Surface-Induced Unfolding of Human Lactoferrin - ACS Publications

We have determined the structural conformations of human lactoferrin adsorbed ... measurements of lactoferrin adsorption in D2O instead of null reflec...
0 downloads 0 Views 214KB Size
3354

Langmuir 2005, 21, 3354-3361

Surface-Induced Unfolding of Human Lactoferrin Jian R. Lu,* Shiamalee Perumal, and Xiubo Zhao Biological Physics Group, School of Physics and Astronomy, the University of Manchester, Sackville Street Building, Sackville Street, Manchester M60 1QD, U.K.

Fausto Miano and Vincenzo Enea S.I.F.I. S.p.A., Via Ercole Patti, 36 - 95020 Aci S. Antonio - Lavinaio (Catania), Italy

Richard R. Heenan and Jeff Penfold ISIS Neutron Facility, Rutherford Appleton Laboratory, CLRC, Chilton, Didcot OX11 0QZ, U.K. Received November 18, 2004. In Final Form: January 28, 2005 We have determined the structural conformations of human lactoferrin adsorbed at the air/water interface by neutron reflectivity (NR) and its solution structure by small angle neutron scattering (SANS). The neutron reflectivity measurements revealed a strong structural unfolding of the molecule when adsorbed at the interface from a pH 7 phosphate buffer solution (PBS with a total ionic strength at 4.5 mM) over a wide concentration range. Two distinct regions, a top dense layer of 15-20 Å on the air side and a bottom diffuse layer of some 50 Å into the aqueous subphase, characterized the unfolded interfacial layer. At a concentration around 1 g dm-3, close to the physiological concentration of lactoferrin in biological fluids, the adsorbed amount was 5.5 × 10-8 mol m-2 in the absence of NaCl, but the addition of 0.3 M NaCl reduced protein adsorption to 3.5 × 10-8 mol m-2. Although the polypeptide distributions at the interface remained similar, quantitative analysis showed that the addition of NaCl reduced the layer thickness. Parallel measurements of lactoferrin adsorption in D2O instead of null reflecting water confirmed the unfolded structure at the interface. Furthermore, the D2O data indicated that the polypeptide in the top layer was predominantly protruded out of water, consistent with it being hydrophobic. In contrast, the scattering intensity profiles from SANS were well described by a cylindrical model with a diameter of 47 Å and a length of 105 Å in the presence of 0.3 M NaCl, indicating a retention of the globular framework in the bulk solution. In the absence of NaCl but with the same amount of phosphate buffer, the length of the cylinder increased to some 190 Å and the diameter remained constant. The length increase is indicative of changes in distance and orientation between the bilobal monomers due to the change in charge interactions. The results thus demonstrate that the surface structural unfolding was caused by the exposure of the protein molecule to the unsymmetrical energetic balance following surface adsorption.

Introduction Lactoferrin is a member of the family of iron-binding proteins. It is widely distributed in the physiological fluids of vertebrates.1-3 As a key protein component in human tears, milk, and blood, its biological functions with regard to its strong bacteriostatic properties depend on its ironbinding properties. The iron sequestering effect of lactoferrin, manifested through a pH-dependent manner, can easily deprive bacteria of iron essential for growth. The same function may also protect cells from free radical damage by binding potentially catalytic free iron. Recent studies have suggested that its iron transport and related biological functions are associated with lactoferrin’s strong membrane affinity.4 Surface immobilized lactoferrin has also been attributed to the causative effect on lung infection.5 Interfacial adsorption can alter its 3D structure which in turn undermines its strength and specificity for * To whom correspondence should be addressed. Phone: 44161-2003926. E-mail: [email protected]. (1) Anderson, B. F.; Baker, H. M.; Dodson, E. J.; Norris, G. E.; Rumball, S. V.; Waters, J. M.; Baker, E. N. Proc. Natl. Acad. Sci. U.S.A. 1987, 84, 1769. (2) Listowsky, I.; Aisen, P. Annu. Rev. Biochem. 1980, 49, 357. (3) Bullen, J. J.; Rogers, H. J.; Leigh, L. Br. Med. J. 1972, 3, 69. (4) Aguilera, O.; Quiros, L. M.; Fierro, J. F. FEBS Lett. 2003, 548, 5. (5) Thomas, L. L.; Xu, W.; Ardon, T. T. J. Immunol. 2002, 169, 993.

iron binding and transport. Thus, interfacial adsorption and associated changes in structural conformation have direct relevance to the understanding of its physiological roles in body fluids and dysfunctions to disease implications. Understanding the in situ structural conformations of lactoferrin will also aid the development of immunotechnology for screening this family of proteins and the exploitation of their use in infection control and wound healing.6,7 The folding and unfolding of proteins under physiologically relevant conditions has challenged scientists for many years.8 Indications for structural unfolding are experimentally difficult to probe. They have to rely on fast spectroscopic methods or mutational analyses to detect transient changes. Measurement of labile hydrogen exchanges has been established as a powerful means to detect structural details associated with folding intermediates and the dynamic pathways.9,10 Techniques such (6) van Berkel, P. H. C.; van Veen, H. A.; Geerts, M. E. J.; Nuijens, J. H. J. Immunol. Methods 2002, 267, 139. (7) Engelmayer, J.; Varadhachary, A. Wounds Compendium Clin. Res. Pract. 2003, 15, 294. See also: Weinberg, E. D. Expert Opin. Invest. Drugs 2003, 12, 841. (8) Rumbley, J.; Hoang, L.; Mayne, L.; Englander, S. W. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 105. (9) Roder, H.; Elove, G. A.; Englander, S. W. Nature 1988, 335, 700. (10) Radford, S. E.; Dobson, C. M.; Evans, P. A. Nature 1992, 358, 302.

