Ligand Presentation Inside Protein Crystal Nanopores: Tunable Interfacial Adhesion Noncovalently Modulates Cell Attachment

Abstract

Protein crystals with sufficiently large solvent pores can non-covalently adsorb polymers in the pores. In principle, if these polymers contain cell adhesion ligands, the polymer-laden crystals could present ligands to cells with tunable adhesion strength. Moreover, porous protein crystals can store an internal ligand reservoir, so that the surface can be replenished. In this study, we demonstrate that poly(ethylene glycol) terminated with a cyclic cell adhesion ligand peptide (PEG-RGD) can be loaded into porous protein crystals by diffusion. Through atomic force microscopy (AFM), force-distance correlations of the mechanical interactions between activated AFM tips and protein crystals were precisely measured. The activation of AFM tips allows the tips to interact with PEG-RGD that was pre-loaded in the protein crystal nanopores, mimicking how a cell might attach to and pull on the ligand through integrin receptors. The AFM experiments also simultaneously reveal the detailed morphology of the buffer-immersed nanoporous protein crystal surface. We also show that porous protein crystals (without and with loaded PEG-RGD) serve as suitable substrates for attachment and spreading of adipose-derived stem cells. This strategy can be used to design surfaces that non-covalently present multiple different ligands to cells with tunable adhesive strength for each ligand, and with an internal reservoir to replenish the precisely defined crystalline surface.

1. Introduction

The arginine-glycine-aspartic acid (RGD) sequence was derived from fibronectin and reported as a cell adhesion peptide nearly 40 years ago.[1] Over the subsequent decades, this peptide and additional adhesion ligands from fibronectin,[2] laminin,[3–6] collagens,[7, 8] and other proteins[9] were discovered. Based on these discoveries, modifying biomaterials with cell-specific peptide ligands has become a ubiquitous strategy for promoting cell adhesion, spreading, and migration. In this work, we use a cyclic RGD-containing peptide, cyclo-RGDfK, shown in Fig. 1.[10]

Fig. 1. Porous protein crystals are capable of loading peptide adhesion ligands and modulating ADSCs’ spreading on the upper surface.

(a) A periplasmic protein, “CJ”, from Camphylobacter jejuni forms porous protein crystals that we stabilize via crosslinking. (b) Typical crystals are hexagonal prisms. (c) Chemical schematic illustration of ADSC attachment to the nanoporous protein crystal through integrin-PEG-RGD complex formation. (d) An AFM tip with covalently attached streptavidin binds a biotinylated integrin-mimetic peptide, to bind the PEG-RGD, enabling measurement of the force required to remove the receptor-decorated tip from the protein crystal pores. (e) A hexagonal array of 13 nm-diameter nanopores runs from the top to the bottom of each crystal. Crystal schematic with nanopore cut away and zoomed-in slice of nanopore side-wall with ionizable amino acids highlighted. Cysteine residues are shown in yellow. Carboxylic acid (Asp, Glu) is shown in red. Arginines are shown in cyan. Lysines are shown in dark blue. Histidines are shown in green. Both the N- and C- terminus contain flexible regions that are not pictured in this crystal structure. Zoomed-in image created using PyMOL. (f) A top view of five adjacent nanopores (PDB code: 5W17, same scale as (d)). (g) Representative illustration of force-distance (F-D) curve of a nanopore center. (h) A schematic of ADSCs’ spreading on the upper surface of nanoporous protein crystal before and after loading with cell attachment modulators (PEG-RGD).

Integrins that bind many of these peptides transduce mechanical signals through the cell cytoskeleton (mechanotransduction). These signals can also alter cell phenotype via signal transduction that modulates gene expression, thereby driving cellular functions ranging from extracellular matrix deposition to stem cell differentiation.[11, 12] The responses of cells to adhesion ligands can be further altered by how the adhesion ligands are presented to cells. For example, cell adhesion peptide gradients on surfaces or in hydrogels have been used to study the responses of cells to varying adhesion ligand concentrations.[13–15] Clustering ligands in nanoscale domains can also enhance signaling and may be necessary to achieve the maximal responses.[16] Because the in vivo cellular microenvironment is not static, techniques to dynamically control peptide presentation through external stimuli have also been developed. Zhao et al. reviewed magnetically responsive, electrically responsive, and thermally responsive materials that can essentially switch between two states of adhesion ligand presentation.[17] In addition, irreversible release or exposure of adhesion ligands has also been triggered by photosensitive, redox, and enzyme-catalyzed reactions to achieve dynamic or spatially patterned ligand presentation.[1] While most reports describe adhesion ligands covalently attached to substrates, Grewal et al. recently reported using non-covalent peptide coiled coil complexes to reversibly and dynamically control the presentation of adhesion ligands from hydrogels and nanofibers.[18] To further elucidate how the properties of substrates and protein presentation influence mechanotransduction, Wen et al. used hydrogels of varying stiffness and porosity to show that that substrate stiffness regulates stem cell fate independent of substrate porosity or the density of links between a protein layer and the underlying substrate.[19] Furthermore, AFM has also been adopted to confirm that material architecture regulates cell adhesion.[20]

The putative isoprenoid binding protein from Campylobacter jejuni (Genebank ID: CJ0420, Protein Data Bank (PDB) code: 5W17) has been modified to form the protein CJ,[21] which forms hexagonal crystals (with P622 space group) with unusually large, 13-nm diameter pores.[21–26] Crosslinked CJ protein crystals are capable of serving as “hosts” to “guest” nanoparticles and macromolecules, including gold nanoparticles,[21, 26] proteins,[22] and oligomeric double-stranded DNA.[23, 27] Furthermore, these large-pore protein crystals have favorable cytocompatibility.[25] The interior surfaces of the protein crystal pores can non-covalently adsorb macromolecules, making the porous protein crystal a capacious reservoir (80% solvent volume) for the presentation of biochemical signals. Here we propose that polymers bound inside the pores very near the crystal surface can be probed by cell surface receptors.

