Synthesis and Characterization of Bioconjugates of S-Layer Proteins

Bioconjugate Chem. 2004, 15, 685−693

685

Synthesis and Characterization of Bioconjugates of S-Layer Proteins Parthasarathy Sampathkumar and M. Lane Gilchrist, Jr.* Departments of Chemical and Biomedical Engineering, The City College of the City University of New York, New York, New York 10031. Received November 6, 2003; Revised Manuscript Received February 11, 2004

The self-assembling proteins that form crystalline surface layers (S-layers) on many microbial species have found numerous applications due to their nanostructured nature. To devise a new method to construct surface displays that exploit S-layer self-assembly activity and nanostructural properties, we have constructed polymer bioconjugates of S-layer proteins. The conjugates formed are similar in function to the monomer alkanethiols that form self-assembled monolayers (SAMs) on gold surfaces. However, the self-assembly is driven by the protein “headgroup” that positions polymer-tethered endgroups on a surface. This paper examines the integration of protein purification, conjugation, and surface assembly that has led to the development of this new method for the formation of nanostructured surfaces. Purified S-layer proteins from Lactobacillus brevis were conjugated with small molecule probes and polymers using amine-based reactions. To keep multiple labeling of protein amine groups to acceptable levels, the conjugations were performed at pH 6.5, allowing for limited yields (24-39%) as determined by mass spectrometry and SDS-polyacrylamide gel electrophoresis. As the presence of high levels of unlabeled S-layer proteins is undesired, we have developed a protocol for further purification that employs monomeric avidin affinity chromatography. The surface selfassembly of the polymer bioconjugates onto amine-terminated microspheres was studied using epifluorescence, confocal, and scanning electron microscopy. The surfaces obtained exhibited homogeneous distributions of tethered molecules. Also, in cases where the modular assembly of two distinct types of tethered endgroups was accomplished, there was no evidence for phase separation in the surfaces. The modular assembly method will provide a potential route to controlling surface display density as the starting assembly conditions guide displayed endgroup concentrations in mixed molecular monolayers.

INTRODUCTION

Microorganisms often employ crystalline surface monolayers of self-assembling proteins for molecular display, exozyme binding, and other extracellular functions (1). These protein monolayers, collectively known as the crystalline surface layer or S-layer, contain subunits with molecular weights ranging from 40 to 200 kDa, depending on the species (1-8). In the assembled form, the S-layer proteins form porous crystalline molecular monolayers with unit cell dimensions on the order of 4-15 nm (5). The regular nanostructure and high degree of longrange order provides favorable attributes for potential uses of the assembled S-layer in nanostructured materials (1, 2, 7, 9). This study concerns the development of a new method to be used for building nanostructured, selfassembled monolayers based on S-layer bioconjugates. Various subtypes of S-layer proteins have been used in wide ranging applications such as ultrafiltration membranes (2), solid supports for lipid bilayer membranes (10), and nanoparticle arrays (11). S-Layer proteins that associate with the peptidoglycan layer of the microbial cell wall, can in general, be disassembled, but not denatured, by treatment with chaotropic agents such as LiCl or guanidine HCl. These proteins can be separated from cells and media by centrifugation and then reassembled in vitro by dialyzing out the chaotrope or salt. Thus, successive steps of disassembly and reassembly can be used to obtain gram quantities of highly purified S-layer proteins. The reassembly of these proteins onto interfaces can be controlled via surface charge or hydro-

