Enzyme-Directed Functionalization of Designed, Two-Dimensional Protein Lattices

Abstract

The design and construction of crystalline protein arrays to selectively assemble ordered nanoscale materials has potential applications in sensing, catalysis and medicine. Whereas numerous designs have been implemented for the bottom-up construction of novel protein assemblies, the generation of artificial functional materials has been relatively unexplored. Enzyme-directed post-translational modifications are responsible for the functional diversity of the proteome and thus, could be harnessed to selectively modify artificial protein assemblies. In this study, we describe the use of phosphopantetheinyl transferases (PPTases), a class of enzymes that covalently modify proteins using coenzyme A (CoA), to site-selectively tailor the surface of designed, two-dimensional (2D) protein crystals. We demonstrate that a short peptide (ybbR) or a molecular tag (CoA) can be covalently tethered to 2D arrays to enable enzymatic functionalization using Sfp PPTase. The site-specific modification of two different protein array platforms is facilitated by PPTases to afford both small-molecule- and protein-functionalized surfaces with no loss in crystalline order. This work highlights the potential for chemoenzymatic modification of large protein surfaces towards the generation of sophisticated protein platforms reminiscent of the complex landscape of cell surfaces.

Graphical Abstract

Introduction

The structural and chemical diversity of proteins positions them as Nature’s premier tools in facilitating a myriad of cellular processes. This diversity, in turn, makes them attractive building blocks for the construction of functional synthetic materials.^1–8 Proteins exhibit functional complexity in a crowded cellular environment due to their self-assembly into multidimensional supramolecular assemblies, giving rise to emergent properties inaccessible to individual proteins.⁹ While the noncovalent self-assembly of multicomponent protein architectures with other components, including small molecule cofactors, inorganic complexes (e.g., metal clusters¹⁰), or other biomolecular species (e.g., nucleic acids¹¹), is programmed at the genetic level, the structural and chemical diversity of the proteome is further enhanced and indeed, enabled, by enzyme-mediated post-translational covalent modifications. Such modifications play vital roles in the biosynthesis of complex biomolecules (e.g., fatty acids, peptides, polyketides)^12,13 or in large-scale cellular functions such as cell-cell signaling and recognition (e.g., via glycosylation, phosphorylation, or membrane anchoring through the attachment of lipid tails).¹⁴ For example, neuronal microtubule filaments are found to be more stable in depolymerizing environments due to polyamination of glutamic acid residues by transglutaminases, which may affect aspects of brain development, neuronal regeneration and aging.¹⁵ Bacterial and archaeal S-layers, prototypical two-dimensional (2D) self-assembling protein lattices, perform critical roles in protection, virulence, cell morphology and surface recognition.¹⁶ For many S-layer proteins, translocation across the cell membrane is facilitated through post-translational modifications such as the N-glycosylation of select asparagine residues or lipid modifications.^16,17 Inspired by diverse structure-function relationships in nature, the hierarchical construction of biological machines to perform challenging biological and chemical tasks remains a prominent goal of bionanotechnology.^6,18 The development of protein-based biomaterials amenable to post-translational modifications^19–21 presents potential applications in diagnostic sensors, vaccines, drug-delivery, or biomineralization matrices.^22,23

Crystalline 2D protein arrays are a promising biotechnological platform due to their high-density display of polypeptides with nanoscale tunability and reconfigurability.¹⁹ Engineering the surfaces of 2D protein lattices could provide an opportunity to selectively organize target molecules or particles of interest in a site-dependent manner. This bottom-up route would furnish the ability to not only create periodic patterns with sub-10-nm precision (i.e., in a diffraction-unlimited fashion) but also to hierarchically assemble complex, multicomponent architectures that cannot be obtained through self-assembly alone. In addition, 2D protein crystals represent robust and inherently functional platforms²⁴ that provide high surface area-to-volume ratios and thus could provide distinct advantages in engineering heterogeneous biocatalysts or immunotherapeutic agents, as well as in the fundamental study of biological/enzymatic reactions on 2D surfaces.

One can envision two major paths to generate functionalized 2D protein materials: repurposing natural protein assemblies like S-layer proteins or designing artificial assemblies from scratch. Naturally occurring extended structures provide a ready-made template for chemical or genetic manipulation; however, structural characterization and facile manipulation of these structures in vitro and in vivo is key to designing function. To this end, Sleytr and coworkers performed formative experiments to demonstrate the potential uses of S-layer lattices, reconstituted in vitro, for applications in membrane filtration²⁵, drug delivery²⁶, and the spatial organization of immunogenic biomolecules or inorganic nanoparticles.^22,27 More recent reports describe genetic modification of S-layer substrates with short peptide tags (e.g., the split-protein system SpyTag-SpyCatcher²⁸) to generate hierarchical 3D materials²⁹ and high-density engineered displays on living cells.³⁰ Many of these examples rely on chemical biology tools to facilitate biomolecule immobilization on solid substrates (e.g., biotinylation, azide-alkyne cycloaddition, gene fusions, incorporation of reactive peptides, split-protein tags).^31–33 However, given a dearth of atomic-resolution structural information on S-layer proteins, it has remained challenging to manipulate them at will to create functional materials, and in vitro manipulation of S-layer proteins requires protein truncations to afford solubility and stability.

In contrast with the limited applicability of reconstituted or genetically altered S-layer proteins in vivo, the bottom-up assembly of protein architectures enables the construction of tailor-made nanomaterials with well-known structures and desired properties. There has been considerable progress in recent years to create artificial 2D protein assemblies using chemical design strategies including genetic fusions of symmetric protein modules,^34–36 computational design of protein-protein interfaces,³⁷ metal coordination,^3,38 or reversible covalent bonding³⁹ towards generating novel biological machines. While a number of these strategies have been employed to generate extended 2D arrays with crystalline order, there has been little exploration thus far in using these arrays as functional platforms.^40,41 Interfacing chemical/biological conjugation tools with such ordered assemblies would allow for the facile incorporation of functionality onto crystalline substrates. In this work, we used two crystalline assemblies previously designed in our lab, RIDC3 and ^C98RhuA, as 2D platforms whose surfaces can be tagged with functional peptides as recognition elements to allow enzyme-mediated modification. Our results show that the incorporation of functional peptides onto the 2D crystal surfaces can be carried out genetically or chemically and does not disrupt the underlying lattice packing, which, in turn, enables site-specific surface modification with synthetic small molecules as well as large proteins.