In previous work, we showed that these large-pore protein crystals can adsorb double-stranded DNA oligonucleotides with high affinity.[23] The DNA could be subsequently removed from the protein crystal pores by attachment to the tip of an atomic force microscope probe, via a disulfide bond formation chemistry. This experimental method enables probing an individual “loaded” pore near the surface of the crystal to determine the aggregate binding energy of a cluster of macromolecules attached to the AFM tip. In the course of our previous study, we serendipitously discovered that the high affinity binding of macromolecular guests was not unique to DNA, but extended to other macromolecules, including poly(ethylene glycol) (PEG).

In the present work, we use a CJ protein variant, which is well-suited for growing large protein crystals (Figure 1 a and b). The modified CJ is sequence variant CJ-A34I-L48F-V50I-V121M-N162C-I163W-V165I. We propose the name CJ^OPT for this optimized large-crystal variant. The nucleic acid sequence (S.1.1), the amino acid sequence (S.1.2), and the mutations (S.1.3) of CJ^OPT protein are outlined in the Supporting Information. Briefly, the CJ^OPT sequence contains the following mutations designed to stabilize the inner hydrophobic cavity: A34I, L48F, V50I, I163W and V165I. Additional mutations include V121M for strengthening a key crystal contact, and N162C to allow covalent installation of guest molecules using thiol chemistry as previously demonstrated.[24] Lastly, the N-terminal 20 residue signaling peptide (KKVLLSSLVAVSLLSTGLFA, UniProtKB Q79JB5)[28] was removed to simplify protein expression as previously described.[21] We hypothesized that an adhesion ligand (RGD) conjugated to PEG could be adsorbed into the protein crystal pores, and that the non-covalent binding of PEG to the protein crystal pore or surface would provide an adhesive strength sufficient to enable cell adhesion and spreading. Because the attachment is non-covalent, but presumably occurs through multiple valency of the PEG chain, the attachment strength could theoretically be tuned by altering the PEG length, so that PEG molecules of different lengths could provide different strengths of receptor-surface binding. Because the PEG-RGD is adsorbed into the volume of the protein crystal, a crystal could provide a reservoir of ligand to replenish the surface if PEG-RGD molecules are removed by cells or are lost through other mechanisms.

To demonstrate this novel concept, we first show by confocal microscopy imaging that PEG-RGD is rapidly taken up by porous protein crystals (within ~10 min), as expected from earlier PEG adsorption studies. We hypothesized that protein crystals loaded with PEG-RGD could be used to present the adhesive peptide, RGD to cells, thereby enabling cell attachment (Fig. 1c). Furthermore, the force required to retract an AFM tip presenting RGD receptors from a single protein crystal pore can be quantitatively measured by chemical force microscopy (Fig. 1d). This experiment simulates the attachment of cell surface integrins and the application of contractile force from the cell cytoskeleton. Chemical force microscopy was performed by decorating an AFM tip with domains that have affinity for the RGD ligand and imaging the protein crystal using the peak-force quantitative nanomechanics (peak force QNM) mode, on a Bruker Bioscope Resolve microscope. Specifically, previous research identified a cyclic peptide (IntP, CWDDGWLC) that binds RGD.[29, 30] We used a biotinylated version of this peptide as a mimic of the RGD-binding site of the integrin β₃ subunit (Fig. S6).

Using the peak force QNM imaging mode, we were able to collect hundreds of thousands of F-D curves (Fig. 1g) during a single imaging experiment without interrupting the imaging process. This imaging mode collects force-versus-distance data at every pixel of an image, enabling us to precisely map the force-distance curves obtained when the AFM tip interacts with a protein crystal pore or the wall of a pore, providing details of the connection between the mechanical behaviors and the morphology of the surface. In this way, we measured the aggregate interactions of activated AFM tips with a PEG-RGD-loaded porous protein crystal. This study thereby provides parameters that may enable future researchers to tune protein crystals and solution conditions for the storage and release of molecular cell modulators. Notably, this current study is not designed to isolate the interaction of a single PEG-RGD chain with the host crystal. Instead, multiple RGD receptors attached to the activated AFM tip mimic the prospective multivalent attachment of cells to protein crystals via PEG-RGD. One caveat is that an IntP-decorated tip may display RGD binding sites at a higher local density than a cell surface can, due to the receptor spacing. The current interaction data and analysis will also guide the design and interpretation of future single-molecule studies of this system. We further show that by adding PEG-RGD to protein crystals, the spreading and cytoskeletal arrangement of ADSCs can be modulated; ADSCs on PEG-RGD-loaded crystals exhibit more well-developed actin stress fibers than cells on non-loaded crystals (Fig. 1h). Since the force and adhesion energy generated by the interaction between tip and the crystal, mediated via streptavidin, biotinylated Int-P, and PEG-RGD, fall within a repeatable and narrow range, PEG-RGD may serve as a cell adhesion modulator in future applications, such as guiding stem cell differentiation. The strategy in this study can also be applied to other more cell type-specific adhesion peptides to prepare surfaces that selectively bind different cell phenotypes with different strengths.

2. Materials and Methods

2.1. Materials.

Polyethylene glycol (5 kDa and 10 kDa) with a methyl terminus at one end, and either a fluorescein or a succinimidyl group at the other end (mPEG₅₀₀₀-FITC, mPEG₁₀₀₀₀-FITC, mPEG₅₀₀₀-succinimidyl ester (SC), and mPEG₁₀₀₀₀-SC) were purchased from Biochempeg (Watertown, MA). Cyclo-RGDfK peptide (Fig. S6 in Supporting Information) was purchased from MedChemExpress (Monmouth Junction, NJ). A biotinylated disulfide-cyclized receptor peptide for the RGD ligand (IntP, CWDDGWLC, Fig. S6 in Supporting Information) was purchased from Genscript (Piscataway, NJ). CJ^OPT protein crystals were obtained and crosslinked according to procedures outlined in the Supporting Information as Section S.1.