10.1021/la047162j CCC: $30.25 © 2005 American Chemical Society Published on Web 03/10/2005

Surface-Induced Unfolding of Human Lactoferrin

as stopped flow methods and high resolution NMR analysis have been widely used to reveal transiently formed submolecular structures. However, these techniques are generally inadequate for studying the structural unfolding of interfacially adsorbed protein layers except that when proteins are adsorbed onto solid particulate dispersions, NMR and circular dichroism (CD) can offer some useful information indirectly about interface-induced structural unfolding.11 Fourier transfrom infrared (FTIR) spectroscopy can similarly reveal signatory features related to substrate surface-induced structural unfolding. Although techniques such as ellipsometry and surface laser scattering can estimate the amount of protein adsorbed at interfaces, they have no sensitivity to reveal any structural changes. Neutron reflectivity (NR) measures not only the amount of surface adsorption but also the thickness of the protein layer with a depth resolution of around 1-2 Å.12,13 NR is particularly effective at revealing any structural inhomogeneity along the surface normal direction. Thus, information about the structural distribution of the protein layer on the surface of water together with its 3D crystalline structure enables us to infer orientational conformations and possible structural unfolding induced by the surface polarization. The high interfacial resolution of neutron reflection stems from the low wavelength of neutron sources which is typically a few angstroms and is comparable to the size of molecules, making the neutron measurement inherently sensitive to molecular dimension. A further benefit of NR is that the neutron signal is sensitive to isotopic substitution. H/D labeling alters the neutron signal generally without affecting the chemical nature of the interface. In the case of protein adsorption on the surface of water, the use of H2O and D2O as solvent helps to highlight the adsorbed protein layer differently. Appropriate implementation of isotopic contrasts can significantly enhance interfacial structural resolution. Lactoferrin is a homomeric glycoprotein with some 700 amino acid residues and a molecular weight of ∼80 000.1,14-16 It is bilobal in structure, and the two halves are symmetrical. Peptide fragmentation and X-ray studies have shown that the polypeptide chain can be cleaved into two almost equal lobes, each carrying one iron-binding site. Amino acid sequence alignments showed that there is some 40% sequence identity between its N- and C-terminal halves. The 3D crystalline structure of lactoferrin has been revealed by X-ray studies with different resolutions.1,14-16 Figure 1 shows a “blurred” schematic aimed at illustrating the bilobal contour shape of the molecule. The most cited dimensions are 152 × 95 × 56 Å3. Each lobe is an ellipsoid with approximate dimensions of 55 × 35 × 35 Å3. The two lobes are connected by an R-helical connection, with their long axes roughly antiparallel. The R-helical connection may offer flexibility for orientational adjustment when the molecule is either in a solution state or adsorbed at the air/water interface. The measurement of the density (11) (a) Neutron reflection study of protein adsorption at the solid/ solution interface. In Biopolymers at Interfaces, 2nd ed.; Malmsten, M., Ed.; Dekker: New York, 2003; Chapter 23, pp 609-640. (b) Zoungrana, T.; Findenegg, G. E.; Norde, W. J. Colloid Interface Sci. 1997, 190, 437. (c) Kondo, A.; Oku, S.; Higashitani, K. J. Colloid Interface Sci. 1991, 143, 214. (12) Lu, J. R.; Thomas, R. K. J. Chem. Soc., Faraday Trans. 1998, 94, 995. (13) Lu, J. R. Annu. Rep. Prog. Chem. C 1999, 95, 3. (14) Peterson, N. A.; Arcus, V. L.; Anderson, B. F.; Tweedie, J. W.; Jameson, G. B.; Baker, E. N. Biochemistry 2002, 41, 14167. (15) Baker, H. M.; Anderson, B. F.; Baker, E. N. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 3579. (16) See the webpages of Brookhaven Protein Data Bank.

Langmuir, Vol. 21, No. 8, 2005 3355

Figure 1. Schematic representation of the bilobal globular framework of human lactoferrin with crystalline 3D dimensions of 56 × 95 × 152 Å3. Each lobe has the dimensions 55 × 35 × 35 Å3. The two halves are jointed by an R-helical connection, affording structural flexibility between the lobes when solubilized in physiological fluid. The dashed lines outline a cylindrical model as revealed by SANS in physiological solution.

Figure 2. Variation of surface tension against time at different protein concentrations at 36 °C with and without 0.3 M NaCl. Lines were drawn to guide the eye.

distribution of the interfacial layer provides a useful indication about what structural conformations the bilobal molecule adopts on the surface of water. Inhomogeneity in the length scale less than the shortest axial length is a strong indication of structural deformation or unfolding of the globular framework. In contrast, the parallel scattering intensity measurement from small angle neutron scattering (SANS) offers a useful estimate of the molecular size and shape in the bulk solution mimicking representative physiological conditions. Strong deformation of the bilobal molecule leading to solution unfolding or possible self-assembly to form an even greater size of protein aggregates can also be detected. Results (A) Surface Tension Measurements. In lung or tear fluids, lactoferrin works in close association with tear lipids. Understanding its interfacial behavior requires detailed knowledge of its adsorption with and without the lipids present. In the initial part of this work, we have set out to understand the surface behavior of this protein in the absence of any lipid. The surface tension measurement was first used to assess the dynamic process of lactoferrin adsorption under well-defined solution conditions. Figure 2 shows the variation of surface tension with time at a fixed protein concentration. All solutions were controlled at pH 7 using phosphate buffer solution (PBS), and the total ionic strength was fixed at 0.0045 M. To assess the effect of salt concentration, we have also measured the surface tension in the presence of 0.3 M

3356

Langmuir, Vol. 21, No. 8, 2005

Lu et al.

