Protein-Crystal Interface Mediates Cell Adhesion and Proangiogenic Secretion

Abstract

The nanoscale materials properties of bone apatite crystals have been implicated in breast cancer bone metastasis and their interactions with extracellular matrix proteins are likely involved. In this study, we used geologic hydroxyapatite (HAP, Ca₁₀(PO₄)₆(OH)₂), closely related to bone apatite, to investigate how HAP surface chemistry and nano/microscale topography individually influence the crystal-protein interface, and how the altered protein deposition impacts subsequent breast cancer cell activities. We first utilized Förster resonance energy transfer (FRET) to assess the molecular conformation of fibronectin (Fn), a major extracellular matrix protein upregulated in cancer, when it adsorbed onto HAP facets. Our analysis reveals that both low surface charge density and nanoscale roughness of HAP facets individually contributed to molecular unfolding of Fn. We next quantified cell adhesion and secretion on Fn-coated HAP facets using MDA-MB-231 breast cancer cells. Our data show elevated proangiogenic and proinflammatory secretions associated with more unfolded Fn adsorbed onto nano-rough HAP facets with low surface charge density. These findings not only deconvolute the roles of crystal surface chemistry and topography in interfacial protein deposition but also enhance our knowledge of protein-mediated breast cancer cell interactions with apatite, which may be implicated in tumor growth and bone metastasis.

Graphical abstract

1. Introduction

Protein-crystal interactions play a fundamental role in many biological processes such as biomineralization [1–4]. For example, bone is a natural protein-crystal composite containing nanoscopic bone apatite crystals highly organized within a fibrillar protein extracellular matrix (ECM) that consists predominantly of collagen I [5]. Apatite surface chemistry and topography (ranging from the nano- to the microscopic level) are important mediators of protein-crystal interactions. Biological bone apatite has a platelet-like shape with (100) being the primary face, which interacts strongly with water [6]. Synthetic and geologic hydroxyapatites (HAP, Ca₁₀(PO₄)₆(OH)₂) are closely related to bone apatite but also develop other faces such as (001) and (101), although (100) is usually the prevalent face due to its highest hydrophilicity and stability in aqueous environment [7,8]. Biological and synthetic nanocrystalline apatites have high surface reactivity, such as ion exchange and protein adsorption, due to the presence of a metastable hydration layer [9]. The interactions between this sub-nanometer-thick hydration layer and crystal surfaces depend on face-specific HAP surface chemistry, and can significantly affect the adsorption of biomolecules [10–12]. A recent study reports that apatite-collagen interaction energies strongly influence the molecular scale organization of collagen on apatite surfaces [13]. Furthermore, interactions between mineral-modulating proteins and biologically relevant crystals are found to be primarily electrostatic in nature [14]. For example, fibronectin (Fn), another key skeletal ECM protein, adsorbs preferentially onto purely ionic crystal surfaces without structural water molecules incorporated in the crystal lattice [15]. On the other hand, both nanoscale and microscale surface topography have been shown to influence Fn adsorption (including total amount, spatial distribution, and molecular rigidity of Fn) as well as subsequent cell adhesion and signaling, although whether Fn conformational changes play a role in this process remains unclear [16–21].

Although the skeletal ECM is primarily composed of collagen I, Fn is the first bone matrix protein synthesized by osteoblasts and is required for subsequent deposition of collagen I [22–24]. In fact, the continuous presence of Fn is essential for maintaining the integrity of the mature bone collagen matrix [25]. Moreover, bone metabolism, including osteogenesis, induction of osteoblast differentiation, and survival of osteoblasts, all depend on interactions between osteoblasts and Fn [26,27]. In addition to cellular Fn originating from osteoblasts, circulating plasma Fn produced by the liver is also incorporated in the bone matrix. Interestingly, circulating plasma Fn represents the predominant source of Fn affecting bone mineralization and matrix properties [28]. Fn is a critical mechanotransducer whose conformational changes (e.g., induced by interactions with crystal surfaces) result in either exposure or disruption of most of its binding sites. Among these sites are binding sequences for integrins, which are cell surface receptors regulating cellular responses to chemical, physical, and mechanical signals from the microenvironment [29–34]. Previous work using self-assembled monolayer substrates has reported that surface charge controls cell adhesion via Fn-mediated integrin engagement [35].

There has been increasing evidence that the nanoscale materials properties of bone apatite crystals likely modulate the pathogenesis of breast cancer bone metastasis [36–38]. Bone metastasis frequently occurs in patients at advanced stage of breast cancer and remains a major source of mortality in these patients [39]. Synthetic HAP has been widely used in previous studies to mimic bone apatite, although bone apatite has lower crystallinity and higher solubility; moreover, the crystal size, chemical composition, and distribution of bone apatite vary as a function of age and disease progression [40–44]. The presence of HAP in mineralized tumor models increases tumor cell adhesion, proliferation, and secretion of both proangiogenic and proinflammatory factors [36]. Furthermore, HAP materials properties including particle size, crystallinity, carbonate incorporation, and morphology, all have combined and/or individual effects on protein adsorption and breast cancer cell behaviors [37,38,45]. However, synthetic HAP nanoparticles usually present multiple faces and tend to form agglomerates, which makes it difficult to deconvolute effects of face-specific surface chemistry and surface topography during interactions with ECM proteins and tumor cells.