2.2. Synthesis of PEG-RGD.

mPEG₁₀₀₀₀-SC (or mPEG₅₀₀₀-SC) (30 mg) was dissolved in DMF to make a 10 mM solution. Separately, 6 mg of cyclo-RGDfK was dissolved in an aqueous Na₃PO₄ solution (0.1 M and pH = 7.2) to prepare 10 mM concentration. The mPEG-SC and cyclo-RGDfK solutions were mixed and reacted with gentle mixing for 12 h at room temperature. The mixture (1 mL) was then transferred into a dialysis cassette (molecular weight cutoff of 7 kDa) and dialyzed against water for two days. White powder products (PEG₁₀₀₀₀–RGD or PEG₅₀₀₀-RGD) were collected after freeze-drying. The 5 kDa PEG has approximately 120 repeat units, and the 10 kDa PEG has approximately 240 repeat units, with corresponding contour lengths of approximately 45 and 90 nm, respectively. On the basis of prior work correlating PEG length with scattering experiment data, the radii of gyration can be approximated as 3.1 and 4.6 nm, respectively.[31, 32] These dimensions can change depending upon solvent conditions, and at surfaces, particularly at high grafting density.[33]

2.3. PEG Loading Into Porous Protein Crystals.

Confocal microscopy was used to monitor the loading of mPEG-FITC. Protein crystals were immobilized on their sides in a microchannel with a cover to prevent evaporation during the experiment. mPEG-FITC with 1 μM concentration in TE buffer (tris and ethylenediaminetetraacetic acid in deionized water, pH = 7.5) was added to the solution surrounding the crystals while z-stack imaging was continued for 30 min. Imaging was performed on a Nikon TiE spinning disc confocal microscope.

2.4. Crystal Immobilization.

For the AFM experiments (described below in Section 2.6), the CJ^OPT protein crystals were immobilized on glass-bottom petri dishes (Willco Wells) with a UV-curable glue (Bondic Inc.). First, the top of a crystal probe (Minitool HR4–217) was used to transfer a drop of UV-curable glue onto the surface of a petri dish (Ted Pella, Inc. 14025–20). The glue was gently and evenly spread on the dish surface to make the layer of glue as thin as possible. CJ^OPT protein crystals were transferred to the glue with a loop. Critically, the crystal was transported inside a tiny drop of buffer, such that the crystal was not desiccated. The UV-glue was viscous and did not noticeably mix with the buffer. The glue was then cured by exposing to UV-light LED (Bondic SK001) from above for 10 s. The glue cured after about 2 min, after which additional drops of buffer (typically ~5 mL) were added to the dish to prevent the crystal from drying. More details of this procedure and the suitability of this method for subsequent AFM imaging of protein crystals are provided in our recent report.*

2.5. AFM Tip Activation.

ScanAsyst Fluid+ AFM tips (Bruker) were modified to noncovalently attach PEG-RGD. The initial tips have a slim shape, a nominal tip radius of 2 nm, a silica surface, and a nominal spring constant of 0.7 N/m. Hydroxyl groups on the tip surface were activated by oxygen plasma, enabling the tip to be modified with the (3-aminopropyl)trimethoxysilane (APTMS) by molecular vapor deposition (MVD).^[34] A monolayer of APTMS would increase the tip radius by less than 0.5 nm. Detailed experimental procedures are outlined in the Supporting Information, Section S.2. 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) and N-hydroxysuccinimide (NHS) were then dissolved in phosphate-buffered saline (PBS) buffer (pH = 7.0) to make 10 mM solution. Then the tips were immersed into the EDC-NHS mixture solution for 60 min at room temperature. Tips were later washed several times with PBS buffer (Gibco, 1×, without calcium chloride or magnesium chloride). Streptavidin was coupled to the tips by incubating with 50 μL of 1 mg/mL streptavidin at room temperature for 60 min. Streptavidin tetramers are about 4–6 nm in diameter.[35] Unbound streptavidin molecules were removed by rinsing with PBS. The tips were finally modified with the biotinylated IntP (5 mg/mL) in solution at 4 °C overnight. Then tips were washed several times with PBS and DI water.

A schematic illustration of AFM tip modification process is shown as Fig. S3. Each step in the surface modification of the AFM tips was evaluated by X-ray photoelectron spectroscopy (XPS). Detailed procedures of XPS characterization and data processing are outlined in the Supporting Information Section S.2.

2.6. AFM Imaging and Force-Distance Measurements.

We operated the AFM (Bruker Bioscope Resolve, mounted on a spinning-disc confocal microscope built around a Nikon Eclipse TiE) in quantitative nano-mechanics (QNM) PeakForce capture mode. All images and force curves were collected using ScanAsyst Fluid+ tips (Bruker). Crystal imaging was performed in Tris-EDTA buffer (5 mM Tris-HCl, 1 mM EDTA, pH = 7.5, dissolved in de-ionized water) with crystals affixed to the bottom of a glass petri dish. The AFM line scan rate was set to 1.0 Hz and the peak force tapping frequency was set to 1.0 kHz. The maximum peak force set point was set to 2.0 nN (activated AFM tip on loaded protein crystals) and 1.5 nN (all other conditions). The cantilever deflection sensitivity was measured and the distance between the tip and the sample was corrected for the cantilever deflection for all measurements. AFM has been used to measure the strength of bonds between biological receptor molecules and their ligands.[36–39] Notably, the retract force measured here (50 – 750 pN, under various conditions) is comparable to the forces used by investigators conducting similar experiments to study the receptor-ligand interactions. Specifically, Beebe et al. used a force of 442 ± 17 pN to measure the ligand-receptor bond-rupture force between the streptavidin/biotin partners.[40, 41] Stayton et al. measured the force of detachment by AFM as 433 ± 33 pN between biotin-functionalized tips and streptavidin-functionalized samples.[42] In the present work, there are likely multiple IntP presented on the tip, enabling multi-valent tip-PEG-RGD interactions, and in our experiments, retraction forces are for multivalent interactions. In addition, as a control experiment, to determine if the observed adhesion energy reflects equilibrated molecular interactions, we also operated some of the experiments under the exact same condition but a much slower tapping frequency of 1 Hz (Supporting Information, Fig. S.2). Tip-sample interactions are measured with pN-resolution by the deflection of the cantilever. Analysis of the AFM data was performed in NanoScope (Bruker, Inc.), Origin (OriginLab, Inc.), Python (Version 2.7), and Matlab (Version 2019).

