Engineering Antiviral Agents via Surface Plasmon Resonance

Summary

The present protocol describes new tools for SPR binding assays to examine CV-N binding to HA, S glycoprotein, related hybrid-type glycans, and high-mannose oligosaccharides. SPR is used to determine the K_D for binding either dimeric or monomeric CV-N to these glycans.

Abstract

Surface plasmon resonance (SPR) is used to measure hemagglutinin (HA) binding to domain-swapped Cyanovirin-N (CV-N) dimer and to monitor interactions between mannosylated peptides and CV-N's high-affinity binding site. Virus envelope spikes gp120, HA, and Ebola glycoprotein (GP) 1,2 have been reported to bind both high- and low-affinity binding sites on dimeric CVN2. Dimannosylated HA peptide is also bound at the two low-affinity binding sites to an engineered molecule of CVN2, which is bearing a high-affinity site for the respective ligand and mutated to replace a stabilizing disulfide bond in the carbohydrate-binding pocket, thus confirming multivalent binding. HA binding is shown to one high-affinity binding site of pseudo-antibody CVN2 at a dissociation constant (K_D) of 275 nM that further neutralizes human immunodeficiency virus type 1 (HIV-1) through oligomerization. Correlating the number of disulfide bridges in domain-swapped CVN2, which are decreased from 4 to 2 by substituting cystines into polar residue pairs of glutamic acid and arginine, results in reduced binding affinity to HA. Among the strongest interactions, Ebola GP1,2 is bound by CVN2 with two high-affinity binding sites in the lower nanomolar range using the envelope glycan without a transmembrane domain. In the present study, binding of the multispecific monomeric CV-N to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike (S) glycoprotein is measured at K_D = 18.6 µM as compared with nanomolar K_D to those other virus spikes, and via its receptor-binding domain in the mid-µ-molar range.

Introduction

Tetherin-associated antiviral activity is induced by interferon-α, and it comprises protein-based tethers, that leads to the retention of fully formed virions on infected cell surfaces¹. The necessity for tetherin glycosylation in the inhibition of virus release remains uncertain, implying the importance of glycosylation patterns on recombinantly expressed glycans for in vitro studies^1,2, which depends on the conformation of (in the case of influenza virus) surface-expressed influenza hemagglutinin HA^3,4. It has been noted that modification of oligosaccharide tethered to N-linked glycosylation is enough for tetherin-mediated restriction of HIV type-1 release², while dimerization plays an essential role in preventing virus release, thereby involving the transmembrane domain or glycosyl-phosphatidyl-inositol (GPI)-anchor for tethering the budding virions⁵. Unique features are described for human and murine tetherin to block multiple enveloped viruses, retroviruses, and filoviruses. BST-2/tetherin is an interferon-inducible antiviral protein of the innate immunity^1,6, acting with broad-spectrum antiviral activity and is antagonized by envelope glycoproteins⁵ to either translocate tetherin or disrupt the structure of tetherin⁶. For example, surface-expressed envelope glycoprotein HA and neuraminidase on influenza A virus are well known for tetherin antagonism in a strain-specific manner⁷, facilitating the recognition of host receptor binding sites⁸. Glycan-targeting antibodies are studied in the stoichiometry of their interactions with the rapidly customizing glycan shields on HA, resulting in binding affinity to influenza A H3N2 and H1N1 subtypes⁴.

To elucidate the binding mechanisms between antiviral agents and virus envelope spikes, i.e., carbohydrate ligands, and complementary immunological and spectroscopic methods, mono-, di- and tri-mannose moieties are chemically synthesized. The mannosylated peptides are created via azido glycosylation of glycosyl {beta}-peracetates to 1,2-trans glycosyl azides transformation⁹, mimicking the typically found N-acetyl glucosamine and high-mannose oligosaccharides on the surface of life-threatening viruses. Triazole bioisosteres are utilized to mimic linkages that form the mannosylated residue of HA peptide¹⁰ and facilitate site-specific interactions with antiviral CV-N derivatives around the second N-linked glycosylation spot on the HA head domain (HA top with 4 N-linked glycans N54, N97, N181, N301)^8,11,12. Interactions between glutamic acid (Glu) and arginine (Arg) and the resulting helix dipole manifestated good stability of both model peptides and proteins but are visualized using SPR. If compared with recognizing a single chemically synthesized glycosylation site on HA¹⁰ by directly inhibiting receptor binding on the glycan moieties, a higher affinity of a four-site mutated Fc structure to its receptor is shown to elicit effector functions in vivo, revealing the unrelated composition of N-linked glycans attached to Fc mutant to be mechanistically determined¹³.

CV-N displays antiviral activity against HIV^14,15, influenza¹⁶, and Ebola virus, which is mediated by nanomolar binding to high-mannose oligosaccharide modifications on envelope spike proteins^12,17,18,19. Influenza HA binding to one high-affinity carbohydrate-binding site (H) in CV-N or two Hs in covalently linked dimeric CVN2 is determined to have equilibrium dissociation constants (K_D) = 5.7 nM (Figure 1A) and K_D = 2.7 nM, respectively. Both CV-N and CVN2 harbor another one or two low-affinity carbohydrate-binding sites (L)s^12,17,20,21. Ebola GP1,2 binds to 2H of CVN2 with affinities in the lower nanomolar range (K_D = 26 nM). CV-N WT binding to Ebola GP1,2 and HA exhibits affinities from K_D = 34 nM to K_D = 5.7 nM (A/New York/55/04)¹². Lectins, such as CV-N, which specifically target high-mannose glycans on the viral envelopes, further inhibit replication of hepatitis C virus, SARS-CoV, herpesvirus, Marburg virus, and measles virus²².

The small CV-N molecule has been studied thoroughly for more than 20 years as it functionalizes to bind a wide range of viruses to inhibit viral entry^16,18. Structural analyses and binding affinity assays indicate cross-linking of two Ls in a domain-swapped CVN2 dimer by bivalent binding in the micromolar range to enhance avidity to viral envelope glycoproteins^10,19. Selective binding of Manα1-2Manα on Man(8) D1D3 arms and Man(9) comprises two binding sites of differing affinities located on opposite protein protomers²⁰, thereby reaching nanomolar binding affinities (Figure 1B). Thus, CVN2 is considered a pseudo-antibody concerning its application to bind epitopes on HIV gp120, similar to virus-neutralizing antibodies¹⁷. Herein, the author is interested in investigating the potential binding of CVN2 to the SARS-CoV-2 spike via its receptor-binding domain (RBD). Binding curves of immobilized human angiotensin-converting enzyme (ACE)-2 with the SARS-CoV-2 RBD result in K_D = 4.7 nM for this biologically relevant binding interaction²³.

