Structure Characterization of Bacterial Microcompartment Shells via X-ray Scattering and Coordinate Modeling: Evidence for Adventitious Capture of Cytoplasmic Proteins

Abstract

Bacterial microcompartments (BMCs) are self-assembling protein shell structures that are widely investigated across a broad range of biological and abiotic chemistry applications. A central challenge in BMC research is the targeted capture of enzymes during shell assembly. While crystallography and cryo-EM techniques have been successful in determining BMC shell structures, there has been only limited success in visualizing the location of BMC-captured enzyme cargo. Here, we demonstrate the opportunity to use small-angle X-ray scattering (SAXS) and pair distance distribution function (PDDF) measurements combined with quantitative comparison to coordinate structure models as an approach to characterize BMC shell structures in solution conditions directly relevant to biochemical function. Using this approach, we analyzed BMC shells from Haliangium ochraceum (HO) that were isolated following expression in E. coli. The analysis allowed the BMC shell structures and the extent of encapsulated enzyme cargo to be identified. Notably, the results demonstrate that HO-BMC shells adventitiously capture significant amounts of cytoplasmic cargo during assembly in E. coli. Our findings highlight the utility of SAXS/PDDF analysis for evaluating BMC architectures and enzyme encapsulation, offering valuable insights for designing BMC shells as platforms for biological and abiotic catalyst capture within confined environments.

Graphical Abstract

INTRODUCTION

Bacterial microcompartments (BMCs) are organelle-like structures found ubiquitously across microbial phyla that consist of a protein shell encapsulating selected enzyme and cofactor cargo, thus permitting catalytic reaction chemistry to be sequestered from the cytoplasm and tuned by the selective permeability and confined microenvironments of the protein shell.^1-3 Structures for intact BMCs and their protein shell constituents have been determined with atomic-scale resolution from both X-ray crystallography⁴ and cryo-electron microscopy.^5,6 The consistency of protein shell subunit structures when crystallized as isolated proteins compared to those within intact BMC shell architectures point toward the building block assembly mechanism that can be variously combined and tailored to house specific reactions and enables predictive modeling of BMC structures based on subunit composition.^4,6 The modularity of shell proteins and their potential for self-assembly in both heterologous expression and in vitro assembly systems present a remarkable opportunity to understand shell protein properties and the principles of self-assembly, as well as to develop a modular tunable framework for spatially confined chemistry that is applicable to both biological and abiotic chemistry.^7-14

Several heterologous expression systems have been developed for BMC shell production as stable shells both with and without targeted enzyme cargo encapsulation.^4,5,7,15-17 While the plasticity of tile–tile building block interactions creates opportunities for mixed-and-matched assembly that can be tuned for selected functional outcomes, it also creates complexity reflected in the pleomorphism of assembled structures.¹⁵ Similarly, while cryo-EM image sorting algorithms have proven to be of critical value in providing a means to achieve atomic-scale structure resolution of BMC shells, there has been less success in using these or other techniques, such as dynamic light scattering, SDS-PAGE, or mass spectroscopy, to obtain a quantitative measure of the amount of enzyme cargo encapsulated by BMC shell assembly or to obtain information on the location of the captured cargo with respect to the shell architecture.

In this report, we implement an in situ structure analysis approach for characterization of a series of BMC shell structures, obtained by heterologous expression of proteins from plasmids carrying the BMC shell genes from myxobacterium Haliangium ochraceum (HO) in E. coli, in solution and under conditions that are directly relevant to biochemical function. In this approach, we calculated both small-angle X-ray scattering, SAXS, and corresponding real-space pair density distribution functions, PDDF, from the atom coordinates of BMC shell reference structures and quantitatively compare these to experimentally measured SAXS and PDDF patterns. The analysis was aided by the use of algorithms that permit desktop SAXS and PDDF calculations from atomic coordinates for large, multiple megadalton-scale BMC shell structures.¹⁸ Comparison of experimental and calculated real-space PDDF curves proved critical for developing and testing of structural models that provide an explanation of variances between the solution X-ray experiment and model coordinate structures.

Notably, we found that this structural analysis approach allows identification of adventitious protein cargo capture during BMC shell assembly in E. coli. In retrospect, while adventitious cargo capture during heterologous expression and assembly could be assumed, there has been no recognition or measurement demonstrating that such capture occurs. In this report, we demonstrate the opportunity to use SAXS/PDDF analyses to quantitate the extent and give information on the localization of adventitious as well as targeted cargo capture during BMC shell assembly. We expect that this approach will prove to be valuable for the development and evaluation of BMC shell architectures for biological and abiotic catalyst capture and as platforms for constructing compartments for catalysis in confinement.

MATERIALS AND METHODS

BMC Preparation.

We used Gibson cloning to insert C-terminal his-tagged HO-BMC-P (Hoch_5814) into a pET11 HO-BMC-H-HO-BMC-T1 coexpression vector (pARH329), described previously,¹⁹ to generate the pET11_HT1Phis vector. BL21(DE3) was transformed with pET11_HT1Phis, and 2L of LB broth was grown at 37 °C until OD₆₀₀ of 0.8. Protein expression was induced by addition of 0.1 mM IPTG, and cells were grown overnight at 22 °C. Cells were harvested by centrifugation at 10,000×g and stored at −20 °C until use. Cell pellets were resuspended in 20 mL of buffer A (20 mM Tris pH 7.4, 50 mM NaCl) with 20 mM imidazole, supplemented with 200 μL of 20 mg/mL DNase and lysed by 2 passages through a French Press at 25,000 psi. The lysate was then centrifuged at 38,000×g for 45 min at 4 °C, and the supernatant was applied on 5 mL of HisTrap HP (Cytiva). The column was then washed with 6cv of wash buffer (20 mM Tris pH 7.4, 300 mM NaCl, 20 mM imidazole) before elution with 2cv buffer A with 300 mM imidazole. The elution was then applied on a HR 16/10 MonoQ anion exchange column equilibrated in buffer A, and proteins were then eluted with a linear gradient to buffer B (20 mM Tris pH 7.4, 1 M NaCl). HPhis BMC shells as verified by SDS-PAGE eluted at 34% buffer B, and HT1Phis shells eluted at 36% buffer B. Peak fractions were pooled and concentrated and buffer exchanged to buffer A using Amicon spin filters (30K MWCO).

SAXS and PDDF Measurements.

