Customized Peptidoglycan Surfaces to Investigate Innate Immune Recognition via Surface Plasmon Resonance

Abstract

Glycan-protein interactions facilitate some of the most important biomolecular processes in and between cells. They are involved in different cellular pathways, cell-cell interactions and associated with many diseases, making these interactions of great interest. However, their structural and functional diversity poses great challenges in studying them at the molecular level. Surface plasmon resonance (SPR) technology presents great advantages to study glycan-protein interactions due to its superior sensitivity, ability to monitor real-time interactions, relatively simple data interpretation, and most importantly, direct measurement of binding without a need for fluorescent labeling. Here, another dimensionality of SPR in studying glycan-protein interactions is demonstrated via examples of binding between human innate immune receptors and their bacterial peptidoglycan ligands. In order to best resemble interactions in solution, a novel strategy of tethering the carbohydrate at different positions to the biosensor surface is applied to represent the potential displays of the carbohydrate ligand to the receptor. Subsequent kinetic analysis provides insights into the optimized configuration of peptidoglycan fragments for binding with its receptors. The manuscript contains a "how-to guide" to help with the implementation of these methods in other glycan-protein binding systems.

1. INTRODUCTION

1.1. Importance of glycan-protein interactions

Glycans are perhaps the most diverse and versatile macromolecules in cells. They are involved in an array of biological processes, including maintenance of tissue structure and integrity, modulation of other macromolecules functions they attach to, cell-cell communications, recognition of foreign agents, and mimicry of host glycans in bacteria and viruses (Varki, 2017). In many cases, these functions are carried out by the interactions between glycans and glycan-binding proteins. An example of glycan-protein interaction is the binding between glycoprotein hemagglutinin (HA) on the surface of influenza A virus and sialylated host glycan. Viral HA recognizes specific glycan patterns on the host surface and binds to these features at high affinity, allowing for viral anchoring and initiation of invasion (Suzuki, 2005). Notably, recent studies of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of coronavirus disease 2019 (COVID-19), have highlighted the crucial role of glycan-protein interactions in viral infection. The glycan "shield" on the spike proteins (S) of coronavirus was found to sterically mask the recognition site for neutralizing antibodies and allow the virus to escape from the host immune response (Grant et al., 2020; Watanabe et al., 2020). Identification and analysis of effective human antibody recognition of these glycan antigens may provide useful information for vaccine and drug development against coronavirus (Pinto et al., 2020; Wang, 2020). On the other hand, glycan derivatives comprised of glycosaminoglycans (GAGs) have been shown to inhibit this process. The Lindhart group reported that the tight binding between heparin derivatives and the S proteins helps block their interactions with host receptors and reduces viral uptake (Tandon et al., 2021). Along with increasing awareness of the critical roles of glycan-protein interactions, extensive research efforts have been devoted to investigating the molecular mechanisms of glycan-protein interactions. The results of these studies will help answer more questions about basic cellular events and promote the development of new therapeutics against glycan-associated diseases.

1.2. Difficulties in studying protein-glycan interactions?

Different forms of glycans, either tethered to the cell surface or intracellular glycans, contain multiple functionalities, including various linkage points, sulfation patterns, and charged groups. These extremely diverse structures implicate a wide range of binding patterns. For example, on the cell surface, a single glycan-protein interaction occurs with relatively low affinity (mM-μM K_d). Higher affinity interactions are often achieved via multivalent interactions or the use of metal ions within protein binding pockets (Collins & Paulson, 2004; Jayaraman, 2009; Mann et al., 1998). On the other hand, soluble peptidoglycans, natural fragments composed of carbohydrates crosslinked to peptides released from the bacterial cell wall; have been shown to bind to protein receptors at nanomolar affinity (Grimes et al., 2012). Due to the multi-dimensionality of glycan features, it is challenging to probe glycan-protein interactions, and a variety of biophysical methods are needed to glean molecular level insight.

1.3. Current toolbox to study glycan-protein interactions

The past decade has seen the emergence of new technologies to study glycan-protein interactions. Traditional methods of measuring the changes in intrinsic properties of proteins or glycans upon binding, such as circular dichroism (CD) or fluorescence polarization (FP), are beneficial in studying a robust system where binding events generate substantial changes in the signal (Kakehi et al., 2001). However, these methods become less reliable for dynamic, low-affinity interactions or for proteins with short half-lives and high non-specific binding with fluorophore molecules. A powerful method to obtain structural information is X-ray crystallography. Many groups have successfully employed a quick-soaking strategy to co-crystallize a glycan ligand and its respective binding protein. Briefly, soaking the carbohydrate into the existing solution of protein crystal promotes the formation of ternary complexes (Albesa-Jové et al., 2019). However, due to inherent flexibility in glycan structures, the final crystal product might not meet the quality needed for subsequent diffraction steps. Even when high-quality crystals are formed, and diffraction is successful, it is difficult to build a correct model of a glycan in a complex due to low electron density maps for carbohydrate chains (Frenz et al., 2019). The bottleneck of this technique is the demand for highly concentrated and homogeneous samples and laborious initial screening and thorough optimization for high-quality crystal formation (García-Nafría & Tate, 2020). Complementary to X-ray crystallography is solution nuclear magnetic resonance (NMR) spectroscopy. Solution NMR has the capability to monitor atomic details of binding events in conditions that closely reflect the native environments in cells, allowing for capturing a dynamic view of the interactions (Bewley & Shahzad-ul-Hussan, 2013; Poveda & Jiménez-Barbero, 1998; Rennella et al., 2012). Despite these advances, NMR technology still faces major challenges: the inability to maintain a high-resolution structure if the protein exceeds 35 kDa and the convoluted data interpretation if the complex contains large oligomers (Frueh et al., 2013). To overcome the requirement for high protein concentration or fluorescent labeling, many surface immobilization techniques have been extensively developed and applied to study these complex molecular interactions.