In this study, we investigated the individual effects of HAP surface chemistry and nano/microscale topography on both Fn molecular adsorption and subsequent cell behaviors. To obtain surfaces with homogeneous structural and chemical properties, Geiger el al. synthesized calcium-(R,S)-tartrate single crystals substrates that were tens of microns large for the study of epithelial cells adhesion [46,47]. Later, geologic calcite single crystals with identical surface chemistry but different surface roughness were used as model substrates for studying nano-topography sensing by osteoclasts [48]. Herein, we adopted a similar strategy by using geologic apatite single crystals to ensure uniform and well-defined surface chemistry (by controlling crystallographic orientation) and tune precisely nano/microscale surface topography (by controlling surface roughness). This strategy allowed us to (i) independently control HAP surface chemistry and topography, and (ii) avoid cytotoxicity associated with cellular uptake of HAP nanoparticles. Both molecular conformation and overall quantity of Fn adsorbed onto HAP facets were determined by Förster resonance energy transfer (FRET). MDA-MB-231 cells were subsequently cultured on Fn-coated HAP facets to evaluate the role of Fn-HAP interactions on cell adhesion and proangiogenic secretion, with likely implications in breast cancer bone metastasis.

2. Materials and methods

2.1 Geologic HAP crystals

Geologic apatite crystals with natural crystal termination faces were from Madagascar (Etsy, Inc). To determine the purities and elemental compositions of apatite crystals, a small portion was cut out of each crystal and ground into powders for powder X-ray diffraction (pXRD) and inductively coupled plasma atomic emission spectrometry (ICP-AES) analysis, respectively. The major phase of each crystal was determined to be HAP by pXRD (Scintag Inc. PAD-X theta-theta X-ray diffractometer, Cu Kα 1.54 Å, accelerating voltage 40 kV, current 40 mA, continuous scan, 1.0 deg/ min). For elemental composition analysis, 3 mg of powders were dissolved in 25 mL 5% HNO₃ at 80 °C overnight. The solutions were then diluted 1:5 and analyzed using an ICP-AES spectrometer (Spectro, Ametek Material Analysis Division).

2.2 HAP facets preparation

HAP crystals were cut along natural faces to generate facets with two types of surface chemistry, (100) and (001), using a wafer cutting diamond saw. Each type of surface chemistry refers to a specific geometrical arrangement of ions and ionic groups, defined by the unit cell and the crystallographic orientation, resulting in specific surface properties (including interfacial energy, surface charge density, etc.) of each crystallographic facet. The crystallographic orientations of the facets were determined via XRD. These facets were then mechanically polished with a precision polisher (Allied High Tech, MultiPrep™ System) to obtain two types of topography/roughness. Micro-rough HAP facets 100M and 001M were polished using 30 µm diamond lapping films, resulting in a uniform array of grooves on the crystal surface. Nano-rough HAP facets 100N and 001N were polished using 30 µm, 9 µm, 1 µm diamond lapping films in sequence, and finished with 20 nm colloidal silica nanoparticles to smoothen any groove left by the lapping films. Lastly the nano-rough facets were polished against a polishing cloth in water for 5 min to remove the remaining silica nanoparticles. The dimension of original apatite crystals was approximately 1 cm. The facets used in this study were typically 6 ~ 8 mm in width and length, 2 mm in thickness. All lapping films, polishing cloths, and colloidal silica nanoparticles were obtained from Allied High Tech, Inc.

2.3 HAP surface chemistry and roughness characterization

The surface chemistry of HAP facets was characterized by measuring their surface zeta potential using phase analysis light scattering with Zetasizer Nano-ZS (Malvern Instruments Ltd. ZEN3600) [49]. The facets were mounted on a dip cell, immersed in aqueous buffer containing polystyrene latex standard at pH 9 (Malvern, DTS1235 ZP Transfer Standard, −42 +/− 4.2 mV), and placed between the electrodes of the cell. The mobility of tracer particles in the vicinity of the charged surface was measured at 5 displacements (located at 125 µm to 625 µm from the HAP surface), with 3 measurements per position, and fitted with linear regression to extrapolate the surface zeta potential at the HAP surface. The roughness of micro-rough HAP facets 100M and 001M were measured with an optical profiler (ADE phase shift microXAM optical interferometric profiler), using a 50x objective. The roughness of nano-rough HAP facets 100N and 001N were measured via atomic force microscopy (AFM, Veeco Dimension 3100), using silicon cantilevers holding tetrahedral silicon tips of radius 7 nm (Olympus, spring constant 26 N/m), in tapping mode, with scanning areas of 10 µm × 10 µm (512 samples/line).

2.4 Fibronectin FRET labeling and adsorption

Fibronectin (Fn) was obtained from Life Technologies, NY. AlexaFluor 488 succinimydyl ester and AlexaFluor 546 maleimide (Invitrogen, CA) were used to label Fn for intramolecular FRET as previously described by Baneyx et al. [50] and Smith et al. [30]. Labeling ratios and Fn concentrations were determined using a DU®730 UV/Vis spectrophotometer (Beckman, IN) at 280 nm, 495 nm, and 556 nm. Calibration of FRET labeled Fn in solution was carried out in guanidine hydrochloride (GdnHCl) solution at concentrations of 0 M, 2 M and 4 M to obtain FRET ratios, defined as acceptor/donor intensity ratios (I_A/I_D), as a function of protein denaturation.

Stock solutions of Fn were diluted to 50 µg/mL using phosphate buffered saline (PBS) and then used for coating HAP facets. For FRET experiments, the diluted Fn solution contained 10% FRET labeled Fn and 90% unlabeled Fn to avoid intermolecular FRET, so that only intramolecular FRET was measured to assess single molecular conformation of Fn. A droplet of 50 µL diluted Fn solution was incubated on each HAP facet (surface area approximately 0.35 cm²) for 24 h at 4 °C. 8-well Lab-Tek chambered borosilicate coverglass (Thermo Scientific, IL) was used as control, where 130 µL diluted Fn solution was added per well for 24 h at 4 °C. After incubation, the HAP facets and control coverglass were washed 3 times with PBS and kept immersed in PBS.