Small-angle X-ray scattering (SAXS) experiments were performed at the 12-ID-B beamline of the Advanced Photon Source, Argonne National Laboratory and beamline 16-ID at the National Synchrotron Light Source II (NSLS-II), Brookhaven National Laboratory. At APS, the X-ray wavelength was set to 0.932 Å (13.3 keV); detection used an Eiger2 9 M X-ray detector (DECTRIS), and the measurements covered a Q-range up to 0.85 Å⁻¹. At NSLS-II, the X-ray wavelength was set to 0.827 Å (15.0 keV) and a Pilatus3 1 M detector (DECTRIS) was used. BMC samples were suspended in 50 mM Tris buffer, pH 7.5 and the protein concentration was adjusted to be in the range of 1–8 mg/mL. To minimize protein damage by X-ray radiation, samples were gently refreshed in a flow cell with a syringe pump.

SAXS patterns, I(Q), were obtained by azimuthally averaging the detector images and binning these as a function of Q = (4π/λ)sin θ, where 2θ is the scattered angle and λ is the X-ray wavelength. I(Q) patterns were obtained for BMC shell samples as the difference between the scattering measured for BMC-containing solutions, I_{BMC solution}(Q), and the scattering measured from the suspending buffer solution alone, I_buffer(Q). Pair distance distribution functions, PDDF, for BMCs were obtained by indirect Fourier transform of the difference I(Q), using the program GNOM.²⁰

SAXS and PDDF Calculations from Coordinate Structures.

SAXS and PDDF patterns from the HO-BMC shell reference structures were calculated using fast SAXS and fast PDDF algorithms in SolX3.¹⁸ These algorithms allow X-ray scattering calculations for large, multiple megadalton molecular assemblies to be accomplished on desktop computers and to be extended to large Q-ranges.¹⁸ For example, the algorithms have to been demonstrated to be accurate for simulation of biomolecular wide-angle scattering (WAXS).^21,22 We found the accuracy of these algorithms at a high angle to be useful for accurate fitting of SAXS experiments that extended to 0.85 Å⁻¹. SolX3 is available for download from https://12idb.xray.aps.anl.gov/solx.html.

To perform SAXS calculations across thousands of BMC shell models, cargo-loaded shells, and molecular dynamics (MD) simulation frames, we used CRYSOL.²³ SAXS profiles were calculated over the Q-range from 0 to 0.2 Å⁻¹, with 401 data points and 30 spherical harmonics, ensuring a small deviation from higher-accuracy tools like SolX3 at this Q-range, while maintaining a high level of computational efficiency, as illustrated in Figure S1. The corresponding PDDF profiles were extracted by evaluating Cα atom pair distances. We note that PDDFs calculated from Cα atoms are essentially indistinguishable from PDDFs calculated from all-atom models, as shown in Figure S2.

BMC shell configurations were assembled using custom Perl scripts that allowed flexible variation in trimer composition (T1, T2, and T3) and the number of pentamers, with a maximum of 12 pentamers per shell and ensuring that the sum of trimers always totaled 20. The script randomly selected trimer and pentamer positions and generated PDB files by merging the components into a complete shell model under the assumption that the overall shell structure remained the same irrespective of the type of trimer and missing pentamer tiles. A total of ~1600 unique shell models were generated to explore how variations in trimer composition and pentamer counts affect the scattering profiles.

To generate protein-filled BMC shells, a second custom Perl script was used to populate the shell interiors with randomly selected E. coli protein structures for the most abundant proteins based on proteomics data (Table S1). Experimental structures with known oligomerization states were used where available. For proteins where experimental structures were not available, we used structures from AlphaFold2²⁴ as provided in the Uniprot database.²⁵ The script ensured that the total weight of the added proteins matched a predefined target while distributing proteins inside the shell to avoid structural overlap with other shells or other proteins in the interior. In more detail, selected E. coli proteins were initially placed at predefined grid positions far enough away from each other to ensure that there was no overlap in the starting positions. A centrosymmetric potential was then applied to guide the proteins to pack within a sphere with a size slightly smaller than the interior diameter of BMC shells during a series of minimization and short molecular dynamics (MD) simulations. The MD simulations were performed by using CHARMM with a minimal Cα-based coarse-grained (CG) model to allow close packing without clashes. The interaction potential for the CG model consisted of Cα–Cα harmonic bond terms, Lennard–Jones interactions (ε = 0.05 for polar residues, ε = 0.2 for hydrophobic residues, σ = 3.8 Å), and weak electrostatic repulsion for all residues (Q = 0.1). The overall structures of folded domains were restrained with a harmonic biasing potential. Optimized CG models were then converted to atomistic models by superimposing the initial atomistic structures on the Cα positions. In this manner, we generated 6750 cargo models for the HT1P and HT1T2T3P shells as well as 420 models for the HP shell, with varying protein weights and spatial distributions. This approach explicitly considers that the measured SAXS profiles may be averaged over heterogeneous ensembles of different models of the cargo-filled shells.

Ab Initio Molecular Envelope Reconstruction from SAXS Profiles.

Three-dimensional (3D) molecular envelopes were reconstructed from SAXS profiles using ab initio programs DENSS²⁶ and DAMMIN.^27,28 DENSS (DENsity from Solution Scattering) utilizes an iterative structure factor retrieval algorithm for calculating 3D particle electron densities from 1D scattering data. The program iteratively alters the voxel electron density values on the predefined grids to match the 1D scattering data. The resulting electron density map delineates both the shape of the particle and the electron density distribution, if not uniform, within the particle. DAMMIN first reads a PDDF file and builds a search space large enough to accommodate the molecular structure under study by taking account of the largest dimension value (Dmax) provided in the PDDF file, and densely fills this volume with small beads of uniform electron density, also known as dummy atoms, Figure S3. The program utilizes a simulated annealing algorithm to search out an ensemble of beads that their SAXS profile fits the input SAXS data. This ensemble of beads, also often called the molecular envelope, represents a structure solution to the SAXS data. If the symmetry of the molecule is known, then both DENSS and DAMMIN can apply it to the 3D reconstruction, which greatly facilitates the molecular envelope search. In this study, icosahedral symmetry was imposed in both DENSS and DAMMIN calculations. In DENSS calculations, 10,000 steps of iterations were performed, and icosahedral symmetry was applied during 3000–9000 every 100 steps, followed by 1000 steps without symmetry restriction to optimize the SAXS data fitting and reduce structural artifacts. Icosahedral symmetry was imposed throughout DAMMIN calculations.