1.4. Surface analysis techniques to study glycan-protein interactions

In the past twenty years, several immobilization methods to analyze glycan-protein interactions have emerged. Generally, these methods involve tethering either the glycan or the protein to the surface and then applying the binding partner. Building upon the successful printed DNA and protein arrays, glycan microarrays display large libraries of glycans on a solid support, enabling high-throughput screening of potential glycan-protein interactions and providing rich information about binding specificity (Richard D. Cummings & Pierce, 2014). One of the latest achievements by glycan microarrays is the comprehensive characterizations of natural antibody specificity for glycan antigens in non-human primates (Nanno et al., 2020). While this is a great tool for quick profiling of interactions between glycans and their binding partners, extracting kinetics information from an existing microarray data set requires additional computational methods (Klamer et al., 2019). Biolayer interferometry (BLI), a more recent technology, also provides a high-throughput platform, either 96- or 384-well format, for measuring biomolecular interactions. Unlike microarrays, BLI employs a “dip and read” method where a ligand is immobilized on a biosensor tip and then dipped into wells containing analyte solutions (Petersen, 2017). For example, Núñez and colleagues tethered an innate immune receptor, NLRP6, to the tip of the BLI sensor. Using purified glycans from the Inohara lab, binding constants were determined (Hara et al., 2018). However, the major drawbacks of BLI are limited chemistries available for immobilization and lower sensitivity compared to a competitive technique - surface plasmon resonance (SPR) (Yang et al., 2016).

Together with glycan microarrays and BLI, SPR offers a different approach for investigating molecular interactions. The surface plasmon resonance phenomenon, the heart of this technology, occurs when incident light hits a metal surface with a defined refractive index. This event excites free electrons (surface plasmons) and induces electron oscillation in parallel with the metal. This oscillation on the metal surface is very sensitive to any event that causes changes to the refractive index, such as the adsorption of molecules like carbohydrates or proteins. The changes in refractive index can be recorded as resonance units (RUs) and monitored in real-time, allowing researchers to investigate the kinetics and thermodynamics of biomolecular interactions (Schasfoort, 2017a; Wood, 1902)(Figure 1). A common SPR biosensor consists of a thin layer of metal deposited onto a glass slide. The most popular metal of choice is gold due to its inherent ability to produce high SPR shift (high sensitivity) and its chemical stability (inert to most materials used in biological applications and resistant to atmospheric contaminants). The plain gold surface also offers opportunities for functionalization with small organic molecules (Schasfoort, 2017b). In the late 1980s, the Whitesides Group introduced the concept of self-assembled monolayers (SAMs) surface functionalization, where ordered monolayers on a gold surface were formed by simply soaking a gold film in a solution containing alkanethiol chains. They showed that strong gold-sulfur (Au-S) bonds drive the adsorption of these chains onto the gold surface and intermolecular forces among spacer groups (methylene) induce a uniform arrangement (Bain & Whitesides, 1988). This discovery has opened up endless possibilities to customize gold surfaces with various functional groups such as alcohols, thiols, carboxylic acids, and aldehydes at the monolayer-water interface (Figure 2A) (Cunningham & Wells, 1993; Devaraj et al., 2005; Hahn et al., 2007; Bo Johnsson et al., 1991). The availability of these groups enables the attachment of biomolecules to the biosensor surface via different capturing methods. Upon binding an analyte to the immobilized ligand, the response signal reflects the binding events in real-time, yielding a sensorgram with rich kinetic information. This article highlights the versatility of the SPR method via a specific application of SPR in probing the interactions of two innate immune receptors and their peptidoglycan ligands.

Figure 1:

Schematic diagram of the SPR system. Left: Sensor chip is made of a gold film coating on a glass slide. When polarized light is shined on the sensor chip, the surface reflects the incident light at a certain angle, resulting in the excitation of surface plasmons and causing a dip in the intensity of the reflected light. Right: Upon the binding of analytes to immobilized ligands, the intensity of surface plasmon resonance changes, resulting in an angular shift. The resonance angle shifts are monitored in real-time and displayed as changes in a response signal (Adapted from (Song et al., 2015)).

Figure 2:

Different types of SPR sensors. A: Plain gold chips are grafted with different functional groups on n-alkylthiols where alcohols form a flat matrix, and other groups are used for ligand coupling (left to right: aldehyde, thiol, carboxylic acid). B: Commercially available chips contain a thick 3D matrix of dextran decorated with different modifications such as carboxyl group (COOH), polycarboxylate (HC), streptavidin (SA), nitrilotriacetic acid (NTA), anti-Glutathione S-transferase (anti-GST), and protein A.

1.5. SPR in probing the mechanism of innate immune recognition

The innate immune system relies on families of structurally conserved receptors to recognize bacterial-derived fragments and mount a proper immune response. Toll-like receptors (TLRs) and nucleotide-binding oligomerization (NOD)-like receptors (NLRs) are responsible for recognizing conserved pathogen-associated molecular patterns (PAMPs) both intracellularly and extracellularly (Janeway & Medzhitov, 2002; Kopp & Medzhitov, 1999). Belonging to the latter family, NOD2 and NLRP1 serve as essential intracellular sensors of the innate immune system (Figure 3A) (Inohara et al., 2005; Kobe & Kajava, 2001). Both proteins contain structurally conserved C-terminal leucine-rich repeat (LRR) domains but differ in their N-terminal effector domains (Kanneganti et al., 2007). Upon binding its ligand to the LRR domain, NOD2 interacts with RICK to initiate a variety of signaling pathways, including the NF-κB pathway and MAPK (Strober & Watanabe, 2011). Meanwhile, NLRP1 forms an inflammasome that leads to activation of the Caspase-1 protease and subsequent production of downstream inflammatory molecules (Chavarría-Smith & Vance, 2015). While both proteins have different modes of action, they have been shown to share a common ligand, bacterial peptidoglycan fragments such as muramyl dipeptide (MDP) (Figure 3B) (Faustin et al., 2007; Girardin et al., 2003). MDP has been shown to be the smallest synthetic fragment of peptidoglycan (PG) representing all bacterial cell walls, from Gram-positive to Gram-negative, and acts as an immunostimulant (Ellouz et al., 1974). The fact that binding of MDP (and other peptidoglycan fragments) to NOD2 and NLRP1 can generate immune responses from separate signaling pathways has raised great interest in mechanistic details behind these interactions (D'Ambrosio et al., 2020; Faustin et al., 2007; Grimes et al., 2012; Lauro et al., 2017; Lazor et al., 2019). Notably, NOD2 and NLRP1 variants are associated with inflammatory diseases such as Crohn’s disease and vitiligo, respectively, and it is postulated that misrecognition of bacterial ligands leads to inappropriate immune responses (J. R. Cummings et al., 2010; Hugot et al., 2001; Ogura et al., 2001). Unraveling the mechanism of peptidoglycan recognition at the molecular level will contribute to a deeper understanding of the innate immune system and provide more information to develop better therapeutics against related diseases.

