Confinement Facilitated Protein Stabilization As Investigated by Small-Angle Neutron Scattering

While mesoporous silicas have been shown to be a compelling candidate for drug delivery and the implementation of biotechnological applications requiring protein confinement and immobilization, the understanding of protein behavior upon physical adsorption into silica pores is limited. Many indirect methods are available to assess general adsorbed protein stability, such as Fourier-transform infrared spectroscopy and activity assays. However, the limitation of these methods is that spatial protein arrangement within the pores cannot be assessed. Mesoporous silicas pose a distinct challenge to direct methods, such as transmission electron microscopy, which lacks the contrast and resolution required to adequately observe immobilized protein structure, and nuclear magnetic resonance, which is computationally intensive and requires knowledge of the primary structure a priori. Small-angle neutron scattering can surmount these limitations and observe spatial protein arrangement within pores. Hereby, we observe the stabilization of fluid-like protein arrangement, facilitated by geometry-dependent crowding effects in cylindrical pores of ordered mesoporous silica, SBA-15. Stabilization is induced from a fluid-like structure factor, which is observed for samples at maximum protein loading in SBA-15 with pore diameters of 6.4 and 8.1 nm. Application of this effect for prevention of irreversible aggregation in high concentration environments is proposed.


Appendix 1. Sample Preparation and Protein Adsorption
Isotherms Bulk lysozyme and myoglobin solutions were prepared by reconstituting lyophilized protein powder in 25 mM pH 7.2 D 2 O phosphate buffer solution. The buffer pH of 7.2 was selected for both samples, as this enabled the investigation of protein packing in two different electrostatic regimes. While the pI of myoglobin is 7.2, which results in a net neutral protein charge, the pI of lysozyme is 11.35, which results in a net positive charge of the protein. This facilitates an electrostatic attraction with the silica surface (pI 2). The pH of 7.2 is also well within the optimal activity range of lysozyme, pH from 6.0 to 9.0, and near enough to the optimal enzymatic activity pH of 6.2 that the structural integrity of the protein is ensured. Protein  Figure 1 of the main text. The sample preparation was guided by previously measured protein adsorption isotherms for each of the SBA-15 pore sizes, respectively ( Figure S1). An initial 18 mg/mL stock protein solution was prepared by reconstituting lyophilized protein powder in 25 mM D 2 O phosphate buffer solution (pH 7.2) and dialysing in a 3.5 kD RC membrane suspended in 1 L phosphate buffer for 20 hours to complete H/D exchange. The SBA-15 stock solutions were prepared as the protein stock solution at a concentration of 8 S3 mg/mL and H/D exchanged in a separate dialysate phosphate buffer solution. The protein, SBA-15, and buffer stock solutions (specific quantities determined via mass balance using the protein adsorption isotherms shown in Figure S1) were then combined in 50 mL centrifuge tubes to prepare samples with the final nominal desired protein concentration adsorbed on the SBA-15 surface; 5, 10, 50 mg/mL pore volume, and max pore loading (maximum pore loading, where max ≡ Q max for all lyz, and max ≡ 0.95Q max for all mb). Each of these samples was then placed on a shaker table rotating at 50 rpm for 48 hrs to facilitate electrostatic physical adsorption while inhibiting the settling of the SBA-15 particles. Next, the samples were centrifuged at 18,000 rcf and the supernatants were decanted and measured by UV-vis to verify the adsorbed amount of protein by mass balance. The samples were then washed by re-suspension, centrifugation, and decantation. The supernatants were measured to ensure no detectable desorption of protein was occurring. The samples were then suspended in a minimal amount of phosphate buffer and sedimented into 2 mm Helma cells at 4 to 6 rcf using a custom centrifuge rotor bucket. Each cell was left with a minimal hydration layer of phosphate buffer above the packed sample. This layer was left to ensure that the samples did not dry out and maintained full hydration during neutron scattering experiments. The Helma cells were then sealed with PTFE caps secured by parafilm. The preparation results in a wet powder of SBA-15 particles with an isotropic distribution of the cylindrical pore orientations. Figure S1: Protein adsorption isotherms on each mesoporous material. The solid lines represent the Langmuir adsorption isotherm of best fit.