Materials and Methods

Protein expression and purification

The ybbR peptide was cloned into a pET20b(+) plasmid housing the RIDC3 vector using Quikchange mutagenesis protocols. RIDC3 and ^ybbRRIDC3 were expressed and purified as previously described.⁴² Briefly, cells harboring the RIDC3 or ^ybbRRIDC3 plasmid were grown in a lysogeny broth (LB, BioPioneer Inc.) medium containing 100 mg/L ampicillin and 34 mg/L chloramphenicol in 2.8-L flasks and shaken at 200 rpm for 16-20 h at 37 °C. Cells were harvested by centrifugation at 5,000 × g for 10 min, resuspended in a buffered solution containing 5 mM sodium acetate (pH 5.0) and 100 mg lysozyme (VWR) and lysed by sonication on ice. The crude lysate was subject to pH titration with 40% sodium hydroxide to pH 10 immediately followed by the addition of 50% (v/v) hydrochloric acid to pH 4.5 and centrifuged at 10,000 × g for 30 min. The cleared lysate was applied to a CM sepharose gravity column and eluted using a stepwise gradient (0 – 500 mM NaCl). Collected fractions were concentrated using an Amicon Stirred Cell (Millipore) and dialyzed against a buffered solution containing 10 mM sodium phosphate (NaPi) (pH 8.0) overnight. The protein was collected and purified using a DuoFlow workstation station equipped with a Bio-Scale Mini Macro-prep High Q-cartridge column (BioRad) and eluted using a linear gradient (0 – 500 mM NaCl). The separation of intact and truncated ^ybbRRIDC3 proteins (Figure S11a) was observed in this ion-exchange purification step, with intact ^ybbRRIDC3 eluting at ca. 80 mM NaCl and truncated ^ybbRRIDC3 eluting at ca. 110 mM NaCl. Protein purity was assessed by ultraviolet-visible (UV-vis) spectroscopy and samples with a RZ ratio (A₄₁₅/A₂₈₀) > 6 were pooled, concentrated and flash frozen for storage at −80 °C.

A pJ414 plasmid housing the ^ybbRRhuA gene was purchased from DNA 2.0. ^ybbRRhuA was expressed and purified in a similar fashion to ^C98RhuA, as previously described.³⁹ Cells were grown in 30-mL LB cultures containing 100 mg/L ampicillin and shaken overnight at 37 °C. From these starter cultures, 5 mL was inoculated into 1-L LB media and shaken at 200 rpm at 37 °C to an optical density (OD) of 0.8 – 1. Protein expression was inducted with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG, BioPioneer Inc.) for 12 h and cells were harvested by centrifugation at 5,000 × g for 10 min. Cells were resuspended in a buffered solution containing 10 mM tris(hydroxymethyl)aminomethane hydrochloride (Tris) (pH 7.5), 10 mM β-mercaptoethanol (βME) and 1 mM ZnCl₂ supplemented with 100 mg lysozyme and 50 mg PMSF and lysed by sonication on ice. Cell lysate was centrifuged at 10,000 × g for 30 min, the cleared lysate was subject to the addition of Polymin-P (Acros) to a final concentration of 0.15% (w/v), and stirred for 30 min at 4 °C prior to centrifugation at 10,000 × g for 30 min. The supernatant was loaded onto a DEAE sepharose gravity column and eluted using a stepwise gradient (0 – 500 mM NaCl). Fractions containing ^ybbRRhuA were subject to the addition of ammonium sulfate to a final concentration of 1.7 M to precipitate the protein and stirred for 30 min prior to centrifugation at 10,000 × g for 30 min. The precipitated protein was resuspended in a buffered solution containing 10 mM Tris (pH 7.5), 10 mM βME, and 1 mM ZnCl₂ and dialyzed against the same solution three times. The protein was collected and purified using a DuoFlow workstation station equipped with a Bio-Scale Mini Macro-prep High Q-cartridge column (BioRad) and eluted using a linear gradient (0 – 500 mM NaCl). Protein purity was assessed by SDS PAGE and pure fractions were dialyzed against a buffered solution containing 10 mM Tris (pH 7.5), 10 mM βME, and 1 mM ZnCl₂ and flash frozen for storage at −80 °C.

B. subtilis Sfp, superfolder GFP (sfGFP), and ^ybbRGFP (EGFP-ybbR N-terminal fusion) variants were expressed and purified as previously described.^43,44 Briefly, cells were inoculated into terrific broth containing 100 mg/L kanamycin and shaken at 200 rpm at 37 °C to an OD of 0.8. Protein expression was induced with 1 mM IPTG and shaken at 200 rpm overnight at 16 °C. Cells were harvested by centrifugation at 5,000 × g for 10 min, resuspended in a buffered solution containing 50 mM Tris (pH 7.5) and 250 mM NaCl supplemented with 100 mg lysozyme, 5 μg/mL DNAse I (Sigma), and 5 μg/mL RNAse (Worthington Biochemical Corp.) and lysed by French pressure cell press (500 – 1,000 psi). The cell lysate was centrifuged at 12,000 × g for 45 min and the cleared supernatant was subject to a Ni-NTA (Sigma) gravity column. The column was washed with a buffered solution containing 50 mM Tris (pH 7.5), 250 mM NaCl and 10 mM imidazole prior to protein elution at 300 mM imidazole (Sigma). Pure fractions were desalted into a buffered solution containing 50 mM Tris (pH 7.5) using a PD-10 desalting column (GE Healthcare Life Sciences), concentrated using an Amicon spin filter (Millipore), exchanged into a buffered solution containing 50 mM Tris (pH 7.4), 150 mM NaCl and 20% glycerol and flash frozen at −80 °C.