phobicity. This process has been studied extensively by Sleytr and co-workers (1, 2, 9). These advances in purification and reassembly have facilitated the ultrastructural analysis of the assembled layers and paved the way for nano- and biotechnological applications (2, 5). The crystalline nanostructure of the S-layer gives rise to protein surface sites that repeat regularly in two dimensions over the structure assembled onto an interface. The arrays of protein reactive sites provided by surfaces of assembled S-layer proteins have been used for the conjugation of antibodies, vaccine adjuvants, and other molecules. Thus, the proteins produce a very highdensity array of displayed molecules that retain the underlying regular spacing of the S-layer (2, 9). A recent major advance involved the successful construction of S-layer fusion proteins containing antibody fragments and streptavidin monomers that have been assembled onto surfaces (12, 13). In previous studies involving surface chemical modification, the conjugation was carried out by first assembling the S-layer and then conjugating a single moiety (9). In contrast, the aim of the research described here is to first build individual bioconjugates of various functionalities tethered by highly hydrophilic poly(ethylene glycol) (PEG) linkers and then assemble them in a modular fashion onto surfaces. The driving force for selfassembly is provided by the S-layer protein, and, in essence, the S-layer proteins serve as headgroups in molecules designed to form self-assembled monolayers (SAMs). The PEG tethers are included so as to provide

10.1021/bc034204r CCC: $27.50 © 2004 American Chemical Society Published on Web 07/03/2004

686 Bioconjugate Chem., Vol. 15, No. 4, 2004

passivation from nonspecific protein adsorption (14-17). The important aspect of SAM formation by this method is that a new route to form surfaces of controlled surface density that contain mixtures of displayed functional groups is envisioned. As this system is built upon the nanometer spacing of the S-layer, larger molecules such as polymers are displayed in adjacent lattice sites with imposed separation. This spacing imparts two major favorable characteristics to such self-assembled monolayers. First, phase separation during assembly of mixed monolayers would be suppressed, leading to uniform surface distributions of displayed molecules of disparate sizes or functionalities (18-20). Second, the intrinsic separation between neighboring endgroups would curb steric hindrance of molecular recognition and block interendgroup interactions that could result in aggregation and loss of functional activity. The overall consequence would be the placement of more active and accessible endgroup molecules per unit area, leading to optimized molecular recognition at interfaces and increased miniaturization and complexity of the resulting molecular array. This paper provides a detailed description of the synthesis and characterization of S-layer protein bioconjugates, spanning from small molecule probes to biotinylated polymers. Reversed-phase high performance chromatography/mass spectrometry (RP-HPLC-MS) and polyacrylamide gel electrophoresis (SDS-PAGE) were used to obtain bioconjugation reaction yields. Gel filtration and monomeric avidin affinity chromatography was used to purify the bioconjugates. S-Layer bioconjugates were assembled onto amine-terminated polystyrene beads, and ligand display and surface distribution were studied using epi-fluorescence, confocal, and electron microscopy. MATERIALS AND METHODS

Materials. S-Layer proteins (SLP) were extracted from the surface of the bacteria Lactobacillus brevis ATCC 8287 that were purchased from ATCC, Manassas, VA. Biotin-PEG3400-N-hydroxysuccinimide (Bio-PEG-NHS), fluorescein-PEG2000-N-hydroxysuccinimide (Fl-PEG-NHS), and vinyl sulfone-PEG3400-N-hydroxysuccinimide (VSPEG-NHS) were purchased from Shearwater Polymers (Huntsville, AL). Fluoroscein isothiocyanate (FITC) was obtained from Sigma, St. Louis, MO. BODIPY 558/568N-hydroxysuccinimide (Bodipy558), D(+)-biotin, and NeutrAvidin-coated TransFluoSpheres fluorescent nanospheres (100 nm nominal size; 488 nm excitation/605 nm emission) were obtained from Molecular Probes, Eugene, OR. Monomeric avidin affinity column gels and Sephadex G-50 and G-25 gel supports were purchased from Sigma, St. Louis, MO. Biotin-N-hydroxysuccinimide (Bio-NHS) and BS3 (bis(sulfosuccinimidyl) suberate) were obtained from Pierce Chemical, Rockford, IL. All other chemicals used to make buffers were purchased from Sigma except for glycine (BioRad Laboratories, Hercules, CA). The regenerated cellulose dialysis tubing used during purification was purchased from Fisher Scientific (nominal molecular weight cut off between 12 000 and 14 000). S-Layer Protein Purification. The bacteria (L. brevis ATCC 8287) were grown unstirred in MRS broth (Difco Laboratories, Sparks, MD) at 37 °C for 2 days after inoculation with 3% (v/v) 2 day culture. The extraction and purification steps are similar to that performed by Lortal and co-workers (21). Exponential-phase cells were recovered by centrifugation at 14 000g for 20 min. The cell pellets were resuspended in distilled water to wash the remaining broth and centrifuged again at 14 000 g