S5
Langmuir Adsorption Isotherm: where Q C bulk ≡ mass of adsorbed protein , Q max ≡ maximum loading concentration , K eq ≡ equilibrium adsorption constant , C bulk ≡ protein concentration in bulk solution . Note: As discussed in the text, K eq is shown here for thoroughness and only approximate. In contrast to Q max , K eq is used as an empirical fitting parameter only.
It is of importance to note that protein adsorption occurs onto a porous, concave surface, which includes both surface and volumetric filling contributions. Therefore, the binding equilibrium constant K eq is only approximate, as the real isotherm is a distortion of the ideal Langmuir model. Additionally, as such, Q max is shown in units of [mg protein / mg material], as it may not only represent surface adsorption.

S6
Maximum Protein Pore Filling Fraction:

Appendix 2. Models and Specification
All SANS scattering models were convoluted with the instrumental resolution according to the collimation distance and detector distance (2 m and 8 m), including wavelength spread (∆λ/λ = 0.2) and aperture sizes (6 mm sample, and 30 mm collimation), by procedures described by Pedersen et al. 2

Scattering of Ordered Mesoporous Particles
For ordered systems of two phases (particles and matrix) that are separated by sharp interfaces, the coherent scattering cross section can be written as is characterized by wavelength λ and scattering angle θ. The outer brackets · e indicate an ensemble average, while the inner F * i (q)F j (q) c represents the average over size and orientation variation of the particles, which are both assumed to be uncorrelated.
The scattering amplitudes F i can be written as where m , the matrix scattering length density, and s , the scattering length density of the particle i with volume V c . Expressing as: together with n = N V and the average structure factor: leads to the coherent scattering expression: The scattering of an isotropic powder of particles containing a mesoporous lattice I lattice (q) = n F (q) c 2 S lattice (q) can be described with the structure factor S lattice for approximately S8 cylindrical mesopores as 4,5 with the peak multiplicities m hk of the corresponding Bragg peaks with Miller indices (hkl), and the peak width σ B of the Gaussian peaks at peak position q hk . The Bragg peaks represented by S lattice (q) describe the order of mesopore rods in reciprocal space. Particle size distribution, surface roughness, and lattice disorder lead to a Debye-Waller like attenuation factor e −q 2 σ 2 DW in the structure factor. 4,5 The individual mesopores are described by a cylindrical form factor with radius R and length L 6 with J 1 as the first order Bessel function of the first kind and α as the angle between q and the cylinder axis. For infinitely long cylinders the form factor simplifies to and restricts the observable scattering intensity, due to lattice contributions, to α = π/2. The first term in Eq. S-8 describes deviations from the lattice structure as translational disorder, micropores, or inhomogeneities, which lead to diffuse scattering. This diffuse scattering is described by a Lorenzian squared term 7 To complete the scattering contribution of the mesoporous particle, (general) Porod scattering of the surface of the particles I P orod = A P q β , with β > 4, and a constant term I inc to account for the incoherent scattering of the samples is included. Finally, the overall scattering of the mesoscopic particles with ordered mesoporosity is described as