CJ^OPT protein crystal samples were also initially characterized by non-modified AFM tips to confirm crystal immobilization and to verify a clean crystal surface. It is important to independently confirm the porous morphology of the protein crystal surface prior to PEG-RGD interaction measurements, because if crystals are poorly prepared and crosslinked, the surfaces can be obstructed with amorphous (non-crystallized) or aggregated protein molecules. After loading the guest molecules, the (10 kDa) PEG-RGD-loaded crystal was then imaged with an activated AFM tip, and the corresponding force distance (F-D) curves were collected at each pixel in the AFM image. AFM imaging and F-D curve collection was conducted for three different experimental conditions, using combinations of un-modified and activated tips and either loaded or unloaded protein crystals. The three experimental conditions are: (condition a) an un-modified AFM tip on an unloaded crystal, (condition b) an activated AFM tip on an unloaded crystal, and (condition c) an activated AFM tip on a (10 kDa) PEG-RGD-loaded crystal. From each AFM image, F-D curves were manually assigned to one of two classes, corresponding to protein crystal surface features: pores and walls.

2.7. Cell Culture and Imaging.

The experimental procedures for culture of human adipose-derived stem cells (ADSCs), and ADSC imaging on the protein crystals are outlined in the Supporting Information, Section S.3.

3. Results and Discussion

3.1. PEG-RGD Loading.

PEG has an affinity for the interior of CJ^OPT protein crystals and was adsorbed by the crystals from solution within minutes. To confirm loading, a crosslinked CJ^OPT protein crystal in AFM imaging buffer was imaged on a Nikon spinning-disc confocal microscope, during loading. The crystal (approximately 50 μm tall and 80 μm in diameter) was placed on one side, and a fluorescein-labeled PEG (10 kDa) was added to the imaging buffer at a final concentration of 1 μM. The crystal was imaged every 30 s, at five different z positions (separated by 5 μm), near the center of the crystal. The 10 kDa PEG rapidly accumulated in the protein crystal interior. Four confocal images from the center of the crystal, showing the first 3 minutes of diffusion are shown in Fig. 2. Additional time points and diffusion of fluorescein-labeled PEG (5 kDa) are also shown in the Supporting Information, Fig. S7.

Fig. 2. Guest molecules can be loaded into the nanopores of protein crystals by diffusion.

A hexagonal prism CJ^OPT protein crystal (DIC image shown in upper left, with a 50 μm scale bar) was placed on one side and imaged during mPEG₁₀₀₀₀-FITC loading, by confocal microscopy. Four confocal microscopy images (bottom row) taken at 1-minute intervals are shown, from a z-plane near the center of the crystal. (See Fig. S7 in Supporting Information for additional images.)

In this work, we confirmed that protein crystals with unusually large pores can adsorb guest biomolecules with high affinity and non-covalently bind the guest molecules within the nanopores. Presumably, interfacial biomolecule interaction between guest molecules and nanoporous protein crystals affects the diffusion of polyethylene glycol inside the nanopores. Guest molecules with larger size and molecular weight likely have larger adhesion energy, modulating intra-pore diffusion kinetics. In this work, the diffusion of 5 kDa, and 10 kDa polyethylene glycol in the nanopores of protein crystals was imaged by confocal microscopy. (See Fig. S.7 in the Supporting Information.) The intensity of fluorescence of guest molecules versus time is also quantitatively measured during the diffusion process by the confocal microscopy.

3.2. Surface Morphology of the Protein Crystals by AFM

AFM imaging shows the details of the porous CJ^OPT crystal surfaces (Fig. 3 (a–c) and Fig. 5 (a–b)). The CJ^OPT protein crystal surface has a regular honeycomb nanopore structure with features that were consistent among different protein crystal samples. We have imaged the surface morphology under the three conditions described in Section 2.6 and Fig. 3 (d–f). For comparison, we also quantitatively measured the interaction between an inactivated AFM tip with a PEG-RGD-loaded crystal (Supporting Information Fig. S1 (b)). The observed surface structure and morphology of CJ^OPT protein crystals was quite uniform and did not change significantly when imaged under different loading and tip modification conditions. It is remarkable that tips covalently laden with streptavidin (~ 6 nm diameter for streptavidin tetramers[35]) could still be used to resolve the crystal surface clearly as well as to penetrate into the major nanopores. Indeed, per Fig. 7 (b), the fully encumbered tips appeared to penetrate more deeply into the nanopores. Surface modification of the AFM tip with streptavidin, biotinylated IntP, and PEG-RGD alters both the adhesion of the tip to the crystal and the mechanical properties of the tip. Particularly, when adding PEG-RGD to the crystal and imaging with an activated tip, a greater peak force set point (2 nN, instead of 1.5 nN) had to be used to obtain image quality capable of resolving the pores and walls of the protein crystal surface.

Fig. 3. AFM enables simultaneous high-resolution imaging of the protein crystal and measurements of tip-surface interaction.

High-resolution AFM images of (a) an unloaded crystal imaged with a non-modified AFM tip, (b) an unloaded crystal imaged with an activated AFM tip, and (c) a crystal loaded with PEG-RGD imaged using an activated AFM tip. Corresponding force microscopy schematic illustrations of AFM tips with porous protein crystals are shown in (d-f). The condition indices a-c in this figure correspond to the conditions described in Section 2.6.