Figure 3. Effect of protein concentration on surface tension with and without the addition of 0.3 M NaCl. Lines were drawn to guide the eye.

NaCl. The addition of salt shortens the time required to reach equilibration. This is evident from the two adsorption curves measured at the lowest concentration 0.01 g dm-3. Figure 2 also shows that, at 1 g dm-3 with salt added, conditions close to the physiological environment, the timedependent drop of surface tension is fastest and it takes seconds for the surface tension to reach equilibration. Change in the equilibrated surface tension is better shown in Figure 3 by plotting the tension values after the first 60 min against protein concentration. It can be seen from Figure 3 that the addition of salt does not cause any further reduction in surface tension. The fast adsorption and its reluctant variation with salt and protein concentration indicate the strong structural flexibility of the protein molecule in retaining its optimal surface energetic balance. (B) Neutron Reflectivity. (B1) In the NRW Subphase. We first show the reflectivity profiles measured from the adsorption of lactoferrin layers at the air/null reflecting water (NRW) interface. The NRW contained 8.1% D2O, and under this isotopic contrast, the water phase was invisible to neutrons. Thus, the reflectivity measured from this interface arose entirely from the adsorbed protein layer. This isotopic contrast allowed us to examine the change in the structure and composition of the protein layer with protein concentration without the contribution of water to the signal, thereby simplifying the data analysis procedure.17-19 All the measurements were made at 36 °C and a pH of 7 with the total ionic strength of the PBS fixed at 4.5 mM. A time lag of 30 min was allowed to ensure that the experimental observations were made on an equilibrated protein layer. Figure 4 shows the measured reflectivity at 0.01 g/L, plotted as a function of wave vector (κ) perpendicular to the reflecting interface where

κ)

4π sin θ λ

(1)

where θ is the incidence angle and λ the wavelength of the incidence neutron beam. In the absence of any interfacial adsorption, a flat background around 4 × 10-6 was detected at the air/NRW interface. This flat background has been subtracted from the data. It can be seen from Figure 4 that the measured reflectivity forms an interference fringe around κ ) 0.07 Å-1. For a good approximation, π/κ at this position offers an estimate for the layer thickness; the value was found to be ∼60 Å. (17) Lu, J. R.; Lee, E. M.; Thomas, R. K. Acta Crystallogr. 1996, A52, 11. (18) Lu, J. R.; Thomas, R. K.; Penfold, J. Adv. Colloid Interface Sci. 2000, 84, 143. (19) Lu, J. R.; Perumal, S.; Powers, E.; Kelly, J.; Webster, J.; Penfold, J. J. Am. Chem. Soc. 2003, 125, 3751.

Figure 4. Neutron reflectivity measured from 0.01 g dm-3 lactoferrin adsorbed at the air/NRW water interface. The dashed line indicates the best uniform layer fit of 80 Å thick, and the continuous line represents the best two layer model with the top dense layer of 15 Å on the air side and the bottom loose layer of 60 Å in the water subphase.

More quantitative information about the thickness and composition of the adsorbed layer can be obtained from model fitting based on the optical matrix formalism. As described previously,13,17,18 we started the modeling process with an assumption of a structural model for the adsorbed layer, followed by calculation of the reflectivity from the model. We then compared the calculated reflectivity with the measured data, and the structural parameters were then varied in a least-squares iteration until a best fit was found. The structural parameters used in the fitting were the number of layers, the thickness (τ), and the corresponding scattering length density (F) for each layer (equivalent to the optical refractive index). For a uniform layer adsorbed at the air/water interface, the area per molecule (A, in Å2) and surface excess (Γ, in mol m-2) for the protein can be calculated from19

A) Γ)

∑bp Fτ

(2)

1020 NaA

(3)

where ∑bp is the total scattering length for the protein and Na is Avogadro’s constant. Due to H/D exchanges, ∑bp varies with the ratio of H2O to D2O. Its value for lactoferrin is 0.190 Å in NRW and 0.315 Å in D2O. The total volume of the protein was estimated by adding the volume of all individual peptides together, giving a total of 93 600 Å3. The scattering length density of the pure protein (Fp) is equal to 2.03 × 10-6 Å-2 in NRW and 3.36 × 10-6 Å-2 in D2O. In these calculations, we assumed the exchanges between labile hydrogens and solvents were complete. The reliability of this assumption will be assessed later. The volume fraction of the protein adsorbed in the layer (φp) at the air/NRW interface was estimated from the ratio of F to Fp. A uniform layer model was first used to fit the measured reflectivity profiles in NRW. It was found that the model did not result in a satisfactory fitting to the measured data. The best uniform layer fit shown as a dashed line in Figure 4 managed to fit the measured reflectivity on the left side of the minimum well but failed to match the skewed reflectivity profile on the right side. The continuous line shown in Figure 4 represents the two layer model that shows a substantial improvement over the uniform layer model, as indicated by a better fit to the measured profile over the entire κ range. The two layer model consisted of a top layer of 14 Å on the air side with F ) 1.1 × 10-6 Å-2 and a bottom layer of 60 Å toward the