The resolution of the molecular envelope relies on the voxel size for DENSS and the bead size for DAMMIN. In DENSS calculations, oversampling ratios of 6 and 256 grid points per dimension were used, which provides a voxel size of about 9 Å. To achieve an optimal resolution for HT1P reconstruction, DAMMIN calculations were run in the Expert mode, which allows an upper total bead number limit around 30,000–40,000 for this program. All DAMMIN calculations started from an icosahedron search space with an outer radius of 190 Å, filled with12 Å diameter beads.²⁰ We note that the 3D structure reconstruction from 1D SAXS and PDDF data is an ill-posed problem. However, successful fitting yields examples of 3D electron density maps that match experimental SAXS and PDDF patterns and can be used for formulate models for the solution state structure.^26,27

Molecular Dynamics Simulations.

To test the hypothesis that the discrepancy between experimental and computed SAXS profiles was due to nanoscale dynamics of the BMC shells, we conducted molecular dynamics simulations starting from the 6mzx cryo-EM structure. As some residues were missing from the experimental structure, AlphaFold structures from Uniprot²⁹ were taken to fill in the gaps and were aligned to the original cryo-EM structure to generate a complete shell model. The protein model was solvated in a 456 Å cubic box, neutralized, and ionized with 0.15 M NaCl in VMD,³⁰ prior to equilibration and simulation with NAMD 3.0.³¹ Noting that water fluxes were imbalanced across the shell, additional water molecules were placed within the shell to balance the volume. Production simulations were carried out for 100 ns to sample structural heterogeneity at short time scales for a thermalized structure. The simulation was performed at 298 K and 1 atm maintained by a Langevin thermostat³² with a friction coefficient of 1 ps⁻¹ and an isotropic Langevin barostat³³ with 2 fs timesteps enabled through the use of the SETTLE algorithm.³⁴ Long-range electrostatics were captured through the particle mesh Ewald method³⁵ with a 1.2 Å grid resolution. The trajectory was analyzed for water flux and the root-mean-square deviation (RMSD) of the protein components. Snapshots were evaluated to calculate SAXS profiles and measure their variation over time.

Proteomic Analysis.

Proteomic analysis was conducted at the MSU Proteomics Core Facility as described previously.³⁶

RESULTS AND DISCUSSION

To investigate BMC shell structures under conditions that are directly relevant to biochemical function, we implemented an approach that compares experimental SAXS and PDDF patterns for BMC shells measured in solution with those calculated from model structures and structure ensembles. Initial structure characterization is performed by a comparison of experimental SAXS and PDDF to patterns calculated from cryo-EM reference structures. Possible sources for deviations between experimental and reference SAXS and PDDF patterns are analyzed by using ab initio 3D structure reconstruction. Since SAXS data alone often cannot tell whether the signals arise from homogeneous structures, SAXS ab initio methods typically assume a single model and reconstruct low-resolution 3D structure from 1D SAXS data, providing decent structural representation for samples with good quality control. Recognizing limitations for 3D structure reconstruction, we then employ large-scale coordinate-based modeling to explore the effects of shell composition, structural dynamics, cargo encapsulation, and possible ensemble structure representation, progressively increasing the model complexity to refine our understanding of cargo localization within the shells.

Figure 1 shows BMC shell reference structures used in the analysis of experimental X-ray scattering for HT1T2T3P, HT1P, and HP BMC shells from Haliangim ochraceum (HO) that were isolated following heterologous gene expression in E. coli. Five structural subunits are used in assembly of HO BMCs: BMC-H, BMC-T1, BMC-T2, BMC-T3, and BMC-P, referred to as H, T1, T2, T3, and P, respectively. H is a hexamer that forms a hexagonal tile, while T1, T2, and T3 are trimers with a duplication in a protein sequence and structure that allows them to form similar hexagonal tiles. T2 and T3 each dimerize to form stacked tiles, while T1 remains unstacked. P is a pentamer that forms the vertices of the icosahedron.¹⁵ By controlling which HO-BMC subunits are present during expression, several distinct HO shell types can be assembled, including a full HT1T2T3P and minimal HT1P shells.¹⁵ Models for the HT1T2T3P shell have been determined from both X-ray crystallography⁴ and cryo-EM.⁶ T2 and T3 are indistinguishable on the level of primary cryo-EM image classification but are readily distinguished from T1. Cryo-EM 3D image analysis categorized assembled shells among 10 structure classes that vary in BMC-T composition, with the main population having a composition of 35.4% T1 and 64.6% T^D, where T^D refers to an unresolved mixture of T2 and T3 dimer tiles.⁶ Cryo-EM image classification with symmetry expansion allowed refinement of a reference structure, 6mxz, that corresponds to a HT^DP shell.⁶ Similarly, a model for the HT1P shell was obtained by substituting the T1-6H shell fragment, 6n06,⁶ into T^D positions in the HT^DP shell structure. Smaller HP shells without trimers are also found to be assembled during full and minimal shell expression and can be isolated separately. A structure has not yet been determined for the HO-HP shell. However, HP BMC shell structures have been determined in other organisms, including the T=3 (6owf) and T=4 (6owg) icosahedral shells from Halothece sp.,³⁷ shown in Figure 1. These reference structures will serve as starting points for analysis of SAXS and PDDF patterns measured for HT1T2T3P, HT1P, and HP HO-BMC shells.

Figure 1.

BMC shell reference structures. HT^DP (6mzx) and model HT1P shells from Haliangium ochraceum and HP T=4 (6owg) and T=3 (6owf) shells from Halothece sp. PCCC 7418. Hexamer subunits are shown in blue, trimers in green, and pentamers in yellow.

SAXS Measurements and Comparison to BMC Reference Structures.

Experimental measurements of SAXS for HO-HT1P and HO-HT1T2T3P shell preparations are shown in Figures 2A and 2B, respectively. These are compared with scattering patterns calculated from the HT1P and HT^DP reference structures. We note that the experimental HT1T2T3P shell will be expected to differ from the HT^DP reference by having approximately 1/3 of the T^D dimer tiles replaced by monomer T1 tiles. An examination of how subunit composition variation in BMC shells affects SAXS patterns will be discussed in the section on coordinate-based model studies below.

Figure 2.