Fig. 5. The mechanical interaction between AFM tip and nanoporous protein crystals are three-dimensional.

(a) 3D rendering of a portion of the height data from Fig. 3 (a), (b) a perpendicular cross-section view of 3D AFM image of the CJ^OPT protein crystal surface from Fig. 3 (a), across the center of multiple pores in one single line, and (c) an illustration showing possible axial elastic deformation of the protein crystal surface (lateral deformation is not illustrated), and the AFM probe penetration into the pores, “d” represents the effective diameter of a nanopore, while “h” represents the actual depth of AFM tip penetration into a pore.

Fig. 7. Adhesion and penetration are different under various imaging conditions:

(a) The average +/− standard deviation of adhesion energy of pore and wall areas by an integrin peptide-activated tip on a (10 kDa) PEG-RGD-loaded crystal, pore and wall areas by an activated tip on an unloaded crystal, and pore and wall areas by an inactivated probe on an unloaded crystal with each individual measurement marked by a symbol. Detailed data are listed in the Supporting Information Table S1. The corresponding statistical analysis is attached in the Supporting Information Fig. S9; (b) probe penetration depth with error bar signifying the mean +/− standard deviation of probe penetration depth into the nanopores (n > 150 per condition). Detailed data are provided in the Supporting Information Table S1; and (c) Adhesion energy in the pores, normalized by the interacting pore area (adhesion energy divided by average area computed from the depth of penetration of the AFM tip into the pore). The experimental conditions here correspond to the conditions described in Fig. 7 (b). (n = 30)

3.3. Modification of AFM Tips.

The spring constant and tip diameter of each AFM tip used for imaging and quantitative measurements were quantified using hardness and surface roughness standards.[43] The spring constant of a fully activated AFM tip used to collect data in this work was 0.89 N/m, with an estimated tip diameter of 5.87 nm (ETD, data from NanoScope).

Detailed XPS results are shown in Fig. 4 (a–h). XPS was used to characterize the modification of AFM tips. In this study, the high-resolution XPS spectra of N1s peaks confirm each step of modification chemistry. Activation of the tip with biotinylated peptide enables the PEG-RGD complexes to bind the AFM tip via the IntP-RGD interaction. This enhances the interaction between the activated AFM tip and the PEG-RGD-loaded protein crystal pores.

Fig. 4. XPS confirms the chemical activation of AFM tips.

High-resolution XPS spectra of ScanAsyst Fluid+ tips at different stages of modification, in the regions of the N1s and Si2s envelopes: (a) confirms that there is no nitride or ammonium prior to reaction with APTMS, (b) +APTMS (399.69 eV amine and 400.83 eV ammonium), (c) +streptavidin-integrin peptide (399.72 eV amine, 400.80 eV amide, and 401.85 eV ammonium), (d) +PEG-RGD (398.52 eV imine, 399.66 eV amine, 400.78 eV amide, and 401.80 ammonium) proving that PEG-RGD complex can be bound to the activated AFM tip and (e-h) the Si2s from silicon on AFM tip surface during each step of the reaction. Modification attenuated the strength of Si2s signal.^{[44, 45]}

3.4. Interaction of PEG₁₀₀₀₀-RGD and Nanoporous Protein Crystals.

We hypothesized that activated AFM tips would have higher adhesion to the PEG-RGD-loaded crystals, due to interactions between PEG-RGD in the crystal and the Int-P on the AFM tip. The force versus distance data provided a means to test this hypothesis. Bruker’s PeakForce QNM imaging mode captures and records one force-distance (F-D) curve at each pixel of the scanned area of the CJ^OPT crystal surface. To measure multivalent interaction effects between PEG-RGD and the CJ^OPT protein crystals, three different combinations of tip activation and crystal loading were used: (Fig. 3 (d–f)): (condition a) an un-modified AFM tip on an unloaded crystal, (condition b) an activated AFM tip on an unloaded crystal, and (condition c) an activated AFM tip on a PEG-RGD-loaded crystal. Since the force-distance data is collected at each pixel in the image, we can classify each force-distance curve as one of two types: “pore” pixels (in which the AFM tip penetrates a 13-nm diameter crystal pore) and “wall” pixels (in which the AFM tip is interacting with a pore wall). Fig. 3 (a–c) shows that loading PEG-RGD into the nanopores of CJ^OPT protein crystals did not obviously change the surface morphology, nor did it change the nanostructure of the protein crystals. While minor differences might be attributed to crystal-to-crystal variation in growth and crosslinking, the regular honeycomb-like lattice of 13-nm diameter pores was clearly visible in all conditions.

In this work, the forces measured during retraction were between 50 and 750 pN. Fig. 6 (a–b) shows representative F-D curves obtained from both “pore” and “wall” pixels, under various conditions. To select pixels representative of the center of a pore (or wall), a pixel representing a local minimum (or maximum) in the height was selected. When an activated AFM tip interacted with a PEG-RGD-loaded crystal, there was little to no measurable adhesion to the “wall” pixels, but a strong adhesion force was obtained for the “pore” pixels, measured by the negative force in the “retraction” curve. This suggests that the activated AFM tip was experiencing a binding force inside the protein crystal pores (Fig. 6a). However, when the activated tip interacted with a protein containing no PEG-RGD, there was very little adhesion to the crystal at either “pore” or “wall” sites (Fig. 6b). Therefore, the adhesion inside the pores noted in Fig. 6a, was due to the presence of the PEG-RGD.

Fig. 6. Adhesion energy shows differences among different crystal surface locations and loading conditions.

Representative measured data point examples of force-distance (F-D) curves, at both “pore” and “wall” pixels, obtained using (a) an activated AFM tip and a (10 kDa) PEG-RGD-loaded protein crystal, (b) an inactivated AFM tip on an unloaded protein crystal.