Preparation of ybbR-N₃ peptide

The desired sequence of the synthesized ybbR-N₃ peptide is: DSLEFIASKLAG-K(N₃). The Fmoc-Lys(N₃) amino acid was purchased separately from Anaspec. Solid phase peptide synthesis using Rink amide MBHA resins was used to generate the peptide. A solution of 20% 4-methylpiperidine in dimethylformamide (DMF) was used for Fmoc deprotection (2 × 5 min) and peptide coupling was performed using 1-[Bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxide hexafluorophosphate (HATU) and N,N-diisopropylethylamine (DIPEA). Deprotection and coupling cycles were repeated for the addition of every amino acid and the final peptide was cleaved from the resin using a solution of trifluoroacetic acid (TFA) and dichloromethane (DCM) at a 9:1 ratio. Peptides were precipitated using cold ether, pelleted by centrifugation at 5,000 rpm for 15 min and dried in vacuo. The peptide was purified by reverse-phase high pressure liquid chromatography (HPLC) using a Hitachi-Elite LaChrom L2130 pump with a binary gradient and a UV-vis detector and eluted using a linear gradient of acetonitrile in water. Peptide purity was assessed by matrix-assisted laser desorption/ionization (MALDI).

Preparation of modified CoA conjugates

TAMRA-CoA was synthesized and purified as previously described.⁴⁵ GFP-CoA was prepared by first exchanging sfGFP into a buffered solution containing 20 mM 3-(N-morpholino)propanesulfonic acid (MOPS) (pH 7.5). Approximately 5 equiv. of sulfosuccinimidyl-4-[N-maleimidomethyl]cyclohexane-1-carboxylate (sulfo-SMCC, Fisher) dissolved in a solution of 100 μL DMF was added to a 1 mL solution containing 100 μM GFP and allowed to react at 37 °C for 1.5 h. The sample was buffer exchanged into a fresh solution of 20 mM MOPS (pH 7.5) using a 10 D/G desalting column to remove any unreacted SMCC and 5 equiv. of coenzyme A, free acid dissolved in a buffered solution containing 20 mM MOPS (pH 7.5) was added and allowed to gently shake overnight at room temperature on an orbital shaker. The sample was buffer exchanged into a fresh solution of 20 mM MOPS (pH 7.5) using a 10 D/G desalting column and characterized by UV-vis spectroscopy.

Chemical conjugation of self-assembled RIDC3 crystals

RIDC3 crystals were self-assembled by first exchanging RIDC3 into a buffered solution of 20 mM 2-(N-morpholine)ethanesulfonic acid (MES) (pH 5.5) using a 10 kDa Amicon spin filter. A concentrated stock solution of RIDC3 was supplemented with a solution of 20 mM MES (pH 5.5) and 5 equiv. ZnCl₂ dissolved in water to a final protein concentration of 100 μM. After mixing with ZnCl₂, the solution turned cloudy within 5 min and crystals matured in 7-10 days. Mature crystals were buffer exchanged × 5 into a fresh solution containing 20 mM MOPS (pH 7.5) by pelleting suspensions of crystals at 3,500 rpm in 30 s bursts for 2 min and carefully pipetting out the clear supernatant. Approximately 10 equiv. of a dibenzocyclooctyne-N-hydroxysuccinimidyl ester (DBCO-NHS) dissolved in DMF (25 mg/mL) was added to a suspension of RIDC3 crystals and allowed to gently shake at room temperature for 2 h. Crystals were buffer exchanged × 5 into a fresh solution containing 20 mM MOPS (pH 7.5) via centrifugation to remove any unreacted DBCO. A stock solution of ybbR-N₃ dissolved in DMF was added to the suspension of RIDC3 crystals to a final ratio of 10:1 peptide:crystals and gently shaken overnight at room temperature. RIDC3-ybbR crystals were again buffer exchanged × 5 into a fresh solution containing 20 mM MOPS (pH 7.5) and analyzed by UV-vis spectroscopy and electrospray ionization mass spectrometry (ESI-MS).

The preparation of RIDC3-CoA crystals was performed similarly to RIDC3-ybbR crystals, starting with buffer exchanging mature crystals into a solution containing 20 mM MOPS (pH 7.5) via centrifugation. Approximately 10 equiv. of a stock solution of sSMCC dissolved in DMF was added to the suspension of crystals and allowed to gently shake at room temperature for 1.5-2 h. Crystals were buffer exchanged × 5 into a fresh solution containing 20 mM MOPS (pH 7.5) via centrifugation and 10 equiv. of coenzyme A, free acid dissolved in a buffered solution containing 20 mM MOPS (pH 7.5) was added and gently shaken overnight at room temperature. Crystals were buffer exchanged one final time and analyzed by UV-vis spectroscopy.

Preparation of ^ybbRRIDC3 and ^ybbRRhuA crystals

^ybbRRIDC3 solutions were stored at 4 °C to limit degradation of the appended peptide. A stock solution of ^ybbRRIDC3 was combined with Zn²⁺ in buffered solutions at a pH range of 5.5 – 8.5. In general, a 50-μL sample contained 100 μM protein at [Zn²⁺]:[protein] ratios ranging from 1:1 to 10:1. Buffered solutions containing 20 mM MES were used at pH 5.5 and pH 6.5, 20 mM MOPS at pH 7.5 and 20 mM CHES at pH 8.5. Crystalline materials were found in solutions at pH 5.5 and pH 7.5. ^ybbRRhuA crystals were prepared by buffer exchanging protein into a fresh solution containing 10 mM T ris, 10 mM βME, and 1 mM ZnCl₂ using a 10 kDaAmicon spin filter to a final protein concentration of 125-150 μM ^ybbRRhuA tetramer. Samples were gently shaken at 4 °C for 1-2 weeks until crystals matured.