Comparison of experimental and calculated X-ray scattering and PDDF for the HT1P and HT1T2T3P BMC shells. Panels (A) and (B) show experimental (black) and calculated (red) SAXS patterns for HT1P and HT1T2T3P preparations and HT1P and HT^DP reference structures, respectively. Scattering for the HT^DP shell was calculated directly from the cryo-EM atomic coordinate model PDB entry 6MZX, while scattering for the HT1P shell was calculated from a coordinate model built from the 6MZX structure, where the HT^D fragments were replaced by the HT1 partial structure, PDB entry 6N06. Panels (C) and (D) show PDDF patterns corresponding to the scattering shown in panels (A) and (B). PDDF from experimental scattering data were obtained from indirect Fourier transform using the program GNOM.²⁶ PDDF were calculated from atomic coordinates using the program SolX3.¹⁸

Here, experimental traces are scaled and plotted with the scattering patterns calculated from the atomic coordinates for the corresponding HO-BMC shell reference structures. Overall, the experimental and calculated scattering patterns show good correspondence. In the small-angle region, Q < 0.005 Å⁻¹, the scattering curves asymptotically approach the limiting value, I(0), a parameter proportional to the squared molecular mass of the shell. I(0) and the radius of gyration, R_g, can be extracted from Guinier plots,^38-40 log[I(Q)] versus Q.² Guinier plots for the HT1P and HT1T2T3P shells, Figures S4A and S4B, respectively, show that the experimentally determined R_g for these shells, 179 (±5) and 181 (±5) Å, are larger than the R_g calculated from coordinates of the corresponding reference structures, 165.1 and 169.9 Å, respectively. The linear Guinier plots for the experimental HT1P and HT1T2T3P data show that these shell preparations are homogeneous in particle size, with no indication of deviations that would be characteristic of significant interparticle interactions or aggregation.

At higher angles, the SAXS patterns for BMC shells show a series of intense inference fringes that are characteristic of approximately spherical polyhedral structures. Calculations using analytical expressions for core–shell structures, Figure S5 and discussed further below, show that the positions and intensity pattern of the interference fringes are highly sensitive to the diameter, wall thickness, and electron density of the interior space. Apart from the first interference feature, the interference fringes measured for the HT1P and HT1T2T3P shells show close correspondences to those calculated from the reference structures, suggesting the appropriateness of the hollow, icosahedral shell reference structures as starting models for the solution state structures. However, the experimental measurement for both HT1P and HT1T2T3P shells shows deviations in the shape and position of the first interference feature, seen in the experimental traces by the minimum near Q = 0.02 Å⁻¹, and by the dampening of the subsequent oscillatory features. In this context, it is interesting to point out that the higher order interference peaks calculated from the HT^DP shell, Figure 2B, are seen to be more strongly damped compared to those calculated from the HT1P coordinate structure, Figure 2A. Presumably, this can be understood to arise from the variation in shell wall thickness produced by the exterior-facing T^D subunits.

Figure 3A shows experimental X-ray scattering data for HO-HP shells compared to scattering calculated from coordinate structures determined for the T=4 (6owg) and T=3 (6owf) HP shells from Halothece sp. PCCC 7418.³⁷ The prominent interference frequency mismatch with the Halothece T=3 HP shell shows that this model structure is much smaller than that of the HO-HP shell. However, scattering and interference frequency pattern calculated from the T=4 HP structure shows good correspondence to the scattering features measured for HO-HP shells, albeit with additional broadening present in the in situ experiment. This shows that the overall shape and dimensions of the HO-HP and Halothece T=4 HP shells are comparable. We can also confirm the HO-HP shell size determination by fitting the interference peak frequency pattern measured in the experiment with those calculated from analytical expressions for hollow spheres and icosahedrons, both of which are closely related geometric models for BMC shells, Figure S6.

Figure 3.

Comparison of experimental and calculated X-ray scattering and PDDF for the HP BMC shells. Panel (A) shows experimental scattering for HO-HP shells (black) compared to a scattering pattern calculated for the T=3 and T=4 HP shells from Halothece sp. PCC 7418, 6owf plotted in blue and 6owg plotted in red, respectively. Panel (B) shows PDDF patterns corresponding to the experimental and calculated scattering curves shown in panel (A). The PDDF from atomic coordinates, 6owf and 6owg, give distance maxima of 223 and 253 Å, respectively.

Experimental scattering for the HO-HP sample shows deviation from the Halothece T=4 HP model in the low Q region, Q < 0.01 Å⁻¹, where the scattering intensity asymptotically approaches a maximum value. Guinier plots, Figure S4C, show that data for the HO-HP shell deviate from the linear plot obtained from the scattering calculated from the Halothece T=4 HP structure. The results are indicative of some form of heterogeneity, aggregation, or interparticle interactions for the HO-HP shell preparation. Further, it is interesting to note that in the experimental HO-HP data and calculated scattering from the Halothece models, interference features are seen to extend to at least Q = 0.8 Å⁻¹, Figure S7, suggesting that it would be possible to use wide-angle scattering data with spatial resolution better than 7.85 Å to guide refinement of coordinate models to fit the HO-HP experimental data.

PDDF Analysis for Comparison to Reference Structures.

A more direct comparison between the BMC shell structure and X-ray scattering can be achieved by transform of the reciprocal space scattering to real-space pair density distribution functions, PDDF. For example, PDDF patterns for HT1P and HT1T2T3P shell preparations are shown in Figure 2C,D, obtained by indirect Fourier transform²⁰ of the experimental SAXS data in the Q-range up to 0.30 Å⁻¹. Experimental PDDF patterns are plotted with atomic pair correlations calculated¹⁸ from the corresponding reference structures. Experimental and calculated PDDF patterns for the HP shells are shown in Figure 3B.

PDDF patterns calculated for each of the BMC shell reference structures show a similar triangular shape that scales with the dimensions of the shell. An overlap of the HP, HT1P, and HT^DP PDDF patterns calculated from the corresponding reference structures is shown in Figure S8. The triangular shapes can be understood to arise from electron density pairs starting from the nearest neighbors and extending across progressively longer chords of the shell until a maximum is reached along the shell diameter at the farthest circumference that maintains the average electron density of the shell. Beyond this, the PDDF falls steeply, reflecting “roughness” of the shell outer edge, due to both the shape of the shell and extensions of the amino acid side chains. The longest atom pair distances for the T=3 HP, T=4 HP, HT1P, and HT^DP reference models are 223, 253, 374, and 416 Å, respectively. The arrangement of T^D tiles that have protein domains protruding on the exterior surface of the HT^DP shell can be seen to add undulations to the triangular PDDF shape in the distance range 50–250 Å and then to extend the pair distance maximum. Figure S9 shows a comparison of scaled PDDF for HT1P and HT^DP reference structures, illustrating the difference in the longdistance edges. The comparison of PDDF for HT1P and HT^DP structures is of interest, since it illustrates how protein attachment to the exterior surface of BMC shells would be recognized. The models show that while the addition of protein protrusions on the exterior of the shell adds pair distance correlations within the diameter of the shell, they most notably alter the shape of the PDDF by extending the long-distance edge.