The adhesion energy at each pixel for the tip interaction with the protein crystal can be calculated from the integral of the area between the extend force curve and the retract force curve. This energy is illustrated by the shaded area for the “pore” shown in Fig. 6a. By precisely mapping these force curves to the “pore” and “wall” areas of the protein crystal image, we find that pixels collected from pore areas of the nanostructure have generally larger maximum retract force, and longer distance of interaction than the pixels collected from wall areas. As a result, the interaction traces collected within the nanopores have higher adhesion energy than the wall areas. Fig. 6 (a) shows that the interaction of the activated AFM tip with protein crystal nanopore in the retract portions of the curve occurs over 0–30 nm in the z-direction. This distance likely indicates the interaction of multiple PEG-RGD bound along the length the AFM tip as the tip is withdrawn from the pore. During this process, the crosslinked protein crystal surface also deforms elastically under the approximately 1.5 to 2 nN forces applied here Fig. 5 (c). Axial and lateral deformation may also contribute to the penetration distance and induce an extra energy increase by performing extra work.

The high adhesion energy recorded during retraction indicates that guest molecules loaded within each nanopore engage with the AFM tip. As the AFM tip is retracted, these guest molecules continue to form adhesive interactions between the AFM tip and the protein crystal. The activated AFM tip interacting with a PEG-RGD-loaded crystal measures an adhesive force of approximately −0.7 nN, during retraction from the center of a pore. This is much stronger than the adhesive force of an activated tip interacting with the pore of an unloaded crystal or an inactivated tip interacting with an unloaded crystal (Fig. 6). As the tip is retracted the adhesive force is steadily reduced over a distance of about 30 nm. The gradual decrease in adhesive force is likely caused by the sum of many small rearrangements of multiple PEG chains bound along the length of the tip, rearranging their interactions with the protein crystal, and by the rupture of multiple IntP-RGD bonds, as the tip is retracted. In some cases, the derivative of the retraction trace indicates one to two major separations after the molecule-laden tip starts retracting from the nanopore. (See 30 individual F-D curves obtained from pore centers for the activated tip and the PEG-RGD-loaded crystal in Fig. S10 in the supporting information.)

In this work, we randomly selected 30 sets of F-D curves on pores and walls respectively for each of the three experiments described in Fig. 3 (d–f) (180 sets of F-D curves total). As shown in Fig. 7 (a), measuring with an activated AFM tip, when PEG-RGD complexes were not preloaded into a nanoporous protein crystal, minimal differences were observed for average total adhesion energy between pore and wall areas. Similarly, when measuring with inactivated AFM tips on an unloaded protein crystal, the adhesion energy of both pore and wall areas was similar and very low. The average total adhesion energy in the pore areas using an activated AFM tip on a PEG-RGD-loaded crystal reached 2.38 × 10⁻² fJ, which was the highest average total adhesion energy among all conditions. In support of the model that the guest PEG-RGD was only accessible within the host crystal nanopore, for comparison, the average total adhesion energy for the wall areas of a PEG-RGD loaded crystal collected by activated tips was only 7.37 × 10⁻³ fJ, which was similar to the average adhesion energy of the wall areas from an unloaded crystal collected by inactivated tips (7.36 × 10⁻³ fJ). The average adhesion energies for pore areas of an unloaded protein crystal by activated and inactivated AFM tips were 9.80 × 10⁻³ fJ and 8.60 × 10⁻³ fJ respectively. The average total adhesion energy of the pore areas significantly increased with the presence of PEG-RGD.

We also observe a large attractive force encountered in the retraction curve for the “pore” areas when activated AFM tips are interacting with PEG-RGD-loaded protein crystals (as shown in Fig. 6 (a)). This attractive force gradually declines in magnitude over a distance of about 30 nm in tip travel as the tip is retracted. The gradual decline in the adhesion force as the tip is retracted is consistent with a sum of multiple small adhesive interactions, rather than with a few strong interactions. These many small interactions may be consistent with both the dynamic equilibrium of RGD-IntP binding/unbinding within the pore and the rearrangement of PEG chains deformed by the tension induced by the AFM tip motion. Each PEG chain can form multivalent interactions with the protein crystal pore wall. As the tip is retracted some IntP-RGD bonds are ruptured so the tip engages fewer PEG-RGD within the pore thereby reducing the adhesive force.

To further investigate this, Fig. 8 shows a map of the adhesion force at each pixel of the AFM height images from Fig. 3. The addition of PEG-RGD to the crystal causes a consistent increase in the adhesion force of an activated tip with a protein crystal pore (Fig. 7A, Fig. 8C), compared to imaging either an unloaded crystal with an un-activated tip, or an un-loaded crystal with an activated tip (Fig. 8A and 8b). During imaging the AFM tip rasters horizontally across the crystal surface collecting a single line of pixels per second (1 Hz) and tapping the surface at a rate of 1 kHz. At the 512 × 512-pixel resolution used here; each pixel is smaller than 1 nm. Since the pores are larger than a single pixel (13 nm diameter), the AFM tip revisits the same pore on multiple line scans, registering a similar value for the adhesion force each time the tip is near the center of a pore. This suggests that AFM tip and protein crystal pore have very similar interactions during separate line scans (separated in time by 1 s at the 1 Hz scan rate used here). The tip has interacted with the crystal 1000 times (at a peak-force tapping frequency of 1 kHz) during the intervening 1 s, after which the tip returns to the same pore in subsequent rows of pixels. This confirms that the tip is retaining its complement of streptavidin and biotinylated IntP, and that the nature of individual pores is also not changing substantially over the multi-second time scale. During retraction, most PEG-RGD likely remain in the pore, and IntP remains bound to the AFM tip. Hence, the measured adhesion force is most likely due to several IntP-RGD bonds. The consistency of both the surface topography and adhesion measured by the probe when visiting the same pore on multiple line scans also confirms that any deformation of the crystal caused by the tip is elastic.