Gel digestion and MS-MS analysis of RIDC3-ybbR crystals

A suspension of RIDC3-ybbR crystals was dissolved using a solution containing 1 mM ethylenediaminetetraacetic acid (EDTA), separated from any possible contaminants on a SDS-PAGE gel and stained with Coomassie Brilliant Blue. The band corresponding to RIDC3-ybbR was cut out into 1 mm × 1 mm cubes and destained × 3 with a solution containing 100 mM ammonium bicarbonate for 15 min followed by the addition of 100 μL acetonitrile (ACN) for 15 min. The supernatant was removed and gel pieces were dried in vacuo and chemically reduced with 200 μL of a solution containing 100 mM ammonium bicarbonate and 10 mm dithiothreitol (DTT) at 55 °C for 30 min. The supernatant was removed and gel pieces were incubated with 200 μL of a solution containing 100 mM ammonium bicarbonate and 55 mM iodoacetamide and incubated for 20 min at room temperature in the absence of light. Gel samples were washed first with a fresh solution of 100 mM ammonium bicarbonate and finally acetonitrile prior to dehydrating gel pieces using a Speedvac. Gel pieces were digested by first covering them in an ice-cold solution containing 50 mM ammonium bicarbonate and trypsin (0.1 μg/μL) and incubated on ice for 30 min. After the gel was completely rehydrated, excess solution was removed and replaced with a fresh solution containing 50 mM ammonium bicarbonate overnight at 37 °C. Peptide extraction was performed by the addition of a 50 μL solution containing 0.2% formic acid (v/v) and 5% ACN (v/v) in water and mixing at room temperature for 30 min. The supernatant was collected and the extraction procedure was repeated again, combining the 2^nd supernatant with the previous solution. Samples were analyzed by liquid chromatography (LC) with tandem mass spectrometry (MS/MS) using electrospray ionization.

Trypsin-digested peptide solutions were analyzed using ultra high pressure liquid chromatography coupled with LC-MS/MS using nanospray ionization. Ionization experiments were performed using a TripleTof 5600 hybrid mass spectrometer (ABSCIEX) interfaced with nano-scale reverse-phase UPLC equipped with a 20-cm 75-micron ID glass capillary packed with 2.5-μm C18 CSHTM beads (Waters corporation). Peptides were eluted using a linear gradient (5-80% ACN) at a flow rate of 250 μL/min for 1 h. The UPLC solutions used were: Buffer A - 98% H₂O, 2% ACN, 0.1% formic acid, and 0.005% TFA and Buffer B - 100% ACN, 0.1% formic acid, and 0.005% TFA. MS/MS data were acquired in a data-dependent manner; the MS1 data was acquired for 250 ms at m/z of 400 to 1250 Da and the MS/MS data was acquired from m/z of 50 to 2,000 Da. The independent data acquisition (IDA) parameters were as follows: MS1-TOF acquisition time of 250 ms, followed by acquisition of 50 MS2 events of 48 ms for each event. The threshold to trigger a MS2 event was set to 150 counts for ion charge states of +2, +3 and +4. The ion exclusion time was set to 4 s. Finally, the collected data were analyzed using Protein Pilot 5.0 (ABSCIEX) for peptide identifications.

Enzymatic labeling

Samples for enzymatic labeling were buffer exchanged via centrifugation (for crystal suspensions) or using Amicon spin filters (for soluble protein solutions) into a solution containing 20 mM MOPS (pH 7.5). Crystal suspensions were dissolved via the addition of a stock solution of EDTA for a final concentration of 1 mM and buffer exchanged × 5 into a fresh buffer (without EDTA) for experiments with soluble proteins. In general, a 20-μL reaction consisted of 10 μM Sfp PPTase, 15 mM MgCl₂, 100-200 μM modified CoA and 50 μM protein-ybbR conjugate in a buffered solution containing 20 mM MOPS (pH 7.5). The solution was gently shaken on a gel rocker for 16-24 h at room temperature. Completed reactions were buffer exchanged × 5 via centrifugation (3500 rpm) to remove unbound dye and enzyme into a solution containing 20 mM MOPS (pH 7.5). A homogeneous solution of proteins (e.g., ^ybbRRIDC3 monomers or ^ybbRRhuA tetramers) were enzymatically labeled in an identical fashion. Completed reactions were buffer exchanged using 10 kDa Amicon spin filters × 5 to remove unbound dye.

Confocal microscopy

A 5-μL suspension of crystals was pipetted onto a glass slide and covered with a cover slip (Fisher), sealing the edges with clear nail polish to prevent sample drying. Samples were imaged with a 100x oil objective on a spinning-disk confocal Zeiss Axio Observer inverted microscope equipped with a pair of Roper Quantum 5125C cameras. Samples were excited at 488 nm for green fluorescence and 564 nm for red fluorescence. Differential interference contrast and fluorescence images were captured at 1-s and 100-ms exposures, respectively. Images were collected in Slidebook 6 (Intelligent Imaging Innovations) and analyzed using Fiji (http://fiji.sc/Fiji).

Negative-stain transmission electron microscopy (ns-TEM)

A 3-μL suspension of crystals was pipetted onto formvar/carbon-coated Cu grids (Ted Pella, Inc.) that had been glow discharged for 45-60 s. Samples were incubated for 5 min, washed with 50 μL of MilliQ water and blotted with filter paper. A 3-μL drop of 1% uranyl acetate in water was pipetted onto the grid and incubated for 30-60 s and blotted dry with filter paper. Grids were imaged using a FEI Sphera transmission electron microscope operating at 200 keV, equipped with a LaB6 filament and a Gatan 4K charged-coupled device (CCD). Micrographs were collected using objective-lens underfocus settings ranging from 250 nm to 2 μm and analyzed using Fiji (http://fiji.sc/Fiji).

Atomic force microscopy (AFM)

A 10-μL suspension of crystals was deposited onto freshly cleaved mica (Ted Pella, Inc.) and incubated for 10 min. The mica disc was gently dried using a stream of nitrogen with care not to push the drop over the edge of the disc. AFM measurements were performed on a Bruker Dimension Icon ScanAsyst atomic force microscope using a ScanAsyst-Air tip (Bruker) operating in tapping mode. Images were analyzed using NanoScope Analysis (v.1.5, Bruker).

Fluorescence microplate reader measurements

Fluorescence measurements were performed using a 96-well plate (Falcon) containing 50 μL solutions of each sample. Excitation/emission wavelengths of 485/510 nm and 557/583 nm were used for green and red fluorescence respectively with a 2 nm slit width, a 0.2 s integration time and a gain of 100.