By contrast, selected immunoglobulin classes recognize specific and consistent structural protein patterns, which impart a substrate for affinity maturation in the membrane-anchored HA regions²⁴. CV-N shows highly potent activity in almost all influenza A and B viruses¹⁶, and it is a broadly neutralizing antiviral agent. Our knowledge is incomplete on the location of targeted epitopes on the stem of HA1 and HA2 that possibly involve epitopic structures for glycan-targeting by highly neutralizing antibodies and as compared with lectin binding²⁵.

Figure 1: Schematic representation of the SPR binding assay for CV-N to virus envelope spikes. (A) SPR Assay for CV-N binding to ligand: HA full-length protein (90 kDa). Kinetic data set (5120, 2560, 1280, 640, 320, 160, 80, 40, 20, 10, 5, 2.5, 0 nM) showing real-time double-referenced binding to influenza HA A/New-York/55/04 (H3N2). (B) CVN2L0 variant V2 binding to immobilized ligand DM within a concentration range of 500 nM to 16 µM. Sequence: L residues are highlighted in yellow. H residues are highlighted in gray. E58 and R73 are a replacement for cysteines in the wildtype protein and make V2 a stable protein fold with three instead of four disulfide bonds Please click here to view a larger version of this figure.

Whereas the glycan shield on the membrane-distal HA top part induces high-affinity binding to CV-N¹², CVN2 binding to HA adjacent to a disulfide bridge of the HA top part has further been observed at its low-affinity sites^10,12. Various polar interactions and interaction sites are identified in carbohydrate-binding by CV-N. These interactions are verified by generating knock-out variants in the binding site to correlate binding affinities to in silico predicted glycosylation¹². Thus, the project aims to compare previously tested chemically mannosylated HA peptides in binding affinity and specificity with short peptide sequences from SARS-related 2019-nCoV spikes and SARS-CoV-2, naturally occurring modified by a small number of different N-linked glycosylation sites and O-linked glycosylation. Using cryo-electron microscopy and binding assays, Pinto and coworkers report a monoclonal antibody, S309, that potentially recognizes an epitope on SARS-CoV-2 spike protein containing a conserved glycan within the Sarbecovirus subgenus, without competing with receptor attachment²⁶. The protocol of this study describes how designing, expressing, and characterizing CV-N variants are important to study how CV-N and CVN2 bind to glycosylated proteins and synthetic mannosylated peptides using the SPR technology^10,12.

Tandem-linked dimer CVN2L0²⁷ and binding-site variants (V2-V5) are recombinantly expressed and variants are with disulfide bond replacements (C58E and C73R) (Figure 2A). Also, a mutant with a single-point mutation E41A is prepared because this position has been seen as an intermolecular cross-contacting residue. This mutant is another interesting molecule for SPR binding measurements between the lectin and high-mannose oligosaccharides deciphering binding domains and allows comparison with the dimeric form. The domain-swapped crystal structure of CVN2 shows a flexible linker, that extends between 49 and 54 residues. The two domains can continue moving around the hinge as rigid bodies, developing either a monomer through intramolecular domain interactions (domain A -residues 1-39;90-101- with domain B -residues 40-89) or a dimer by intermolecular domain swapping [domain A (of the first monomer) with domain B (of the second), and domain B (of the first monomer) with domain A (of the second copy)]. There are no close interactions between the two protomers' A and B domains, except for Glu41²⁸. The gene for CV-N can be developed using a repetitive PCR method with 40-mer synthesized oligos²⁹ and is then subcloned into the NdeI and BamHI sites of pET11a for transformation (electroporation) into electrocompetent cells as described by Keeffe, J.R.²⁷. The protein, which is used for achieving the respective crystal structure (PDB ID 3S3Y), includes an N-terminal 6-histidine purification tag followed by a Factor Xa protease cleavage site. Site-directed mutagenesis is utilized to make point mutations, switch codons, and insert or delete single or multiple bases or codons for amino acid exchange. These transformations provide invaluable insight into protein function and structure. Recombinantly expressed and purified CV-N, CVN2, and CVN3 have been biophysically well studied^20,21,27, are cheap to produce, and therefore used to characterize binding assays to glycans immobilized onto SPR sensor chips. Conventional enzyme-linked immunosorbent assay (ELISA) provides less reproducibility concerning the immobilization technique of glycan ligands and transforms real-time binding of various binding-site variants, which is shown for SPR, into endpoint assays.

Binding-affinity variant CVN2L0-V2 (an intact fold of homodimeric CV-N with a disulfide bridge substitution¹⁰) is expressed with a His-tag in Escherichia coli (E. coli), purified over Ni-NTA column applying affinity chromatography and tested for binding to HA (H3N2), monomannosylated HA-peptide and dimannosylated HA-peptide using SPR. Chemically mannosylated peptides, or HA and S protein, all are ligands and amine coupled to the hydrophilic chip surface via reactive esters or biotin-streptavidin protein engineering. The same procedure of sequential runs is applied to those ligands, injecting various dilutions of CV-N and variants of CV-N (and CVN2) to obtain kinetic information for the molecular interaction analyses as described below³⁰. RBD-immobilized SPR sensor chip is used for binding studies on CV-N to S peptides, and affinities are compared to SARS-CoV-2 binding with the human ACE2.

Protocol

For the present study, a CVN-small ubiquitin-like modifier (SUMO) fusion protein has been used in enzyme-linked immunosorbent assays instead of CV-N and is suitable for cell-based assays. Recombinant full-length influenza A virus HA H3 protein is obtained commercially (see Table of Materials) or expressed in mammalian HEK293 cell lines and baculovirus-infected insect cells according to standard protocols¹². Wuhan-1 spike protein is expressed in mammalian HEK293 cells. The synthesis of monomannosylated peptide (MM) and dimannosylated peptide (DM) allows the detection of homogeneous ligands to CVN2 and monomannosylated small molecule¹⁰.