Fig. 8.

Adhesion maps of the CJ^OPT crystals for all imaging conditions, corresponding to the high-resolution AFM images in Fig. 3 (a–c): (a) an unloaded crystal imaged with a non-modified AFM tip, (b) an unloaded crystal imaged with an activated AFM tip, and (c) a crystal loaded with PEG-RGD imaged using an activated AFM tip. The highest adhesion force occurs at the locations of the pores. In panel C the purple arrows indicate locations in the lattice where a pore is expected, but where there is no apparent local maximum in the adhesion force. These correspond to locations in Figure 3C where the tip did not penetrate into a pore, due to pore blockage.

To our knowledge the equilibrium binding affinity for cyclo-RGDfK and IntP has not been quantitatively established, nor the binding/unbinding rate constants, nor the structure of the complex. RGD binding to authentic integrin domains relies on metal ion-dependent adhesion sites, primarily requiring divalent Ca²⁺ and Mg²⁺.[46] Equilibrium binding experiments for IntP and PEG-RGD in solution suggested a relatively weak interaction (Kd between 5.6 mM and 11.1 mM). (See Table S.2, Fig. S5 and Fig. S6 in the Supporting Information.) While our experiments with crystals also lacked added divalent cations, we did not use chelating agents sufficient to guarantee the absence of trace divalent cations and the environment inside CJ crystals may be enriched in counterions. Despite this caveat, weak observed PEG-RGD-to-IntP adhesion in solution is consistent with the model where PEG-RGD remains within host crystals rather than being fully extracted.

Each 500 nm × 500 nm image in Figs. 3 and 8 contains over 700 nanopores. A small number of locations in the lattice where a pore is expected (six locations in Fig. 8C, marked with purple arrows) have no apparent increase in the adhesion force. These locations correspond to locations where the tip did not penetrate into a pore (compare to Fig. 3C). This can be attributed to obstruction of the pore. Pore obstruction is likely caused by protein molecules adsorbed and crosslinked in off-lattice positions (e.g., surface defects in the protein crystal). These rare cases in Fig 8C where there is no apparent adhesion also are consistent with the tip and each pore retain their respective receptor (IntP on the tip) and ligand (PEG-RGD) in the pore, throughout the imaging experiment.

To further interrogate the underlying biophysics, we collected force-distance curves at two oscillation speeds. When imaging with the PeakForce QNM mode, the tip oscillates with a sinusoidal z-position profile, reaching a minimum in velocity (dz/dt = 0) when the tip is fully extended. At 1 kHz, with a total travel of 200 nm, the tip remains within 1 nm of the fully extended position for 52 μs.[23] This dwell time in the pore should be sufficient for IntP presented by the tip to interact with PEG-RGD molecules, which are themselves sampling the local crystalline environment to find energetically favorable bound conformations. To confirm that the IntP-modified tip can probe the adhesion of the PEG to the protein crystal pores during the 1 kHz oscillation, we also collected force-distance curves using approach-retract experiments with a linear ramp at a 1000 × slower (1 Hz frequency) using an activated AFM tip from PEG-RGD-loaded protein crystal, and with an unloaded protein crystal. Here, the frequency of 1 Hz means that the entire approach-retract cycle takes 1 second. While the speed would technically depend upon the retraction distance, to a first approximation the slow-moving tip would remain within 1 nm of the fully extended position for 52 ms. Despite significantly increasing the time for PEG-RGD to bond the activated tip, no obvious change was observed in the adhesion energy (Supporting Information Fig. S2 (a–b)).

One additional challenge to interpreting these data is that the apparent height difference between “pore” and “wall” pixels varies substantially depending upon the condition. In our previous work, we defined the “penetration depth” as the difference in heigh between a pore (relative height minimum) and a wall (relative height maximum) in a line crossing multiple pore centers. [23] Considering the different distances of tip penetration into the pores, the average adhesion energy within the nanopores can be normalized by the average individual pore area of interaction, more specifically, calculated by approximating the area of a 13-nm diameter cylindrical pore [21–24, 26] $\left(A_{\mathit{\text{pore}}\mathit{\text{surface}}}=13nm\cdot h\cdot \pi \right)$ at the local depth of penetration $h$ of the AFM tip (as shown in Fig. 5 (c)). We measured pore depths in this work on over 150 pores per condition. In this work, an average of normalized adhesion energy for the pore areas of the PEG-RGD-loaded CJ^OPT protein crystals, collected by an activated AFM tip, was 1.14 × 10⁻⁵ fJ/nm², with an average probe penetration depth of 15. 8 ± 2.7 nm. For comparison, the average penetration depth on unloaded crystals was respectively 4.5 nm and 8.0 nm for AFM tips before and after activation. The normalized adhesion energy for pore areas of unloaded protein crystal by activated and inactivated AFM tip reached 1.64 × 10⁻⁵ fJ/nm² and 8.10 × 10⁻⁶ fJ/nm² respectively. This somewhat counterintuitive result suggests that favorable interactions are a more important driver for probe depth than steric repulsion, since a steric model would predict a reduced penetration depth for AFM tips that are encumbered with bulky streptavidin tetramers as well as bound PEG-RGD. Penetration depth will make a difference on the effective interacting area when the tip penetrates the pore, and thereby affect the adhesion energy of interaction. The uncertainties on these calculated normalized adhesion energies are large (Fig. 7c), but we can conclude that the increased adhesion energy observed for the activated tip with the PEG-RGD-loaded crystal is also associated with an increase in the approximate surface are over which the tip may be interacting with the protein crystal pore. Single-molecule studies in the future could further elucidate the interactions of individual PEG-RGD with CJ^OPT nanoporous protein crystal nanostructures.