2.5 FRET data acquisition

Samples were imaged with a Zeiss 710 confocal microscope (Zeiss, Munich, Germany). 16-bit z-stack images were acquired using the C-Apochromat water-immersion 40x/1.2 objective, a pinhole of 2.4 AU (2 µm section), 488 nm laser with 5% laser power, pixel dwell time of 6.3 µs, PMT1 and PMT2 gains of 700 V, and z step size of 1 µm. FRET labeled Fn molecules were excited with a 488 nm laser line, and emissions from donor and acceptor fluorophores were simultaneously collected in the PMT1 channel (514–526 nm) and the PMT2 channel (566–578 nm), respectively, while associated brightfield images were acquired in the T-PMT channel (transmitted light detector). These z-stack images were analyzed with a customized Matlab code to generate FRET ratio (I_A/I_D) images and histograms. Only Fn regions in focus were analyzed in each z slice to calculate mean FRET ratio for each field of view.

2.6 Cell culture and seeding

We developed a HAP/Polydimethylsiloxane (PDMS) system for cell culture studies (Fig. 1). First, HAP facets were embedded inside PDMS slabs and inserted into PDMS sample holders. Next, PDMS wells containing HAP/PDMS slabs were incubated with SuperBlock (Thermo Scientific, IL) for 1 h at room temperature to block nonspecific binding, washed 3 times with DI water, and dried in air. Then, each HAP surface was exposed by removing the PDMS thin layer (approximately 5 mm x 7 mm in size, 250 µm in thickness) that covered the HAP crystal in the PDMS slab. After sterilizing the HAP/PDMS system with UV light for 1 h, HAP surfaces were washed with sterile PBS three times, and a 50 µL droplet of Fn solution (50 µg/mL, unlabeled Fn) was incubated on each HAP facet for 24 h at 4 °C. For control coverglass, 130 µL Fn solution (50 ug/mL, unlabeled Fn) was added per well. After washing 3 times with PBS, these Fn coated surfaces were used for cell culture experiments.

Fig. 1.

Preparation of HAP/PDMS system for cell culture studies. Briefly, HAP facets were embedded inside PDMS slabs and inserted into PDMS sample holders. PDMS wells containing HAP/PDMS slabs were incubated with SuperBlock for 1 h to block nonspecific binding. HAP surface was then exposed by removing the PDMS thin layer that covered the HAP crystal in the PDMS slab. After sterilizing the HAP/PDMS system with UV light for 1 h, HAP surfaces were incubated with Fn solution for 24 h before seeding MDA-MB-231 cells. Extra media were added 30 min after initial seeding and cells were cultured for either 6 h or 24 h.

MDA-MB-231 breast cancer cells (ATCC® HTB26™) were cultured in α-MEM media supplemented with 10% FBS and 1% P/S at 37 °C in 5% CO₂, with media refreshed every 48 h. At approximately 80% confluency, cells were detached with trypsin-EDTA and used for cell seeding. A 50 µL droplet of cell solution containing 12,000 MDA-MB-231 cells was added onto each Fn coated HAP facet (seeding density circa 3 × 10⁴ /cm²). After 30 min, additional 650 µL α-MEM media supplemented with 1% FBS and 1% P/S was added into each well, resulting in a total volume of 700 µL media per well. For control coverglass, 114 µL cell solution containing 27, 000 MDA-MB-231 cells were added per well (seeding density circa 3 × 10⁴ /cm²); after 30 min, additional 286 µL media was added leading to a total volume of 400 µL media per well. After culturing at 37 °C in 5% CO₂ for either 6 h or 24 h, cell viability and adhesion were analyzed, and culture media were collected for secretion quantification.

2.7 Cell viability and adhesion

Breast cancer cell viability was assessed with a LIVE/DEAD® Viability/Cytotoxicity Kit (Molecular Probes, OR) based on the simultaneous determination of live and dead cells with calcein AM and ethidium homodimer (EthD-1), respectively. Nonfluorescent cell-permeant calcein AM converts to intensely green fluorescent calcein, which is membrane impermeable, after cleaved by intracellular esterase in live cells. EthD-1 can only enter damaged membranes of dead cells and produces a bright red fluorescence upon binding to nucleic acids. After collecting culture media, samples were gently washed once with warm PBS, and incubated in Live/Dead working solutions (PBS containing 2 µM calcein-AM and 4 µM EthD-1) for 30 minutes at 37 °C in 5% CO₂. Then tiling images were acquired in phenol red free DMEM/F12 media with an inverted Zeiss LSM880 confocal/multiphoton microscope (Zeiss, Munich, Germany), using a Fluar 5x/0.25 objective. The entire surface area of each sample was imaged and the number of live cells was used to quantify cell adhesion.

2.8 VEGF and IL-8 secretions

The media collected after culturing for 6 h or 24 h were centrifuged for 15 min at 13500 rpm and supernatants free of cell debris were utilized to quantify VEGF and IL-8 contents with commercial VEGF and IL-8 ELISA kits (Life Technologies). The media collected at 6 h time point were concentrated 6 fold with 10K centrifugal filter devices (Amicon® Ultra, 10,000 NMWL) for measuring VEGF, or 3 fold with 3K centrifugal filter devices (Amicon® Ultra, 3,000 NMWL) for measuring IL-8. The media collected at 24 h time point were used without further concentrating for measurement. VEGF and IL-8 amounts were normalized by the number of live cells for each corresponding sample.

2.9 HAP facets cleaning

After each experiment, cells were removed from HAP facets by washing three times with PBS and incubating in 0.5% trypsin in PBS at 37 °C for 30 min. Then, after washing three times with PBS, remaining cell residues were cleaned by cell scrapers and HAP facets were immersed in PBS containing 5% sodium dodecyl sulfate (SDS) overnight, sonicated in 0.1 M NaOH, and rinsed with DI water and ethanol.

2.10 Statistical analysis

One-way ANOVA with Tukey’s post test and Student’s t-test were used to determine statistical significance between conditions in GraphPad Prism (GraphPad Software, Inc., CA). In all cases, p < 0.05 is indicated by a single star, p < 0.01 by two stars, and p < 0.001 by three stars.