Results and Discussion

Enzymatic labeling of peptide-tagged RIDC3 arrays

As our first 2D platform, we chose RIDC3, a variant of the monomeric protein cytochrome cb562, which we previously engineered for assembly into 1-, 2- and 3D protein lattices via Zn²⁺ coordination (Figure 1a).³ Self-assembled RIDC3 arrays have been previously shown to tolerate a broad pH (pH 5 – 9), temperature (4 – 80 °C) and solvent range (H₂O, tetrahydrofuran, isopropanol), providing a robust platform for the surface display of biomolecules.²⁴ To selectively functionalize RIDC3 array surfaces, we chose 4’-phosphopantetheinyl transferase (PPTase), a post-translational modification enzyme. Such enzymes perform selective labeling of biological architectures while maintaining efficiency and specificity under mild solution conditions (e.g., pH 7 at 25 °C in aqueous buffers), all desirable characteristics for interfacing nanomaterials with biological molecules. Native enzymes have been employed to direct protein-protein ligation or surface immobilization in aqueous environments⁴⁶ (e.g., sortase^47,48, PPTase⁴⁹, farnesyl transferase⁵⁰, transglutaminase⁵¹) to functionalize polymeric nanoparticles, gold surfaces, hydrogels and even to tailor cell surfaces. In particular, PPTases have shown remarkable flexibility as a tool for site-selective attachment of chemical probes onto proteins or peptides.⁵² PPTases covalently modify acyl carrier proteins (ACPs) at a conserved serine residue via the transfer of phosphopantetheine (PPant) from coenzyme A (CoA), which serves a crucial role in various biosynthetic pathways.¹³ Previous studies by Walsh and coworkers report the use of phage display to discover an 11 amino acid peptide (ybbR: DSLEFIASKLA) as a surrogate for the native ACP.⁵³ The ybbR peptide acts as a minimal recognition sequence for the surfactin phosphopantetheinyl transferase (Sfp) from Bacillis Subtilis and can be used as a short peptide tag for site-specific protein labeling (Figure 1b).^43,49 Sfp is known to be promiscuous towards many substrates and is capable of enzymatically transferring different sets of biomolecules tethered to CoA in a site-specific manner to a serine residue of ACP or ybbR (Figure 1c, Figure S1).^45,54,55 The promiscuity of Sfp for a range of CoA analogs, and its recognition of short peptide substrates, provides a model system to explore the enzymatic surface functionalization of 2D crystalline protein arrays. As described below, we developed two strategies for the incorporation of a functional peptide handle onto RIDC3 crystals (Figure 1d, e): chemical modification of pre-formed 2D arrays with the peptide and the genetic incorporation of the peptide onto RIDC3 monomers followed by 2D self-assembly.

Figure 1.

Design of peptide-tagged RIDC3 arrays for enzymatic labeling. (a) Structure of RIDC3 with salient metal-binding residues (H63, H73, H77, E81) shown as blue sticks. (b) Chemical structure of ybbR with the site of enzymatic modification highlighted in green. (c) Cartoon schematic showing site-specific modification of ybbR with modified CoA substrates using the PPTase Sfp. 3’, 5’-PAP is 3’,5’-phosphoadenosine phosphate. The chemical conjugation (d) or genetic incorporation (e) of a functional peptide onto RIDC3 to generate peptide-tagged RIDC3 arrays.

We first chose to chemically modify RIDC3 array surfaces with a synthesized ybbR peptide (1) to verify that crystallinity was retained upon the addition of the peptide and (2) ensure that chemoenzymatic labeling of ybbR remained possible on the surface. Suspensions of pre-formed RIDC3 crystals were treated with a 10-fold excess of a dibenzocyclooctyne-N-hydroxysuccinimidyl ester (DBCO-NHS) to modify surface-exposed lysine (Lys) residues (Figure 2a).⁵⁶ The strained cyclooctyne, DBCO, was used to avoid using Cu(I) with Zn-coordinating RIDC3 arrays. Following treatment with DBCO, 10 equiv. of a synthesized ybbR peptide with a Gly spacer and azide-terminated non-natural amino acid, DSLEFIASKLA-G-K(N₃) (Figure S2), was added to the RIDC3-DBCO crystal suspension and allowed to react overnight. Negative-stain transmission electron microscopy (ns-TEM) snapshots of chemically modified RIDC3 arrays confirmed that crystallinity was retained (Figure S3). To quantify ybbR conjugation, suspensions of RIDC3-ybbR crystals were washed and dissolved with the addition of the metal chelator ethylenediaminetetraacetic acid (EDTA) prior to characterization with electrospray ionization mass spectrometry (ESI-MS) and UV-vis spectroscopy (Figure 2b, c). ESI-MS revealed peaks corresponding to 1, 2, and 3 additions of DBCO and ybbR per RIDC3 monomer. These results were corroborated by UV-vis spectroscopy measurements monitoring the addition and consumption of DBCO (ε₃₀₉ = 12000 M⁻¹ cm⁻¹)⁵⁷, with 2.5 ± 0.35 ybbR peptides added per RIDC3 monomer (Figure S3c). A close examination of the RIDC3 crystal packing (PDB ID: 3TOM) revealed up to 7 surface-exposed Lys residues accessible for chemical modification (Figure S4). Thus, we performed tandem mass spectrometry (MS/MS) analysis of trypsin digested RIDC3-ybbR samples and identified three sites of modification on RIDC3: Lys19, Lys28, Lys85 (Figure S4, S5). Spectroscopic and ESI-MS data closely matched MS/MS results and confirmed the covalent attachment of ybbR onto RIDC3 crystals at surface-exposed Lys residues.

Figure 2.

Characterization of RIDC3-ybbR arrays. (a) Cartoon representations showing the generation of RIDC3-ybbR arrays from pre-formed RIDC3 crystals. (b) ESI-MS characterization of DBCO and ybbR additions onto RIDC3. Multiple species are observed in the sample indicating the presence of up to 3 labels per protein. (c) UV-vis characterization of the samples shown in (b).