Surface-Induced Unfolding of Human Lactoferrin

Langmuir, Vol. 21, No. 8, 2005 3357

Figure 5. Neutron reflectivity measured from 2 g dm-3 lactoferrin adsorbed at the air/NRW interface. The dashed line indicates the best uniform layer fit of 96 Å thick, and the continuous line represents the best three layer model with the top dense layer of 26 Å on the air side, the middle layer of 52 Å, and the outer diffuse layer of 40 Å in the water subphase. Table 1. Two Layer Model Fits to Reflectivity Profiles Measured under NRW in PBS concn/g dm-3

τ1/Å (3

φp1 (0.03

τ2/Å (3

φp2 (0.03

0.01 0.1 1 2

14 13 20 26

0.54 0.54 0.53 0.54

60 62 60 70

0.20 0.25 0.27 0.27

water subphase with F ) 0.4 × 10-6 Å-2, giving φp values of 0.54 and 0.20, respectively. Because the thickness of the dense top layer is significantly smaller than the shortest axial length 35 Å, the result suggests strong deformation or structural unfolding of the globular lobes. To substantiate these findings, we have performed the reflectivity measurements over a range of concentrations under otherwise the same solution conditions with the main parameters listed in Table 1. Figure 5 compares the one and three layer model fits with the measured reflectivity obtained at 2 g dm-3, the highest concentration for surface adsorption studied. The best uniform layer model gave a thickness of 95 Å with F ) 0.7 × 10-6 Å-2, suggesting an average φp value of around 0.34. It can be seen from Figure 5 that this fit only reproduced the measured data over the lowest κ range. A two layer model similar to that described previously offered significant improvement. The continuous line shown in Figure 5 represented the best three layer model fit that included a very diffuse layer further toward the bulk water phase. The top layer on the air side is 26 Å with F1 ) 1.1 × 10-6 Å-2 (φp ) 0.54), the middle layer is 52 Å with F2 ) 0.55 × 10-6 Å-2 (φp ) 0.27), and the outer diffuse layer is 40 Å with F ) 0.15 × 10-6 Å-2 (φp ) 0.07). These data show that although the thickness of the top layer is increased, its protein volume fraction remained the same as that at 0.01 g dm-3. The middle layer is still the thickest, and its volume fraction has increased slightly. The outer diffuse layer of some 40 Å contains some 10% protein that is 10fold the bulk concentration, and the data analysis indicated the sufficient sensitivity of the reflectivity to its presence. Without incorporating this layer, the fitting to the low κ range would still be rather poor. (B2) In the NRW Subphase with 0.3 M NaCl. To assess the effect of electrolyte on the structural distribution of the protein layer across the interface, we have also performed the adsorption measurements in NRW by adding 0.3 M NaCl into the PBS. We show in Figure 6 the reflectivity profile measured at 1 g dm-3 lactoferrin. It can be seen from Figure 6 that the shape of the curve

Figure 6. Neutron reflectivity measured from 1 g dm-3 lactoferrin under NRW and 0.3 M NaCl. The continuous line represents the best two layer model with the top dense layer of 24 Å on the air side and the bottom loose layer of 55 Å in the water subphase. Table 2. Two Layer Model Fits to Reflectivity Profiles Measured under NRW and 0.3 M NaCl concn/g dm-3

τ1/Å (2

φp1 (0.03

τ2/Å (3

φp2 (0.03

0.01 0.1 1

16 22 24

0.49 0.50 0.57

50 50 55

0.10 0.12 0.14

looks broadly similar to those measured under NRW without NaCl addition. This might indicate that salt addition did not alter the shape of protein distribution, but the data analysis showed a substantial reduction in the adsorbed amount. Again, the best fit required two sublayers to describe the structural distribution: the top layer was 24 Å with F ) 1.15 × 10-6 Å-2 (φp ) 0.57), and the bottom layer was 55 Å with F ) 0.28 × 10-6 Å-2 (φp ) 0.14). Measurements were also made at lower lactoferrin concentrations, and the results are listed in Table 2. These parameters show that the top dense layer has a thickness between 15 and 24 Å with its volume fraction between 0.49 and 0.55, again indicating structural deformation and unfolding. In contrast, the bottom diffuse layer is thinner and less dense than the corresponding ones in the absence of NaCl, indicating that salt addition has reduced the total layer thickness and suppresses the amount of protein adsorption. (B3) In D2O. Selected reflectivity measurements have also been made at the air/D2O interface. At this isotopic contrast, the reflectivity contained contributions from the adsorbed layer, the D2O, and the interference between them. Thus, these measurements serve to not only verify the protein distributions obtained from NRW but also reveal the relative location between the protein and solvent at the interface. The starting point for analyzing the reflectivity profiles in D2O is to use the same volume fraction distribution as obtained from the respective NRW one, except that other than under the NRW the interface is now visible to the neutron beam and due consideration needs to be taken of the extent of layer immersion in water. Note that, due to the exchange of labile hydrogens, the scattering length density needs to be adjusted to take into account the effect. For the part of the protein layer immersed in water, the average layer scattering length density can be calculated from the following equation19

F ) φpFp + φwFw

(4)

where Fp and Fw are the scattering length densities of protein and water and the volume fraction of the water φw ) 1 - φp. This means that the only parameter unknown to the fitting of the reflectivity profiles in D2O is the extent

3358

Langmuir, Vol. 21, No. 8, 2005

Figure 7. Reflectivity measured from surface layer adsorbed from 0.1 g dm-3 lactoferrin with 0.3 M NaCl. The continuous line represents the fit taking 18 Å out of 22 Å of the top layer out of water with the rest of the layer immersed. In contrast, the dashed line assumes the entire layer is fully immersed in water.