Figure 3:

NOD2 and NLRP1 share a conserved leucine-rich repeats domain at the C-terminus and bind to a common ligand – muramyl dipeptide (MDP). A: Structure alignment of rabbit NOD2-LRR (PDB 5IRN, residues 765-1040, grey) and human NLRP1-LRR (PDB 4IM6, residues 791-990, red). B: Structure of MDP. MDP consists of N-acetyl muramic acid (blue) (MurNAc) linked to a short peptide chain of L-Ala-D-isoGln (orange).

Several factors must be considered to design an SPR experiment for studying these interactions. First, access to both soluble protein and glycan must be available. While lysate can be used for binding studies, high background signals from non-specific interactions or other protein-glycan interactions might complicate data interpretation. Therefore, purified proteins with substantial purity are preferable for kinetics studies. In this case, LRR domains of NOD2 and NLRP1 are expressed and purified from E. coli at high concentration and high purity. Detailed biochemical studies have shown that the LRR domain is sufficient to bind bacterial ligands (Lauro et al., 2017; Ting et al., 2010). The second consideration is choosing a binding partner to tether to the sensor surface. Since the SPR signal depends on the mass changes on the surface, immobilizing the lower molecular weight molecule and using the larger molecular weight protein as an analyte to flow over will help maximize the changes in refractive index upon binding. This strategy also requires only a small amount of analyte to generate a visible signal. Third, the choice of biosensors should be determined empirically. By far, the most common SPR chip is a dextran sensor surface. This sensor contains a flexible and non-crosslinked mesh of high molecular weight dextran (~500 kDa) on top of SAMs. Because of the natural 3D structure of dextran, this layer acts as a barrier to prevent non-specific adsorption and allows higher quantities of biomolecules to be coupled to the surface (Figure 2B) (B. Johnsson et al., 1995). Despite these advantages, the dextran matrix has drawbacks for certain applications, especially for glycan-binding proteins. In this case, LRR proteins are known to bind to peptidoglycan molecules, and dextran is made of glucose; they are likely to nonspecifically adsorb on this matrix and cause a low signal-to-noise ratio, as observed when the first PG-SAM was created (Grimes et al., 2012). Another option is to use plain gold sensor chips and the SAM chemistry described above to decorate the surface with different chemistries (Figure 2A). A well-established strategy reported by Whitesides groups to functionalize plain gold chips with mixed SAMs of carboxylic acid-terminated and alcohol-terminated thiols was implemented to study the LRR-PG binding (Grimes et al., 2012; Lahiri et al., 1999; Mrksich et al., 1995). The choice of a functional group on the SAMs also leads us to the final consideration: the immobilization method. Direct coupling methods covalently attach the ligand molecule to the sensor surface. This approach creates a more stable surface since drifting of the ligand is omitted and allows for decoration with a very high ligand density. In thinking about coupling the carbohydrate to the chip, it is important to pick a modification on the sugar that does not alter the biological activity. We have found that simple amine substitution for specific groups is useful for coupling the highly functionalized glycan. Conveniently, there are methods to activate the carboxylic acid groups on the chips using NHS chemistry to form a stable amide between the surface and the carbohydrate. To study LRR-MDP interactions, MDP is tethered to surface carboxylic acids using an amine coupling strategy (Figure 4A). It is noteworthy to mention that one major disadvantage of SPR is that ligand immobilized might not represent the actual binding conformation present in solution, and observed binding with analyte might not reflect the actual native interactions. Here we show that this limitation can, however, be overcome by tethering the molecule in various confirmations. Amine groups were synthetically introduced at different positions of the peptidoglycan fragments, allowing for MDP ligand presentation at different orientations. Kinetic data for each orientation provides information about the affinity between MDP and these LRR receptors and reveals which "faces" of MDP are important for binding (Figure 4B). The following section describes each step of an SPR various display experiments in detail, from protein construction and purification to SPR surface assembly to data analysis.

Figure 4:

Functionalization and immobilization of a plain gold chip with bacterial peptidoglycan fragments. A: SAMs with terminated carboxylic acid are activated with EDC/NHS and subsequently coupled with amine-modified MDP. B: A library of amino-functionalized MDP derivatives. 6-amino-MDP (1), 6-amino-GlcNAc (2), L-Ala-D-isoGln peptide (3), 2-amino-MDP (4), amino-terminated at D-isoGln MDP (5), 1-amino-MDP (6).

2. PROTOCOL

2.1. Express and purification of NLRP1-LRR and NOD2-LRR

NLRP1-LRR

Details of NLRP1-LRR (residues 791-990) cloning can be found in (D'Ambrosio et al., 2020). Briefly, the gene encoding for NLRP1-LRR was amplified using primers listed in table 1 and was cloned into pGEX-6P-1 between BamHI and XhoI restriction sites. The sequence was confirmed by Sanger sequencing and transformed into BL21 RIPL-codon plus cells (Agilent Technologies). The expression vector pGEX-6P-1 contains an N-terminal GST tag followed by a recognition site for PreScission protease, allowing the cleavage of the GST tag at the end of purification to achieve a tag-free protein.

