. That maximum is at a mole fraction of ∼0.55. Inaddition, the scattering intensity at the maximum is 46 timeshigher than that of lactotransferrin. Since the weight concen-tration is almost constant during titration (BCN is 16.1 g/L andlactotransferrin is 14.3 g/L) it means that the molar mass ofthe aggregate is approximately 46 × 78 kDa = 3588 kDa. Ifthe aggregate contains n molecules, than (0.45 × n × 78 kDa) +(0.55 × n × 24 kDa) = 3588 Da and it follows that n is about 74.Hence, each complex contains 33 lactotransferrin and 41 BCNmolecules. This is an interesting result, as the number of BCNmonomers in a pure BCN micelle is on the order of 30 to40.32,34,53 This result seems to suggest that lactotransferrin binds into“the hairs’’ of the BCNmicelle (Figure 8 gives an artist impression).

Figure 5. Plot of the fit parameters (of Figure 3, BCN) as a function oftemperature. For the 4 g/L (left pane) and 16.1 g/L (right pane): 0,amplitude of the Guinier function;Δ, Rg of sphere;●, amplitude Debyefunction; +, Rg Debye. The amplitude of the power law function was setto zero.

Figure 6. Plot of the scattering spectra on titrating BCN intolactotransferrin at 25 C. For comparison we plotted the scatteringspectra of BCN and lactotransferrin as well. Lines are fits through thedata. Low Q scattering is maximal at equimolar composition. Fitparameters are presented in Figure 7.

Figure 7. Plot of the fit parameters of eq 4 (spectra of Figure 6) as afunction of the mole fraction BCN.○, amplitude Guinier;△, Rg1 (nm);□, amplitude erf1; + amplitude second Guinier level; ▲, Rg2 (nm); ●,amplitude erf2.

Langmuir Article

dx.doi.org/10.1021/la402236f | Langmuir 2013, 29, 10483−1049010487

This is similar to the interaction depicted by Jackler et al.60 forbovine serum albumin when bound to spherical polyelec-trolyte brushes composed of a polystyrene core and apoly(acrylic acid) shell.We measured the changes in scattering intensity at mole

fraction x ≈ 0.5, so close to the electroneutrality point, as afunction of temperature (Figure 9). The data fitting was good

with the exception at the lowest Q-values. Clearly the Beaucagefunction cannot “depress” scattering intensity at lowQ due to theparticle interactions which are incorporated in the structurefactor. Nevertheless, it is of interest to present the fit data inFigure 9 together with the experimental low Q scattering (weaveraged the 4 lowest Q-values).The averaged experimental low Q values show a clear dis-

sociation of the micelles when going to lower temperature(Figure 10). Curve fit is a sigmoidal function. Of the 6 fitparameters of the two-level Beaucage function, we notice that thefirst amplitude follows the experimental data, but due to the poorfit, at low Q it is above the experimental amplitude. The Rg1 is12 nm and larger than the value from the combined Beaucage-Debye fit. It illustrates that parameters are correlated and dependon the chosen fit function.Obviously the complexes dissociate and probably into

monomeric complexes, i.e., one BCN plus one lactotransferrinat lower temperatures. The micellization of BCN is promoted bythe binding of the lactotransferrin as the dissociation temper-ature is shifted to lower temperatures after binding. For pureBCN (Figure 4), the inflection point is at 14 °C, but in the pres-ence of lactotransferrin, the inflection point is at 9 °C (Figure 10).

Lactotransferrin into KCN.We titrated lactotransferrin intoKCN and the results were very similar to those for lactotransferrininto BCN. Therefore, we will not repeat the discussion, but pres-ent the results only. The scattering spectra are given in Figure 11,

and in Figure 12 we present the fit coefficients as a function ofmole fraction.The scattering intensity at the max (x ≈ 0.83) is about

80/2.783 = 28.7 times higher than for lactotransferrin. So, themolar mass of the complex is 78 × 28.7 = 2242 kDa. The molarfraction of KCN in the complexes is 0.83, so there are about5 KCN molecules per lactotransferrin molecule. It follows that5 × n × 19 + n × 78 = 2242, which gives n = 13. Thus, we have65 KCN molecules binding with 13 lactotransferrin proteins.The molar mass of a KCN micelle would then be 1235 kDa,compared to 840 kDa for BCN. The Rg of KCN and BCN isvirtually the same at 14 nm. This implies that KCN micelles aremore compact than BCN micelles. The micellization of KCN isnot temperature dependent while that of BCN is, and so onewould expect a stronger interaction and more compact structurewithin the KCN micelle.

