Creating cellular patterns using genetically engineered, gold- and cell-binding polypeptides

The ability to arrange cells in customizable patterns on substrates with tunable surface properties enables fundamental investigations in cell biology such as modulation of cell–cell interactions and cell morphology,¹ single cell assays,² directed stem cell phenotype studies³ and analysis of 2D cell migration.^4,5 Methods for creating cellular patterns on surfaces include micromolding of self-assembled monolayers (SAMs) in capillaries,⁶ and photolithographic or soft lithographic patterning of SAMs,^7–11 polyethylene glycol (PEG),^12–14 protein resistant PEG-like amphiphilic comb polymers^15–17 and other polymer brushes.^18–21 These patterning methods have been applied to a range of substrate materials including gold, glass, and polymers.^22–25 

These methods often rely on the selective chemical coupling of cell adhesive ligands or the physical adsorption of cell adhesive proteins to a substrate for biofunctionalization. A facile approach to direct surface attachment and orientation of adsorbed proteins is to artificially incorporate bioadhesive motifs along with motifs that direct binding to a material surface. For example, cell attachment to a graphene surface was enhanced by precoating the surface with an adsorbed recombinant polypeptide comprising end-pendent graphene-binding and cell-binding domains.²⁶ It has been shown that the cells can interact with an adsorbed protein layer rather than the surface itself, enabling the cell membrane-bound receptors to bind with programmable specificity.²⁷ This strategy could easily be extended to cell patterning by immobilizing cell-adhesive proteins directly on a surface with spatial control of their ligand accessibility and hence cell-binding behaviors.

We hypothesized that by biasing material affinity of a recombinant polypeptide, we could generate cell patterns through directed ligand accessibility on a heterogeneous micropattern consisting of two dissimilar materials, gold on silicon, resulting in either enhanced or impeded local cell adhesion. We designed a bispecific, hetero-end-functional elastinlike polypeptide (ELP) fusion protein incorporating both N-terminal cell-binding (GRGDS) and C-terminal gold-binding (cysteine) domains RGD-ELP-Cys. We engineered the cysteine-rich domain to enhance C-terminus binding to gold via chemisorption of thiols on gold.²⁸ We chose GRGDS as a cell adhesive ligand at the N-terminus of the polypeptide to direct the surface adhesion and spreading of human umbilical vein endothelial cells (HUVECs).^29–31 Intrinsically disordered ELPs (Refs. 32–35) (Val-Pro-Gly-Xaa-Gly repeats, Xaa = guest residue), derived from tropoelastin, are characterized by their lower critical solution temperature behavior,^36,37 where above their cloud point temperature (T_t), they collapse to expel water and form polypeptide-rich sediments. Genetic incorporation of a substrate-binding domain into biopolymers enables direct surface immobilization, obviating the need for ancillary surface modification or chemical conjugation processes.^38–41 In addition, it potentially provides some modulation of biopolymer orientation to either expose or conceal the desired functional groups (e.g., cell-binding domains) for specific biointeraction.^42,43

We programmed cell-binding domains onto biopolymer-modified surfaces with spatial control over the accessibility of these domains, leading to formation of well-defined cellular micropatterns. These results suggest the orientation of adsorbed RGD-ELP-Cys (C-terminal thiols to gold) enhances the accessibility of terminal-end GRGDS domains and promotes cell adhesion within the gold regions. This is in contrast to RGD-ELP-Cys adsorbed on silicon oxide regions, which showed limited biointeraction for cell adhesion. While both surfaces adsorb RGD-ELP-Cys from solution [as measured by quartz crystal microbalance (QCM)], enhanced cell-adhesion is only observed on the gold-modified regions. This further suggests that distinct ligand accessibility of the RGD-ELP-Cys on either gold or silicon plays a key role in the observed dissimilar cell binding activity in each region.