1. Creating CV-N constructs

1. For each of the CVN2 variants and CVN2L0 protein (PDB ID 3S3Y), obtain the gene construct with an N-terminal pelB leader sequence and His-tag in pET27b(+) vector from commercial sources (see Table of Materials, Supplementary File 1).
2. Obtain CVN2L0 and its variants (V2, V3, V4, and V5; Figure 2A,C) in the background of a CVN2L0 template gene consisting of two distinct DNA sequences for each CV-N repeat.
3. Dissolve the lyophilized plasmid DNA in sterile deionized distilled water (ddH₂O) to a final concentration of 100 ng/µL.

2. Preparation of LB-agar plates with plasmid DNA transformed cells

1. Prepare culture medium LB-Lennox by dissolving 10 g/L of peptone, 5 g/L of yeast extract, and 5 g/L of NaCl in ddH₂O (see Table of Materials), and adjust the pH to 7.4. Perform the transformation into competent E. coli BL21 (DE3) for each variant (V2-V5) by chemical method following a previously published report¹⁰.
2. Split the solution (900 µL and 100 µL), transfer 100 µL on LB-agar plates (50 µg/mL kanamycin), and gently use a sterile cell spreader. Incubate the agar plates overnight at 37 °C.

3. Cloning

1. Subclone the gene for CV-N into the NdeI and BamHI sites of pET11a (see Table of Materials) for transformation (electroporation) into electrocompetent cells following reference²⁷.

4. Site-directed mutagenesis

1. To generate CVN2L0²⁹ and mutant CVN-E41A in the background of a CVN2L0 template gene containing two distinguished DNA sequences for each CV-N repeat²⁷.
2. Make mutations using a site-directed mutagenesis kit (see Table of Materials) and specific mutagenic primers 5'-gagaaccgtcaacgtttgcgataacagagttcagg-3' and 5'-cctgaactctgttatcgcaaacgttgacggttctc-3' for running the PCR³¹.
   1. Start a series of sample reactions using multiple concentrations of double-stranded DNA (dsDNA) template ranging from 5-50 ng (e.g., 5, 10, 20, and 50 ng of dsDNA template). Keep the primer concentration constant.
      NOTE: The PCR mix and thermal cycling protocol are generally used as described in the instruction manual for the site-directed mutagenesis kit³².
3. Add the Dpn I restriction enzyme (1 µL, 10 U/µL, see Table of Materials) below the mineral oil overlay. Thoroughly and gently mix reactions, spin down in a table-top microcentrifuge for 1 min, and incubate immediately at 37 °C for 1 h for digesting the parental supercoiled dsDNA.

5. Transformation of bacterial cells

1. Thaw the XL1-Blue supercompetent cells (see Table of Materials) gently on ice. To transform each control and sample reaction, aliquot the supercompetent cells (50 µL) to a prechilled polypropylene round-bottom tube (14 mL).
   1. Transfer 1 µL of the Dpn I-treated single-stranded DNA (ssDNA) from each control and sample reaction (mutated ssDNA) to separate aliquots of the supercompetent cells, which synthesize the complementary strand. Swirl the transformation reactions carefully to mix and incubate the reactions on ice for 30 min.
      NOTE: Before transferring the Dpn I-treated DNA to the transformation reaction, it is recommended to remove any remaining mineral oil carefully from the pipette tip. As an optional control, the transformation efficiency of the XL1-Blue supercompetent cells needs to be checked by mixing 0.1 ng/µL of the pUC18 control plasmid (1 µL) with a 50 µL aliquot of the supercompetent cells.
2. Apply heat pulse to the transformation reactions at 42 °C for 45 s, and then place the reactions on ice for 2 min.
   NOTE: The applied heat pulse has already been optimized for the mentioned conditions in polypropylene round-bottom tubes (14 mL).
3. Add 0.5 mL of NZY+ broth (containing per liter: 10 g of NZ amine (casein hydrolysate), 5 g of yeast extract, 5 g of NaCl, 12.5 mL of 1 M MgCl₂, 12.5 mL of 1 M MgSO₄, 10 mL of 2 M glucose, pH 7.5, and preheated to 42 °C) and incubate the transformation reactions at 37 °C with shaking at 225-250 rpm for 1 h. Plate the correct volume of each transformation reaction (5 µL from control plasmid transformation; 250 µL from sample transformation) on LB-ampicillin agar plates.
   NOTE: For the transformation controls and mutagenesis, spread cells on LB-ampicillin agar plates having 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal, 80 µg/mL) and isopropyl-1-thio-β-D-galactopyranoside (IPTG, 20 mM) (see Table of Materials). Inoculate 50 mL of the cell cultures with a single colony of transformed E. coli cells to purify mutated plasmid DNA for analyses. Mutagenesis is confirmed by DNA sequencing at an external facility.