Charge Neutrality of the Complexes.We downloaded theamino acid composition of BCN and KCN from the SwissProtein Data Bank.28Wemodified the sequence in that we added

Figure 8. Cartoon of the binding of lactotransferrin to BCN micelles.Lactotransferrin is depicted as a dumbbell according to protein structurein the Swiss Protein Data Bank.28

Figure 9. Scattering spectra of lactotransferrin-BCN complexes as afunction of wave vector at temperatures between 5 and 30 °C. Thesample was equimolar (x ≈ 0.5). A few fit curves are included.

Figure 10. Scattering intensities at the 4 lowestQ-values as a function oftemperature, and the values of the fit parameters from eq 4. The samplewas equimolar (x ≈ 0.5) lactotransferrin and BCN. Panel A, all data;panel, B detail of panel A.○, experimental scattering intensities at lowestQ-values; □, amplitude of Guinier fit; ◊, Rg1; □, a2, amplitude powerlaw; ▼, a3, amplitude of Guinier 2 fit; ⧫, a4, Rg2; ■, a5, amplitude powerlaw 2.

Figure 11. Scattering spectra of LF-KCN complexes and fits of eq 4. Forcomparison also the scattering functions of KCN and LF are presented(the lower scattering spectra).

Langmuir Article

dx.doi.org/10.1021/la402236f | Langmuir 2013, 29, 10483−1049010488

a typical phosphorylation to the serines. In addition we addedglycosylated threonil groups to KCN. Subsequently, we cal-culated the number of charges as a function of pH using theSCRIPPS protein calculator (Figure 13). For lactotransferrin we

find that at pH 6.5 there would be 21 net positive charges. ForBCN there are 14 net negative charges and for KCN 0 to 6 netnegative charges depending on phosphorylation and glycosyla-tion levels. It is interesting to note that the hydrophilicN-terminus half of the BCN molecule contains all the net negativecharges, while the hydrophobic half is charge-neutral.The lactotransferrin−BCN complex contained 41/34 = 1.2

BCN/lactotransferrin. So indeed the charges on BCN justcompensate the surface charges on lactotransferrin. The averagecharge density of KCN is somewhat uncertain, but somewherebetween 0 and −6 at pH 6.5. From the molecular mass calcula-tions we find 5 KCN/lactotransferrin at charge neutrality and thissuggests that KCNcarries 4.2 negative charges at pH 6.5, which is arealistic value. These charges are in the C-tminus of the molecule.Summary. BCN is mainly monomeric at 2 °C and pH 6.5.

On increasing temperature, micellar structures are formed, whichcontain 40 monomers. Addition of lactotransferrin, a strongcationic protein, leads to the formation of protein−protein com-plexes. At low temperature the complexes stay in solution. Athigher temperature the BCN forms micellar structures and thelactotransferrin is attached to the micelle structure.

From SAXS measurements we concluded that 1.2 BCN bind 1lactotransferrin when charge neutrality is reached. Themicellization of BCN is promoted by the binding of the lacto-transferrin as the dissociation temperature is shifted to lowertemperatures. KCN micelles bind lactotransferrin as well and inway similar to BCN. However, KCN carries only about 4 netnegative charges and therefore 5 KCN are needed to neutralizethe 21 positive charges of lactotransferrin. It is not easy toimagine how the 5 KCN molecules are arranged on a relativelysmall protein measuring about 5 nm, even more so because KCNis arranged in micellar structures. In a study of the binding of LFto native casein micelles we found (JAFC 2013) that by increasingLF concentration from1% to 2%w/w the ratio increased from1 to 2.Adding higher levels may have increased the ratio further.Now, one may ask (as one of the reviewers rightfully did)

what’s the use of all this? We think it is extremely relevant tosee that a pair of structured and unstructured proteins behavesquite similarly to a structured protein and a (large) polelectrolytemolecule, e.g., gum Arabic, carrageenan, and DNA. Since thereare many unstructured proteins in body fluids (e.g., saliva) thereis a good reason to assume that these proteins bind to globularproteins as well and therewith serve protective functions, e.g.,against enzymatic degradation which is often based on dockingof charged groups on the protein. We are hesitant to call this achaperone activity as no folding is involved. More generally, thecomplexes could be used to protect proteins during processingand drug delivery, even more so because the unstructuredproteins can be cross-linked easily.