3. Results and discussion

3.1 Geologic apatite crystal: mimic of bone apatite

Geologic apatite is ubiquitous in various rocks because of its overall stability in geological processes. Many minor and trace elements can substitute in the apatite lattice, and the degree of substitution affects the crystal color and lattice stability [51,52]. The elemental compositions of apatite crystals were characterized by ICP-AES (Supplementary Fig. 1A). Major elements in all three apatite crystals used in this study, named G1, G2, and G3, were found to be Ca and P. The Ca/P ratios were determined as 1.74, 1.70, and 1.64 for G1, G2, and G3, respectively, which were comparable to the Ca/P ratio of stoichiometric HAP (1.67). Less abundant elements such as Si, S, and Mg were also present. Elements with concentrations lower than 0.005 wt% are not shown in the graph. All apatite crystals also contained rare earth elements (Ce, La, Yb, and Y) that could result in laser-induced luminescence (data not shown) [53]. The intensity of such autofluorescence was quantified for all three apatite crystals used and found negligible compared with signal from FRET labeled Fn used in our FRET analysis.

The purities of these apatite crystals were determined by pXRD (Supplementary Fig. 1B). The pXRD pattern of each crystal was identified to match with HAP, as compared with other apatitic phases, such as chlorapatite and fluorapatite, suggesting that the major phase was HAP, although chloride and fluoride substitutions were also likely present. All peaks matched with HAP except the peak at 29.4°, indicating the presence of a minor phase, probably calcite (CaCO₃), whose strongest peak is located at 29.5° (ICDD PDF no. 01-072-1650). The intensity of this peak was also proportional to the Ca/P ratio obtained from ICP-AES, i.e., the crystals with more Ca had a stronger calcite peak. The low peak intensity suggested that this minor phase was only present in very small amount, resulting in maximum 4% increase in Ca/P ratio as compared with stoichiometric HAP. Moreover, it likely exists as crystal inclusions locally trapped within the apatite crystal. Thus we do not expect the presence of this minor phase to have any major (systematic) effect on Fn adsorption.

3.2 Hydroxyapatite facets with controlled surface chemistry and nano/micro topography

Four types of facets with two types of surface chemistry, (100) and (001) crystallographic orientations, and two levels of surface roughness, nano-rough and micro-rough, were generated from geologic apatite crystals. The primary natural termination faces of these crystals were identified by XRD to be (100) and (001), and HAP facets were cut along these two types of natural faces. These HAP facets were then polished to generate two types of topography/roughness. Nano-rough facets 001N and 100N were characterized by AFM (Fig. 2A and Supplementary Fig. 2A), while micro-rough facets 001M and 100M were analyzed by optical profilometry (Fig. 2B and Supplementary Fig. 2B). The surface chemistry of all HAP facets was assessed by measurement of their surface zeta potential in aqueous buffer. Average roughness and surface zeta potential results are summarized in Table 1.

Fig. 2.

Representative (A) AFM image of a nano-rough (001) HAP facet and (B) optical profiler image of a micro-rough (001) HAP facet, with the line profile of a cross section marked in green shown at the bottom of each image. Fn conformation schematics show that the nano-roughness level is comparable to the diameter of an individual Fn module, while the micro-roughness level is comparable to the contour length of a full Fn molecule in extended conformation. Corresponding AFM and optical profiler images for (100) facets are shown in Supplementary Fig. 2.

Table 1.

Summary of surface roughness and surface zeta potential of HAP facets.

Sample ID	Average Roughness Ra [nm]a	Surface zeta potential [mV]b
100N	0.71 ± 0.38	−42.4 ± 3.0
001N	0.54 ± 0.26	−32.1 ± 3.5
100M	198.3 ± 35.4	−28.1 ± 2.8
001M	233.7 ± 34.4	−21.8 ± 1.7

^aAll data represented as means and standard deviations, N = 7 for nano-rough facets 100N and 001N, N = 14 for micro-rough facets 100M and 001M.

^bSurface zeta potential uncertainty was defined as the 95% confidence interval of the Y-intercept (apparent zeta potential) in each linear regression fit (Supplementary Fig. 4). Experiments were repeated three times showing consistent trends, and only results from one experiment are shown here.

Nano-rough facets 100N and 001N both had average roughness values below 1 nm. This sub-nanometric roughness was chosen because it compared with the size of individual beta sheet modules (2 to 3 nm in diameter) that comprise Fn molecules (Fig. 2A, schematics). Micro-rough facets 100M and 001M both had average roughness values around 200 nm, with rough textured polishing grooves covering their entire surfaces oriented along the rotating direction of the platen underneath the sample surface during polishing. Such roughness was picked as it was on the order of the contour length of an extended Fn molecule (Fig. 2B, schematics). To further illustrate the topographical features of polishing grooves, we acquired AFM images of micro-rough facets (scan size 30 µm x 30 µm, Supplementary Fig. 3) and used line profiles perpendicular to the polishing direction to quantify the peak-to-peak distance: average distance was 0.64 ± 0.39 µm for 100M, and 0.68 ± 0.40 µm for 001M (Mean ± SD, N = 3, n = 140 and 141 for 100M and 001M, respectively).

In terms of surface chemistry, (100) facets had more negative surface zeta potentials than (001) facets, regardless of roughness, indicating denser surface charge of (100) facets. Moreover, nano-rough facets also showed more negative surface zeta potentials than their micro-rough counterparts, regardless of crystallographic orientation. The roughness-induced changes in surface zeta potential are likely associated with the heterogeneity of surface chemistry of micro-rough facets, caused by the presence of polishing grooves. To further characterize the effect of surface roughness on hydrophobicity, we performed contact angles measurements (in deionized water) on micro- and nano-rough facets of identical surface chemistry (100). Contact angles of 72.1 ± 3.0 and 44.8 ± 2.8 degrees measured for 100N and 100M, respectively (Mean ± SD, N = 3, sessile drop method), indicated that increased roughness resulted in lower contact angle (hence decreased surface hydrophobicity). To illustrate the inherent difference between (100) and (001) facets, because of the known instability of (001) face in water [11], we refer to previous computation study by Lin et al, in which it was shown that (010), which is equivalent to (100) in hexagonal lattice, had higher immersion energy (or affinity for water) when compared with the (001) face [7]. It should be noted that although mechanical polishing introduces new nano-facets on the surfaces, the surface properties of the crystallographic planes along which they were polished were still maintained (at least to some extent), as supported by the different surface charge densities measured for these facets. Thus, on average, the surface chemistry of each facet was still different rather than randomized even after polishing.