TABLE 1:

Primers used for the cloning of NLRP1-LRR and NOD2-LRR

Primer ID	Primer sequence
NLRP1-LRR forward	5’TTTTTTGGATCCCCAGTCACAGATGCCTATTGG3’
NLRP1-LRR reverse	5’TTTTTTCTCGAGACTTGGTTTCCGTCTGC3’
NOD2-LRR forward	5’GATCGGATCCATCTGAAGAACGTCTGGCAC3’
NOD2-LRR reverse	5’TTTTTTCTCGAGTTACAGCAGCAGGCCGGT3’

• Reagents:

  • Luria-Bertani (LB) broth (Fisher Scientific), 25 g/L, sterilized by autoclaving
  • Isopropyl ß-D-1-thiogalactopyranoside (IPTG) (Goldbio) dissolved in filtered water to make a 1 M stock
  • Carbenicillin (Goldbio) dissolved in sterile water at 100 mg/mL stocks
  • Bradford assay kit (Bio-rad)
  • cOmplete, EDTA-free protease inhibitor cocktail (Roche)
  • PreScission protease (Cytivia)
  • Glutathione Sepharose 4 Fast Flow (Cytiva)
  • 4-20% Mini-Protean TGX precast protein gels (Bio-rad)
• Equipment:

  • 4-liter glass flasks
  • Large-scale shaking incubator
  • Spectrophotometer (Eppendorf)
  • Econo-Column Chromatography columns 2.5x10 cm (Bio-rad)
  • Mini-Protean tetra cell for protein gel electrophoresis (Bio-rad)
• Buffer preparation: (all buffers are kept on ice)

  • Lysis buffer: 50 mM Tris-HCl, 150 mM NaCl, add fresh 1 mM DTT, 1 mM EDTA (final concentration), and 2 protease inhibitor tablets, pH 7.4
  • Wash buffer: 50 mM Tris-HCl, 150 mM NaCl, add fresh 1 mM EDTA (final concentration), pH 7.4
  • High salt buffer: 50 mM Tris-HCl, 500 mM NaCl, add fresh 1 mM EDTA (final concentration), pH 7.4
  • Elution buffer: 50 mM phosphate, 150 mM NaCl, add fresh 1 mM EDTA, pH 7.4
• Expression and purification:

  • Prepare overnight culture a day before expression (in a sterile environment): Use a sterile pipette tip or loop to scrape some frozen bacteria from glycerol stock and add to 10 mL of sterile LB medium containing 100 μL/mL of Carbenicillin. Allow shaking overnight at 37°C at 200 rpm.
  • Next day: Inoculate 1 L of sterile, fresh LB medium containing 100 μL/mL of Carbenicillin with 10 mL of the overnight culture.
  • Allow shaking at 200 rpm at 37°C. Check OD₆₀₀ every hour using a spectrophotometer.
  • When OD₆₀₀ reaches 0.9, induce protein production by adding 1 mM IPTG (final concentration) to the culture and shake at 200 rpm at 16°C for 18-20 hours.
  • Harvest cells by centrifugation at 10800xg at 4°C for 30 minutes. Combine pellets from 2 L culture and keep at −80°C for long-term storage.
  • The day of purification: Thaw one pellet at 4°C and resuspend in 40 mL of lysis buffer. Vortex to achieve a homogeneous solution.
  • Lyse cells on ice using sonication set at 50% amplitude, pulsing 2 seconds on and 2 seconds off for 1 minute then let lysate rest for 1 minute. Repeat 4 times so the total sonication time is 4 minutes, not including rest time.
  • Clarify lysate by centrifugation at 25000xgfor 25 minutes at 4°C.
  • Pack 5 mL of glutathione resin evenly in an Econo column. Remove ethanol storage solution by gravity. Equilibrate the resin with 3 column volumes (CV) of wash buffer.
  • Collect clarified supernatant and run through nylon 0.22 μm filters. Transfer filtered supernatant to a pre-packed resin and incubate on a shaker at 4°C for 1.5 hours.
  • Collect flow-through and wash column with 3 CV of wash buffer, 3 CV of high salt buffer, and 3 CV of elution buffer.
  • Add 15 mL of elution buffer containing 100 units of PreScission protease to the column and let shake overnight at 4°C to maximize completion of GST tag cleavage.
  • Next day: Collect flow-through (E1) containing tag-free NLRP1-LRR
  • Analyze purified protein on a 4-20% SDS-PAGE gel. Measure the concentration of purified protein using Bradford assay (follow the manufacturer's instructions). Protein in solution is stable at 4°C up to 3 days.

NOD2-LRR

Details of NOD2-LRR (residues 765-1040) cloning can be found in (Lauro et al., 2017). Briefly, the gene encoding for NOD2-LRR was amplified using primers listed in table 1 and was cloned into pMAL-c5x vector between BamHI and XhoI restriction sites. The sequence was confirmed by Sanger sequencing and transformed into BL21 RIPL-codon plus cells (Agilent Technologies). The vector pMAL-c5x was modified to contain a 6xHis tag followed by the maltose-binding protein (MBP) tag. According to previous studies, MBP-tagged NOD2-LRR can be purified from E. coli and still retain full functions.