Figure 8. SANS scattering intensity (I) plotted against wave vector (κ) from 1 wt % lactoferrin in PBS (]) and the subsequent measurements at 1 (+), 0.5 (O), and 0.25 wt % (0) in PBS with 0.3 M NaCl.

of protein molecule exposed to air. The continuous line shown in Figure 7 is the best fit to the D2O reflectivity profile measured from 0.1 g dm-3 lactoferrin adsorption under 0.3 M NaCl. It can be seen from Table 2 that at 0.1 g dm-3 the top dense layer is 22 Å. The best fit to the corresponding D2O profile assumed that 18 Å is projected in the air and 4 Å is immersed in water alongside the 50 Å loose layer under water. Figure 7 also shows an alternative calculation shown as a dashed line. This model assumes that the entire protein layer is fully immersed in water. The poor fit from the fully immersed model confirms the extent of layer immersion. It should be noted that analysis of the D2O measurement in the absence of salt also indicated the predominant exposure of the top dense layer to air. (C) Small Angle Neutron Scattering (SANS). To support the hypothesis that the unfolding described above was induced by surface adsorption, it was necessary to confirm that the globular structure was intact in the corresponding bulk solution. SANS was used to determine the solution structure of lactoferrin at pH 7 with its concentration varied between 0.2 (2 g dm-3) and 1 wt % (10 g dm-3). As can be seen from Figure 8, this range of concentration gives an adequate scattering signal but is just above the physiological level of the protein present in tears. However, it is sufficiently dilute for us to neglect the intermolecular scattering. Data analysis under this condition offers direct and reliable information about the contour size and shape of the molecule in water, thereby

Lu et al.

providing an easy assessment of any major structural unfolding. As in the case of NR work, phosphate buffer at a total ionic strength of 4.5 mM was again used to control the solution pH. To examine the effect of salt, we have performed parallel SANS measurements with and without 0.3 M NaCl, with the protein concentration fixed at 1 wt %. Figure 8 shows that the scattering profiles could be described with a cylindrical model as depicted in Figure 1. The optimal diameter in both cases was 47 Å. It was however found that the length of the cylinder dropped from 190 Å without NaCl to 105 Å when the salt was added. We subsequently tested the sensitivity of the fitted parameters by varying each of them below and above the optimal values while keeping the other parameters constant. We found that the model is very sensitive to the variation of the diameter, within (3 Å in both cases. However, the sensitivity to the length is much less. In the case of the fitting to the measurement without NaCl, we found that, for noticeable deviations to occur between the measured and calculated data, it could be varied from some 180 to 230 Å. Nevertheless, it can be concluded from the data that salt addition reduces the height of the cylinder and hence the repulsive force between the two lobes. Further SANS measurements at the lower lactoferrin concentrations were obtained in the presence of 0.3 M NaCl to substantiate the model obtained. For comparison, these data are also shown in Figure 8. We found that the SANS data measured over the lower concentration range could be fitted to the same cylindrical model as that described for 1 wt % lactoferrin in the presence of salt. At 0.2 wt %, the optimal fitting led to a diameter of 47 Å and a height of 108 Å. At 0.5 wt %, the diameter was also found to be 45 Å with a height of 104 Å. It is thus clear that, within the experimental error, the protein molecule in the aqueous solution could be represented by a cylindrical contour with a diameter of 47 ( 3 Å and a height of 105 ( 5 Å under the solution conditions mimicking the physiological fluid. These results show that the size and shape of the molecule did not vary with dilution. As the parameters revealed from SANS are comparable to the overall contour dimension of the bilobal protein molecule, it can be concluded that in the aqueous solution the protein retains its globular framework. The volume fraction of protein inside the cylinder as modeled from the SANS data can be estimated from Vp/ Vc, where Vp is the volume of the protein obtained by summing up all the peptides in the molecule, Vc ) πR2L, with R denoting the radius and L denoting the height. Thus, φp inside the cylinder was found to be 0.52 in the presence of 0.3 M NaCl and 0.28 in the absence of the salt, showing that the size of the cylinders was 2 and 3 times greater than Vp, respectively. The remaining space was filled by D2O, and the average Fp values under the association of D2O in the cylindrical model were found to be 4.8 × 10-6 and 5.4 × 10-6 Å-2 with and without salt, respectively. Another parameter that could be derived from the SANS data analysis is the scale factor (sf), where sf can be expressed as

sf ) 10-24φ(Fp - Fw)2

(5)

where φ is the solution protein volume fraction in the form of a water-filled cylindrical molecule. The values of sf were found to be very close, 4.1 × 10-6 with salt and 3.5 × 10-6 without salt, giving φ values of 0.017 and 0.039. These values are compared well with the values 0.019