Sampathkumar and Gilchrist

for 20 min. After two such wash cycles, 10-15 mg of moist pellet was suspended per milliliter of 5 M lithium chloride (LiCl) (Sigma) and incubated for 30 min at 4 °C to extract the S-layer protein. The extract was then centrifuged at 14 000g for 20 min, and the supernatant was designated as crude extract. The crude extract was purified using a gel filtration chromatography column, 2.5 × 50 cm, with a uniform packing of Sephacryl S200 HR (Sigma) gel, equilibrated with 5 M LiCl. The protein was eluted at 30 mL/h, and 1.5 mL fractions were collected. In some runs, the crude extract was concentrated prior to injection into the column. They were concentrated 10-fold using Millipore Ultrafree-15 centrifugal filter device (Sigma) having a 30 kDa cut off. The device was first washed for 10 min with deionized water, and then the protein was added (in 5 M LiCl solution) and spin-filtered at 4000g until the final concentration was reached. In some other runs, the protein was first dialyzed against water overnight, and the assembled protein was resuspended in 5 M LiCl before injection. Construction of Bioconjugate Modules. Purified protein suspended in 5 M LiCl was dialyzed overnight against water. The assembled proteins were dissolved separately in 50 mM sodium phosphate buffers at pH 6.5 to obtain a final protein concentration of 5-10 mg/mL. The probes were then dissolved in deionized water and immediately added to the protein solution for conjugation. All reactions were carried out at 4 °C in darkness, in a flask shaker. Twenty-fold molar excess of small molecule probes (FITC, Bodipy558, and Bio-NHS) and 10 molar fold excess of Fl-PEG-NHS, Bio-PEG-NHS, and VS-PEG-NHS were used for conjugation, with 24 and 12 h reaction times, respectively. The reactions were terminated with 150 mM hydroxylamine (Sigma) at pH 7. Unreacted probes were removed by gel filtration either through Sephadex G-25 (Sigma) (small molecule probes) or Sephadex G-50 (PEG-tethered probes) gravity columns that have been equilibrated with 50 mM, pH 6.5 sodium phosphate buffer. The gels were hydrated for 1 h and 3 h, respectively, and decanted for any floating particles prior to loading. The S-layer protein small molecule probe conjugates were analyzed with HPLC in tandem with mass spectrometry (HPLC-MS) to identify their respective yields. PEGylated S-layer protein yields were obtained from gel desitomograms of 4-15% Tris SDSPAGE (Bio-Rad Laboratories, Hercules CA) runs carried out in a minigel apparatus with Coomassie protein staining. Gel densitomograms were obtained using a 16 bit Umax flatbed scanner, with image processing carried out using Igor Pro 4.0. Monomeric Avidin Affinity Column Purification. To separate the biotinylated conjugates from the unlabeled S-layer protein, we used affinity columns containing monomeric avidin. These columns selectively retain the biotinylated proteins during disassembly process or 5 M LiCl wash. The required biotinylated and fluorescently labeled protein monomers were eluted with free biotin buffer (50 mM sodium phosphate, 150 mM sodium chloride containing 1 mg/mL D(+)-biotin). The affinity gels after packing were washed with four column volumes of 50 mM sodium phosphate with 1 mg/ mL of D-biotin, pH 7. Subsequently, the columns were washed with 8 to 10 volumes of 100 mM glycine‚HCl, pH 2, to displace biotin from exchangeable sites. The resins were then reequilibrated with 50 mM sodium phosphate, 150 mM sodium chloride, pH 7, prior to sample injection. Samples of volume approximately 10% the column volume were injected to saturate the resin. The resins were