3.3 Molecular conformation of fibronectin adsorbed onto hydroxyapatite nano- and micro-rough facets

To investigate the effect of HAP surface chemistry and topography on Fn adsorption, FRET imaging was performed to quantify both the molecular conformation and the amount of Fn adsorbed onto the four types of HAP facets described in Table 1. A diluted Fn solution carrying trace amount of FRET labels (50 µg/mL, 10% FRET labeled) was used for incubation to ensure that only intramolecular FRET was measured to assess single molecular conformation of Fn [30]. Donor and acceptor fluorophores were imaged (confocal z-stacks) simultaneously (Fig. 3A, B), while associated brightfield images were recorded in the transmitted light channel (Fig. 3C). The donor channel collects the fluorescence emitted by donor fluorophores (green in panel A) when excited directly by a green (488 nm) laser, while the acceptor channel collects the fluorescence emitted by acceptor fluorophores (red in panel B) excited indirectly by non-radiative energy transfer (through dipole-dipole interactions) from nearby donors. The transmitted light channel shows the surface features of the HAP facets. FRET ratio was defined as acceptor/donor intensity ratio (I_A/I_D), and mean FRET ratio was calculated for each image using color-coded FRET ratio maps and FRET ratio histograms (Fig. 3D, E). As determined in our FRET calibration (Supplementary Fig. 5) and correlated with previously published Fn circular dichroism data [30], FRET ratios were high when Fn had compact conformations (0 M GdnHCl), decreased as Fn became extended (0 M ~ 2 M GdnHCl), and deceased further when Fn started losing tertiary structure (2 M ~ 4 M GdnHCl), i.e. when some Fn type-III beta sheet modules (magenta ovals/lines in Fig. 3E schematics) started unfolding.

Fig. 3.

Confocal images of a micro-rough (001) HAP facet coated with FRET labeled Fn after 24 h incubation at 4 °C: (A) donor channel, (B) acceptor channel, and (C) transmitted light channel. (D) Color-coded FRET ratio map, with high FRET ratio color coded in red (compact Fn) and low FRET ratio in blue (unfolded Fn). (E) FRET ratio histogram with schematics of Fn conformations correlated to the FRET calibration values reported in Supplementary Fig. 5 (compact, loss of quaternary structure when extended, and loss of tertiary structure when type-III modules represented by magenta ovals/lines start unfolding). Scale bars 50 µm. Confocal images and FRET analysis for a nano-rough (100) HAP facet and a micro-rough (100) facet are shown in Supplementary Fig. 6 and Supplementary Fig. 7, respectively.

We first assessed the molecular conformation of Fn adsorbed onto HAP facets via FRET (Fig. 4A). FRET ratios of Fn adsorbed onto (100) facets were higher than those onto (001) facets regardless of roughness, indicating that Fn adsorbed in more compact conformations onto (100) oriented HAP facets. Additionally, FRET ratios increased with increasing roughness regardless of crystallographic orientation, suggesting that Fn conformation became more compact when adsorbed onto rougher HAP surfaces. Lab-Tek™ chambered coverglass with sub-nanometer surface roughness was used as control in all experiments.

Fig. 4.

(A) FRET ratios and (B) quantities of Fn adsorbed onto HAP facets with varied surface chemistry and topography, after incubation at 4 °C for 24 h. Data shown were obtained from 9 to 10 fields of view per sample. Coverglass was used as control in all experiments. In all cases, p < 0.05 is indicated by a single asterisk, p < 0.01 by two asterisks, and p < 0.001 by three asterisks.

We next quantified the amount of Fn adsorbed per unit volume onto HAP facets by measuring the sum of donor and acceptor fluorescence intensities (Fig. 4B). The sums of donor and acceptor intensities were larger for micro-rough facets than nano-rough facets, regardless of surface chemistry, suggesting that more Fn molecules adsorbed onto micro-rough HAP facets. Additionally, the amount of Fn adsorbed seemed to be independent of surface chemistry (no significant difference was noted). Importantly, the results in Fig. 4B reflect the amount of Fn adsorbed per unit volume, and may not be relevant to Fn adsorbed per unit surface area. The specific surface area per unit volume of micro-rough facets was characterized from AFM images (Supplementary Fig. 3, scan size 30 µm × 30 µm). Specific areas of 0.67 ± 0.06 µm⁻¹ and 0.58 ± 0.05 µm⁻¹ were measured for 100M and 001M, respectively (Mean ± SD, N = 3). On nano-rough facets, the roughness-induced increase in surface area per unit volume was found negligible. Hence the higher level of fluorescence intensities measured on micro-rough facets could result from the larger surface area of HAP per voxel (0.4×0.4×1.0 µm³), rather than from the larger quantity of Fn adsorbed per unit surface area. Nevertheless, our Fn conformation data (Fig. 4A) suggest that Fn adopted more compact conformations and thus could occupy less space when adsorbed onto rougher facets, implying that more Fn could be adsorbed per unit surface area onto micro-rough facets. Collectively, our FRET analysis show that (i) more compact Fn molecules adsorbed onto (100) than onto (001) facets, and (ii) larger amounts of more compact Fn adsorbed onto micro-rough HAP facets than onto their nano-rough counterparts.