Surface-Induced Unfolding of Human Lactoferrin

Langmuir, Vol. 21, No. 8, 2005 3359

and 0.036 as converted from φp, the fraction of protein in the respective cylinders. The consistency in the mass balance was subsequently checked for the two lower concentrations in the presence of salt. The scale factors were found to be 2 × 10-6 and 0.78 × 10-6 and were well in agreement with the expected decrease of protein concentration, confirming the high degree of quantitative consistency of the SANS data. Discussion All the structural profiles obtained under different conditions from NR are consistent with a two layer model, a top dense layer of 10-20 Å with some 50% polypeptide and a bottom loose layer of 40-70 Å with 10-20% polypeptide. The thickness of the top layer is in all cases smaller than the shortest axial length 35 Å for each lobe. This, together with the presence of a very loose and diffuse tail distributed deeply into the aqueous solution, indicates unfolding of the bilobal framework into polypeptide configurations on the surface of water. The unfolded structural pattern is robust to lactoferrin concentration or salt addition, which is consistent with the relatively small variations in surface tension under these conditions, as shown in Figures 2 and 3. In addition, the surface tension and the structure of the adsorbed layers reach equilibrium rather fast when compared with other proteins, for example, lysozyme and human serum albumin (HSA), indicating a rather distinct mode for lactoferrin adsorption. In our previous neutron reflection studies, we found that lysozyme adsorption at the air/water interface formed a uniform layer over a wide range between 10-3 and 1 g dm-3. The layer thickness showed a steady transition from 30 to 45 Å around 0.1 g dm-3, indicating that the ellipsoidal molecules switched from sideways-on adsorption (with its short axial length perpendicular to the surface) to headways-on adsorption (with its long axis perpendicular to the surface).20,21 Within experimental error of (2 Å, the thickness measured matched the axial lengths well, indicating little structural deformation. In contrast, while the surface adsorption of bovine serum albumin (BSA) and HSA also formed uniform layers, the layer thickness showed a steady increase from some 25 to 40 Å.22,23 This range of layer thickness change indicated the formation of sideways-on conformation without structural unfolding into polypeptide distributions, and the steady thickness increase was indicative of structural deformation over the concentration range studied. In fact, it was observed previously that the adsorption of globular proteins at the hydrophobic solid/water interface generally causes some structural unfolding.24 In this content, the unfolding observed for lactoferrin adsorption at the air/water interface resembles the general phenomena we have observed at the hydrophobed solid/water interface. There is however insufficient data available on different systems at this point to generalize the effect of surface chemical nature and the pattern relating to the type and origin of different proteins. It is also worth noting that although protein adsorption as a topic has been actively pursued and many publications have been produced, few tech(20) Lu, J. R.; Su, T. J.; Thomas, R. K.; Penfold, J.; Webster, J. J. Chem. Soc., Faraday Trans. 1998, 94, 3279. (21) Lu, J. R.; Su, T. J.; Howlin, B. J. Phys. Chem. B 1999, 103, 5903. (22) Lu, J. R.; Su, T. J.; Thomas, R. K. J. Colloid Interface Sci. 1999, 213, 426. (23) Lu, J. R.; Su, T. J.; Penfold, J. Langmuir 1999, 15, 6975-6983. (24) Lu, J. R.; Su, T. J.; Thomas, R. K.; Rennie, A. R.; Cubit, R. J. Colloid Interface Sci. 1998, 206, 212.

Figure 9. Surface excess obtained from one layer (O) and two layer (4) model fits to NRW measurements in PBS at pH 7. Two layer model fits to the parallel NRW measurements with 0.3 M NaCl (+) are also shown for comparison. Lines were drawn to guide the eye.

niques offer the sort of in situ structural resolution comparable to the NR.25,26 The surface excess of lactoferrin has been calculated from eqs 2 and 3 using the structural parameters obtained from NRW contrasts. For NRW measurements in the absence of NaCl, surface excess values were obtained from one and two layer model fits, and the comparison is given in Figure 9. It can be seen from Figure 9 that the agreement is exceedingly good between the two parallel model treatments, given that the uniform layer model did not fit the shape of the NRW reflectivity profiles well. This supports the previous observation that the uniform layer model is sufficiently accurate to estimate surface excess even though it cannot reproduce the whole measured reflectivity profile. The surface excess from NRW measurements in the presence of NaCl is also shown in Figure 9. It can be seen that salt addition reduces surface adsorption. This is again consistent with our previous observations for other proteins such as lysozyme and HSA. The reduced adsorption may arise from the association of salt ions with protein molecules, or the improvement of the solvent property, resulting in the increased solubility of the protein. It is likely that salt addition causes stiffening of the charged polypeptide tails, leading to the formation of very diffuse layers deep into the aqueous solution. This possibility was examined during the data fitting and was found not to be the case for the adsorption of lactoferrin. Although the entire signal in NRW arises from the adsorbed protein, it is in general too weak to study the diffuse layer distribution. Therefore, the use of D2O as an alternative isotopic contrast serves to highlight the entire protein distribution relative to the solvent across the interface, especially the diffuse region extending into the bulk solution. Data analysis showed that each pair of reflectivity profiles measured from NRW and D2O at a given concentration could be fitted well by an identical volume fraction profile, with due consideration of the extent of immersion of the layer. Such good agreement is largely attributed to the relatively high volume fraction of the polypeptide in the diffuse layer, as evident from Tables 1 and 2. In all cases, the volume fraction in the main diffuse layer is close or above 10%. The data analysis also shows that the dense top layer is predominantly out of the water, indicating the relatively hydrophobic nature of the polypeptide within this region. (25) Lu, J. R.; Su, T. J.; Georganopoulou, D.; Williams, D. E. J. Phys. Chem. B 2003, 107, 3954-3962. (26) Lu, J. R.; Thomas, R. K. In Physical Chemistry of Biological Interfaces; Baszkin, A., Norde, W., Eds.; Dekker: New York, 2000.