Bioconjugates of S-Layer Proteins

mildly agitated once to ensure proper mixing of the contents and incubated in dark for 30 min. The buffer was then replaced with 5 M LiCl, and the resin was again incubated in dark for 1 h to facilitate complete disassembly of the S-layer protein multimers. Following the incubation, there was 5 M LiCl wash to remove unlabeled S-layer proteins, followed by a neutral buffer wash and free biotin elution of the conjugates at 15 mL/h flow rate with all the fractions being collected. The affinity-purified biotinylated conjugate is recovered during the free biotin buffer elution (50 mM sodium phosphate with 1 mg/mL of D-biotin, pH 7). All the peak fractions were pooled and collected for further analysis. Microsphere Surface Assembly Experiments. Purified S-layer bioconjugates were assembled onto amineterminated polystyrene microspheres (6 µm nominal size; Polysciences, Inc, Warrington, PA) using the following procedure. Purified S-layer bioconjugates at 2 mg/mL concentration were centrifuged at 14 000g to remove assembled fragments. The resulting supernatant containing disassembled S-layer conjugates and unlabeled S-layer protein was used in surface assembly. To conduct modular self-assembly, desired proportions of individual conjugates were mixed and added to amine-terminated S-layers in large excess to total surface area. The selfassembly was conducted at pH 8 in phosphate buffer for 24 h at 4 °C. Subsequently, the assembly solution was removed by 50g centrifugation-induced settling of the microspheres. The microspheres were then gently washed with pH 8 buffer to remove any remaining unassembled protein, and then the amine-reactive homobifunctional cross-linker BS3 was added at a concentration of 10 mM for a 30-min reaction in 50 mM sodium phosphate buffer, pH 8 at 4 °C. The cross-linking reaction was quenched with hydroxylamine (150 mM), followed by a washing step with pH 8 buffer to remove any remaining deactivated BS3. At this stage the microspheres assembled with biotin-PEG3400-S-layer conjugates were incubated overnight with and excess of NeutrAvidin-coated TransFluoSpheres fluorescent nanospheres at 4 °C, followed by washing with pH 8 buffer to remove any excess unbound nanospheres. Mass Spectrometry. S-Layer protein samples and their small fluorophore bioconjugates were analyzed using reversed-phase HPLC (C-18 column) in line with an electrospray ionization mass spectrometer. The studies were carried out using an HPLC-ESI-MS Agilent Chem Station at the Hunter College of the City University of New York Mass Spectrometry facility. The samples were prepared to give an average protein content of 1015 µg in a 10% acetic acid solution. The samples injected into the C18 column of the HPLC-MS were eluted at the rate of 0.5 mL/min with a gradient flow of acetonitrile. Confocal, Epi-Fluorescence, and Electron Microscopy. Samples were imaged using a Leica TCS-SP four channel confocal microscope at the Mt. Sinai School of Medicine, using 488 nm excitation for fluorescein and 568 nm for the TransFluoSpheres fluorescent nanospheres. The transmitted light channel was used to position the imaging plane at the center of the microspheres. Epifluorescence microscopy was conducted using a Nikon TE300 inverted microscope with B-2A and G-2E/C filter cubes for fluorescein and TransFluoSpheres fluorescent nanospheres, respectively. Samples for scanning electron microscopy were dried, sputter-coated with gold, and observed with a JEOL 35 scanning electron microscope at the Mt. Sinai School of Medicine. Image analysis, including fluorescence emission intensity profiling, was accomplished with IgorPro 4.0 and NIH Image 1.63.