• Reagents:

  • Luria-Bertani (LB) broth (Fisher Scientific), 25 g/L, sterilized by autoclaving
  • Isopropyl ß-D-1-thiogalactopyranoside (IPTG) (Goldbio) dissolved in filtered water to make a 1 M stock
  • Carbenicillin (Goldbio) dissolved in sterile water at 100 mg/mL stocks
  • Chloramphenicol dissolved in 100% ethanol at 34 mg/mL stocks
  • Bradford assay kit (Bio-rad)
  • cOmplete, EDTA-free protease inhibitor cocktail (Roche)
  • HisPur Ni-NTA resin (Thermo Fisher)
  • 7% protein gels
• Equipment:

  • 4-liter glass flasks
  • Large-scale shaking incubator
  • Spectrophotometer (Eppendorf)
  • Econo-Column Chromatography columns 2.5x10 cm (Bio-rad)
  • Mini-Protean tetra cell for protein gel electrophoresis (Bio-rad)
  • Amicon Ultra-15 spin column with 30 kDa cutoff limit (Millipore)
• Buffer preparation: (all buffers are kept on ice)

  • Lysis buffer: 50 mM phosphate, 150 mM NaCl, 15 mM imidazole, and 2 protease inhibitor tablets, pH 7.4
  • Wash buffer: 50 mM phosphate, 500 mM NaCl, 25 mM imidazole, pH 7.4
  • Elution buffer: 50 mM phosphate, 150 mM NaCl, 250 mM imidazole, pH 7.4
  • Dialysis buffer: 50 mM phosphate, 150 mM NaCl, pH 7.4
• Expression and purification:

  • Prepare overnight culture a day before expression (in a sterile environment): Use a sterile pipette tip or loop to scrape some frozen bacteria from glycerol stock and add to 10 mL of sterile LB medium containing 100 μL/mL of Carbenicillin and 34 μL/mL of chloramphenicol. Allow shaking overnight at 37°C at 200 rpm.
  • Next day: Inoculate 1 L of sterile, fresh LB medium containing 100 μL/mL of Carbenicillin and 34 μL/mL of chloramphenicol with 10 mL of the overnight culture.
  • Allow shaking at 200 rpm at 37°C. Check OD₆₀₀ every hour using a spectrophotometer.
  • When OD₆₀₀ reaches 0.6-0.8, induce protein production by adding 1 mM IPTG (final concentration) to the culture and shake at 200 rpm at 16°C for 18-20 hours.
  • Harvest cells by centrifugation at 10000xg at 4°C for 30 minutes. Combine pellets from 2 L culture and keep at −80°C for long-term storage
  • The day of purification: Thaw one pellet at 4°C and resuspend in 40 mL of lysis buffer. Vortex to achieve a homogeneous solution
  • Lyse cells on ice using sonication set at 50% amplitude, pulsing 2 seconds on and 2 seconds off for 1 minute then let lysate rest for 1 minute. Repeat 4 times so the total sonication time is 4 minutes, not including rest time.
  • Clarify lysate by centrifugation at 25000xg for 25 minutes at 4°C.
  • Pack 5 mL of Ni-NTA resin evenly in an Econo column. Remove ethanol storage solution by gravity. Equilibrate the resin with 3 column volumes (CV) of lysis buffer.
  • Collect clarified supernatant and run through nylon 0.22 μm filters. Transfer filtered supernatant to a pre-packed resin and incubate on a shaker at 4°C for 2 hours.
  • Collect flow-through and wash column with 10 CV of wash buffer
  • Add fractions of 5 mL of elution buffer to elute 6xHis-MBP-tagged NOD2-LRR proteins
  • Elution fractions are pooled together and concentrated using a 30 kDa spin column.
  • Dialyze concentrated NOD2-LRR into a dialysis buffer overnight. Change buffer at least 2 times to achieve maximum dialysis efficiency.
  • Analyze purified protein on a 4-20% SDS-PAGE gel. Measure the concentration of purified protein using Bradford assay, following the manufacturer's instructions.

2.2. Functionalize plain gold chips with SAMs

Plain gold chips can be functionalized with SAMs of alkanethiols (structures below) at an appropriate molar ratio. Previous studies showed that increasing the molar ratio of carboxylic acid-terminated alkanethiol ((χ(1)) will increase the density of SAM and might increase non-specific adsorption of proteins (Bain et al., 1988; Collins et al., 2004). Therefore, the desired χ(1) should be determined empirically for a specific binding experiment. Here, a mixed SAM solution with a low mole fraction of 1 (χ(1) = 0.01) is used.

• Reagents and equipment:

  • 100% ethanol, pure 200 proof (Fisher)
  • 23-(9-Mercaptononyl)-3,6,9,12,15,18,21-heptaoxatricosanoic acid monothiol PEG terminated carboxylic acid, compound 1 below, dissolved in ethanol at 50 mM, stored at −80°C (CAS 221222-49-7, Toronto Research Chemicals)
  • Triethylene glycol mono-11-mercaptoundecyl ether (monothiol PEG terminated alcohol, compound 2 below, dissolved in ethanol at 50 mM, stored at −80°C (CAS 130727-41-2, Sigma Aldrich)
  • Handheld torch
  • Puretane, triple rinsed butane (butanesource.com)
  • Glass vial with a tight cap
  • Gold Chip (supplier can vary based on the instrument: here two are presented Biacore (Cytiva) and BI-4500 (Biosensing)
• Pre-cleaning: Plain gold chips should be cleaned properly before functionalization. Cleaning procedures can be different depending on the manufacturer. The following instructions only apply to gold chips from Cytiva and Biosensing Instrument companies. The steps below are performed under a clean hood to avoid the deposition of contaminants on the gold surface.

  • Cytiva gold chips: Use a tweezer to carefully lift a gold chip from a sticky pad in the storage container at 4°C. Invert the chip so that the gold side will face upward. A scratch test can be performed at a corner of the chip if the gold side needs to be determined.
  • Biosensing Instrument gold chips: Cover a flat glass surface (heat-resistant) with a sheet of aluminum foil. Remove the paper wrap and place the chip with the gold side facing upward on the aluminum surface. Light a torch filled with triple refined N-butane and hold it at a 45° angle so that the tip of the butane flame is about 5 cm away from the gold surface. Move the flame quickly in a zig-zag pattern for 5 secs, rest for 5 secs and repeat 3 times. By this time, evaporation of water deposition on the gold surface can be observed, indicating that contaminations are successfully removed. Quickly move the chip into the hood and let it rest for 1 minute (Figure 6).