We propose that the face-specific HAP surface chemistry contributes to different Fn conformation and deposition onto (100) and (001) facets. Fn adsorbed in more compact conformations onto (100) than onto (001) facets (Fig. 4A). This observation can be partially attributed to the lower energy barrier for water to diffuse away allowing for charged side chains of Fn molecules to penetrate the hydration layer and interact with surface ions of (100), as compared with (001) facets [11]. Thus it is likely that more Fn molecules could adsorb per unit area onto (100) in a given time due to favored kinetics, so that Fn conformation would be constrained by the smaller area available per molecule on the (100) surface. Whereas on (001), adsorption kinetics could be slower and might lead to less Fn molecules per unit area, so that they have enough space to lay down on the (001) surface and adopt more extended conformations. From a thermodynamic perspective, the exposure of the hydrophobic core of higher numbers of type III modules (comprising each Fn molecule) may also serve as a driving force for Fn to extend and/or unfold onto the more hydrophobic (001) facets. Additionally, local electrostatic interactions may also affect the availability of binding sites on Fn. For example, the Fn-III₉ and III₁₀ modules (cell binding domains essential for cell adhesion) carry more negative charge than the surrounding domains, and thus could be more accessible when adsorbed onto densely negatively charged surfaces due to local electrostatic repulsion [54]. As surface roughness increases, both structure and chemistry of the surface become heterogeneous, which could result in smaller differences in interfacial energy and surface charge density between facets, yet the mechanisms proposed above could still help to explain the different Fn conformation adsorbed onto 100M and 001M.

The amount of Fn deposited is influenced by factors including electrostatic interactions, dipolar polarization, and mechanical entrapment within surface topographical features. As Fn carries a net negative charge in PBS at pH 7.4 [55], the macroscopic electrostatic interaction discourages Fn adsorption onto negatively charged surfaces. However, nano- to microscopic polarization of protein modules, Debye force, and/or other short-range interactions such as hydrogen bonding, can facilitate Fn adsorption (e.g., via its numerous positively charged lysine and polar residues) onto negatively charged surfaces. Previous work by You et al shows that the amount of Fn adsorbed increased with increasing surface charge density of self-assembled monolayers within a range of −28 to −121 mV, and decreased at even higher surface charge density. Their work implies that, within such range of surface charge density, microscopic polarization rather than macroscopic electrostatic repulsion plays a dominant role in Fn adsorption [56]. Although the polarization effect is weaker on micro-rough facets due to lower surface charge density, lower electrostatic repulsion combined with higher probability of mechanical entrapment of Fn molecules within surface features can lead to higher Fn adsorption onto micro-rough facets than onto their nano-rough counterparts. Furthermore, stronger protein-protein interactions at high surface coverage could promote more compact Fn conformation observed on micro-rough facets [57]. In fact, FRET ratios were found to increase with increasing z depth (distance from peak to valley) for Fn adsorbed onto micro-rough facets, suggesting more compact Fn conformations trapped within surface features (Supplementary Fig. 8).

3.4 Breast cancer cell viability and adhesion

To determine whether HAP surface chemistry and topography affect breast cancer cell functions and whether HAP-cell interactions are mediated by Fn, we developed a customized HAP/PDMS system (Fig. 1) that enables long-term cell studies in controlled conditions, and seeded MDA-MB-231 breast cancer cells onto the four types of Fn-coated HAP facets. MDA-MB-231 cells were chosen because they are known to be highly invasive in vitro [58]. We first quantified cell viability via Live/Dead assay. The entire sample surface seeded with cells was imaged to count the number of live (stained in green) and dead (stained in red) cells for each HAP facet (Supplementary Fig. 9). Our data show that, on average, 95% cells were alive on all five samples including control coverglass up to 24 h after initial cell seeding (Supplementary Fig. 10). Fn coated coverglass (rather than bare HAP or coverglass) was chosen as control, as it allowed us to generate an initially well-defined single protein layer prior to cell incubation. In contrast, bare HAP or coverglass would have led to surface adsorption of multiple proteins (e.g., albumin) present in the FBS used to supplement the culture media for breast cancer cells, which would have complicated the interpretation of our results. Average cell viability decreased to circa 75% after 48 h on nano-rough facets, but remained at circa 95% on micro-rough facets (data not shown). We chose the 6 h and 24 h time points and excluded the 48 h time point for subsequent adhesion and secretion experiments to avoid the complexity of data interpretation due to poor viability.

We next investigated whether Fn adsorbed onto HAP facets regulates cell adhesion. The number of live cells per unit area extracted from Live/Dead assay was also used to quantify cell adhesion or rather live cell density at surfaces. Live cell density was independent of underlying Fn conformation or HAP surface properties, although there were slightly more cells present on micro-rough HAP facets, probably due to larger amounts of Fn adsorbed and increased ligand availability and/or density (Fig. 5A, B) [35]. Cells typically express multiple types of integrin receptors and some receptors share the same ECM ligands with different affinities. Conformational changes occurring in Fn can significantly affect the spatial arrangement and accessibility of its binding sites for integrins. Two predominant integrins are involved in cell binding to ECM molecules containing the RGD (tripeptide Arg-Gly-Asp) binding motif: α₅β₁ integrins require the close vicinity of both the RGD sequence on Fn-III₁₀ and the synergy PHSRN site on Fn-III₉ for dual engagement and activation, while α_vβ₃ integrins only require the RGD sequence to bind to Fn [59]. In addition, the activation state of α₅β₁ also affects its binding affinity for Fn [60,61]. Therefore, cells can utilize α_vβ₃ integrins but not α₅β₁ integrins to bind to extended/unfolded Fn when the spatial arrangement (relative angles and distance) between RGD motif and synergy site is disrupted [62]. Nevertheless, previous work has demonstrated that MDA-MB-231 cell adhesion onto Fn-coated 2D surfaces actually increased modestly after depletion of either β₁ or β₃ integrins, suggesting that both integrins can compensate for the loss of each other in ligand-binding [63]. Thus, our observation that MDA-MB-231 cell adhesion (or rather live cell density) is not sensitive to Fn conformation may be explained by high levels of expression of both β₁ and β₃ integrins in these cells, which are able to compensate for each other when binding to a wide range of Fn conformations. However, this varied engagement of integrins with Fn likely propagates downstream and triggers differential intracellular signaling cascades that subsequently alter secretion, among other cell functions.