■ AUTHOR INFORMATIONCorresponding Author*E-mail: C.G.deKruif@UU.NL (C.G.(Kees) de Kruif).NotesThe authors declare no competing financial interest.

■ REFERENCES(1) Tiebackx, F. W. Gleichzeitige Ausflockung zweier Kolloide. Z.Chem. Ind. Kolloide 1911, 8, 198−201.(2) Bungenberg de Jong, H. G.; Kruyt, H. R. Coacervation (partialmiscibility on colloid systems). Proc. K. Akad. Wet-Amsterd. 1929, 32,849−856.(3) Lvov, Y.; Decher, G.; Mohwald, H. Assembly, structuralcharacterization, and thermal-behavior of layer-by-layer depositedultrathin films of poly(vinyl sulfate) and poly(allylamine). Langmuir1993, 9, 481−486.(4) Izumrudov, V. A. Soluble polyelectrolyte complexes ofbiopolymers. Polym. Sci., Ser. A 2012, 54, 513−520.(5) Rubinstein, M.; Papoian, G. A. Polyelectrolytes in biology and softmatter. Soft Matter 2012, 8, 9265−9267.(6) de Kruif, C. G.; Weinbreck, F.; de Vries, R. Complex coacervationof proteins and anionic polysaccharides. Curr. Opin. Colloid Interface Sci.2004, 9, 340−349.(7) Schmitt, C.; Turgeon, S. L. Protein/polysaccharide complexes andcoacervates in food systems. Adv. Colloid Interface Sci. 2011, 167, 63−70.(8) Kayitmazer, A. B.; Seeman, D.; Minsky, B. B.; Dubin, P. L.; Xu, Y.Protein-polyelectrolyte interactions. Soft Matter 2013, 9, 2553−2583.(9) de Vries, R.; Weinbreck, F.; de Kruif, C. G. Theory ofpolyelectrolyte adsorption on heterogeneously charged surfaces appliedto soluble protein-polyelectrolyte complexes. J. Chem. Phys. 2003, 118,4649−4659.(10) Weinbreck, F.; Rollema, H. S.; Tromp, R. H.; de Kruif, C. G.Diffusivity of whey protein and gum arabic in their coacervates.Langmuir 2004, 20, 6389−6395.(11) Bouhallab, S.; Croguennec, T. Spontaneous assembly andinduced aggregation of food proteins. Adv. Polym. Sci. In Press.

Figure 13. Net charge of the proteins as a function of pH using theSCRIPPS protein calculator. Lactotransferrin, +. The primary sequenceof BCN was modified by adding 5 phosphate groups (●). Similarly forKCN (△, primary chain) we added 1 phosphate and 5 sugar groups toKCN (○). The primary sequence contains the amino acids only.

Figure 12. Fitted parameters of eq 4 as a function of composition(mol fraction KCN). ○, a0, amplitude Guinier1; △, Rg1; ▲, Rg2. Otheramplitudes are small and constant.