6. Expression and protein purification

1. For a large-scale culture, inoculate a small amount of LB (containing ampicillin) with a single colony from the transformed plate.
2. Using the overnight culture, inoculate the expression culture with additives, such as 10 mM of MgCl₂, 10 mM of MgSO_4, and 20 mM of glucose, diluting the seed culture to 1/100.
   1. Grow cells with vigorous shaking at 37 °C. Grow cells to an Abs 600 nm between 0.4-0.6 (mid-log phase) before cooling the cells to 20 °C. Induce with 1 mM IPTG and grow overnight.
   2. Then, harvest the cells by centrifuging at 4,000 x g for 15 min at 4 °C, and discard the supernatant with a pipette.
3. Resuspend the cell pellet in phosphate-buffered saline (PBS) buffer and re-centrifuge at 4,000 x g for 15 min at 4 °C. Then, discard the supernatant with a pipette. Resuspend the remaining pellet in 10 mL of lysis buffer and incubate the suspension for 1 h at 37 °C.
   NOTE: Composition of lysis buffer: (50 mM of NaH₂PO₄, 300 mM of NaCl, 2% Triton-X100, 500 ng/mL of lysozyme, 1 mM of phenylmethylsulfonylfluoride (PMSF), 1 mM of dithiothreitol, 1 mM of MgCl₂, pH 8, see Table of Materials).
   1. Subject the mixture to two freeze-thaw cycles (-80 °C). Separate soluble and insoluble fractions by centrifugation at 4,000 x g for 15 min at 4 °C and analyze them using polyacrylamide gel electrophoresis (PAGE), in particular, sodium dodecyl sulfate (SDS)-PAGE³³(Figure 2B,C).
      NOTE: Cells can be lysed in several ways, such as freeze-thaw, sonication, homogenization, enzymatic lysis, or a combination of these methods. Purification from inclusion bodies is recommended for collecting high protein yields²⁶.
4. Purify proteins using Ni-NTA column chromatography (see Table of Materials). Load the soluble fraction onto a regenerated Ni-NTA bead column (1 mL/min). Wash the system with TBS buffer (50 mM of Tris, 150 mM of sodium chloride, pH 7.5) before starting a gradient (0-100% 500 mM imidazole in TBS over 60 min) and collecting fractions (1 mL/min). Dialyze the purified proteins for biochemical characterization against 100 mM PBS (Figure 2C,D).
5. Alternatively, use His-Select Ni²⁺ affinity gel (see Table of Materials) in 14 mL tubes to bind and re-suspend his-tagged recombinantly expressed CV-N in buffer solutions with 20 mM imidazole and 250 mM imidazole, respectively. Incubate in batch for at least 30 min.
   NOTE: Apply these semi-purified proteins to a single-use prepacked column for buffer exchange and cleanup of biological samples, for example, carbohydrates and proteins, which can load 1-1.5 mL eluate from immobilized metal affinity chromatography.
6. Transfer the protein solutions to centrifugation tubes with a 10 kDa cut-off filter (see Table of Materials) and concentrate them by centrifuging for 10 min at 4,500 x g and 4 °C. For SPR measurements, exchange the analyte solutions to 10 mM HEPES, 150 mM sodium chloride, 3 mM ethylenediaminetetraacetic acid (EDTA), and 0.05% Tween20, pH 7.4 (HBS-EP(+), see Table of Materials).
   1. Add this SPR running buffer to a dilution factor of 1:10 and centrifuge four times to the initial volume for 10 min at 4,500 x g and 4 °C.
7. Determine the protein concentration at 280 nm using a NanoDrop UV-Vis spectrophotometer (see Table of Materials) based on the calculated extinction coefficient (20,440 M^-1 cm^-1) for the main protein CVN2L0 showing a size of 23,474 Da³⁴. Use PBS (100 mM, pH 7.0) or SPR buffer as a blank, and measure the protein concentration at three dilution steps (1:1, 1:10, and 1:100).

Figure 2: CV-N sequences and expression. (A) CVN2 without a linker between each CV-N repeat (101 amino acids each) and four disulfide bridges is expressed in pET11a vector in E. coli. (B) Expressions of two independent colonies for CV-N (monomer) and CVN2 (dimer). (C) Disulfide bond variants are purified and analyzed on SDS-PAGE. A low molecular weight marker (6 µL) is used as a reference. WT = CVN2L0 bearing four disulfide bridges as marked in (A). V2 is a variant with a disulfide bond replacement by polar residues at positions 58 and 73. V3-V5 are variants with two remaining S-S bonds and either polar (C58E-C73R) or non-polar (C58W-C73M) substituting amino acids or a combination of these residue pair substitutions. (D) HPLC chromatograms of purified CVN2L0 are elutated at a flow rate of 1 mL/min with a linear gradient from 5%-65% buffer B in buffer A over 30 min. Buffer A is: 0.1% (v/v) trifluoroacetic acid in ddH₂O, buffer B is: 0.08% (v/v) trifluoroacetic acid in acetonitrile. Protein is analyzed on a high-performance silica gel 300-5-C4 (150 x 4.6 mm) column at 214 nm and 280 nm. Please click here to view a larger version of this figure.

7. SPR spectroscopy

1. Use the Dual Channel SPR system (see Table of Materials) with running buffer HBS-EP(+) and 10 mM glycine HCl pH 1.5-1.6 as the regeneration buffer. Turn on the instrument, degasser, auto-sampler, and pump and wash the entire system with ddH₂0 for 1 h. Place ready-to-use running buffer in a separate bottle.
2. Drop emersion oil onto the detector and mount a glass sensor chip (see Table of Materials) coated with a thin gold film and on the upper side functionalized with carboxymethyldextran hydrogel directly onto the detector below the three-port flow cell. Fix the setting by pulling down the handling.
   NOTE: C19RBDHC30M 200 nm streptavidin derivatized carboxymethyldextran hydrogel with a medium density of biotinylated severe acute respiratory syndrome coronavirus-2 RBD protein, is a ready-to-use sensorchip with the pre-immobilized ligand.