It is worth noting that our method for cell patterning is not limited to recombinant proteins, but should also be amenable to synthetic polymers with material- and cell-binding domains. While a number of polymers may be applicable to our patterning approach, ELPs offer the following features: (1) a bioinert backbone for incorporation of bioactive peptide domains;⁴⁴ (2) monodispersity;⁴⁵ (3) simple purification without the need for chromatographic methods;⁴⁶ and (4) high yield expression (e.g., grams/liter).⁴⁷ Beside cell-adhesive ligands, other bioactive domains (e.g., growth factors, protease cleavable sites, etc.) can be potentially included into the biopolymer-modified surfaces with precise locations and spacing, providing state-of-the-art surface engineering approaches to direct cell behavior on patterned surfaces.

Silicon wafers were obtained from University Wafer (Boston, MA). All solvents and reagents were purchased from commercial sources and were used as received. Milli-Q water with a resistivity greater than 18.2 MΩ cm was used for all experiments.

The ELP constructs used in this study consist of 40 repeating pentapeptides (Val-Pro-Gly-Val-Gly) with a trailer sequence of (Gly-Gly-Cys)₈, serving as a gold binding domain. Synthetic genes encoding for the ELP with a C-terminal (GGC)₈ trailer (ELP-Cys) were available from previous studies.^48,49 Synthetic oligonucleotides encoding for both strands of an integrin binding peptide (SGRGDSGRGDS) with BseRI and NdeI compatible sticky ends were synthesized by Integrated DNA Technologies, Inc. (IDT, Coralville, IA). It is worth noting that we selected GRGDS as the cell-binding motif of the multifunctional ELP because GRGDS-modified surfaces show significantly enhanced binding affinity of cells compared to RGDS, GRGD, and RGD sequences.⁵⁰ To assemble the gene of the ELP fusion protein (RGD-ELP-Cys), the pET 24a plasmids encoding the ELP-Cys construct were digested with BseRI and NdeI (New England Biolabs, Beverly, MA), then enzymatically dephosphorylated using calf intestinal alkaline phosphatase (CIP; New England Biolabs). The linearized ELP vectors were separated from other deoxyribonucleic acid (DNA) fragments by gel electrophoresis with low melting agarose, AquaPor LM (National Diagnostics, Atlanta, GA) and were purified using a Qiaquick Gel extraction kit (Qiagen, Inc., Germantown, MD). The purified ELP vectors were ligated with the RGD encoding oligonucleotides to create a plasmid-harboring gene for RGD-ELP-Cys. The gene encoding RGD-ELP-Cys was confirmed by DNA sequencing. The complete amino acid sequence of the RGD-ELP-Cys is MSGRGDSGRGDSYGSKGPG(VGVPG)₄₀ (GGC)₈WP. All ELPs used in this study were expressed in Escherichia coli and purified by inverse transition cycling.⁵¹ These purified ELP-Cys and RGD-ELP-Cys fusion proteins were characterized by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE; BioRad, Inc., Hercules, CA). The molecular weights of purified ELP proteins were confirmed by matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS), which was performed on a PE Biosystems Voyager-DE instrument equipped with a nitrogen laser (337 nm). The ELP samples for MALDI-MS measurements were prepared in an aqueous 50% (v/v) acetonitrile solution containing 0.1% (v/v) trifluoroacetic acid, using a sinapinic acid matrix. The measured molecular weights were within 0.4% of the calculated value.

To enable fluorescent labeling, RGD-ELP-Cys was dissolved at 150 μM in tris(2-carboxyethyl)phosphine buffer, and Alexa Fluor 488^® C5-malemide (Life Technologies, Carlsbad, CA) was added to a final concentration of 3 mM. The reaction was allowed to proceed for 1 h at room temperature, then unreacted dye was removed by chromatography over Sephadex G25 (GE Healthcare Life Sciences, Pittsburgh, PA). The purified samples were concentrated with an Amicon 3000 molecular weight cutoff spin filter (Millipore, Billerica, MA). Labeling efficiency was determined spectrophotometrically using a NanoDrop 1000 Spectrophotometer (Thermo Scientific, Waltham, MA).⁵² 