Figures 2C,D and 3B show that the PDDFs measured for HO-BMC shell preparations deviate significantly from those calculated from reference structures. Characteristic features of these deviations are prominent addition of pair electron density at distances shorter than the diameter of the shells and only relatively minor deviations in long-distance edges of the PDDF. The extent of the added pair electron density between the experiment and calculations appears to vary among the different shell types. However, we observed that the PDDF pattern measured can vary significantly among different HT1P preparations, Figure S10, with differences detected mainly in the amount of added density at distances shorter than the shell diameter. The HT1P preparations show equivalent chromatographic purity, implying that the variation detected in the PDDF patterns is integral to the shell structure. Further preparation-dependent studies will be needed to determine extents of PDDF pattern variability in HT1T2T3P and HP shell types.

BMC Shell Structure Reconstruction from Ab Initio Fitting to SAXS.

To understand how the structure might correlate to the discrepancies between the experimental and reference structure SAXS and PDDF, ab initio 3D molecular envelope reconstruction was performed from experimental and reference HT1P SAXS patterns using the programs DENSS²⁶ and DAMMIN.²⁷ Both programs were found to accurately reconstruct the empty, hollow core–shell HT1P reference structure using the SAXS pattern calculated from atomic coordinates as an input fitting target, as judged by the overlap of the 3D reconstruction with the HT1P reference coordinate structure. SAXS and PDDF in Figure 4A,B and Figure S3A show different views of the structures derived DENSS and DAMMIN fitting, respectively. Least square residuals for fitting input SAXS profiles were found to be generally smaller using DENSS compared to DAMMIN. This is presumably linked to a smaller size and variable electron density of the voxels used in DENSS compared to the uniform electron density and larger bead size accessible in DAMMIN. The comparison of DENSS and DAMMIN ab initio fits to the reference structure HT1P SAXS pattern is shown in Figure 4C and Figure S3B, respectively. To facilitate 3D structure reconstruction from 1D SAXS pattern fitting, icosahedral symmetry constraints were applied as described in the Materials and Methods section, with additional details described in the legends of Figure 4 and Figure S3. Despite the differences in goodness of fit, the overlap between reconstructed and coordinate structures was found to be excellent, as shown by the fitting results from DENSS and DAMMIN in Figure 4D and Figure S3A, respectively. Similarly, a close match is found between the PDDF calculated from the reference coordinates and the reconstructed structures, as shown in Figure 4D, produced from the DENSS fitting protocol.

Figure 4.

DENSS ab initio reconstruction of the HT1P structure from SAXS calculated from the reference atomic structure. (A) Overall DENSS molecular envelope/electron density map. (B) Cross-section view for the overlay of DENSS (gray) and the reference structure (green). (C) DENSS SAXS fitting. The SAXS standard deviation (σ, in gray color in the top panel) was set as 30% of the intensity value. chi² was calculated as the average of the square of ΔI(Q)/σ. ΔI(Q) is the difference between input SAXS data and DENSS fitting. (D) PDDFs deviated using GNOM from curves in panel (C) with the same color coding.

In contrast to the hollow shell envelopes reconstructed by fitting the SAXS pattern calculated from the HT1P reference structure, the DAMMIN and DENSS fittings to experimental HT1P SAXS data shown in Figure 2A yielded more complex shell structures. Both fitting protocols produced having a significant amount of additional molecular mass in the shell interior, associated both with the shell wall and placement within the interior space. Results from DAMMIN fitting are presented in Figure S3. Results from DENSS fitting with forced icosahedral symmetry produced shells with spoke-like internal structure, Figure S11, similar to reconstructions produced in DAMMIN fitting, Figures S3, but yield poor fits to experimental data, Figure S11. However, by implementing fitting protocols that removed forced symmetry in the final iteration steps as described in the Materials and Methods, DENSS fitting of shell structures with asymmetric distribution of added electron density within the shell and near-exact fits to data, Figure 5.

Figure 5.

DENSS ab initio reconstruction of the HT1P structure from the SAXS experiment. (A) Exterior view, (B) cross-section view, and (C) DENSS fitting of the experimental SAXS data shown in Figure 2A. Q-Dependent standard deviation, σ(Q), of SAXS intensity is plotted in gray in the upper part of panel (c) but are largely obscured by the SAXS data line thickness.

A key feature of ab initio fitting to experimental SAXS/PDDF measurements for HT1P shells is that experimental SAXS data from Figure 2A are found to be fit by shell structures similar to that of the HT1P reference structure but which include additional electron density within BMC shell interior and on the exterior shell surface. For example, in considering the higher resolution 3D electron density reconstruction from DENSS, Figures 6A and 6B show the structure envelopes achieved from fitting the HT1P reference (gray) and experimental (cyan) SAXS data, respectively. The overlap of the two structures is shown in Figure 6C. The cyan protrusions reflect the additional electron density added to the exterior of the HT1P shell exterior. In addition, the crosssectional view in Figure 6D shows accompanying additional electron density added to the inside shell wall and interior volume. Further, we did DENSS fitting to the SAXS measured for the HT1P preparations #2 and #3 shown in Figure S10.

Figure 6.

Comparison of DENSS ab initio reconstructions of the HT1P structure from the reference structure (gray) and experimental (cyan) SAXS. (A, B) Individual exterior view, overlay exterior view (C), and cross-section view (D).

These are plotted in Figure 7A and 7B, respectively. The corresponding 3D electron density reconstructions for preparations #1, #2, and #3 are shown in Figure 7C, from exterior (top row) and cross-sectional (bottom row) views. The cross-section views include the structure reconstructed from fitting the calculated HT1P reference structure SAXS. The reconstructed structures in Figure 7C are colored to correlate with the SAXS and PDDF plots from different HT1P preparations shown in Figure S10. In each case, DENSS fitting is found to accurately fit the SAXS and PDDF patterns. A trend can be seen in which the amount of added PDDF amplitude at distances shorter than the shell diameter relative to that of the hollow shell reference structure, Figure S10, is correlated to the amount of additional electron density placed by DENSS in the HT1P shell interior. The DENSS fits for each HT1P preparation are also shown as radial electron density plots in Figure S12 and are compared to the radial electron density reconstructed from fits to SAXS calculated from the reference structure. Again, these plots show that the different HT1P preparations are distinguished by the amount of electron density added to the shell interior. The exterior electron density protrusions found from fitting SAXS measured for HT1P preparations can be understood to correlate with the long-distance tails in the experimental PDDF patterns.

Figure 7.