3.5. Modulation of Cell-adhesion to Protein Crystals.

We hypothesized that protein crystals loaded with PEG-RGD could be used as an unusual platform for the display of the RGD peptide to cells, thereby modulating cell attachment and spreading behavior. To investigate this possibility, we cultured adipose-derived stem cells (ADSCs) on protein crystals with and without guest PEG-RGD. Via confocal microscopy (Fig. 9) we observed that the unloaded protein crystal permits cell attachment to the surface (Fig. 8, left column). While the number of cells attached was similar, we did observe differences between control crystals and PEG-RGD-loaded crystals. Most clearly, cells on the PEG-RGD-loaded crystals, loaded with either PEG₅₀₀₀-RGD or PEG₁₀₀₀₀-RGD exhibited more well-developed actin stress fibers. Although the RGD is not covalently bound to the protein crystals, the force required to remove the guest molecules from the protein crystals is sufficiently high that the cells can form adhesive contacts. The formation of actin stress fibers is important for multiple cellular processes, including cell migration and cellular morphogenesis. Furthermore, stress fiber formation is evidence that the adhesive contacts provided by the PEG-RGD are sufficiently strong to enable the cells to form cell-ECM-like contacts with the modified protein crystal surface, that can endure the tension necessary for cell spreading. We therefore propose that PEG-RGD-loaded porous protein crystals could be used to tune important cell behavior, by tuning the strength of the PEG-RGD binding to the protein crystal. This could be accomplished by varying the PEG length, rather than by varying the surface density of adhesion ligands. RGD is an adhesion ligand for multiple cell-surface receptors, including integrins that permit cell attachment and mechanotransduction. This strategy could be expanded to other more specific cell adhesion peptides to prepare surfaces that selectively bind different cell phenotypes with different strengths. Adhesion ligands conjugated to shorter PEG chains would provide weaker adhesive contacts, or would prohibit mechanotransduction of some cell types, by breaking free from the surface when pulled. In contrast other adhesive ligands conjugated to longer PEG chains would provide stronger adhesive contacts. In this way the same surface could be tuned to selectively present different adhesiveness to multiple cell types.

Fig. 9. Loading of guest molecules mechanically modulates the spreading of attached ADSCs.

Confocal microscopy images of ADSCs cultured on large protein crystals (original magnification is 60 ×). Cells are cultured on either an unloaded protein crystal (left column), a protein crystal loaded with PEG₅₀₀₀-RGD (middle column), or PEG₁₀₀₀₀-RGD (right column). The top row shows merged images of the red (rhodamine phalloidin, for F-actin) and blue (DAPI for cell nuclei), channels, in the middle and bottom rows, respectively. The protein crystal exhibits some auto fluorescence, appearing also in the blue channel. Between 2 and 5 protein crystals were imaged for each condition, and representative images are shown, with between 30 and 50 cell nuclei on each crystal surface. Scale bars on all images represent 50 μm.

Next, we sought additional evidence to support the hypothesis that cell adhesion and spreading was dependent on the RGD domain (as opposed to the presence of PEG). Passivation of the protein crystal surface with covalent attachment of PEG results in no cell adhesion (not shown).

Cells use adhesive contacts with their surrounding matrix to probe the local mechanical properties and thereby make cell fate decisions. For example, substrate or matrix mechanical properties are a key driver of mesenchymal stem cells toward adipogenic or osteogenic differentiation,[47] the transdifferentiation of vascular smooth muscle cells in arterial medial calcification,[48] prometastatic signaling in tumors,[49] and macrophage polarization.[50] To achieve precise control over the downstream effects of adhesion ligand presentation, it is therefore essential to control the restoring force with which the surrounding matrix responds to the cell-imposed contractile forces. We propose that materials with tunable nanostructures containing non-covalently attached adhesion ligands can provide signals to adhered cells; rather than modifying the stiffness of the substrate itself, non-covalent attachments can be tuned to rupture if cells pull with too much force, thereby obviating undesirable stiffness-induced responses, such as metatstatic transition, calcification, or inflammation. In principle, if the contractile force imparted by the cell on a specific ligand is greater than the force required to rupture the non-covalent bond between the ligand and the material, then the cell would be unable to attach. By decoupling the mechanotransduction from the material mechanical properties, different apparent stiffness values could be presented to different cell types through specific receptor-ligand interactions using non-covalent attachments of different strengths. The presentation of ligands with tunable and cell-specific elasticity or rupture energy will offer a new dimension through which peptide ligands can be used to control cell behavior, or to simultaneously control multiple behaviors of cells that express different receptors.

Conclusions

In this study, the nanoporous surface morphology of CJ^OPT protein crystal was observed, and imaged in the liquid phase via high-resolution AFM, revealing details of the nanoporous crystal surface. PEG-RGD as a cell adhesion ligand, was successfully loaded into the nanoporous protein crystals through diffusion, confirmed by the confocal microscopy results. As hypothesized, activated AFM tips that display a peptide that mimics integrin receptors (IntP) had a dramatically stronger interaction with crystal nanopores that were pre-loaded with PEG-RGD. Future work could further elucidate the molecular details underlying this interaction. In particular, it would be useful to conclusively determine whether PEG-RGD molecules remain with the host crystal or the AFM tip during extraction. Future control experiments where the PEG-RGD is covalently attached to either the tip or the host crystal would be particularly useful. Nonetheless, the optimized AFM characterization in this work quantitatively measures and analyzes the mechanical behaviors, as well as the nanoscale variations in the adhesion energy, between the interactions of PEG-RGD and nanoporous protein crystals. Furthermore, ASDCs were successfully attached to the surface of nanoporous protein crystals. Confocal imaging results also suggested that non-covalently attached PEG-RGD can alter the spreading behavior of ADSCs.

Data Availability

The processed data required to reproduce these findings are available to download from datadryad.org. Kipper, Matt J. et al. (2023), Poly(ethylene glycol) terminated with a cell adhesion peptide: Interactions with a highly porous protein crystal, Dryad, Dataset, https://doi.org/10.5061/dryad.931zcrjrb.