Photomask patterns were designed with AUTOCAD 2014 and printed on photopaper (CAD Art Services, Bandon, OR). Silicon wafers were cleaned with 70% ethanol and deionized water and blown dry with nitrogen gun. A layer of negative photoresist (NFR016D2; JSR Micro, Inc., Sunnydale, CA) was spin coated (Headway spinner; Headway Research, Inc., Garland, TX) onto silicon wafers at a ramp speed of 1000 rpm and a spin speed of 3000 rpm for 30 s. The samples were baked in an oven at 90 °C for 2 min and a photoresist layer approximately 3.9 μm thick was formed. The wafers were aligned with photomasks using a mask aligner (Karl Suss MA6/BA6; SUSS MicroTec, Garching bei München, Germany) and subsequently irradiated with ultraviolet (UV) light (365 nm, 12.0 mW/cm²) with constant intensity and hard contact mode for 12 s. After exposure, the samples were baked in oven again for 2 min and developed with Microposit™ MF-319 Photoresist Developer (Dow Chemical, Co, Midland, MI) at room temperature for 60 s. Photoresist residues were washed off with deionized water and blown dry with nitrogen gun. The patterned photoresist layer served as a mask on the silicon wafers to prevent the deposition of metals onto nontarget silicon oxide regions. A thin layer of titanium (50 nm, 5 Å/s) was deposited onto surfaces first as an adhesive layer to aid gold layer binding, and a layer of gold (200 nm, 5 Å/s) was deposited onto the surfaces afterward. Both metal layers were deposited using an E-beam evaporator (Kurt Lesker PVD 75; Kurt J. Lesker Company, Jefferson Hills, PA). After metal layer deposition, the samples were placed in a beaker filled with 1165 stripper (NMP; Dow Chemical) overnight to remove the protective photoresist layer. The metals deposited on photoresist masked regions were stripped from the surface and washed off thoroughly with isopropyl alcohol (Sigma Aldrich, Co, St. Louis, MO) and deionized water, and blown dry with nitrogen gun. The metals deposited directly onto unmasked silicon regions remained on the silicon wafers and formed the patterned gold layers. The silicon wafers with patterned gold layers were cleaved into 0.5 × 0.5 cm square pieces with a diamond scribe. All cleaved samples were cleaned with 70% ethanol and deionized water, and blown dry with a nitrogen gun before treatment with ELP solutions.

To modify the gold patterned silicon wafer surfaces with peptides, solutions of RGD-ELP-Cys (labeled with Alexa Fluor 488) at a concentration of 150 μM were added to the substrates for 20 min and subsequently washed with phosphate-buffered saline (PBS). An ELP without a RGD peptide (ELP-Cys, labeled with Alexa Fluor 488) was used as a control. The ELP-modified surfaces were visualized using fluorescence microscopy (Axioimager; Carl Zeiss Microimaging, Inc., Jena, Germany) through 10×, 20×, and 40× objectives. At least seven images were captured at various known radial distances from the center using a CCD camera (Axiocam; Carl Zeiss, Inc., USA). Axiovision software was used for further data analysis.

We also used Q-Sense E4 QCM-D instrument (Q-Sense AB, Västra Frölunda, Sweden) to investigate the in situ adsorption of the RGD-ELP-Cys onto gold and silicon oxide substrates. Shifts in the crystal frequency for the third overtone were continuously monitored using the Q-Soft software (Q-Sense AB). Adsorption of RGD-ELP-Cys was performed on the silicon oxide- and gold-coated quartz crystals with varying adsorption times to achieve stabilization of frequency. Once the system reached equilibrium, loosely bound peptides were removed by washing with PBS buffer. All the measurements were taken at 25 °C at a flow rate of 50 μl/min. These experimental data were fitted to the Voight-based viscoelastic model.^53,54 The thickness of the peptide layer is obtained by solving the Voigt model at different oscillation frequencies (overtones) of the QCM-D using Q-Tools software (Q-Sense AB). Prior to use, silicon dioxide coated sensor crystals (QSX 303, Q-Sense AB) were exposed to UV/ozone for 10 min, followed by immersion in 2% SDS for 2 h. These crystals were then rinsed thoroughly with deionized water and dried by nitrogen gas. Gold coated sensor crystals (QSX 301, Q-Sense AB) were pretreated with piranha solution (75% H₂SO₄/25% H₂O₂, v/v), then rinsed with deionized water and dried with nitrogen gas.