3360

Langmuir, Vol. 21, No. 8, 2005

The corresponding broad and diffuse layer immersed in water is most likely to be of a hydrophilic nature consisting of the polar and charged amino acids. This general pattern of the structural distribution of the polypeptides is entirely consistent with the energy polarization of the air/water interface. A close inspection of the primary sequence of lactoferrin reveals the presence of many short peptide stretches that are either hydrophobic or hydrophilic. These stretches vary in different lengths. Examples of hydrophobic stretches include AAVA, AVAVV, LAVAVV, and LVFLVLLFLGALGLCLA. Those with a hydrophilic nature include KGKK, GRRRR, GDEQGENK, and EATKCFQWQRNMRKVR. Upon unfolding of the globular framework at the air/water interface, these stretches or fragments tend to segregate, resulting in heterogeneous polypeptide segment distribution along the surface normal. It should also be noted that stretches such as ALLCLDGKRK contain adjacent hydrophobic and hydrophilic parts and would adsorb effectively like surfactants themselves. A piece of relevant information is the exchanges of labile hydrogens between the protein and the bulk water. As outlined previously,27-29 time-dependent H/D exchange processes have been used to probe the kinetic pathways of structural unfolding of wild type and mutant proteins. It is hence of general interest to explore if neutron reflection can in this case provide useful information for us to assess the unfolding mechanistic process separately. There are two main types of labile hydrogens: those associated with the polypeptide backbone and those associated with the polar and charged amino acids. Hydrogen bonding and hydrophobic encapsulation may slow or prevent the H/D exchanges. However, surface adsorption disrupts the stability of the globular structure and this process will promote the exchange. The strong structural unfolding of the adsorbed lactoferrin means that there was little structural barrier inhibiting the labile hydrogen exchanges and that the exchange process must be fast and complete within minutes, a time scale beyond the current capability of neutron reflection. It should be noted at this point that we have previously assessed the H/D exchanges from lysozyme adsorption by comparing surface adsorption from freshly prepared and wellequilibrated lysozyme samples in D2O. No measurable difference in the adsorbed amount was detected, showing that the H/D exchange process was aided by structural disruption upon interfacial adsorption. Sethuraman et al.30 have very recently characterized the structural unfolding of interfacially adsorbed lysozyme using attenuated total reflection Fourier transform infrared (ATRFTIR) spectroscopy. They observed a two step sequential dynamic change in the secondary structure, with the first one involving a fast conversion of R-helix to random/turns within the first minute and the second step involving a slow conversion (1-1200 min) from R-helix to β-sheet without change in random/turns. The results together indicate that the fast secondary structural changes within the first few minutes of adsorption are accompanied by fast H/D exchanges. Given that the structural unfolding of lactoferrin is highly evident from its interfacial structural profile, its H/D exchange process must also be fast and complete. (27) Su, T. J.; Lu, J. R.; Thomas, R. K.; Cui, Z. F. J. Phys. Chem. B 1999, 103, 3727. (28) Su, T. J.; Lu, J. R.; Thomas, R. K.; Cui, Z. F.; Penfold, J. J. Phys. Chem. B 1998, 102, 8100. (29) Su, T. J.; Lu, J. R.; Thomas, R. K.; Cui, Z. F.; Penfold, J. J. Colloid Interface Sci. 1998, 203, 419. (30) Sethuraman, A.; Vedantham, G.; Imoto, T.; Przybycien, T.; Belfort, G. Proteins: Struct., Funct., Bioinformatics 2004, 56, 669.

Lu et al.

The neutron scattering data from SANS revealed that protein molecules retained their globular structure under the same solution conditions as those for NR study, thereby indicating that the unfolding at the air/water interface was caused by adsorption. The globular structure in the aqueous solution was well represented by a cylindrical shape with a diameter between 46 and 48 Å. The change of cylindrical length with salt concentration reflects the effect of solution ionic strength on the distance between the two lobes. This observation is broadly consistent with the reduced surface layer dimension upon salt addition, though the detailed molecular formations are different as described previously. As shown schematically in Figure 1, the two monomers are attached by an R-helical connection. Given the large structural flexibility between the two lobes, it is entirely reasonable to expect the structural adjustment in response to solution ionic strength. It should also be noted that if the two lobes are aligned in parallel to each other to some degree, we should expect an interference peak to occur around κ ) 0.1 Å-1, with the exact peak location dependent on the distance. The absence of such a peak suggests that the two lobes are rather misaligned with each other in the aqueous solution. This model is consistent with the fitting of cylindrical shape to the molecule. Given that the diameter of the cylinder is little affected by salt addition, it will be particularly interesting to examine how the length of the cylinder varies with solution ionic strength in a more systematic manner in our future work. Further detailed experimental work combined with computer modeling will lead to a better understanding of the structural location between the two lobes and the screening effect of salt addition. Such experiment will benefit from the measurement of neutron scattering to lower wave vector range using SANS instruments such as D22 in Institut Laue Langevin (ILL), Grenoble. The data to be obtained over the low wave vector range will help constrain the fitting of the cylindrical length, leading to a more reliable determination of the distance between the two lobes. There is immense interest in fundamental research to understand the structural pathways of the conversion of native lactoferrin into proto-fibrils and insoluble fibril deposits. Emerging evidence has indicated the possible linkage between surface activity and solution aggregation of protein and its amyloidogenicity.31,32 Thus, characterization of surface adsorption and solution aggregation may provide clues about conditions leading to the formation of proto-fibrils. Nilsson et al.33 have examined the stability of lactoferrin under different pH, temperature, and solvent conditions. They found that the protein was structurally less stable between pH 4 and 5. The structural unfolding was also revealed in urea and upon heating. They further found that the protein was more prone to forming aggregates under an elevated temperature at pH 7-8. Further neutron studies may help to elucidate the structure associated with lactoferrin adsorption and aggregation under different pHs and temperatures, thereby contributing to the understanding of the relationship between surface activity and amyloidogenicity. Conclusions The structural conformations of lactoferrin adsorbed at the air/water interface and in the bulk solution are schematically outlined in Figure 10. NR shows that the (31) Kelly, J. W. Curr. Opin. Struct. Biol. 1998, 8, 101. (32) Dobson, C. M. Philos. Trans. R. Soc. London, Ser. B 2001, 356, 133. (33) Nilsson, M. R.; Dobson, C. M. Biochemistry 2003, 42, 375.