DENSS ab initio structure reconstruction from different preparations of E. coli expressed HO-HT1P shells. Panels (A) and (B) show experimental and DENSS fit SAXS for preparations labeled #2 and #3 from Figure S10. The SAXS data and DENSS fit for preparation #1 are shown in Figure 5C. The standard deviation (σ) of experimental SAXS data is plotted in gray. Panel (C) shows exterior (top row) and cross-section (bottom row) views of the DENSS 3D reconstructed electron density maps for preparation #1 (replotted from Figure 6C,D), #2, and #3. In the bottom row, each structure is overlaid with the shell reconstructed from DENSS fitting of SAXS calculated for the HT1P reference structure (in magenta). The magenta density map is the same one in gray color in Figures 4 and 6. The colors of the structures correlate with the colors of the PDDF lines plotted in Figure S10.

We note that the ab initio 3D electron density map reconstruction from 1D SAXS has two limitations. First, because 3D structure reconstruction is an underdetermined problem, fitting does not produce a unique fit.^26-28 Instead, a variety of structures can be generated depending on how multiparameter search spaces are entered and traversed. However, recovery of common structural features from different fitting approaches can serve as the basis for establishing models for in situ structure. One such feature reconstructed from both DENSS and DAMMIN fitting of SAXS patterns from HO-BMC shells is the inclusion of variable electron density within the shell interior. Such structures are suggestive of the capture of cytoplasmic constituents during shell assembly. Best fits from DENSS are associated with shell structures having an asymmetric distribution of cargo in the interior. Further fitting analysis will be required to determine how much reliance to place on this conclusion.

A second limitation of 3D structure reconstruction from 1D SAXS or PDDF data is that it fits experimental scattering data with a single structure, effectively assuming that the experimental populations are homogeneous. Cryo-EM image processing identified 10 structure classes of HT1T2T3P shells that vary in their BMC-T composition.⁶ In addition, inspection of TEM images for E. coli expressed HO-BMC shell preparations also shows evidence for minor fractions, at the level of 1–5 particles per 100 intact icosahedral shells, of structures that appear to be collapsed or broken, elongated shell shapes (for example see Figure S9A, ref 4) and that no longer fit icosahedral symmetry. Such structures can be expected to contribute to the long-distance tails seen in experimental PDDF patterns and possibly contribute features that required an asymmetric distribution of electron density as a compensating structure feature in ab initio SAXS fitting as well. We anticipate that the development of methods that combine the TEM structure and SAXS form factor classifications will provide approaches to achieve more definitive structure reconstruction.

Coordinate-Based Modeling.

To investigate how variations in the BMC shell structure and presence of encapsulated cytoplasmic cargo correlate to SAXS/PDDF profiles, we carried out modeling studies calculating SAXS and PDDF patterns for around 15,000 different HO-BMC models, systematically varying the composition of the BMC shell subunits, molecular dynamics, and inclusion of a portfolio of E. coli cytoplasmic proteins within HT1P shell interiors as candidates for adventitiously encapsulated cytoplasmic cargo.

Effect of BMC Shell Composition.

HO HTP BMC shells exhibit structural heterogeneity, in part due to the presence of three naturally occurring trimer species. The most significant distinction is that T1 trimers are single-layered, while T2 and T3 are double-stacked trimers, as shown in Figure 1. Additionally, “wiffle ball” structures without pentamers have been observed,⁴¹ suggesting the possibility that fully assembled shells might have holes due to missing pentamer tiles. When we altered the trimer composition in our models, Figure 8A,D, we observed minor changes in both the SAXS and PDDF profiles. Increasing the proportion of T3 trimers led to a reduction in the depth of the minima in the SAXS profile and a general increase in intensity due to the larger system size. This was also noted above in Figure 2, comparing SAXS for HT1P and HT1T2T3P shells. While this affected the absolute values in the PDDF, Figure S8, normalizing the PDDF profiles at a common distance showed that the trimer ratio changes resulted similar profiles with the most notable variation seen by the double layered T3 contributions to pair distances above ~300 Å, Figure 8D. Therefore, variations in trimer composition alone cannot account for the experimental PDDF curves, which show significantly higher relative intensities below ~300 for HT1P and HT1T2T3P shells. Consequently, any smaller proteins decorating the shell exterior are unlikely to significantly alter the overall SAXS and PDDF profiles, aside from contributing to the increased values beyond 300 Å. Still, the PDDF and ab initio DENSS models suggest that additional decoration of the outer shell (or potentially structural heterogeneity) is required to fully explain the experimental data. Similarly, changes in pentamer content had a limited impact on SAXS and PDDF profiles, Figures 8B and 8E, respectively. Reducing pentamer content caused a dampening of the oscillatory SAXS features but only minor changes in the PDDF’s triangular shape.

Figure 8.

Sensitivity of SAXS and PDDF profiles to structural changes and local dynamics. Panels (A)–(C) compare experimental (black) with calculated (colored) SAXS profiles. Panel (A) illustrates the effect of replacing a fraction of T1 trimers with T3 trimers in the HT1P shell. Panel (B) highlights the impact of varying pentamer content on the SAXS profile of the HT1P shell. Panel (C) shows the changes of the SAXS profile over a 200 ns molecular dynamics (MD) simulation of the HT1T2T3P shell. Panels (D)–(F) show the corresponding PDDF profiles, normalized at 310 Å. (D) PDDF profiles for varying trimer ratios, (E) influence of pentamer content, and (F) changes in the PDDF profiles during the MD simulation. Overall, local structural conformational changes have a limited impact on SAXS and PDDF and are not sufficient to explain the differences with the experimental data.

Effect of BMC Shell Dynamics.

Additionally, we tested the impact of local dynamics through additional molecular dynamics (MD) simulation with the HO-BMC shell. To explore the role of protein dynamics, we performed an atomistic MD simulation started from the 6mzx cryo-EM structure and calculated the SAXS and PDDF profiles across different frames. Given the large system size (around 10 million atoms for a fully solvated shell), the length of the simulation was limited to 100 ns, sufficient to capture local fluctuations but not large-scale structural changes. During this time frame, only minor damping of the first well in the SAXS pattern was observed (Figure 8C), leading to small changes in PDDF profiles (Figure 8F). We conclude from this analysis that structural changes in the trimer composition, pentamer content, and local dynamics are insufficient to explain the discrepancies between the model and experimental data.

Effect of HP and HTP Shell Mixtures.