Bioconjugate Chem., Vol. 15, No. 4, 2004 687 RESULTS AND DISCUSSION

Purification of Lactobacillus brevis S-Layer Protein. Surface layer proteins range in size from 40 to 200 kDa, depending on the species of origin. Our rationale for the choice of the smaller Lactobacillus variant S-layer protein for these studies was based on suitability for facile characterization by both mass spectrometry and eventually high-resolution solid-state NMR. The Lactobacillus brevis S-layer protein was obtained at yield of greater than 100 mg/L of microaerobic culture in MRS broth. At these high yields a 10 g pellet of the bacterial cells will conservatively yield ∼0.5 g of S-layer protein. The protein purification was monitored using SDSPAGE, shown in Figure 1A. The S-layer protein crude extract is run through two cycles of (1) dialysis (assembly: 30 kDa Mw cutoff), (2) 5 M LiCl disassembly, and (3) dialysis once again to bring the purity up to >50%, shown in Figure 1, Lane 2. Three main impurities remain with estimated molecular weights of 23, 27, and 35 kDa. Gel permeation chromatography was used to isolate the L. brevis S layer protein with a final purity of >95%, shown in Lane 3. The target protein for the bioconjugation studies was first cloned and sequenced by Vidgren and Palva (22). Their nucleic acid sequence analysis revealed a 465 residue open reading frame with a capacity to encode a 48 159 Da protein. This sequence contains a 30 residue signal peptide common to Grampositive species. Analysis of the purified protein with reversed-phase HPLC electrospray mass spectrometry (RP-HPLC-MS) gave a molecular weight of 45 238 Da. The molecular mass of the L. brevis S-layer protein corresponds to the 465 residue open reading frame with the 30 residue signal peptide removed, resulting in a 435 residue protein. This purified S-layer protein is the starting material for the bioconjugation studies outlined below. Construction of Lactobacillus brevis S-Layer Protein Bioconjugates. The goal of this work was to produce functional S-layer bioconjugates by tethering molecules of various sizes while retaining the selfassembly “activity” of the S-layer protein. As this protein contains no cysteines and many (over 30 each) primary amines, carboxylates, and hydroxyls, when using standard coupling strategies, the challenge is to limit multiple labeling. To achieve this outcome, we have used aminebased coupling chemistry, targeting the R-amino group of the amine terminus by running the coupling reaction at below neutral pH to suppress -amino lysine reactivity. An additional, S-layer specific consideration in the conjugation is the multimeric state of the protein. In these studies the S-layer protein is undergoing the bioconjugation reaction in the assembled form. Thus, the majority of the protein is assembled into multimeric sheet structures with up to micron length scales (as evidenced by visual light scattering and fluorescence microscopy; data not shown). In this state, the larger multimers in the solid state are in dynamic equilibrum with a soluble form of the protein, likely composed of monomers or smaller multimers. Conjugation in the assembled form is dictated by the fact that the fully disassembled or soluble form of the protein is obtained only at high concentrations of LiCl or guanidine hydrochloride, solution conditions incompatible with bioconjugation reactions. An inherent advantage is that reactive sites left exposed when the protein is in the assembled form will most likely lead to conjugates that retain the ability to self-assemble. To initiate the studies and probe the reactivity of the S-layer protein amines at lower pH, we have employed number of amine-reactive labels for bioconjugation, rang-

688 Bioconjugate Chem., Vol. 15, No. 4, 2004

Figure 1. Gel electrophoresis results for the purification of L. Brevis S-layer protein and synthesis of poly(ethylene glycol)based bioconjugates (SDS-PAGE 4-15% Gel). Lane 1 is from the molecular weight markers. Lane 2 is from proteins extracted from the surface of the cells after two cycles of: (1) dialysis (30 kDa Mw cutoff), (2) washing, (3) 5 M LiCl disassembly followed by concentration to give 10 mg/mL. Lane 3 is after prep-scale gel permeation chromatography using sephacryl SH-200 as the support. Lanes 4 and 5 were obtained after conjugation of vinyl sulfone-PEG3400 and fluorescein-PEG2000, respectively. The bottom figure shows gel densitomograms of lanes 4 and 5 in traces A and B, respectively, zoomed into the relevant molecular weight range.