Enzymatic labeling of RIDC3-ybbR crystals was performed using Sfp PPTase and fluorescent CoA analogs, which provide a facile visual handle for positively identifying crystal modification.⁵⁸ We first tested a small molecule CoA analog (TAMRA-CoA) in a one-pot reaction^43,59, adding the dye with Sfp and MgCl₂ to RIDC3-ybbR arrays and incubating the mixture overnight at room temperature with gentle shaking to facilitate enzymatic conjugation. Control samples were similarly prepared in the absence of Sfp and all solutions were thoroughly washed with buffer to remove any unbound dye (see Materials and Methods). Confocal microscopy measurements after enzymatic labeling showed brightly fluorescent RIDC3-ybbR crystals when incubated with Sfp and CoA whereas no fluorescence signals were detected in a negative control sample devoid of Sfp (Figure 3a). Importantly, differential interference contrast (DIC) images show a perfect overlap between areas of fluorescence and RIDC3-ybbR crystals in the field-of-view (Figure 3a). Corresponding ns-TEM micrographs of the labeled samples confirm that the arrays remain intact after enzymatic treatment and retain crystallinity (Figure S6). Furthermore, PPTases are promiscuous in their recognition of CoA analogs, so we tested enzymatic labeling using a larger biomolecule, GFP-CoA. In order to enable the enzymatic transfer of a GFP-PPant group onto ybbR, a GFP-CoA analog was generated by first incubating sfGFP with the bifunctional linker, sulfosuccinimidyl 4-(N-maleimidomethyl) cyclohexane-1-carboxylate (sSMCC)⁶⁰, to label Lys residues with a NHS-ester moiety followed by the addition of CoA-SH for thiol-maleimide coupling (Figure S7a). CoA addition was quantified using UV-vis measurements, yielding 1.4 CoA per sfGFP (Figure S7b). As with TAMRA-CoA modification, incubation of RIDC3-ybbR crystals with Sfp and GFP-CoA resulted in brightly fluorescent crystals (Figure 3a) and the absence of Sfp showed minimal fluorescence at the location of the arrays. Based on the successful chemoenzymatic transfer of a small molecule dye (TAMRA-PPant) and a protein (sfGFP-PPant) onto RIDC3-ybbR crystals, we posited that a solution containing TAMRA-CoA and GFP-CoA could be used to colocalize both labels on the arrays. Suspensions of RIDC3-ybbR crystals incubated with both CoA substrates and Sfp displayed distinct fluorescence at TAMRA and GFP wavelengths, and an overlay of both images showed an unambiguous overlap of both signals on the crystal surface (Figure 3b).

Figure 3.

Enzymatic labeling of RIDC3-ybbR arrays. (a) Cartoon representation of enzymatic labeling and confocal microscopy images showing TAMRA- and GFP-modified RIDC3-ybbR arrays. Control samples devoid of Sfp PPTase are not fluorescent. (b) Confocal microscopy images of RIDC3-ybbR arrays co-labeled with TAMRA and GFP. The fluorescence overlay of the red and green fluorescence confirms the addition of both labels. (c) Fluorescence profiles of RIDC3-ybbR arrays after enzymatic modification with TAMRA-CoA, GFP-CoA, or both. Control experiments were incubated with both modified CoA substrates in the absence of Sfp. N = 3 for fluorescence measurements. Scale bars in (a) and (b) are 10 μm.

While identifying fluorescently labeled crystals was straightforward using confocal microscopy, quantification of enzymatic labeling of RIDC3-ybbR crystals proved challenging. RIDC3-ybbR crystals after enzymatic modification could be dissolved using EDTA to yield soluble protein amenable for UV-vis spectroscopic measurements; however, the absorbance values for covalently tethered TAMRA or GFP were too low relative to that of RIDC3 for unambiguous detection. This is due in part to the intense absorbance of the c-type heme covalently bound to the RIDC3 monomer at the Soret and Q bands (415 and 527 nm, respectively), partially overlapping with the absorption maxima of GFP (485 nm) and TAMRA (554 nm) (Figure S8a). Instead, we directly probed crystal suspensions using a microplate reader to measure fluorescence intensities for labeled vs. unlabeled samples (Figure 3c). A standard curve of [fluorophore] vs. fluorescence intensity was prepared using TAMRA-CoA or GFP-CoA solutions to determine [RIDC3-ybbR + TAMRA-PPant] or [RIDC3-ybbR + GFP-PPant] based on fluorescence values (Figure S9a). Crystals were dissolved after fluorescence measurements to obtain protein concentrations via UV-vis spectroscopy. Our calculations suggest <0.5% of RIDC3-ybbR is enzymatically modified by Sfp using either CoA analog (Figure S9c). This unexpectedly low labeling efficiency prompted further analysis.

Since enzymatic reactivity at the surface may be lower than in solution, we obtained RIDC3-ybbR monomers (from dissolving chemically modified RIDC3-ybbR crystals with EDTA). Functionalization of these monomers revealed approximately 0.33 TAMRA per RIDC3, significantly higher than labeling on the crystal surface (Figure S8b). Previous studies on RIDC3 arrays have shown that solution-assembled crystals are multilayered²⁴, likely limiting enzyme access to “interior” RIDC3 proteins within a 3D stack. Indeed, dry-state AFM characterization of RIDC3-ybbR crystals confirms that they consist of ~4 nm thick 2D layers that stacked to an average height of 133 ± 40 nm (Figure S9b). Considering that only ybbR peptides decorating the exterior of RIDC3-ybbR crystals are likely available sites of modification using Sfp, we used a combination of spectroscopic data (to quantify ybbR conjugation onto RIDC3) and AFM measurements (to gather crystal dimensions) to determine the enzyme-accessible fraction of the average RIDC3-ybbR crystal. From 46 crystals, we determined average dimensions of a = 3.2 ± 0.84 μm, b = 8.1 ± 2.4 μm, c = 0.133 ± 0.04 μm (Figure S9b, c). We next used the RIDC3 crystal packing to determine the protein “step” in each crystal dimension: 1 protein per 3.79 nm x 3.46 nm x 2.3 nm volume (PDB ID: 3TOM). Based on these measurements, we calculated the surface coverage of RIDC3 proteins to be ~3.5% of total protein contained in a given crystal. Correlating fluorescence intensities to absorbance values, we computed ~3% and ~10% enzymatic labeling of RIDC3-ybbR crystal surfaces for GFP-CoA and TAMRA-CoA respectively. Although these values are lower than those observed for labeling of free proteins in a homogenous solution, our experiments show that it is possible to enzymatically tailor the surfaces of “solid”, micron-scale 2D protein materials with non-negligible yields. To our knowledge, this is the first demonstration of the generation of functional protein arrays directed by enzymatic processes with the use of two distinct covalent tags.