Surface-Induced Unfolding of Human Lactoferrin

Langmuir, Vol. 21, No. 8, 2005 3361

protection could be activated. We suggest that one possible mechanism is the competitive adsorption of other proteins bearing a higher surface activity (e.g., lipocalins). Another hypothesis contemplates that the presence of the lipid layer secreted by the meibomian glands may reduce the adsorption of the protein at the interface. Further experimental NR work will be necessary to verify if lactoferrin will retain its conformation at the interface in the presence of a lipid layer spread upon the aqueous phase. Experimental Section

Figure 10. Schematic diagram of the physical states of lactoferrin adsorbed at the air/water interface and in the bulk solution.

unfolded lactoferrin is characterized by two main regions, a top layer of 10-20 Å with a high polypeptide volume fraction of 0.5 on the air side and a bottom layer of 50-80 Å with a low polypeptide volume fraction of 0.2 immersed in water. The top dense layer is predominantly exposed to air, indicating the strong hydrophobic nature of the polypeptide chain in this region. The remaining part of the unfolded polypeptides is immersed in water. This segregation results from the polarized nature of the air/ water interface and reinforced the strong role of the interface on protein structural unfolding. As the air can be regarded as a hydrophobic continuum in general, the finding here has direct implication for protein structural unfolding at the hydrophobic lipid membrane/solution interface. The addition of salt reduces protein adsorption, but the structural distributions of the polypeptide chains across the interface remain similar. At the same time, the surface tension shows a weak variation with salt or protein concentration. These observations indicate that the adsorbed protein molecules readily adjust to changes of external environment. The parallel SANS measurements reveal the preservation of the globular structure of lactoferrin in aqueous solution, thereby confirming that structural unfolding was induced by surface adsorption. The results discussed above have an important bearing on the behavior of these proteins in biological fluids, especially when their concentration is high and the environment is characterized by a high surface-to-volume ratio, as it is the case, for instance, for lactoferrin in tears. The tear film contains 2 g dm-3 lactoferrin dissolved in a 10 µm thick aqueous layer that is spread beneath a mucin layer, and it is surmounted by a thin lipid layer. Calculating the amount of lactoferrin that is adsorbed at the air/water interface, one observes that at the high surface-to-volume ratio 10% of the total amount of protein will unfold and denaturate unless a mechanism of

Neutron reflection experiments were performed at the ISIS Neutron facility, Rutherford Appleton Laboratory near Oxford, U.K., using the reflectometer SURF. The reflectometer used a white neutron beam with a wavelength from 0.5 to 6.5 Å. Measurements were made using a trough carrier containing five Teflon troughs to facilitate the parallel runs at the air/solution interface. The trough carrier was mounted on an antivibration system to eliminate mechanical vibration of the solution surface. The neutron beam was defined by two sets of horizontal and vertical slits placed before the sample container, providing a typical beam illuminated area around 10 cm × 3 cm. Each reflection experiment was carried out at three incidence angles of 0.5, 0.8, and 1.5°, and the resulting reflectivity profiles combined to cover a momentum transfer (κ) between 0.015 and 0.5 Å-1. The scaling factor was obtained from fitting to the pure D2O profile measured at an incidence angle of 1.5°. Constant background was subtracted using the average reflectivity between 0.3 and 0.5 Å-1. This was found to be around 4 × 10-6 in NRW and 2 × 10-6 in D2O. SANS measurements were performed on LOQ at ISIS using neutrons of wavelengths of 2.2-10 Å separated by time of flight. The 64 cm square detector was at a distance of 4.1 m, giving a κ range of 0.006-0.28 Å-1. Samples were contained in 2 mm path fused silica cells. Data were corrected for the wavelength dependence of the incident spectrum, the measured sample transmission, and relative detector efficiencies, prior to subtraction of the respective D2O buffers. Absolute scaling was obtained by comparison to the scattering from a partially deuterated polystyrene standard. Human milk lactoferrin purchased from Sigma was used as supplied. The protein contained some 30% iron ions equivalent to the binding sites available. Information about the crystalline structure of lactoferrin was obtained from www.expasy.org (trfl_human P02788, PDB number 1CB6, with 711 amino acids and Mw ) 78 338). The isoelectric point of the protein was around 8.6. Phosphate buffer solution (PBS) with a total ionic strength of 4.5 mM was used to control the pH at 7 throughout the entire experiment. To study the effect of salt, additional 0.3 M NaCl was added into the protein solution. After protein dissolution at high protein concentration, the solution pH was checked, and where necessary, the pH was adjusted back to 7 before the experiment started. The D2O used was purchased from Sigma and was used as supplied. Its D content was 99.9%. The surface tension at ambient temperature was found to be close to 70 mN/m using a Kruss K11 tensiometer. Ultrapure water (Ulgastat ultrapure, UHQ) was used throughout for final glassware and trough rinsing and solution preparation. The NaCl used was AR grade from Aldrich and was used as supplied. All the measurementssreflection, SANS, and surface tensionsfor studying surface adsorption and solution structure were made at 3536 °C to mimic the physiological conditions in tears.

LA047162J