8. SPR binding assay for CV-N binding to HA, S protein, and RBD

1. Immobilize the proteinaceous ligands to sensor chips following the steps below.
   1. Open a run table by clicking on Form in the menu bar and Run Table Editor in the integrated SPRAutoLink software (see Table of Materials). Choose and click on BASIC_Immobilization from the list of available run tables and follow the steps of the experimental procedure on the computer screen. The respective Sample Editor used is shown in the right upper corner.
   2. Click on Sample Set Editor in the Form section to fill out the reagents list for two racks placed in the autosampler for further analyses. Click on Autosampler Direct Control as a "Tool" in the menu bar to bring the racks forward or back home. Choose 4 °C as the operating temperature.
      NOTE: The SPR software allows for "SPR Instrument Direct Control" and "Pump Direct Control" via selecting the corresponding tools, as well as autosampler-handling, and by clicking on Form; also Run Table Editor, Data-Plot, or Post-Processing can be chosen to perform data analysis. Files are directly saved in the default directory and exported as scrubber.files from the Post-Processing window.
2. Start the pump to infuse ddH₂0 by clicking on Tools and Pump Direct Control and record data by clicking on SPR Instrument Direct Control, and each time Start in the newly appearing windows. Put the coupling reagents (step 8.3) in 300 µL vials, put them into the autosampler racks and start the run table by clicking on Run.
   NOTE: Chip surfaces are either conditioned with 10 mM glycine buffer pH 9.0 or may have been flushed with 1 M sodium chloride, 0.1 M sodium borate buffer pH 9.0 to condition carboxyl derivatized chip surface for EDC/NHS activation mix³⁰.
3. For this simple protein-protein interaction, use the CMD500D chip (see Table of Materials) to generate a micro-refractive index units (µRIU) = 2500 - 3000 flow cell with immobilized HA and µRIU = 400 flow cell with spike protein. At an infuse flow rate of 15 µL/min, inject an aqueous and equal mixture of 0.4 M N-ethyl-N'-(dimethylaminopropyl) carbodiimide hydrochloride (EDC*HCl) and 0.1 M N-hydroxysuccinimide (NHS) by applying the following sequential steps.
   1. Refill pump refill at 25,000 µL/min, perform baseline adjustment for 30 s, inject 90 µL of sample activation solution (EDC/NHS) over 6 min contact time, and then hold for another 5 min.
   2. Repeat this cycle after baseline running for 1 min at 10 µL/min on only the left flow cell (blue) to inject and immobilize chemically synthesized peptides¹⁰, HA, and spike protein at 20 µg/mL, and allow for the subsequent baseline adjustment with ddH₂O for 1.5 min before quenching the activated chip surface with 1 M ethanolamine HCl pH 8.5.
4. Switch the tubes from the liquid sampler to the degasser from ddH₂0 into the bottle with HBS-EP(+) (Supplementary Figure 1).
5. Analyze the SPR sensorgrams.
   1. To perform kinetic studies, use various analyte concentrations (10^-5-10^-8 M), with a regeneration step after each injection and blank measurements after different analytes. Change the flow rate to 10 µL/min and start injections for 4 min contact time, then 5 min baseline generation, and two regeneration steps of 2 min each with an interval of 30 s.
   2. Inject the buffer solution for blank measurements whose sensorgrams are subtracted from sample runs to normalize different protein concentrations.
6. Click on Form, scroll down, and change to "Post-Processing" by clicking this operation mode. Click on Add to select binding curves generated over time in the Data Plot Form for each flow cell, and export the overlay as a scrubber file (.ovr). Click on File to open the file saving options. Obtain response curves by aligning left and right curves and subtracting signals of the second reference channel from those of the ligand channel.
   NOTE: Data is operated in "Post-Processing" by defining sensorgrams computationally. It is put into an overlay of sensorgrams from left and right flow cells, or represented as sensorgrams from the difference of both channels.
7. Clean the entire fluidics with 50-100 mM glycine buffer pH 9.5, water, and 20% ethanol before and after binding measurements to remove traces of salt or any protein contamination, or more stringent, with 0.5% SDS and glycine.
   NOTE: To prevent instrument damage, it is recommended to check the mechanical stability of the glass chip before re-inserting the chip cartridge into the instrument if chips have been stored under buffer or at 100% humidity.

Representative Results

A dimeric domain-swapped CVN2L0 molecule is tested for binding to the HA top region in three separate SPR experiments and binding affinity is presented in K_D values. Domain B is assumed to comprise H-binding sites, which are impacted by replacing a disulfide bond into ionic residues, and domain A forms L^10,18. Single injections of CVN2L0 and variants V2 (three disulfide bridges) and V5 (two disulfide bridges) are first tested for binding to the HA-coupled sensor chip to estimate binding capacities before setting up the kinetics measurement using the autosampler and running table editor (Figure 3A and Supplementary Figure 2). Repeated injections are automated for a series of CVN2L0, V2, and V5 at various concentrations in the nanomolar and µ-molar range in the system over time (Figure 3B-D). Correlating the number of intact Cys-Cys bonds, CVN2L0 is binding immobilized HA at K_D = 255 nM when reaching good saturation. In this case, both K_D values, calculated either from kinetic data or by fitting the equilibrium data to the Langmuir binding isotherm, are in moderate agreement (Table 1 and Supplementary Figure 3, Supplementary Figure 4).

Figure 3: SPR binding assay for ligand binding via multivalent interactions. CVN2 (WT) and binding site-variants V2 and V5 with disulfide replacements in the high-affinity carbohydrate-binding pocket are tested for binding covalently immobilized HA. (A) Left and right flow cell binding curves of CVN2, V2, and V5 can be extracted from the SPR run protocol, and sensorgrams are summarized in an overlay. (B) Sensorgram from conventional SPR experiment shown as kinetic analyses for binding of CVN2L0 to HA, injecting analyte concentrations from 10^-4 to 10^-8 M. CVN2L0 is measured up to the highest concentration at 1.5 µM (top line) and up to 500 nM (bottom line), for which no saturation on the chip surface nor equilibrium binding is achieved. RU = Response units. A CMD500D sensor chip is used. (C) SPR sensorgram for V2 binding to HA. (D) V5 binding to HA. The figure (B-D) is reproduced with permission from reference¹⁰. Please click here to view a larger version of this figure.

	ka (M-1s-1)	kd (s-1)	KD [HA] Kinetic	KD [HA] Equilibrium
CVN2	5.1 e3	1.3 x10-3	255 nM	246 nM
V2	4.0 e3	1.1 x10-3	275 nM	255 nM
V5	1.2 e3	5.8 x10-2	5 µM	4.9 µM
CVN2L0	2.07 e4	3.0 x10-3	147 nM	600 nM
	2.27 e4	3.0 x10-3	133 nM	490 nM

Table 1: Binding affinities of CVN2 and variants to HA measured via SPR. Scrubber 2.0 (a BioLogic software) is utilized to fit real-time SPR association and dissociation curves and to calculate the equilibrium dissociation constants (K_D) from kinetic and equilibrium data. Ka = association rate constant. Kd = dissociation rate constant.

Whereas K_D values for binding of CVN2L0 and V2 to HA are both in the nanomolar range, V5 is diminished in binding (Table 1). In V5, two disulfide bonds are replaced by ion-pairing residues that retrain potential ionic interactions between mutated Glu and Arg residues (Figure 2A,3D). Primary data is analyzed following the integrated rate of association and rate of dissociation equations, which describe the binding kinetics of the interaction of soluble CV-N with immobilized ligand and the dissociation of the formed complex from the chip surface. Two independent measurements are performed, and sensorgrams, equivalent to those obtained from replicates are analyzed. K_D is calculated as the quotient of k_off/k_on (Supplementary Table 1)^10,35. However, K_D is related to the equilibrium data that fit the Langmuir binding isotherm³⁶, where the equilibrium binding response is plotted as a function of the analyte concentration (Supplementary Figure 3A,4A,3B,4B). In a concentration-dependent manner, dissociation rate constant k_d is determined post analyses and incorporated into association phase fitting (k_on) to estimate the association rate constant k_a. (Table 1, Supplementary Figure 3C, Supplementary Figure 4C).