Fig. 5.

MDA-MB-231 breast cancer cell adhesion (live cell density) after initial seeding for either (A) 6 h or (B) 24 h on Fn-coated HAP facets and control coverglass, as quantified by the number of live cells per unit area via Live/Dead assay. Data collected from 4 to 5 independent experiments. VEGF and IL-8 secretions by MDA-MB-231 cells were analyzed via ELISA assays of media collected after initial seeding for either 6 h, shown in (C) and (E), or 24 h, shown in (D) and (F), respectively. Secretion values were normalized by the number of live cells per sample. Quantitative ELISA tests were carried out in duplicate or higher and all experiments were independently repeated at least twice. Mean ± SD. In all cases, p < 0.05 is indicated by a single asterisk, p < 0.01 by two asterisks, and p < 0.001 by three asterisks.

3.5 Vascular endothelial growth factors (VEGF) and interleukin-8 (IL-8) secretions

We then evaluated the proangiogenic and proinflammatory capabilities of MDA-MB-231 cells via quantification of VEGF and IL-8 secretions. During tumor progression, the formation of new blood vessels, a process called angiogenesis, is critical for primary tumor growth and metastasis [64]. Tumor cells promote angiogenesis essentially by up-regulating proangiogenic factors such as VEGF [65]. High Fn levels detected in tumors have been associated with increased mortality among patients with breast and prostate cancers, and circulating plasma Fn was reported to enhance blood vessel formation and facilitate tumor growth by increasing soluble VEGF contents and enhancing subsequent VEGF-mediated signaling [66]. To assess secretion per cell, all our data (VEGF and IL-8 levels) were normalized by the number of live cells for each sample. After 6 h, VEGF secretion by tumor cells tended to be higher on nano-rough than on micro-rough facets, regardless of crystallographic orientation (Fig. 5C). After 24 h, VEGF levels were not only significantly higher on nano-rough than on micro-rough (001) facets, but they also tended to be higher on 001N than on 100N, suggesting an additional effect of surface chemistry on proangiogenic secretion (Fig. 5D). Furthermore, all secreted VEGF amounts drastically increased from 6 h to 24 h, especially for HAP facets. After 6 h, VEGF level on HAP facets was similar to that of control; whereas after 24 h, it reached 4 to 6 times that of control. Combined with our cell adhesion results, these data suggest that tumor cells interacting with HAP surfaces proliferate slower but exhibit noticeably higher proangiogenic capability than those seeded onto control coverglass, although both the conformation and amount of Fn adsorbed were similar for control and micro-rough HAP facets (Fig. 4).

Differential engagement of integrins with Fn may also explain the higher VEGF secretion levels measured when cells interacted with nano-rough than micro-rough HAP facets (Fig. 6). Dysregulated cell-ECM interactions and associated modified integrin signaling have been previously implicated in alterations of VEGF signaling [67–69]. We propose that the more extended/unfolded conformations of Fn adsorbed onto nano-rough facets could enhance VEGF secretion, as cells preferentially utilized α_vβ₃ integrins to compensate for α₅β₁ binding inability. More specifically, on micro-rough facets coated with compact Fn, cells may predominantly engage Fn via α₅β₁ integrins; while on nano-rough facets coated with extended/unfolded Fn, the disrupted RGD/synergy binding site would disable α₅β₁ binding and favor α_vβ₃ binding. In fact, higher engagement of α_vβ₃ has been reported to increase VEGF secretion in many types of cells [31,67,68]. Thus HAP-induced alterations of Fn mechanobiology (both mechanosensing and mechanotransduction) may regulate proangiogenic signaling pathways that promote tumor progression through enhanced angiogenesis.

Fig. 6.

Mechanisms by which Fn-HAP nanoscale interactions likely mediate breast cancer cell adhesion and subsequent secretions. (A) The more compact Fn conformations induced by HAP facets with high surface charge density, i.e. (100), and micro-rough topography are associated with low cancer cell secretions. (B) In contrast, the unfolded Fn conformations induced by HAP facets with low surface charge density, i.e. (001), and nano-rough topography enhance both proangiogenic (VEGF) and proinflammatory (IL-8) secretions by cancer cells, possibly via differential integrin engagement with underlying Fn.

Although a clear trend was visible when comparing VEGF levels on (100) with (001) facets, this functional difference was not significant possibly due to additional sequestration of VEGF within the adsorbed Fn layer [70]. Furthermore, the heterogeneity of integrin expression level in MDA-MB-231 population should be noted, with 25~50% of the population expressing α_vβ₃ integrins [71]. Overall, these results suggest that VEGF secretion by MDA-MB-231 cells is more sensitive to differential Fn adsorption induced by HAP surface topography than by surface chemistry. There might exist a critical extent of Fn unfolding that triggers the integrin-mediated proangiogenic pathway.