Scattering of Protein Filled Ordered Mesoporous Particles
For high protein volume fractions, the form factor of the cylindrical pores needs to take into account the spatial arrangement of the protein. Starting with Eq. S-5 and assuming a S9 constant protein scattering length density results in the scattering amplitude: with the scattering length s of the solvent and p (r) as the excess scattering length density of each protein with center of mass positions, R p . The protein arrangement is virtually extended over the cylinder compensating for the increase in volume with a box like function C(r), which describes the shape of the cylinder with C(r) = 1 in the cylinder and zero elsewhere, leading to: where the second term corresponds to F cyl (q). The influence of the surface onto the protein arrangement is neglected. By substituting r for R p + r in p (r), the scattering amplitude is described as: where C(q) = F cyl (q) = 2J 1 (qR) / (qR) for an infinitely long cylinder. According to the convolution theorem the Fourier transform of the product results in the convolution F C(r) p (r) = F C(r) * F p (r) = C(q) * p (q), which simplifies F p (q) to: Applying the ensemble average · eP over fluid like protein arrangements the mixed terms vanish due to the e iqRp terms, which average to zero. The form factor reads The last term is independent of the ensemble and represents the scattering of the cylinder.
In the first term, e iq(R p −Rp) describes non-vanishing particle correlations. This resembles Eq. S-4 as a particle distribution with form factors and structure factors. Using the procedure resulting in Eq. S-8 results in the scattering form factor of a protein filled cylinder Here, the fluid like Percus-Yevick structure factor can be used to describe the protein structure factor S P (q) for a protein number density n P . The convolution C(q) * p (q) eP depends on the correlation of protein orientation and cylinder axis. For example, when describing proteins as ellipsoids of revolution, the axis of rotation may be oriented along the cylinder axis. In such case, the protein orientation would need to be taken into account when determining the ensemble average.
Even when contrast matching the solvent with the matrix, the influence of the cylinder geometry is still present for protein scattering, as only the second term in Eq. S-19 vanishes.
Alternatively, if the cylindrical diameter is increased, C(q) approximates a delta function resulting in the conventional description of particles in a solvent. A long narrow cylinder of dimensions similar to the protein, filled with stacked protein, enables the potential separation of the cylindrical axial and radial directions allowing the usage of a one dimensional Percus-Yevick structure factor for S p (q). 8 The scattering of protein filled cylinders in an ordered mesoporous lattice needs to additionally account for the interference of proteins with lattice Bragg scattering. As the ensemble average · c is applied with the scattering amplitude F p (q) in Eq. S-17, which includes protein scattering contributions, the absolute intensity of a specific Bragg peak is related to the average contrast of solvent and protein, which is described by the scattering amplitude in Eq. S-17. Although, the cylindrical mesopore scattering contributions in Eq. S-19 scale by cylindrical volume squared, while protein contributions only scale linearly with volume n P V . Consequently, for long cylinders, the protein contribution is negligible.
Correspondingly, if Eq. S-17 is used to describe the mesopore form factor amplitude, additional protein diffuse scattering contributions originate in the F (q) c . Therefore, the additional diffuse scattering contributions from the S11 protein in ordered mesopores I dif f,P (q) with pore density n C is expressed as: Finally, the scattering of cylindrical oriented mesopores filled with protein is described as: Eq. S-21 is valid for isotropically oriented powders of ordered domains if a corresponding structure factor, such as Eq. S-9, is used. The model may also be extended to oriented samples by taking into account possible correlations of protein orientation with the cylinder axis in Eq. S-19, with a 2D structure factor for oriented domains.
The diffuse scattering term in Eq. S-21 can result from various types of pore inhomogeneities, such as pore size distribution, pore wall microporosity, pore surface roughness, and lattice distortions. As shown in Figure S2 and Figure 2 of the main text, the scattering contribution due to diffuse scattering is relatively smooth, and solely broadens the main peaks which are attributed to the cylindrically ordered structure and protein structure factor. This demonstrates the limitation of the SANS method to exhaustively discriminate further potential influences on the observed fluid-like arrangement of protein, e.g. pore wall corrugation, microporosity, and surface roughness. S12 Figure S2: Contributions to the scattering of ellipsoid packed cylindrical pores approximating the myoglobin sample at maximum concentration (R = 5.4 nm, L = 1000 nm, V ellipsoid = 18.2 nm 3 , ε = 0.43, volume fraction = 0.4). The coherent scattering of a packed cylindrical pore I cyl+P (q) is dominated by the scattering of the cylinder and shows negligible change due to mesopore filling. Conversely, diffuse scattering I dif f,P (q) of ellipsoids packed in cylindrical pores presents an additional peak originating from the ellipsoid packing structure factor at low q-values (2 nm −1 ), due to convolution with the cylindrical form factor. The clear peak of the structure factor S P (q) (with a maximum value ≈ 2 nm −1 ) is smoothed, and, as observed in this example, the peak maximum can be difficult or impractical to accurately determine. The extent of the peak suppression observed depends on the relative strength of both contributions, although, in general, peak strength is reduced.