Langmuir Article

dx.doi.org/10.1021/la402236f | Langmuir 2013, 29, 10483−1049010489

(12) Anema, S. G.; de Kruif, C. G. Co-acervates of lactoferrin andcaseins. Soft Matter 2012, 8, 4471−4478.(13) Anema, S. G.; de Kruif, C. G. Coacervates of lysozyme and β-casein. J. Colloid Interface Sci. 2013, 398, 255−261.(14) Pan, X.; Yu, S.; Yao, P.; Shao, Z. Self-assembly of β-casein andlysozyme. J. Colloid Interface Sci. 2007, 316, 405−412.(15)Widom, J. Structure, dynamics, and function of chromatin in vitro.Annu. Rev. Biophys. Biomol. Struct. 1998, 27, 285−327.(16) Park, S. Y.; Bruinsma, R. F.; Gelbart, W. M. Spontaneousovercharging of macro-ion complexes. Europhys. Lett. 1999, 46, 454−460.(17) Arcesi, L.; La Penna, G.; Perico, A. Generalized electrostaticmodel of the wrapping of DNA around oppositely charged proteins.Biopolymers 2007, 86, 127−135.(18) Wittemann, A.; Ballauff, M. Interaction of proteins with linearpolyelectrolytes and spherical polyelectrolyte brushes in aqueoussolution. Phys. Chem. Chem. Phys. 2006, 8, 5269−5275.(19) Voets, I. K.; de Keizer, A.; Cohen Stuart, M. A. Complexcoacervate core micelles. Adv. Colloid Interface Sci. 2009, 147−48, 300−318.(20) Chiappisi, L.; Hoffmann, I.; Gradzielski, M. Complexes ofoppositely charged polyelectrolytes and surfactants - recent develop-ments in the field of biologically derived polyelectrolytes. Soft Matter2013, 9, 3896−3909.(21) Leunissen, M. E.; Christova, C. G.; Hynninen, A. P.; Royall, C. P.;Campbell, A. I.; Imhof, A.; Dijkstra, M.; van Roij, R.; van Blaaderen, A.Ionic colloidal crystals of oppositely charged particles.Nature 2005, 437,235−240.(22) Howell, N. K.; Yeboah, N. A.; Lewis, D. F. V. Studies on theelectrostatic interactions of lysozyme with α-lactalbumin and β-lactoglobulin. Int. J. Food Sci. Technol. 1995, 30, 813−824.(23) Kim, A. Y.; Berg, J. C. Fractal heteroaggregation of oppositelycharged colloids. J. Colloid Interface Sci. 2000, 229, 607−614.(24) Rasa, M.; Philipse, A. P.; Meeldijk, J. D. Heteroaggregation,repeptization and stability in mixtures of oppositely charged colloids. J.Colloid Interface Sci. 2004, 278, 115−125.(25) Biesheuvel, P. M.; Lindhoud, S.; Stuart, M. A. C.; de Vries, R.Phase behavior of mixtures of oppositely charged protein nanoparticlesat asymmetric charge ratios. Phys. Rev. E 2006, 73.(26) Farnaud, S.; Evans, R. W. Lactoferrin - a multifunctional proteinwith antimicrobial properties. Mol. Immunol. 2003, 40, 395−405.(27) Jenssen, H.; Hancock, R. E. W. Antimicrobial properties oflactoferrin. Biochimie 2009, 91, 19−29.(28) Boeckmann, B.; Bairoch, A.; Apweiler, R.; Blatter, M. C.;Estreicher, A.; Gasteiger, E.; Martin, M. J.; Michoud, K.; O’Donovan, C.;Phan, I.; Pilbout, S.; Schneider, M. The SWISS-PROT proteinknowledgebase and its supplement TrEMBL in 2003. Nucleic AcidsRes. 2003, 31, 365−370.(29) Payens, T. A. J.; Heremans, K. Effect of pressure on temperature-dependent association of β-casein. Biopolymers 1969, 8, 335−345.(30) Kegeles, G. Shell-model for size distribution in micelles. J. Phys.Chem. 1979, 83, 1728−1732.(31) Thurn, A.; Burchard, W.; Niki, R. Structure of casein micelles 0.1.Small-angle neutron-scattering and light-scattering from β-casein and κ-casein. Colloid Polym. Sci. 1987, 265, 653−666.(32) O’Connell, J. E.; Grinberg, V. Y.; de Kruif, C. G. Associationbehavior of β-casein. J. Colloid Interface Sci. 2003, 258, 33−39.(33) Leclerc, E.; Calmettes, P. Structure of β-casein micelles. Physica B1998, 241, 1141−1143.(34) Mikheeva, L. M.; Grinberg, N. V.; Grinberg, V. Y.; Khokhlov, A.R.; de Kruif, C. G. Thermodynamics of micellization of bovine β-caseinstudied by high-sensitivity differential scanning calorimetry. Langmuir2003, 19, 2913−2921.(35) Kegeles, G. The critical micelle condition revisited. Indian J.Biochem. Biophys. 1992, 29, 97−102.(36) Portnaya, I.; Cogan, U.; Livney, Y. D.; Ramon, O.; Shimoni, K.;Rosenberg, M.; Danino, D. Micellization of bovine β-casein studied byisothermal titration microcalorimetry and cryogenic transmissionelectron microscopy. J. Agric. Food Chem. 2006, 54, 5555−5561.