Figure 6:

Pre-cleaning process to de-contaminate a plain gold chip from Biosensing Instrument. A: Flame-annealing of the gold chip. This process can be done on a lab bench. B: The flame-annealed chip is rested inside a clean hood. During this time, the SAMs solution in ethanol can be prepared in an amber bottle. C: The cleaned chip is gently dropped into the SAMs solution with the gold side facing up.

NOTE: Since extended flame-annealing time might cause delamination of the gold layer, it is critical to perform this procedure in short intervals (5 secs on 5 secs off). If in-house gold chips are used, it is critical to characterize the gold chip surface to ensure a flat and uniform layer of gold on top of the glass layer.

NOTE: A water drop test can be performed both pre- and post-cleaning to assess the cleanliness of the gold surface. This test can be done by placing a drop of deionized water on the surface. If the water spreads out and does not form into droplets, the surface is considered free of contamination. If a droplet formation is observed, repeat the cleaning procedure suggested above (Figure 7).

Figure 7:

Water drop test on a gold chip from Biosensing Instrument. Left: Water forms a droplet on the gold film before the cleaning process, indicating the deposition of particles contaminations on the surface. Right: After butane flame-annealing, water drop spreads out when in contact with the surface, confirming that contaminants have been removed and the chip is ready for functionalization.

• Functionalization: After cleaning, the gold chip should be ready for functionalization. In a glass vial, prepare a solution containing 2 mM thiol in total in which χ(1)=0.01 and χ(2)=0.99 by adding the appropriate volumes from 50 mM stocks. Then add ethanol up to 2 mL total volume. Use a clean tweezer, carefully transfer the pre-cleaned gold chip with the gold side facing upward to the glass vial. Let the gold chip soak in the SAM solution for 18-72 hours.

NOTE: It was observed that the Cytiva chips only require overnight soaking time while chips from Biosensing Instrument need 72 hours incubation in SAM solution for full functionalization.

2.3. Immobilization of 6-amino-MDP onto a functionalized gold chip

• Prepare the instrument: Well-maintained instrument will give high-quality data. Frequent cleaning and priming procedures must be performed pre- and post-experiment. Before any experiment, run a quick cleanup sequence by injecting degassed deionized water (diH₂O), 20% ethanol, 0.05 M NaOH, then diH₂O at a high flow rate (100 μL/min) to remove any "sticky" contaminants from the previous experiment.

• Materials for immobilization:

  • PBS running buffer: 25 mM phosphate, 150 mM NaCl, pH 7.4
  • Water for initial cleaning
  • All buffers must be filtered and vacuum degassed thoroughly
  • Prepare 2 mL of 0.4 M N-(3-Dimethylaminopropyl)-N'-ethylcarbodiimide hydrochloride (EDC) and 0.2 M N-hydroxysuccinimide (NHS) (CAS: 6066-82-6) solution: make fresh and use immediately.

    • Dissolve 76.68 mg of EDC in 1 mL of water in a 1.8 mL vial
    • Dissolve 11.5 mg of NHS in 1 mL of water in a 1.8 mL vial
    • Mix well by vortex and sonication
      NOTE: To avoid introducing air bubbles during sonication, use degassed water and make the solution in a vial with a maximum volume just above 2 mL so that there is minimal room for air.
  • 1 M ethanolamine in water at pH 8.5 for capping/blocking unreacted carboxylic acid groups
  • 10 mg/mL sodium dodecyl sulfate (SDS) solution in water (avoid stirring vigorously to prevent the formation of bubbles)
  • MDP compounds: The compound solution is made by diluting 20 mg/mL stock solution (in water) with PBS buffer to 2 mg/mL. Additionally, saturated NaHCO₃ is added to make the solution basic (coupling reaction is more efficient at high pH). Litmus paper should be used to test for the final pH of the compound solution. Detailed syntheses of MDP with an amino-functional group at different positions can be found in previous literature (D'Ambrosio, Bersch, Lauro, & Grimes, 2020; Grimes et al., 2012; Lauro et al., 2017; Lazor et al., 2019; Melnyk, Mohanan, Schaefer, Hou, & Grimes, 2015; Schaefer et al., 2017).
• Protocol:***********

  • The functionalized chip is removed from the SAM solution, washed with 100% ethanol, and dried under nitrogen.
  • For each experiment, at least one channel/lane should be used as a negative control where no compound will be immobilized. Instead, all available carboxylic acid groups will be capped/blocked using ethanolamine.
  • Table 2 below describes the detailed steps for immobilization of MDP compound on functionalized gold chips in two SPR systems: Biacore 3000 (Cytiva) and BI-4500 (Biosensing Instrument). While the main steps are the same, each system's commands, flow rate, and volumes of solution are different.
  • To this step, either immediately continue the experiment by flowing protein over the chip or keep the chip hydrated by running PBS over until ready to analyze protein binding (Figure 8).

TABLE 2:

Steps for immobilization of MDP compounds

Step	Biacore 3000	BI-4500
Mount the chip	“Dock”: for docking the chip “Prime”: for equilibrating the system with running buffer PBS at 5 μL/min	“Startup”: for cleaning the prism stage, mounting the chip, and purging the system with running buffer at 60 μL/min
Make a 1:1 EDC/NHS solution	“Dilute”: to mix NHS solution with EDC solution	“Add” and “Stir”: to add NHS solution to EDC-containing vial and mix well
Inject EDC/NHS to activate carboxylic acid groups to amine-reactive NHS-esters	“Inject”: pass EDC/NHS solution through the gold chip for 7 minutes	“Inject”: flow 200 μL of EDC/NHS solution over the chip
Wash with PBS to remove residual EDC/NHS	“Inject”: add PBS for 2 mins	“Inject”: to inject 200 μL of PBS
Coupling reaction between MDP containing an amino group and NHS-esters results in the formation of an amide bond.	“Inject”: flow MDP solution over the chip for 7 minutes	“Inject”: add 200 μL of MDP solution
Wash	“Inject”: wash with PBS for 2 minutes	“Inject”: to inject 200 μL of PBS
Deactivate unreacted carboxylic acid group with ethanolamine (EA)	“Inject”: flow 1M EA solution over the chip for 7 minutes	“Inject”: inject 200 μL of 1M EA
Wash with SDS	“Inject”: flow SDS buffer over the chip for 3 minutes	“Inject”: inject 200 μL of SDS buffer
Wash with PBS	“Unclog”: flow PBS buffer over the chip to wash away all residues at a high flow rate	“Inject”: inject 400 μL of PBS

Figure 8:

Representative sensorgram for activation, coupling, and ethanolamine (EA) capping.