ing from fluorescent probes to polymers. RP-HPLC-MS was used to calculate the yields for the fluorescent probes;

Sampathkumar and Gilchrist

for the polymer conjugates, SDS-PAGE densitometry was employed. After exploring a range of conditions in preliminary bioconjugation studies, we converged upon two sets of conditions for low molecular weight probes and polymers. The results of bioconjugation at these two sets of conditions are tabulated in Table 1. In general, the reactions were conducted at S-layer protein concentrations of greater than 5 mg/mL with 20-fold and 10fold molar excess of reactant for low molecular weight probes and polymers, respectively. However, presumably only low levels of reactive N-hydroxysucccinimate and isothiocyanate moieties remained at these 12-24 h long reaction times. For all cases, the yields of singly labeled bioconjugates ranged 24-39%. In all cases except for the FITC coupling, doubly labeled species were obtained, with yields ranging from 4 to 16%. While formed in substantial amounts, these multiply labeled bioconjugates are not expected to limit the applicability of the method as long as self-assembly activity is maintained. For the polymer bioconjugates, the yield of the reactions was estimated from gel densitomograms, shown in Figure 1B. Trace A, obtained from the vinyl sulfone-PEG3400SLP bioconjugates, contains the unlabeled protein (denoted with an asterisk) and two peaks of successively smaller areas from the bioconjugates. The peaks are Gaussian in line shape, leading to straightforward peak fitting to calculate the yields. For the case of the fluorescein-PEG2000-SLP bioconjugate, the peak resulting from the conjugates is more complex and complicated by the fact that the migration in the gel is less for the smaller polymer bioconjugate (2000 vs 3400 Da for the vinyl sulfone-PEG) (23). To limit the yield of multiply labeled species, we have conducted the conjugation at pH 6.5 in order to protonate the -amino groups of surface lysines and thus lower their reactivity. One consequence of this reaction condition is that the majority of the protein remains unlabeled and thus must be separated from the conjugates. In the more common conjugation scenario where multiple labeling is desired, repeated cycles of labeling are carried out to increase the yield. However, in this case this strategy is expected to give unwanted higher levels of multiple site labeling that could potentially affect self-assembly. Since the conjugation is conducted while the S-layer protein is in the multimeric assembled form, this provides an added purification obstacle. The conjugated proteins are likely to be present in multimeric structures and surrounded by unreacted protein monomers. The logical separation strategy would be to disassemble and purify; however, the conditions for disassembly are expected to have deleterious effects on the conjugated target protein or group. For the case of the poly(ethylene glycol) conjugates, our observation is that precipitation or “salting out” occurs at levels of LiCl necessary for disassembly. To explore a means for obtaining high purity S-layer protein bioconjugates, we have employed avidin affinity chromatography, in which the specific binding of biotin ligands to the protein avidin is used to bind the target biotinylated bioconjugate to the column via the formation of an high-affinity ligand-protein complex (KD ≈ 10-15 M). The monomeric avidin support was selected over tetrameric avidin as the conditions for retrieving the final product are milder and higher yields can be obtained. The method we have employed is shown schematically in Scheme 1. In the first step, crude S-layer bioconjugates are loaded onto the support. As shown in panel A, the conjugates bind to the column via biotins on the surface of target protein (or polymer). This is followed by a washing step to remove the unlabeled S-layer protein

Bioconjugates of S-Layer Proteins

Bioconjugate Chem., Vol. 15, No. 4, 2004 689

Table 1. Types and Yields of S-Layer Protein Bioconjugates Constructed % yield (pH 6.5) reactant fluorophores/probes (