Generation of RIDC3 arrays with CoA tags

In addition to chemically conjugating ybbR tags onto 2D RIDC3 arrays as a means for CoA-mediated enzymatic labeling with small fluorophores, we wondered whether the 2D protein surface could first be functionalized with CoA molecules instead and subsequently labeled with guest proteins bearing ybbR tags. Thus, RIDC3-CoA arrays were prepared using a similar conjugation strategy to that used for generating GFP-CoA (Figure 4a). Pre-formed RIDC3 arrays were first treated with sSMCC to covalently modify Lys residues with the NHS-ester moiety followed by the addition of CoA-SH (unmodified, natural substrate) to react with the maleimide portion of SMCC. CoA addition was quantified via UV-vis spectroscopy λ_max = 259 nm) and determined to be ~0.8 per RIDC3 (Figure 4b). Although a 25-fold excess of CoA was added after SMCC coupling, lower efficacy of thiol-maleimide chemistry at the array surface (especially compared to the strained cyclooctyne click chemistry used for ybbR conjugation) and the potential for CoA self-dimerization perhaps contributed to a lower CoA:RIDC3 ratio. Nevertheless, RIDC3-CoA crystals were incubated with Sfp and a genetically encoded ^ybbRGFP (EGFP-ybbR N-terminal fusion) to generate fluorescent GFP-labeled arrays (Figure 4c, d). Control experiments in the absence of PPTase or ^ybbRGFP showed no fluorescence, indicating that there is little to no nonspecific association of GFP to the RIDC3-CoA arrays. Interestingly, ^ybbRGFP-labeled RIDC3-CoA arrays consist of punctate patches of GFP forming an outline along the edges of the array (Figure 4d) unlike RIDC3-ybbR samples, which showed uniform fluorescence across the surface. This could arise from a combination of (1) slower reaction kinetics for maleimide-thiol coupling (in contrast to the rapid copper-free click chemistry employed for ybbR conjugation previously) resulting in (2) a greater density of CoA molecules covalently tethered onto the edges of the array and (3) easier access of Sfp and ^ybbRGFP to the edges relative to the interior. Quantification of ^ybbRGFP labeling onto RIDC3-CoA labels revealed ~8% of surface exposed proteins were labeled, on par with that of RIDC3-ybbR arrays (Figure 4c, Figure S10). These findings illustrate the feasibility of a tripartite protein material platform, in which the surface of the 2D arrays of one protein (RIDC3) can be processed by a second, catalytic protein (Sfp) to site-specifically attach a third protein (GFP).

Figure 4.

Characterization and enzymatic labeling of RIDC3-CoA arrays. (a) Cartoon schematic for the generation of CoA-tagged RIDC3 arrays using sSMCC (bifunctional linker) and CoA-SH, and subsequent enzymatic labeling with ^ybbRGFP. (b) UV-vis characterization of RIDC3 and RIDC3-CoA normalized to the RIDC3 absorption maximum at 415 nm. The increase in absorbance at 259 nm is indicative of CoA conjugation onto RIDC3. (c) Fluorescence profiles for RIDC3-CoA arrays in the presence and absence of Sfp PPTase and ^ybbRGFP. (d) Confocal microscopy images of RIDC3-CoA samples incubated with ^ybbRGFP in the presence and absence of Sfp PPTase. Scale bars are 10 μm.

Design and characterization of 2D protein arrays with genetically incorporated ybbR tags

Our results thus far indicated that enzymatic labeling only engages a fraction of the surface sites containing ybbR or CoA tags. While this can, in large part, be attributed to the inefficient mass transport of crystalline substrates and the immobility of the proteins embedded therein, we posited that the non-optimal positioning of surface attachment sites may also play a role. Thus, we constructed ^ybbRRIDC3 fusions with the ybbR peptide genetically appended to the polypeptide termini of RIDC3 to allow for (1) a uniform distribution of peptide upon forming crystals and (2) to bypass the post-assembly conjugation of ybbR or CoA to generate enzymatically-addressable 2D arrays (Figure 1e).

Upon close inspection of the N- and C-termini of RIDC3, we determined that only the C-terminus of RIDC3 faces the exterior of the crystal surface. Thus, RIDC3 was genetically modified at the C-terminus with a Gly-Gly spacer followed by the ybbR peptide (RIDC3 – GG – DSLEFIASKLA) to generate the ^ybbRRIDC3 plasmid, which was expressed and purified in the same manner as RIDC3. However, in the course of extensive studies, we found that ^ybbRRIDC3 fusions underwent a degradation process (of unknown mechanism) in which the final four residues (SKLA) of ybbR fusion were cleaved and thus rendered inactive for Sfp-mediated labeling (Figure S11). Although the fraction of uncleaved ^ybbRRIDC3 fusions was amenable to enzymatic labeling and self-assembly into extended arrays (Figure S11c, d), the cleavage process persisted even upon self-assembly. Thus, we set out to test the genetic-tagging strategy on another protein building block, with the expectation that an altered protein environment may prevent ybbR auto-cleavage.