Finally, we evaluated the proinflammatory and osteolytic capability of MDA-MB-231 cells via quantification of IL-8 secretion. The proinflammatory factor IL-8 was reported to promote angiogenic responses in endothelial cells, increase tumor cell survival and proliferation, and enhance tumor cell migration and invasion [72,73]. Moreover, IL-8 acts as a mediator of osteolysis induced by bone metastasis, and is dramatically up-regulated in breast cancer cells that preferentially metastasize to bone [36,74]. In fact, a strong correlation has been established between IL-8 expression and metastatic potential of breast cancer cells [75]. Our data show that IL-8 secretion followed a similar yet more pronounced trend as VEGF secretion, with higher IL-8 levels secreted by tumor cells observed on nano-rough facets (Fig. 5E, F). However, after 24 h, a significantly higher level of IL-8 was measured on 001N than on 100N, suggesting that Fn conformation played a key role in regulating IL-8 secretion by tumor cells on nano-rough HAP facets. This surface chemistry effect was not observed when comparing 100M with 001M, probably due to the chemical heterogeneity of micro-rough facets. Additionally, IL-8 secretion increased drastically from 6 h to 24 h, especially for HAP surfaces, similarly to what was observed for VEGF secretion. The overall amount of IL-8 secreted by tumor cells was also much higher than that of VEGF (circa four times that of VEGF at 24 h).

Elevated IL-8 secretion by breast cancer cells can also be attributed to varied integrin engagement with Fn (Fig. 6). The highest level of IL-8 secretion was measured on 001N, which is associated with the lowest FRET intensity detected, i.e., the most unfolded Fn conformation. The observation that VEGF secretion is not as dramatically affected as IL-8 secretion by microenvironmental cues, in particular integrin engagement, has already been observed in other 2D and 3D culture platforms [69]. Together, the up-regulation of both IL-8 and VEGF secretion levels, and their interplay, are likely critical in (i) promoting tumorigenesis through increased angiogenesis and (ii) facilitating bone metastasis through enhanced osteolysis.

Interestingly, the nano-rough facets, which promote proangiogenic and proinflammatory secretions suppress cell viability after prolonged incubation (48 h time point, data not shown). Fn conformational changes can affect both secretion activities and cell viability via integrin-mediated pathways. Integrin α₅β₁ has also been shown to promote survival of growth-arrested breast cancer cells, potentially by negatively modulating apoptotic response via signaling pathways [76]. Therefore, it is likely that the extended/unfolded Fn conformation triggers an integrin switch from α₅β₁ to α_vβ₃, subsequently enhancing cell secretion activities while suppressing cell viability on nano-rough facets after prolonged incubation. However, further investigations are required to confirm whether secretion activity and cell viability are correlated via integrin-mediated pathways.

Collectively, our results indicate that essential breast cancer cell functions such as cell adhesion and proangiogenic and proinflammatory secretions can be mediated by a nanometer-thick film of Fn, whose conformation is controlled by the underlying HAP surface properties. We suggest that differential integrin engagement of cancer cells with the extended/unfolded conformations of Fn present at the (001) and nano-rough HAP facets is a likely mechanism responsible for altered VEGF and IL-8 secretions (Fig. 6). Our results are consistent with previous studies, where more crystalline and larger HAP nanoparticles were found to induce more unfolded Fn conformations and correlate with higher IL-8 secretion of breast cancer cells [37,45]. The importance of integrins in regulating tumor progression makes them appealing targets for cancer therapy, and some integrin antagonists have been tested in clinical trials with promising results [77]. It needs to be noted that surface properties such as surface ion species and hydration layer are probably different between our geologic HAP crystals and biologic HAP nanoparticles [78]. Additionally, the thin Fn film adsorbed onto HAP surfaces is likely remodeled by cancer cells, especially at the 24 h time point. Thus our FRET results are mainly relevant when interpreting initial cell response to the underlying Fn layer. Finally, cellular behaviors in 2D stiff culture platforms can be very different from those in 3D compliant scaffolds, although micro-rough facets also present 3D microstructured features [63,69,79]. In the metastatic microenvironment, there are other types of cells in addition to breast cancer cells. In fact, the ECM in tumors is primarily deposited by cancer-associated fibroblasts and adipogenic precursors [80,81]. The early interaction of cancer cells with Fn could affect subsequent ECM deposition activities of these cancer-associated cells. For example, soluble factors secreted by tumor cells have been shown (i) to promote unfolding and stiffening of Fn deposited by adipose progenitor cells and (ii) to alter subsequent Fn-collagen interplay, which further dysregulates VEGF secretion and ECM sequestration [82,83]. Due to the lack of cancer-associated cells in our experiments, longer-term cell culture results would therefore be hardly relevant for interpreting cell fate in the metastatic microenvironment. Additionally, our 2D platforms are not adapted for long-term studies due to proliferation-induced cell apoptosis. Therefore, this work has focused on the early events occurring between cancer cells and Fn (in presence of HAP) and the likely cell-matrix alterations that could propagate downstream in the more complex tumor environment. Future work utilizing 3D HAP-collagen scaffolds (in presence or absence of Fn) may help provide better implications for in vivo cell behaviors and are currently under development in our group.

4. Conclusions

While our previous work has shown that HAP combined materials properties affect Fn adsorption, we now deconvolute individual effects of HAP surface chemistry and nano/microscale topography on Fn conformational variations and link these changes to altered proangiogenic and proinflammatory capabilities of breast cancer cells, with likely implications for tumor angiogenesis and bone metastasis. Our data indicate that HAP surface properties induced changes not only in Fn molecular conformation (ligand availability) but also in the overall amount of Fn adsorbed (ligand density). Among all four types of HAP facets investigated, the nano-rough (001) facet coated with the most unfolded Fn triggered the highest levels of VEGF and IL-8 secretions by breast cancer cells. Collectively our findings suggest that Fn conformation regulates early cell signaling independently of other variables typically associated with altered ECM deposition (e.g., composition, rigidity), and that altered integrin binding specificity may underlie these changes. While our study focused on breast cancer cell behaviors, other cell types composing the metastatic microenvironment (e.g., osteoblasts, endothelial cells, immune cells) may be similarly responding to Fn conformational changes and will be evaluated in future studies. The simplicity and high control achieved in our 2D model systems allowed us to deconvolute the effects of HAP surface chemistry and nano/microscale topography on Fn-mediated breast cancer cell functions, enhancing our knowledge of apatite-controlled cell-ECM early interactions that may be implicated in tumor growth and bone metastasis.