Given that hexamers and pentamers of the HO HTP shells can form smaller HP shells, we examined whether a mixture of HTP (HT1P or HT1T2T3P) shells with the T=4 version of HP shells could theoretically account for the discrepancies, Figure S14. The SAXS profiles for modeled mixtures of HTP and HP shells, Figure S14A,B, show that incorporating 20% HP content makes the minima significantly shallower. As the proportion of HP shells increases, we observe a notable shift in the position of the first minimum. For an ~40% HP content, the position of the first minimum matches the HT1P shell, though this is not the case for HT1T2T3P shells. These simulations show that even by introducing 20–40% HP content in HTP shell mixtures, the impact on the PDDF profiles Figure S12C,D is not sufficient to account for the deviation from the experimental PDDF. The presence of an HP content is seen to increase the pair distance distribution below ~200 Å but has no effect on the distribution between 200 and 300 Å. As a result, the calculated PDDF profiles still show large deviations from the experimental data. Therefore, a mixture of HP and HTP shells is insufficient to fully explain the discrepancies between the experimental scattering and distance distribution patterns. Further, as discussed above, considering the possibility of purified HT1P and HT1T2T3P shell samples as mixtures with significant mole fractions of multiple shell types can be ruled out from Guinier and TEM⁶ analyses.

Effect of Adventitious Cargo Capture on SAXS Profiles.

Lastly, we tested the effect of encapsulated cargo proteins, as shown in Figures 9 and Figures S15 and S16. We note that adventitious capture of cargo during assembly of recombinant HO-BMC shells could include the full range of biomolecules present within the cytoplasm, presumably in proportion to their concentration in the cytoplasm and gated by accessible volume in the BMC shell interior. However, since SAXS from individual molecules captured within BMC shells is weighted by the product of each molecule’s electron density contrast with solvent, Δρ_i, and its volume, V_i, squared: I_i(q)_, = (Δρ_iV_i)²P_i(q), and where P_i(q) is the scattering form factor associated with the molecular shape,^39,42,43 SAXS from cytoplasmic constituents captured within BMC shells can be expected to be dominated by contributions from large macromolecules and biopolymers such as polynucleotides and proteins. We note that purified BMC shell samples have a UV absorption exhibiting a peak at 278 nm that is characteristic of proteins and distinct from the ~260 nm UV absorption peak that is characteristic for polynucleotides. For these reasons, in this first approach to molecular modeling of adventitious capture of cargo during assembly of recombinant HO-BMC shells for BMC shells, we chose to investigate the influence of encapsulated E. coli proteins on the SAXS and PDDF profiles of HT1P shells.

Figure 9.

Impact of cargo enzymes on SAXS and PDDF profiles in HT1P shells. Panel (A) compares the experimental (black) and calculated (colored) SAXS profiles for HT1P shells containing cargo proteins. The blue-to-red profiles represent varying cargo-occupied volume fractions, and the yellow line indicates the best fit using an ensemble of 10 structures. Panel (B) shows the corresponding PDDF profiles normalized at 306 Å. Panel (C) illustrates the estimated average spatial distribution of cargo proteins within the HT1P shell, revealing that most of the proteins are concentrated near its outer surface. Panel (C) illustrates the strong connection between the cargo-occupied volume fraction and ratio of the area above and below the line between 0 and 306 Å. This relationship allows the PDDF profile to serve as a direct estimator of the cargo-occupied volume. Panel (D) presents a selection of several cargo-filled shell structures from the fitted ensemble, showing a high diversity in cargo packing within the shell. The values above the structures are the respective cargo-occupied volume. Panel (E) illustrates the strong connection between the cargo-occupied volume fraction and ratio of the area above and below the line between 0 and 306 Å. This relationship allows the PDDF profile to serve as a direct estimator of the cargo-occupied volume.

As BMCs function to enclose various enzymes, we investigated the influence of encapsulated cargo proteins on the SAXS and PDDF profiles of the HT1P shells. Using untargeted proteomics,³⁶ we identified cytoplasmic E. coli proteins in our samples, with around 10% of the HTP tile abundance (Figure S13). Based on the most abundant proteins from the proteomics data (Table S1), we generated 6750 cargo models with varying amounts and spatial distributions of proteins within the HT1P shell. Figure 9A shows the SAXS profiles of all models with a blue color gradient indicating different occupied volume fractions of cargo proteins. Protein-occupied volume fractions add electron density distribution to the shell interior volume, analogous to the electron density volume element fitting used in the DAMMIN and DENSS programs described above, but accomplished here with ensembles of explicit, discrete protein structures. As the volume fraction of proteins in the interior increases, the oscillatory features are dampened due in large part to the lowered symmetry of the protein-loaded shell, and the position of the first minima is shifted. A similar shifting of the first inference peak was noted from calculations using analytical form factor expressions for spherical shells and comparing the effect of increasing the electron density of the shell interior, Figure S5.

Effect of Adventitious Cargo Capture on the PDDF Profile of HT1P.

In correspondence to the alteration of SAXS profiles induced by the inclusion of varying amounts and spatial distributions of proteins in the HT1P interior, characteristic changes to the PDDF profiles are also seen, Figure 9B, with higher cargo volume fractions resulting in increased distance distributions in the 0–300 Å range, most prominently between 125 and 275 Å. The PDDF profiles also transitioned from a triangular to a more bell-shaped form.

No single cargo model fully reproduces the experimental SAXS and PDDF profiles. Instead, we average over an ensemble of 10 structures to provide a closer approximation of the experiment (yellow line). Notably, a wide range of ensembles can achieve similar agreement with the experimental profiles. Despite the considerable differences in individual conformations and compositions within these ensembles, the spherically averaged distributions of protein density is similar. We observe that the cargo proteins predominantly localize near the inner surface of the shell (Figure 9C), consistent with the DENSS model. Figure 9D presents the best fit ensemble, highlighting the diversity in the protein number and packing arrangements within the shell. Most conformations contain relatively little encapsulated cargo, while a small fraction exhibits a densely packed arrangement, with over 25% of the inner volume occupied. The best fit ensemble showed significantly better agreement with the experimental PDDF profile, especially in the flatter region between 175 and 310 Å. The ensemble also shifted the first minimum in the SAXS profile to match the experimental minimum and dampened the oscillatory features. However, adventitious cargo alone cannot fully explain the experimental SAXS and PDDF data, particularly deviations at distances above 310 Å. The DENSS model suggests that this discrepancy may arise from a small fraction of cytoplasmic proteins docked on the exterior surface or structural heterogeneity within the shell construct.

Assuming that the primary deviation between experimental PDDF profiles and empty shell models is due to the presence of cargo proteins, we derived a method to directly estimate the occupied volume fraction from experimental PDDF data, Figure S9E. Empty shells form a triangular-shaped PDDF curve, peaking at the average shell diameter. By analyzing the ratio between the area (${\mathrm{A}}_{\text{ratio}}$) above and below the HTP empty shell PDDF from 0 to ~306 Å, we found that the cargo-occupied volume fraction, $V_{\text{frac}}$, can be estimated as follows:

$$V_{\text{frac}}=0.26\cdot A_{\text{ratio}}^{0.90}$$

The volume fraction estimated through this formula produces results consistent with the ensemble-averaged values.