HUVECs were purchased from Lonza and were cultured in EBM-2 medium (Lonza, Basel, Switzerland). We chose HUVECs as a model cell line, because HUVECs abundantly express numerous integrins (e.g., α₅β₁ and α_vβ₃) that are recognized by RGD peptides.^55,56 The cells were cultivated at 37 °C in a humidified atmosphere with 5% CO₂ and passaged regularly. HUVECs were used in passages 4–6. These cells were then labeled with 1 μM calcein red-orange AM (Thermo Scientific) in 1% sterilized bovine serum albumin (BSA) in PBS for 0.5 h and rinsed three times with PBS.

Here, 200 000 HUVECs (2 × 10⁷ cells/ml, stained with calcein red-orange AM) in a 1% BSA solution were seeded onto each patterned surface (0.5 × 0.5 cm) and incubated in PBS for 3 h at room temperature. We include BSA to increase the fidelity of our approach because BSA may promote RGD-integrin binding.⁵⁷ The washing conditions (i.e., fluid flow rates) were precisely controlled by Harvard syringe pumps (Harvard Apparatus, Holliston, MA). Two syringe pumps are used to wash the patterned surface with PBS at constant speed (9 ml/min) after the cell immobilization step (Fig. S1).⁷³ We then cultured the adhered HUVECs on RGD-ELP-Cys coated substrates at 37 °C with repeated media exchange every 3 days. The adhered HUVECs were restained with calcein red-orange AM after 60 days. In addition, the HUVESs of the same density were seeded on a patterned RGD-ELP-Cys (green) substrate at 37 °C for 3 h. After washing in PBS at 37 °C, the substrates were then maintained at 4 °C for 10 min and room temperature for 3 h. All the samples were observed using the fluorescence microscope.

We designed and synthesized a polypeptide (RGD-ELP-Cys) containing an RGD cell adhesion motif (GRGDSGRGDS) at the N-terminus, followed by a thermally responsive ELP block and a cysteine rich domain (GGC)₈ at the C-terminus [see amino acid sequence in Fig. 1(a)]. To fabricate cell patterns, we first used photolithography to generate gold micropatterns on silicon oxide surfaces [Fig. 1(b)]. These heterogeneous surfaces were coated by exposing them to a solution containing RGD-ELP-Cys. It has been reported that immobilization of proteins/peptides via terminal end thiol groups can lead to uniform orientation of molecules on the surface,^58,59 due to the high specificity of the thiol-gold bond. Thus, we hypothesized that, aided by the C-terminal cysteine-rich domain, the polypeptides can adsorb to the gold regions in a directed orientation, whereby terminal-end GRGDS domains are more accessible for cell binding than for RGD-ELP-Cys adsorbed in the silicon oxide regions [depicted schematically in Fig. 1(c)]. After seeding the RGD-ELP-Cys coated micropatterns with cells, the dissimilar accessibility of GRGDS domains on patterned gold/silicon surfaces is evidenced by enhanced cell binding only on the coated gold regions [Fig. 1(d)].

Differences in adsorption behavior of a folded, globular protein, adsorbed on distinct materials but at similar amounts, can lead to marked differences in cell binding behavior (i.e., facile and strong attachment versus attachment resistance).⁶⁰ We asked if an intrinsically disordered protein with chemically dissimilar end groups could simultaneously enhance or resist cell attachment by distinct adsorption behaviors on regions of a gold patterned silicon substrate. In this case, the ELP serves not only as an inert biolinker for precisely spacing active peptide domains in the chain, but acts as a dual-purpose molecule to modulate cell response on different materials.