(37) de Kruif, C. G.; Tuinier, R.; Holt, C.; Timmins, P. A.; Rollema, H.S. Physicochemical study of κ- and β-casein dispersions and the effect ofcross-linking by transglutaminase. Langmuir 2002, 18, 4885−4891.(38) O’Connell, J. E.; de Kruif, C. G. β-Casein micelles; cross-linkingwith transglutaminase. Colloids Surf., A 2003, 216, 75−81.(39) Ossowski, S.; Jackson, A.; Obiols-Rabasa, M.; Holt, C.; Lenton, S.;Porca, L.; Paulsson, M.; Nylander, T. Aggregation behavior of bovine κ-and β-casein studied with small angle neutron scattering, light scattering,and cryogenic transmission electron microscopy. Langmuir 2012, 28,13577−13589.(40) Panouille, M.; Durand, D.; Nicolai, T. Jamming and gelation ofdense β-casein micelle suspensions. Biomacromolecules 2005, 6, 3107−3111.(41) Zhang, X. F.; Fu, X. M.; Zhang, H.; Liu, C.; Jiao, W. W.; Chang, Z.Y. Chaperone-like activity of β-casein. Int. J. Biochem. Cell Biol. 2005, 37,1232−1240.(42) Thorn, D. C.; Meehan, S.; Sunde, M.; Rekas, A.; Gras, S. L.;MacPhee, C. E.; Dobson, C. M.; Wilson, M. R.; Carver, J. A. Amyloidfibril formation by bovine milk κ-casein and its inhibition by themolecular chaperones αs- and β-casein. Biochemistry 2005, 44, 17027−17036.(43) Wu, F.-G.; Luo, J.-J.; Yu, Z.-W. Unfolding and refolding details oflysozyme in the presence of β-casein micelles. Phys. Chem. Chem. Phys.2011, 13, 3429−3436.(44) Livney, Y. D. Milk proteins as vehicles for bioactives. Curr. Opin.Colloid Interface Sci. 2010, 15, 73−83.(45) Tuinier, R.; de Kruif, C. G. Stability of casein micelles in milk. J.Chem. Phys. 2002, 117, 1290−1295.(46) Vreeman, H. J. The role of κ-casein in dairy technology. OfficieelOrgaan FNZ 1973, 65, 1077−1080.(47) Vreeman, H. J.; Both, P.; Brinkhuis, J. A.; van der Spek, C.Purification and some physicochemical properties of bovine κ-casein.Biochim. Biophys. Acta 1977, 491, 93−103.(48) Vreeman, H. J. Association of bovine SH-κ-casein at pH 7.0. J.Dairy Res. 1979, 46, 271−276.(49) Vreeman, H. J.; Brinkhuis, J. A.; van der Spek, C. A. Someassociation properties of bovine SH-κ-casein. Biophys. Chem. 1981, 14,185−193.(50) de Kruif, C. G.; May, R. P. κ-Casein micelles: structure, interactionand gelling studied by small-angle neutron scattering. Eur. J. Biochem.1991, 200, 431−436.(51) Leaver, J.; Law, A. J. R. Preparative-scale purification of bovinecaseins on a cation-exchange resin. J. Dairy Res. 1992, 59, 557−561.(52) Pedersen, J. S. A flux- and background-optimized version of theNanoSTAR small-angle X-ray scattering camera for solution scattering.J. Appl. Crystallogr. 2004, 37, 369−380.(53) de Kruif, C. G.; Grinberg, V. Y. Micellisation of β-casein. ColloidsSurf., A 2002, 210, 183−190.(54) Beaucage, G. Approximations leading to a unified exponentialpower-law approach to small-angle scattering. J. Appl. Crystallogr. 1995,28, 717−728.(55) Leclerc, E.; Calmettes, P. Interactions in micellar solutions of β-casein. Physica B 1997, 234, 207−209.(56) Menon, S. V. G.; Manohar, C.; Rao, K. S. A new interpretation ofthe sticky hard-sphere model. J. Chem. Phys. 1991, 95, 9186−9190.(57) Sood, S. M.; Chang, P.; Slattery, C. W. Interactions in humancasein systems: Self-association of fully phosphorylated human β-casein.Arch. Biochem. Biophys. 1985, 242, 355−364.(58) Sood, S. M.; Chang, P.; Slattery, C. W. Interactions in humancasein systems: Self-association of nonphosphorylated human β-casein.Arch. Biochem. Biophys. 1988, 264, 574−583.(59) Pignon, F.; Belina, G.; Narayanan, T.; Paubel, X.; Magnin, A.;Gesan-Guiziou, G. Structure and rheological behavior of casein micellesuspensions during ultrafiltration process. J. Chem. Phys. 2004, 121,8138−8146.(60) Jackler, G.; Wittemann, A.; Ballauff, M.; Czeslik, C. Sphericalpolyelectrolyte brushes as carrier particles for proteins: An investigationof the structure of adsorbed and desorbed bovine serum albumin.Spectroscopy 2004, 18, 289−299.

Langmuir Article

dx.doi.org/10.1021/la402236f | Langmuir 2013, 29, 10483−1049010490