Next, the binding of CV-N to HA is compared to its interaction with RBD. CV-N WT binding to SARS-CoV-2 RBD is measured at weaker affinity (K_D = 260 µM, Figure 4A) as compared with binding to HA (K_D = 5.7 nM)^8,10,12 and binding to S protein (K_D = 18.6 µM, Figure 5). Concentration versus response is plotted for CV-N WT binding to RBD, assuming specific targeting of possibly conserved carbohydrates on the RBD S1 subunit (Figure 4A). The same affinity plot is shown for CVN2L0-V4 binding to DM that is no longer analyzable due to the replacement of disulfide bonds which interfere with high-affinity binding to small glycosylated peptides (Figure 4B).

CVN2L0-V4 (2 disulfide bonds opposite to unsymmetric replacement of cysteine residues by either Glu-Arg residues or amphipathic amino acids Trp and Met) can form non-polar hydrogen bond networks around the pseudo-domain B carbohydrate-binding site¹⁰. Higher density of glycans on HA protein reaches binding with polar Glu-Arg residues instead of cystines (Supplementary Table 1), and an association with mannosylated peptides (K_D = 10 µM for DM) between CVN2L0 and modifications into Glu-Arg, but shows discontinuous dissociation of variants from this monoglycosylated peptide, in particular when H is modified on both monomers. For the interaction between CVN2L0-V4 with DM, the response of 56 µM CVN2L0-V4 injection yields a higher response than R_max. (Figure 4B). V5 (2x Glu-Arg) fails in dissociation from surface-bound DM (data not shown). In sum, response curves of lower concentrations decline before ending the injection for all sensorgrams on CVN2 binding to peptide without native H and monomer binding to RBD.

Figure 4: SPR analyses of (A) CV-N WT monomer binding to RBD. K_D is calculated by fitting kinetic data (left side, K_D = 260 µM) and compared with affinity data (right side) using a simple biomolecular model for 1:1 binding between ligand and analyte (K_{D equil} = 274 µM). Equilibrium binding response or binding capacity is plotted as a function of concentration in comparison with the SPR binding curves (0-605 µM CVN). H = High-affinity binding site. L = Low-affinity binding site. Both are found on CV-N and CVN2L0. (B) Dimeric CVN2L0-V4 binding to DM, assuming 2L without analyzable K_D. CVN2L0-V4 concentrations are 56 µM, 28 µM, 14 µM, 7 µM, 3.5 µM. Please click here to view a larger version of this figure.

Figure 5: Mutant CVN-E41A and CV-N WT binding to (A) S glycoprotein. Arrows indicate the beginning of the association and dissociation phase. (B) CVN-E41A binding to RBD on SARS-CoV-2. The ligands are pre-immobilized onto HC polycarboxylate and CMD500D sensor chips. Covid-19 S1 subunit [Pinto, D. et al.²⁶; and Barnes, C. et al.³⁷] is used for RBD immobilization. Flow rate: 30 µL/min, 25 °C; plotted raw data. In comparison, the binding of human blood serum antibodies to RBD with a dilution factor of 1:200 is depicted. (C) SPR assay of CV-N WT binding to S glycoprotein that is immobilized to a response level of 400 µRIU to capture CV-N. Please click here to view a larger version of this figure.

It has been shown before that monomeric CV-N with a single L cannot sufficiently bind gp120, but that CV-N's neutralization ability requires cross-linking of two carbohydrate-binding sites and is predominantly restored by dimerization to functionalize with 2H^12,19. Hence, a monomeric mutant CVN-E41A is expressed, which is suspected to destabilize pseudo-domain B or can interrupt connectivity with the second domain A, such as found in CV-N WT. CVN-E41A monomer reveals stability upon binding to S protein similar to CV-N monomer. Although SPR response for 605-680 µM analyte concentrations results in high response units and typical SPR sensorgrams for CV-N WT and E41A binding, respectively, the mutant is unstable when diluted (Figure 5).

Supplementary Figure 1: SPR sensorgram for capturing DM mimicry as part of influenza HA top. Screenshot: SPR data plot showing the SPR run protocol for the immobilization of DM onto SPR sensor chip CMD500D, which is coated with 500 nm carboxymethyl dextran hydrogel and suitable for kinetics analyses of low molecular weight analytes on high ligand density. Sensorchips are directly mounted onto the detector with emersion oil and fixed below a three-port flow cell. After quenching the activated chip surface with 1 M ethanolamine HCl pH 8.5, CVN2 is injected twice (concentration =1 µM and 2 µM, respectively). Purple: Difference between flow cell 1 and flow cell 2; response is recorded at an interval of 0.2 s. Blue: Flow cell 1: 3000-4000 micro-Refractive Index Units (µRIU) coated from a 400 nM solution of the ligand DM to an activated carboxylated surface. Red: Reference flow cell 2. Please click here to download this File.

Supplementary Figure 2: Manual run on HA-functionalized CMD500D sensor chip showing binding of dimeric CVN2L0-V5, V2, and CVN2L0. At infuse flow rates from 30-50 µL/min, CVN2L0 and disulfide bond variants expressed with a His-tag and purified over Ni-NTA are injected in the mid-µM-range and tested for binding to HA with an association phase of 3 min and dissociation phase of 2 min. Injections are V5 (15 µM), V2 (2.4 µM), 0 µM, CVN2L0 subcloned from a SUMO-fusion protein (1 µM), and CVN2L0 (2 µM). Running buffer at physiological pH is used for all experiments at 25 °C. All solutions are degassed and filtered (0.2 µm) before injection into the system. Purple: Difference between flow cell 1 and flow cell 2; response is recorded at an interval of 0.2 sec. Blue: Flow cell 1: 2540 µRIU HA. Red: Reference flow cell 2. Please click here to download this File.

Supplementary Figure 3: CVN2 binding to HA. Analyses of sensorgrams when curves are auto-aligned. (A) Sensorgrams zeroed and aligned using Scrubber software (left) and binding isotherm showing concentration-dependent response curve to bound analytes (right). (B) Binding isotherm expressed in percent capacity. (C) Computationally fitted theoretical association and dissociation curves. (D) Kinetics data on SPR sensorgrams without fitted curves. Please click here to download this File.