To construct polypeptide-modified patterned surfaces, solutions of RGD-ELP-Cys (150 μM) in PBS were incubated with gold patterned silicon surfaces for 20 min and subsequently washed with PBS. Figures 2(a) and 2(b) show images of fluorescently labeled (Alexa Fluor 488) RGD-ELP-Cys bound to different architectures of gold patterned silicon surfaces. The fluorescence intensity line profile for substrates with alternating gold stripes shows ∼20% higher fluorescence intensity in the gold regions as compared to the silicon oxide regions (Fig. S2), consistent with a ∼20% increase in the amount of RGD-ELP-Cys bound to gold surfaces [as measured by QCM, Fig. 2(c)]. In addition, the RGD-ELP-Cys formed slightly thicker peptide layers on gold substrates than that on silicon oxide substrates (Fig. S3).

Cell seeding on the ELP-modified patterned surfaces was accomplished by incubating the substrates with HUVECs suspended in serum free culture media (1% BSA). The HUVECs (labeled with calcein red-orange AM) were seeded at 2 × 10⁷ cells/ml. To investigate cell interaction with the ELPs, cells at the same density were also incubated on surfaces patterned with ELP-Cys lacking an RGD cell-binding domain. After 3 h incubation at 25 °C, the surfaces were washed twice with PBS. Adherent cells were counted using fluorescence microscopy and the average number per area for both gold and silicon oxide regions were determined. HUVECs adhered in appreciable numbers to the RGD-ELP-Cys modified patterned surfaces [Fig. 3(a)], while few HUVECs adhered to ELP-Cys modified surfaces [Fig. 3(b)] or native substrate materials lacking ELPs [Fig. 3(c)]. A significantly larger number of cells adhered onto RGD-ELP-Cys modified areas compared to the ELP-Cys surfaces lacking RGD [p < 0.05; Fig. 3(d)]. This demonstrates that ELPs provide an inert background for programming activity via functional peptide segments in the chain, and suggests that cell adhesion on the RGD-ELP-Cys modified surfaces is achieved through RGD-integrin interactions.^55,56

Gold microislands of variable sizes on silicon surfaces were constructed via photolithography [Figs. 4(a)–4(c)]. After immobilization of fluorescently labeled RGD-ELP-Cys, the HUVECs were incubated with the patterned substrate at room temperature for 3 h. Unbound cells were washed away with PBS, and the remaining adherent cells were visualized by fluorescence microscopy. Figures 4(d)–4(i) show images of HUVECs adhering to RGD-ELP-Cys modified patterns with gold microislands of different sizes (i.e., 30, 40, and 50 μm). It is evident that HUVECs adhered appreciably on RGD-ELP-Cys-modified gold regions and minimal HUVEC adhesion was observed on silicon oxide regions containing adsorbed RGD-ELP-Cys. These observations strongly suggest that dissimilar accessibility of RGD domains on gold and silicon oxide regions leads to clear differences in cell adhesion on the heterogeneous surfaces. Through thiol-gold interactions, the polypeptides are likely oriented with the C-terminus located near the gold surface with enhanced accessibility of N-terminus RGD domains, promoting cell adhesion in the gold regions. In contrast, adsorption of RGD-ELP-Cys on silicon oxide regions may not result in a favorable orientation with accessible N-terminal RGD domains, impeding the RGD-integrin cell binding.