Adventitious Cargo Capture with HT1T2T3P and HP Shells.

We applied this approach for the 6750 cargo models with the HT1T2T3P shells (Figure S15), observing a similar cargo effect as seen in the HT1P shell. For the HT1T2T3P shell, a cargo-based ensemble achieved high agreement with the experimental SAXS and PDDF profiles. The key difference, however, was that the estimated volume fraction was approximately 16%, nearly double the volume fraction found for the HT1P shell (~9%). Notably, we observe significant differences in the PDDFs of different HT1P preparations (Figure S10), suggesting that further research is needed to determine whether this discrepancy arises from the intrinsic shell property variations.

For the HP shell, Figure S16, combining cargo-based ensembles with a mixture of T=4 and T=3 shells produced a comparable level of agreement with the experimental data. This was likely due to the relatively small size differences between the T=3 and T=4 HP shells, making the shell mixing and cargo loading effects harder to distinguish in the SAXS and PDDF profiles. Still, given the evidence of the HT1P shells, it seems likely that the HP shells also capture cargo advantageously.

CONCLUSIONS

Both ab initio and molecular modeling point toward the finding that adventitious capture of E. coli cytoplasmic proteins during BMC shell assembly is the most plausible explanation for the deviations between experimental SAXS and PDDF profiles and those calculated from empty shell models. Molecular modeling demonstrated that changes in shell composition, such as variations in trimer ratios or missing pentamers, as well as local fluctuations, had only a minimal effect on the SAXS and PDDF profiles. While mixtures of shell structures and sizes were found to impact SAXS profiles by dampening the oscillatory patterns, quantities of contaminating shell structures compatible with SAXS and TEM analyses of homogeneity were found to have a limited influence on PDDF profiles. In contrast, the presence of plausible amounts of cytoplasmic proteins was found to substantially influence both SAXS and PDDF profiles by reducing the overall size, shifting the position of the minima, and increasing the distance distribution in the critical midrange region in accord with experimental data. This highlights the significant role that cargo plays in shaping the experimental scattering patterns, making it the key factor in explaining the observed deviations.

While we demonstrated that the PDDF can be linked to the volume fraction occupied by adventitiously captured cytoplasmic cargo, significant uncertainty arises from variations in protein packing density. As an alternative and more universally applicable approach, it is possible to estimate the quantity of cargo captured during shell assembly directly from the difference of the integrated PDDFs because the square root of the PDDF area is proportional to the total electrons or the molecular weight (MW) of the molecular assembly.^38-40 For example, from the experimental PDDFs for HT1P, HT1T2T3P, and HP shells that are scaled to those of reference structures in Figures 2C, 2D, and 3B, we obtain the area ratios between experiment and reference PDDFs of 1.37, 1.56, and 1.25, respectively. From the square-rooted area ratios and the reference theoretical MWs, we obtain cargo protein per shell that are 0.9, 1.5, and 0.2 MDa for HT1P, HT1T2T3P, and HP, respectively. However, adventitious cargo loading seems random and varies among preparations. For example, estimates of the extent of adventitious cargo loading in E. coli expressed HT1P shells varied from 0.46 to 1.2 MDa in three preparations presented in Figure S10B.

That BMC shells without recognition tags adventitiously capture significant amounts of cytoplasmic cargo has not been previously recognized. Understanding this phenomenon could be critical for the design of BMC shells that target selected enzyme capture. Given the extremely restricted volume within nanoscale BMC shells, the presence of adventitious cargo would compete with the enzyme targets for a limited amount of interior space. For example, in shells with recognition tags for selected enzyme encapsulation, SDS-PAGE, immunoblot analysis, and mass spectroscopy have demonstrated elevated presence of targeted enzymes in purified shell samples, although TEM and cryo-EM have had only limited success in identifying the location of these captured proteins.^5,16,34,35 In the case of cryo-EM measurements, variability in the images showing electron density in the shell interior prevented image indexing and construction of maps for captured enzyme localization.⁵ We note that the nanoscale BMC shells are below the resolution limits for fluorescence imaging microscopy. Similarly, while dynamic light scattering is widely used for BMC shell characterization, it is sensitive to the shell size but not the interior content.

Hence, we find that the presented method of comparison of experimental PDDF to reference patterns calculated from reference structures provides a possibly unique opportunity to characterize the extent of protein cargo captured during the BMC shell assembly. One of the advantages of this approach is that SAXS measurements are a high-throughput assay that can be directly applied to the BMC shell solutions used for biochemical enzyme activity assays. We expect that this approach will prove to be valuable for the development and evaluation of BMC shell architectures for biological and abiotic catalyst capture and as platforms for constructing compartments for catalysis in confinement.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsabm.4c01621.

Figure S1, SAXS measured for a HT1P shell sample and compared to scattering calculated from atomic coordinates using spherical harmonic expansions in CRYSOL and atomic form factors as implemented in SolX3; Figure S2, PDDF profiles calculated for a HT1P shell using all atoms versus only C_α; Figure S3, views of the 3D structure reconstruction search space and the resulting molecular envelopes, SAXS, and PDDF patterns retrieved from DAMMIN fitting of HTIP reference structure and scattering experiments; Guinier plots for HT1P, HT1T2T3P, and HP HO-BMC shells compared to reference structures; Figure S5, X-ray scattering features of solid sphere, and hollow core–shell spheres; Figure S6, experimental and simulated X-ray scattering profiles for the HP BMC using the atomic structure and two closely related geometric models; Figure S7, comparison of wide-angle scattering for HO-HP shells measured experimentally with scattering patterns calculated from the Halothece sp. T=3 shell (6owf) and the T=4 shell (6owg); Figure S8, comparison of PDDF patterns calculated from the BMC shell reference structures; Figure S9, comparison of PDDF calculated from the HT^DP and HT1P reference structures; Figure S10, X-ray scattering data and PDDFs measured from different E. coli expressed HO-HT1P shells preparations; Figure S11, DENSS reconstruction with forced icosahedral symmetry from experimental HT1P SAXS data; Figure S12, radial density distribution from DENSS fitting of HT1P SAXS data; Figure S13, untargeted proteomic analysis of purified HTP shells; Figure S14, effect of mixing HTP and HP shells on SAXS and PDDF data; Figure S15, impact of cargo enzymes on SAXS and PDDF profiles in HT1T2T3P shells; Figure S16, effect of cargo loading and T=3/T=4 mixtures on SAXS and PDDF profiles of HP shells; and table S1, Summary of the untargeted proteomic analysis of purified HTP shells (PDF)