Supplementary Figure 4: CVN2 binding to HA. Analyses of sensorgrams when curves are manually aligned. (A) Sensorgrams zeroed and aligned using Scrubber software (left) and binding isotherm showing concentration-dependent response curve to bound analytes (right). (B) Binding isotherm expressed in percent capacity. (C) Computationally fitted theoretical association and dissociation curves. (D) Kinetics data on SPR sensorgrams without fitted curves. Please click here to download this File.

Supplementary Table 1: Kinetic data obtained from SPR sensorgrams for the binding of CVN2L0 and Cys-Cys bond variants V2, V4, and V5 to HA using a Langmuir 1:1 binding model. K_D [M] = k_off/k_on or k_d/k_a. All data is generated at 25 °C in HBS-EP(+) buffer.: Please click here to download this Table.

Discussion

CV-N's binding affinity is correlated with the number of functional binding sites [2H on domains B, and 2L on domain(s) A when engineered as domain-swapped dimer]. A variant with an altered binding affinity (CVN2L0-V2, a homodimeric stable fold of CV-N comprising a disulfide bridge knock-out) is expressed in E. coli, purified, and positively tested for binding to HA-protein (H3N2) using SPR¹⁰, and shows a conformational change upon binding HA with either H or L carbohydrate-binding sites and K_D1 = 49 nM and K_D2 = 8 µM³⁸. As the number of disulfide bridges near the glycan-targeting pocket in CVN2 decreased from 4 to 2, binding affinity to HA protein decreased (Figures 3B-D and Supplementary Table 1). Disulfide bond variants are created by substituting Cys and inserting polar residue pairs Glu - Arg with slightly decreasing stability. The variants otherwise represent the same molecule, and the four binding sites are functional, although multisite binding on the chip is affected for all variants based on two disulfide substitutions. A non-polar substitution of both disulfide bonds in the two high-affinity binding pockets makes variant CVN2L0-V3, which actively binds the HA-peptide in both its monomannosylated and dimannosylated form. The interactions with CVN2L0-V3 are confirmed for micromolar concentrations and binding with the synthetic peptides (MM and DM) in solution via STD-NMR. Notably, CVN2 SPR binding experiments to MM are not analyzable due to insufficient calculation of R_max, a general drawback of the SPR technique, and weak binding affinities to the immobilized MM¹⁰. Increased ligand density of this peptide and other peptides reduces binding selectivity. Using SPR, dimeric CVN2L0 (2H+2L) with high stability shows multivalent binding to HA, which is expressed in insect cells, and specificity for dimannose on a single chemically engineered biomimicry for high-mannose oligosaccharides, assuming a binding model 1:1. Binding to oligo-mannosides is usually measured by isothermal titration calorimetry, which requires a higher concentration of ligand to be tittered against the lectin¹⁷.

Competitive binding to monomannose or any glycan smaller than Man(7) has not been found before either¹⁵, indicating the involvement of the chemical linkage with the peptide backbone in recognition of these glycan moieties. The number of mannose-mannose linkages in the target was varied, deciphering tryptophan interactions in the high-affinity glycan pocket. The type of modification of peptides with triazole-linked monomannose or triazole-linked dimannose may determine K_D values, in particular to H. Concerning binding affinities of CVN2L0 (2H+2L) and CVN2L0-V2 (1H+2L) to DM, k_off is to the maximum 10-fold higher concerning CVN2L0-V2 dissociation from DM versus HA¹⁰. Besides measuring kinetic rate constants, SPR allows to estimate the equilibrium constants for very slowly equilibrating systems without actually reaching equilibrium (Table 1). By contrast, k_on expresses good relative response values for all binding curves of domain-swapped CVN2L0 and its disulfide bond variants (V2, V4, and V5) to HA (Supplementary Table 1), or DM, while the dissociation rate constant k_off shows faster off-rates for V4 and V5, two variants bearing two functional L, and V4 with non-analyzable K_D concerning DM (Figures 4B). Assigning domain B to H and domain A the respective L is a previous finding that basis for the disulfide bond variants to reveal different binding affinities in SPR, specified, to HA^18,28.

SPR is one of the leading tools for biomolecular interaction analysis in biomedical research³⁵. If a specific, timed control of the bulk analyte concentration is feasible in the SPR instrument, the quantitative evaluation of the kinetics of binding progress between macromolecules is possible, with increasing interest in analyzing multi-protein complexes³⁹. The challenge in its use is the controlled positioning of one of the interaction components on the widely used carboxymethyl dextran, or HC polycarboxylate hydrogel surface, which captures small and large biomolecules. To optimize immobilization and binding capacity of the chip surface, streptavidin-biotin sandwich immobilization and immobilization via protein His-tag to an anti-his antibody are available⁴⁰. The biophysical characteristics of proteins or functional groups of small molecules might interfere with immobilization results, but any difficulties are experienced with the herein described glycoproteins and glycopeptides that are covalently bound via EDC/NHS chemistry. The immobilized glycans, either glycoprotein or the synthetic homogenously mono- or dimannosylated glycopeptides, are used on re-usable chips to screen various engineered CV-N variants, and binding assays are applied to fully purified and recombinantly expressed lectins. Impurities in injected protein solution are damaging the SPR channel system. Although experimental procedures are described on an in vitro system in this study, the glycosylation pattern on viral spikes may be recognized selectively by each analyte concentration of this small model protein as injected over time with optimized mass transfer limitation³⁶. The size of CV-N is ~11 kDa and supports reproducible SPR data for its binding to viral spike proteins.

CV-N's binding specificity to high-mannose glycans is confirmed by the bivalent binding interaction with the dimannosylated peptide mimetic rather than binding to structurally functional glycans and the mono-mannosylated peptide using isothermal titration calorimetry (data not shown). A variant with the point mutation Glu41Ala expresses the monomeric 101-residue protein with two intertwined carbohydrate-binding sites on each protomer. Therefore, this mutation site abolishes a contacting residue between the high- and low-affinity carbohydrate sites and reduces molecular binding strength to N-acetyl-D-glucosamine. Binding of CVN-E41A to SARS-CoV-2 spike protein, bearing complex-type N-linked glycosylation and O-glycosylation, is achieved in the SPR, but CV-N binding to this spike is for both the spike protein and RBD without physiological relevance, as measured at micromolar concentrations (Figure 4,5).