2.4. Binding experiment

A successful experiment is a result of careful preparation of samples and a well-written method. Protein samples should be examined for correct folding, and any aggregation should be removed either by filtration or centrifugation ahead of time. Since dilutions will be used later to calculate kinetics parameters (i.e., K_d), a reliable method to measure protein concentration such as Bradford assay or UV 280 nm should be used, and careful calculation of protein dilutions must be carried out. Having all of the materials in hand, one can create a specific method for each experiment using the control software provided by the manufacturer. Immobilization methods as above can also be set up using the software. When writing an SPR method, one should consider many factors, including flow rate, injection order, control, regeneration buffer, association and dissociation time, and volume or time of injection. Below is the specific method to measure the binding between NOD2-LRR and NLRP1-LRR proteins with immobilized MDP.

• Materials:

  • PBS running buffer: 25 mM phosphate, 150 mM NaCl, pH 7.4, filtered and degassed
  • Protein samples: at least 8 dilutions of purified LRR proteins ranging from 20 nM to 3000 nM are prepared in PBS buffer.
    NOTE: it is recommended to have at least 8 dilutions ranging from 0.1xK_d to 10xK_d to generate sufficient data points for accurate K_d calculation. If K_d value is unknown, a scouting experiment with a wider range of protein concentrations should be done to estimate this value.
  • 10 mg/mL sodium dodecyl sulfate (SDS) solution in water (avoid stirring vigorously to prevent the formation of bubbles)
  • Flow rate for the binding experiment is set at 3 μL/min for Biacore 3000 and 60 μL/min for BI-4500
• Protocol:

  • As mentioned above in the immobilization step, it is important to have at least one channel serving as a negative control. This blank lane accounts for any non-specific adsorption of proteins on the matrix.
  • Protein samples should be injected from low to high concentrations to avoid non-specific adsorption of high protein concentration carrying over to the next injection.
  • Before each injection of samples, the chip is washed with 200 μL of running buffer to stabilize the baseline.
  • For each injection, 200 μL of sample is used. An association curve with increasing response units should form during this process (Figure 8). After each injection, set a delay time to ~300 seconds to allow for dissociation to occur. This event should be observed from the drop in responses at the end of the injection.
  • After sample injection with a delay, inject an excess amount of regeneration buffer (SDS solution (~300-400 μL)) to denature any bound protein and completely disrupt the binding between LRR and MDP (Figure 9).
    NOTE: To choose an appropriate regeneration buffer, preliminary experiments must be conducted from the beginning. A working regeneration condition should remove all analytes from the surface but leave the ligands undamaged and functional. A mild buffer is usually preferred if a protein is tethered to the surface. However, a more stringent condition might be a better choice if the ligand is a small molecule or less labile to chemical treatments in general. In our case, the sugar moiety of peptidoglycan MDP is not stable in highly acidic or basic conditions, so we want to choose a more neutral buffer. SDS solution was found to be the best choice as SDS can linearize analyte proteins and disrupt their interactions but does not alter the structure of ligand MDP.
  • After the regeneration step, a wash step is set where 200 μL of PBS is flowed over to remove residual protein or SDS. This step ensures that the surface is clean, stabilized, and ready for the next sample injection.
  • At the end of each experiment, save the data. Then perform a long wash with PBS buffer at a high flow rate to flush all bound analytes and contaminants out of the system and help return the surface to the initial condition. If the chip will be used several days after in another set of experiments, keep passing the buffer over the chip at a low flow rate. If no experiment is planned within a week, stop the flow system, perform a proper cleanup procedure, remove the chip, and turn off the instrument.

Figure 9:

Representative sensorgrams of binding events between NOD2-LRR and 6-amino-MDP. Data from Biosensing software.

2.5. Data processing and analysis

After collecting data, open the data file in an appropriate processing program (BIAevaluation for Biacore 3000 and BI-KA for BI-4500).

• Step1: Subtract the responses in the control channel from the reaction channels using the "Reference" or "Subtraction" command in the software so that non-specific adsorption of protein to the matrix can be removed. If this action results in negative responses, the analyte protein probably binds more tightly to the matrix than the ligand, and one should re-consider the choice of matrix or alternative methods for measuring interactions.

• Step 2: Select regions for analysis for each protein concentration. Generally, the start time for each region should be a little bit before each injection, where the responses are at baseline, and the end time should be extended over the dissociation time a little bit. To make subsequent analysis easier, make sure these regions have the same length (length = end time − start time).

• Step 3: Export these regions to the analysis software (Scrubber 2) to perform kinetics or steady-state analysis. Following the instructions integrated in the software, raw binding curves must be cleaned up before fitting and calculation. First, enter the concentrations in the order of injections in the “Data” section so that each curve is assigned to one concentration. Next, click on the “Zero” tab and choose the flat portion before the injection to be zeroed. This step sets the baseline of pre-injection to be zero. Next, click on the “Crop” tab and select the portion needed for kinetics run, usually including pre-injection, association, and dissociation parts. Next, align curves using the “Align” tool so that injection time for every curve is equal. If a blank run of buffer only is performed during the experiment, click on the “Blank” tab and choose to subtract the blank curve from the kinetic curves. Then enter the information of analyte and ligand in the “Compound” tab. If analytes with different molecular weights are used, enter the molecular weight of each analyte in the form. Responses can be normalized by dividing the actual responses by MW of an analyte and multiplying by 100. To this step, data are ready for fitting and (apparent) K_d calculation (Figure 10).