Contrast Variation
For scattering experiments in general, the visibility of a sample is determined by the contrast of the scattering length density of the sample relative to the scattering length density of the solvent. Specifically regarding SANS, the contrast of protein is approximately twice that of silica, as shown in Eq. S-22. S13 SANS Contrast of Protein Relative to Silica: where ρ protein = 2.3 , Normally, contrast variation experiments aim to determine an optimal (matching) condition, such as when matrix scattering contributions are minimized by matching the scattering length density of the solvent to that of the matrix. Typically, this is achieved by changing the ratio of protonated solvent to deuterated solvent (water). However, using contrast variation for sample measurement is not always an advantageous approach, as incoherent scattering

Ellipsoidal Scattering
Ellipsoid of revolution form factors were calculated using Eq. S-23, 9 where F e (q, z) is defined as the scattering amplitude, and (R, R, εR) are defined as the ellipsoidal semi-axes. For oriented ellipsoids, the bounds of integration are adjusted accordingly.

PDB Crystal Structure Scattering Prediction
The atomistic form factors for lysozyme and myoglobin were calculated using Eq. S-24 with the Protein Data Bank (PDB) crystal structures 2lyz and 1mbn, respectively. 10 Atomistic Protein Form Factor: where b c ≡ coherent scattering length , r ≡ atom position .

Appendix 3. Mesoporous Material Characterization
Independent material characterization was completed using the combination of SANS, smallangle x-ray scattering (SAXS), nitrogen physisorption, and TEM imaging.   Figure S5: Bulk protein scattering profiles for lysozyme and myoglobin after background correction and scaled by concentration (lysozyme is scaled up by a factor of 5 for clarity).
Higher protein concentration samples are scaled by the effective concentration (as indicated in the legend) to result in a good overlap at larger q-values. The solid lines show the protein form factors based on the atomic crystal structures found in the Protein Data Bank (PDB IDs: 2lyz and 1mbn). The inset shows the bulk protein structure factors, S(q), calculated as the ratio of the concentration scaled scattering intensity at highest concentration to lowest concentration. The myoglobin samples are fit with the 3D Percus-Yevick structure factor 11,12 and present with an effective radius of 1.38 nm and volume fraction of 12% . The Lysozyme structure factor shows an upturn at low q-values which is characteristic for an attractive component of the protein interaction. Therefore, the lysozyme samples are fit with a square well potential adding an attractive well to the Percus-Yevick hard core. 13 This results in an effective radius of 1.24 nm and volume fraction of 12% . The potential well has a depth of 1 kT and a width of 0.32 nm, representing the attractive component. It should be emphasized that the resulting parameters are effective parameters that incorporate the non-spherical shape of the proteins and the complex patchy character of the protein surface.  431) ). All contributions are smeared by the resolution of the SANS instrument, which also results in a break at approximately 0.4 nm −1 due to a change in detector distance. Figure S7: Debye-Waller like protein contributions, σ DW , diffuse fluctuation correlation length, ξ dif f , and diffuse scattering amplitude, A dif f , with respect to loading concentration. As discussed in the text, the low signal to noise ratio of the parameters and lack of definitive trends prevents the justifiable proposal of further correlative assertions.