In a recent report from our laboratories, a C₄-symmetric protein – L-rhamnulose-1-phosphate aldolase (RhuA) – was modified with cysteine (Cys) residues at its corners (^C98RhuA) to facilitate the self-assembly of unsupported 2D arrays in solution under controlled oxidation conditions via disulfide bond formation.^39,61 The ^C98RhuA proteins tessellate to form an alternating arrangement of tetramers in the crystalline lattice. Furthermore, rotations about the individual axes of symmetry at the flexible disulfide linkages create a coherently dynamic lattice that display auxeticity (i.e. a longitudinal expansion when stretched in the transverse direction) with a Poisson’s ratio of −1. Aside from these unique dynamic properties, the larger size (274 residues per monomer vs. 106 residues in RIDC3) and inherent symmetry of ^C98RhuA could provide a more robust scaffold for the genetic incorporation of ybbR while retaining its self-assembly properties. As with RIDC3, ybbR was genetically appended to the C-terminus of ^C98RhuA with a short spacer (^C98RhuA – SGSG – DSLEFIASKLA) so that the peptide is displayed at the surface of 2D arrays minimally interfering with disulfide formation (Figure 5a, b). Importantly, no truncation of ybbR was observed during the purification and isolation in this construct (Figure S12a). ^ybbRRhuA self-assembles into a suspension of ordered crystals at the same solution conditions as ^C98RhuA (20 mM Tris pH 7.5, 10 mM βME, 1 mM ZnCl₂), maintaining crystallinity and lattice dynamics (Figure 5c). Furthermore, ^ybbRRhuA tetramers are successfully modified after incubation with Sfp PPTase and TAMRA-CoA, containing ~0.22 TAMRA per polypeptide (i.e. 0.89 TAMRA per^ybbRRhuA tetramer) after enzymatic labeling (Figure 5d). Enzymatic labeling of ^ybbRRhuA crystals with Sfp PPTase and modified CoA was observed via confocal microscopy and ns-TEM. Brightly fluorescent crystals were found with both TAMRA-CoA and GFP-CoA-incubated solutions whereas ^ybbRRhuA arrays in the corresponding DIC images were harder to discern likely due to the increased porosity and minimal 3D layering of 2D crystals relative to RIDC3 arrays (Figure 5e). TEM analysis of the labeled samples confirmed that arrays remained intact after enzymatic modification (Figure 5f). Confocal microscopy and ns-TEM analysis of control samples devoid of Sfp PPTase were minimally fluorescent and remained crystalline (Figure S12b). Quantification of ^ybbRRhuA labeling was performed using fluorescence measurements on a plate reader as previously described (Figure S12c). We calculated ~0.7% and ~5.6% enzymatic labeling of ^ybbRRhuA crystal surfaces for GFP-CoA and TAMRA-CoA respectively. As seen with chemoenzymatic labeling of a solution containing ^ybbRRhuA tetramers, labeling of the array surfaces was less efficient than with their RIDC3 counterparts. However, in contrast to genetically modified ^ybbRRIDC3, the ^ybbRRhuA construct was found to be intact and stable and could be enzymatically modified without loss in crystallinity. Our results thus illustrate the robust enzymatic labeling of 2D protein arrays that are genetically modified with peptide recognition tags.

Figure 5.

Characterization and enzymatic modification of ^ybbRRhuA. (a) Surface and cartoon representations of ^C98RhuA. C98 residues are shown as black spheres. (b) Cartoon schematic of ^ybbRRhuA and self-assembly into 2D crystals. (c) TEM micrographs of 2D ^ybbRRhuA crystals. FFT inset shows an identical pattern to that of ^C98RhuA crystals. Scale bars are 5 μm (left) and 1 μm (right). (d) UV-vis characterization of ^ybbRRhuA and TAMRA-CoA in the presence and absence of Sfp. Samples are normalized to the ^ybbRRhuA absorption maximum at 280 nm. Confocal microscopy (e) and TEM characterization (f) of ^ybbRRhuA incubated with Sfp PPTase and either TAMRA-CoA or GFP-CoA. Scale bars are 10 μm (e) and 5 μm (f).

Conclusions

In this study, we have illustrated several strategies for the site-selective chemoenzymatic functionalization of artificial protein assemblies. Previous studies have investigated bioconjugation using Sfp on polymeric nanoparticles⁵⁴, hydrogels⁶², antibodies⁵⁵, protein-DNA chimeras⁶³ or in conjunction with sortases for dual labeling capabilities.⁶⁴ In our own work, we have utilized bacterial enzymes of this kind, and their peptide substrates, to assemble amphiphilic nanomaterials.⁶⁵ Our study now expands this strategy to artificial, 2D crystalline protein arrays. Using two different protein systems that we previously designed for 2D self-assembly, we demonstrated that a) molecular or peptidic molecular tags could be attached to the 2D arrays both before or after self-assembly, b) the tag attachment could be done both chemically and genetically, and c) the selective enzymatic labeling of these recognition tags could proceed with small molecules as well as with proteins as substrates. Importantly, in all cases, the structural fidelity of the proteins as well as the crystalline order of their 2D assemblies were maintained. Although the overall chemoenzymatic labeling yields of the 2D protein arrays are far from quantitative, this is possibly due to the inherent physical limitations posed by a ternary reaction involving an enzyme, a suspended “solid-state” protein array with limited mobility as a template, and large fluorescent molecules (even a protein) as a substrate. This poses an interesting challenge for future studies in which we can envision optimizing chemoenzymatic labeling by engineering the protein array surfaces, the peptide substrates (leveraging novel machine learning and optimization tools⁴⁵), and the Sfp enzyme for complementary electrostatic interactions or improved reaction sterics. Another exciting possibility is the use of secondary enzymes (like acyl carrier protein hydrolases),^43,66 which have been shown to selectively cleave ybbR peptide tags, opening the path for the reversible tailoring of protein array surfaces.

Enzyme-mediated post-translational modification of proteins is ubiquitous in nature and greatly increases the complexity of the proteome. Here, we have reported one of the first examples for leveraging biological enzymes to selectively modify solid substrates with small molecules and proteins, with an eye toward the hierarchical construction of multi-component protein systems. The versatility and inherently modular nature of the strategies described herein offer promise in generating functional and hybrid materials for use in sensing, catalysis, immotherapeutics, or lab-on-a-chip design frameworks.