• Step 4: Analysis:

  • Steady-state analysis: This analysis can only be applied to binding curves that level out at the end of injections, indicating that the reaction has reach equilibrium. At this stage, the concentration of the complex is directly proportional to the response (R_eq) and the concentration of analyte (C). Click on “Bound” tab and choose the plateau regions of responses and click “calculate”. R_max values of each curve based on the chosen regions will be generated and plotted on a RU vs analyte concentration graph. It is recommended to zoom in and choose the flattest region to obtain accurate R_max values. Move to the "Affinity" tab, click "Fit" to generate a non-linear regression curve to fit the available data points, and K_d value is calculated as ½ analyte concentration required to reach R_max.
  • Kinetics analysis: This approach should be used when not all binding curves reach a plateau. There are three models to choose: k_d, k_a k_d, and k_m k_a k_d depending on the obtained data.

    • Analyze dissociation curve to obtain dissociation rate constant k_d: this approach should only be attempted if the dissociation curve at least 5% decreases from the plateau region. The ideal situation for using this model is that responses drop after the end of injection and reach the baseline, suggesting that the dissociation rate follows single exponential decay. Dissociation rate k_d can easily be extracted from the equation:

      $$R_t=R_0{e}^{-k_d(t-t_0)}$$
      In this tab, enter the end injection time for all curves and choose fitting.
    • Usually, if the association is strong enough to generate a curve reaching a plateau and the dissociation shows clear decaying, one should apply a 1:1 model. However, in many cases, especially when the interaction between analytes and ligands has slow dissociation, the rate no longer follows a simple exponential decay, and accurate k_d cannot be extracted using the previous model. Instead, a more general 1:1 binding model should be applied. In this tab, one should manually enter individual start and end injection times or assume a global injection time for all curves and choose fitting. Analysis software will display results including k_a, k_d, and K_d.
    • If the 1:1 model does not fit the actual binding curves, interactions between analyte and ligand may be limited by mass transfer. This phenomenon occurs when the rate of association is much faster than the rate of analyte diffusion, leading to insufficient analyte needed for saturating ligands at the earlier stage of binding. Hence, the binding kinetics might be a result of both mass transfer and analyte-ligand binding. To account for the effect of mass transfer, k_m or the mass transport rate is added to the equation. One should only choose this model if no other model fits the binding curves and should later try to limit the effect of mass transfer by optimizing experimental design.

Figure 10:

Steady-state analysis of NOD2-LRR and 6-amino-MDP interactions. Binding curves are generated using the one-site binding model.

SUMMARY

The utilization of SPR as an optical biosensor first introduced by Liedeberg and Nylander in 1983 has become extremely useful in both academia and industry (Liedberg, Nylander, & Lunström, 1983). Here we provided another dimension of the SPR -- varied surface display (D'Ambrosio, Bersch, Lauro, & Grimes, 2020). In varying the linkage point of the small molecule to the chip, one can glean information about binding preferences for different receptors. For example, here the method was used to demonstrate that two seemingly very similar innate immune receptors, NOD2 and NLRP1, which both utilize an LRR domain to recognize bacterial peptidoglycan, actually "view" different profiles of the same bacterial ligand. In lieu of a crystal structure, SPR is capable of giving molecular-level insight into binding preferences for difficult-to-study glycan-protein binders.

TROUBLESHOOTING

Problem	Solution
Spikes and disturbance in sensorgrams	Degas buffer more thoroughly. Ideally, fresh buffers should be made before each experiment. Otherwise, buffers can be prepared ahead of time, 0.22 μM filtered, and stored in a clean glass bottle at room temperature. If surfactants like Tween-20 are added, carefully and slowly dissolve the reagent in a degassed buffer to avoid introducing bubbles.
Sudden jumps or drops in responses between steps	Usually caused by buffer mismatch where the sample buffer and running buffer contain different components. It is recommended to dialyze protein samples against the running buffer and change the dialysis buffer at least 2 times. If dialysis is not an option, dilute concentrated samples in a large volume of running buffer so that the mismatching components of original sample buffer is at minimum.
Negative signals	Several things could have caused a negative binding curve. If the negative signals only appear at the beginning of an injection, it is likely that buffer mismatch is the problem. If binding curves show negative signals after reference subtraction, the analytes (proteins in this case) might have bound to the matrix more tightly than to the ligand (small molecule MDP). From our experiences, the level of non-specific binding of analytes to the matrix is lower compared to the binding with ligands. Consider remaking the biosensor if the reverse is true or changing the buffer composition (i.e. adding BSA or Tween-20) to reduce non-specific interactions.
No binding observed or minimal binding signal is observed when analytes are flowed over	Analytes, which are proteins in this case, might have lost activities and ability to bind to ligands. To avoid this problem, always use freshly purified proteins for binding experiments (within 3 days after purification). Also adjust and check pH of buffers to ensure an optimal condition for purified proteins. Another explanation for this observation is that there is no ligand on the surface due to a failed coupling reaction. Check activation and coupling solutions to ensure an optimal basic condition for immobilization reactions.
Low LRR protein expression	Check expression protocol and purification scheme. Many factors can affect the efficiency of protein expression in E. coli. Make sure IPTG stock is not degraded, OD600 reaches the desired value (~0.9 for NLRP1-LRR and ~0.6 for NOD2-LRR), and the incubator is at the correct temperature during expression. If the same problem persists after optimizing protein expression, check for errors during purification. Stored buffers should be checked for bacterial contamination (discoloration or cloudiness) before experiments. Next, all buffers should be adjusted to correct pH at the operating temperature of protein purification. Add temperature/air sensitive reagents (i.e. DTT) to buffers right before purification and use these buffers immediately. Sometimes, GST-tagged NLRP1-LRR protein is stuck in the flowthrough fraction and little protein is observed in the elution fraction. This occurs when the total amount of expressed protein exceeds the binding capability of the column. The problem can be solved by adding the flowthrough fraction back to a new glutathione column and performing the same purification protocol.

Figure 5:

Alkanethiol compounds for functionalization