We also observed various degrees of cell spreading and flattening of cells in gold regions on patterned surfaces [see inset of Figs. 4(g)–4(i)]. In addition, the cells were confined primarily to the RGD-ELP-Cys-coated regions. Interestingly, in long term studies in which HUVECs were seeded at room temperature and cultured for 60 days at 37 °C with repeated media exchange every 3 days, the cells remain viable and confined within the RGD-ELP-Cys modified microislands. The cells were stained with calcein red-orange AM, a live cell tracking dye to confirm the viability of the cells [Figs. 4(j)–4(l)]. These results are consistent with other studies that demonstrate cells on patterned surfaces can remain viable for many weeks. For example, mesenchymal stem cells retained their stem-cell phenotype and maintained stem-cell growth on a nanostructured surface for eight weeks⁶¹ and neurons remained viable for up to four weeks on poly-d-lysine patterned surfaces.^62,63

We seeded cells on hydrated RGD-ELP-Cys at room temperature, well below the solution T_t of concentrated RGD-ELP-Cys (∼32 °C, see Fig. S4). The adhered cells were subsequently cultured at 37 °C (T > T_t) on collapsed RGD-ELP-Cys. Our results indicate that cells remain attached to the patterned surfaces despite triggering the phase transition of the ELPs [Figs. 4(j)–4(l)]. No significant changes in the average cell counts per area were observed (see Fig. S5). Therefore, the RGD-ELP-Cys coated gold surface is able to promote cell adhesion on both hydrated and collapsed RGD-ELP-Cys surfaces, which suggests that end-pendent RGD domains are accessible above and below the T_t of the RGD-ELP-Cys. Thus, thermal phase transition of the ELPs does not interfere with the RGD-integrin cell binding in the polypeptide coated gold regions.

In addition, we seeded cells on heated ELP-coated substrates at 37 °C (T > T_t) and asked if the initial hydration state of the ELP (collapsed versus hydrated) would influence patterning resolution. We found enhanced binding in the coated silicon oxide regions at a seeding temperature T > T_t, resulting in loss of pattern fidelity [Fig. 5(a)]. This suggests that the N terminal binding domains in RGD-ELP-Cys are more accessible on silicon oxide surfaces when the adsorbed ELPs are in a heated state. We then cooled the substrate to T < T_t and found that after washing, most of the cells (∼77%) bound to the coated silicon oxide regions detached, while most cells (∼73%) remained adhered to the coated gold islands [Figs. 5(b) and 5(c)]. This result may have implications for use of modified ELPs for cell release after culture.^64–66 In general, our results indicate seeding cells at room temp (T < T_t) followed by cell culture at 37 °C (T > T_t) provides the best patterning fidelity for the substrate/polypeptide combination presented in this study.

We have genetically engineered polypeptides containing N-terminal cell binding and C-terminal gold binding domains to create cellular patterns on gold patterned silicon surfaces. The cysteine-rich gold binding domains enable direct chemisorption of polypeptides onto gold regions via the C-terminal thiols with enhanced accessibility of N-terminal cell binding domains, promoting cell adhesion and spreading in gold regions. In contrast, RGD-ELP-Cys adsorbed on silicon oxide regions hinders the accessibility of cell binding domains, resulting in minimal cell adhesion within those regions. The likely dissimilar accessibility of cell binding domains on gold and silicon oxide regions leads to formation of well-defined cell patterns. The polypeptide-patterned surfaces are capable of immobilizing cells, and maintain cell viability and pattern integrity for up to eight weeks. This study demonstrates a simple approach for fabricating cellular patterns that potentially enable many applications, such as study of intercellular communication, single-cell analysis, and drug discovery. In addition, genetic incorporation of material-binding domains provides a direct and controlled method to attach proteins on surfaces. A library of artificial peptides with affinity for particular substrates (e.g., silica, gold, zinc, graphene, and titanium) has been identified through combinatorial screening approaches.^43,67–71 By using these material binding peptides, proteins can be biased to a variety of substrates with high affinity.^40,72 Besides material-binding domains, other functional domains can also be precisely incorporated onto biopolymer-modified surfaces, such as growth factors, protease cleavable sites, and crosslinking sites, providing innovative surface engineering approaches to fabricate bioactive substrates for a variety of applications.

This work was supported by the National Science Foundation's (NSF's) Research Triangle Materials Research Science and Engineering Center (MRSEC, DMR-1121107) and the National Institutes of Health (R21GM111584).