Mannosylated peptides developed and generated in this study are used as protein scaffolds to screen the antiviral agents' binding characteristics by SPR and NMR, as they represent invariable ligands for carbohydrate-binding assays attached to a defined amino acid sequence¹⁰. Moreover, STD-NMR spectroscopy is a label-free technique, such as SPR, which allows the characterization of carbohydrate-binding by conformational selection^10,41. In such settings, STD-NMR allows for the development of rapid screening methods for a large library of compounds to identify bioactive ligands via binding, epitope mapping, and direct determination of K_D under different experimental conditions. Accurate K_D values can be obtained by analyzing the protein-ligand association curve utilizing STD values at the limit of zero saturation time^41,42. Therefore, the interrogation of protein-ligand interactions by the STD-NMR method was accomplished, but K_Ds are calculated on the SPR system.

Human 2019-nCoV (Wuhan-Hu-1-2019 novel coronavirus) is estimated to differ among 11 (out of 18) predicted N-linked glycosylation sites to SARS-CoV⁴³. Epitope/paratope mapping reveals a few interplays with host-derived N-glycans and negligible contributions of antibody somatic hypermutations to epitope contacts³⁷. The ability of antiviral lectins and broadly neutralizing antibodies to bind highly conserved epitopes on the influenza HA glycoprotein is most important to the rational design of vaccines for preventive and therapeutic use. Further investigation, however, is needed to evaluate the effect of immune-suppressive N-linked glycosylation around the host receptor binding site. The data may provide insight into the potential of virus spike proteins to escape from antibodies elicited during infection or to bind non-immunogenic CV-N and thus guide structure-based computational protein design for new optimized antiviral CVN2 variants. By contrast, the affinity of glycosylated Fc amino acid mutant to its receptor to elicit effector functions in vivo, uses a composition of glycosylation sites, and the number of glycans involved, may therefore be important to binding affinity but may not be determined for neutralization ability.

Taken together, fewer rotamers are predicted to bind carbohydrate ligands to proteins by computational protein design than protein-protein interactions alone. Domain-swapped CVN2, however, bearing various numbers of carbohydrate-binding sites, provides the necessary stability and flexibility to reveal different binding-affinities attributed to H and to form the multivalent interactions between oligomannose clusters on viral envelope spikes and neutralizing antibodies (NAb), thus allowing for inhibition assays to be experimentally validated.

Materials

Name	Company	Catalog Number	Comments
Äkta primeplus	Cytiva
Amicon tubes	Merck	C7715
Ampillicin	Sigma-Aldrich	A5354
Beckmann Coulter Cooler Allegra X-30R centrifuge	Beckman Coulter	B06320
Cell spreader	Sigma-Aldrich	HS86655	silver stainless steel, bar L 33 mm
Custom DNA Oligos	Sigma-Aldrich	OLIGO
Custom Gensynthesis	GenScript	#1390661	cloning vector: pET27b(+)
Cytiva HBS-EP+ Buffer 10, 4x50mL	Thermo Scientific	50-105-5354
Dionex UlitMate 3000	Thermo Scientific	IQLAAAGABHFAPBMBFB
Dpn I restriction enzyme (10 U/μL)	Fisher Scientific	ER1701
DTT	Merck	DTT-RO
EDC	Merck	39391
EDTA	Merck	E9884
Eppendorf Safe-Lock Tubes	Eppendorf	30120086
Eppendorf Safe-Lock Tubes	Eppendorf	30120094
Eppendorf Minispin and MiniSpin Plus personal microcentrifuge	Sigma-Aldrich	Z606235
Ethanol	Merck	51976
Ethanolamine HCl	Merck	E6133
Falcon 50mL Conical Centrifuge Tubes	Fisher Scientific	14-432-22
Falcon 14 mL Round Bottom Polystyrene Test Tube, with Snap Cap, Sterile, 25/Pack	Corning	352057
Glucose	Merck	G8270
Glycine HCl	Merck	55097
HA H3 protein	Abcam	ab69751
HEPES	Merck	H3375
His-select Ni2+	Merck	H0537
Imidazole	Merck	I2399
IPTG	Merck	I6758
Kanamycin A	Sigma-Aldrich	K1377
Kromasil 300-5-C4	Nouryon
LB agar	Merck	52062
LB agar	Merck	19344
LB Lennox	Merck	L3022
Lysozyme	Merck	10837059001
Magnesium chloride	Merck	M8266
Magnesium sulfate	Merck	M7506
NaH2P04	Merck	S0751
NanoDrop UV-Vis2000c spectrophotometer	Thermo Scientific	ND2000CLAPTOP
NaOH	Merck	S5881
NHS	Merck	130672
NZ amine (casein hydrolysate)	Merck	C0626
PBS	Merck	806552
PD MidiTrap G-10	Sigma-Aldrich	GE28-9180-11
Peptone	Merck	70171
pET11a	Merck Millipore (Novagen)	69436
PMSF	Merck	PMSF-RO
QIAprep Spin Miniprep Kit (1000)	Qiagen	27106X4
Reichert Software Package Autolink1-1-9	Reichert
Reichert SPR SR7500DC Dual Channel System	Reichert
Scrubber2-2012-09-04 for data analysis	Reichert
SDS	Merck	11667289001
Site-directed mutagenesis kit incl pUC18 control plasmid	Stratagene	#200518
Sodim chloride	Merck	S9888
Sodium acetate.Trihydrate	Merck	236500
SPR sensor chip C19RBDHC30M	XanTec bioanalytics	SCR C19RBDHC30M
SPR sensor chip CMD500D	XanTec bioanalytics	SCR CMD500D
Sterilin Standard 90mm Petri Dishes	Thermo Scientific	101R20
TBS	Merck	T5912	10x, solution
Triton-X100	Merck	T8787
Tryptone	Merck	93657
Tween20	Merck	P1379
Vortex-Genie 2 Mixer	Merck	Z258423
X-gal	Merck	XGAL-RO
XL1-Blue Supercompetent Cells	Stratagene	#200236
Yeast extract	Merck	Y1625