Fluorinated Carbohydrates as Lectin Ligands: Dissecting Glycan-Cyanovirin Interactions by ¹⁹F-NMR

Abstract

NMR spectroscopy and ITC are powerful methods to investigate ligand-protein interactions. Here, we present a versatile and sensitive fluorine NMR approach that exploits the ¹⁹F nucleus of ¹⁹F labeled carbohydrates as a sensor to study glycan binding to lectins. Our approach is illustrated with the 11 kDa Cyanovirin-N, a mannose binding anti-HIV lectin. Two fluoro-deoxy sugar derivatives, methyl 2-deoxy-2-fluoro-α-D-mannopyranosyl-(1→2)-α-D-mannopyranoside and methyl 2-deoxy-2-fluoro-α-D-mannopyranosyl-(1→2)-α-D-mannopyranosyl-(1→2)-α-D-mannopyranoside were utilized. Binding was studied by ¹⁹F-NMR spectroscopy of the ligand and ¹H-¹⁵N HSQC NMR spectroscopy of the protein. The NMR data agree well with those obtained from the equivalent reciprocal and direct ITC titrations. Our study shows that strategic design of fluorinated ligands and fluorine NMR spectroscopy for ligand screening holds great promise for easy and fast identification of glycan binding as well as for their use in reporting structural and/or electronic perturbations that ensue upon interaction with a cognate lectin.

Introduction

The growing realization of the enormous potential of glycans in information coding, embodied by the term “sugar code”,^[1] has led to define oligosaccharides as functionally active units and provides the rational for identifying new targets in drug design. Aided by the advances in chemical synthesis of such determinants,^[2] detection of glycan receptors (lectins) and the analyses of the specificity/mechanisms of protein-carbohydrate interactions are key steps in this endeavor.^[3] With glycan-receptor recognition critically involved in viral and bacterial cell attachment, as well as in host defense, immune(dys)regulation and tumor progression,^[4–8] versatile and sensitive analysis/screening tools in this area are urgently needed. NMR can satisfy this need, having proved its value by providing high-resolution molecular structures in solution and valuable information on ligand binding.^[9–12]

Most biomolecular NMR applications employ proton/carbon/nitrogen experiments and 2D heteronuclear NMR is a powerful method to map ligand binding sites on proteins and to quantify protein-ligand interactions.^[9–12] It has, however, some limitations. It is best suited for medium size proteins (< 50kDa) since amide resonance assignments are a prerequisite. In addition, for titration experiments with low affinity ligands, large amounts of ligand are needed to reach saturation, introducing uncertainty in the estimation of affinity parameters, especially for synthetic compounds of limited water-solubility. Thus, ligand-centered methods hold some promise to overcome these problems.

Among ligand-detected NMR approaches, only few exploit the favorable properties of the fluorine nucleus.^[13,14] Mono- or trifluoro N-acetylglucosamines were used to measure affinity constants and map their binding environment.^[15–18] Using fluorinated sugars,^[19–21] saturation transfer difference (STD) experiments in which magnetization was relayed to the vicinal and geminal F-atoms in a monosaccharide were used in a competition screening assay in the search for lead compounds in inhibitor design.^[22,23] Analogous to ITC, where in addition to titrating ligand into the protein solution, a second, reverse titration can be performed in which protein is titrated into the ligand solution, ligand-based NMR titrations provide the reciprocal data to information commonly extracted from 2D ¹H,¹⁵N heteronuclear correlation spectra of the protein. If the correct model is used for data fitting, identical thermodynamic parameters should be obtained, regardless of the direction of the titration.^[24] Here we present a NMR study in which the ¹⁹F-NMR signal of a singly fluorinated small molecule ligand is monitored upon protein addition. We compare these data to results obtained by recording changes in ¹H,¹⁵N amide protein resonance frequencies upon ligand binding. For our study we selected the 11 kDa antiviral lectin cyanovirin-N (CV-N) as the sugar receptor, since its structure and glycan binding properties have been carefully analyzed by NMR spectroscopy, crystallography, and titration calorimetry.^[25–28] CV-N harbors two carbohydrate-binding sites in its pseudo symmetric, bilobal structure, site 1 in domain B (residues 39–89) and site 2 in domain A (residues 1–38/90–101).^[25–28] The minimal glycan recognition unit is a α1,2-linked dimannoside, either as the individual oligosaccharide or as part of the terminal arms of the branched Man-8 and Man-9 structures.^[29,30] Domain A exhibits a slight preference for trimannoside and domain B for dimannoside.^[28,31,32] The availability of detailed STD data for these ligands as well as results with specific deoxy sugars allowed us to identify the pivotal hydroxyl groups responsible for binding.^[33,34] In particular, we determined that the 3′- and 4′-hydroxyl groups are essential for the interaction, whereas the 2′- and 6′-hydroxyl groups of the Manα(1–2)Manα (Man2) terminal unit are not directly involved in CV-N binding.^[34] We therefore prepared methyl 2-deoxy-2-fluoro-α-D-mannopyranosyl-(1→2)-α-D-mannopyranoside (¹⁹F-Manα(1–2)Manα:¹⁹F-Man2) and methyl 2-deoxy-2-fluoro-α-D-mannopyranosyl-(1→2)-α-D-mannopyranosyl-(1→2)-α-D-mannopyranoside (¹⁹F-Manα(1–2)Manα(1–2)Manα:¹⁹F-Man3) (Supplementary Scheme S1) for use in our current work.

We show that fluorination at the 2′ or 2″ position of the sugar does not curtail its bioactivity in a cell-binding assay in which ¹⁹F-Man2 and ¹⁹F-Man3 were used as inhibitors for CV-N’s interaction with the cell surface. We determined affinity constants and exchange rates for ¹⁹F-Man2 and ¹⁹F-Man3 interaction with two of CV-N variants in which one of the binding sites had been impaired,^[35–37] and compared the results to values derived from titration calorimetry.

In addition, due to the inherent high sensitivity of the fluorine chemical shift to its local environment, it was possible not only to distinguish between free and bound signals for ¹⁹F-Man2 and ¹⁹F-Man3, but also to identify two different modes of binding for ¹⁹F-Man3.

Results and Discussion

NMR is becoming a standard approach for studying protein-ligand interactions since it provides powerful means for mapping the structural details of binding sites on the protein as well as the ligand. In addition, affinity parameters can be determined. Here, we employed ¹⁹F-NMR and fluorinated ligands to evaluate the interactions between di- and trimannoside with CV-N and demonstrate that, based on the favorable properties of the ¹⁹F nucleus, novel features can be discerned.

Bioactivity of the ¹⁹F-containing glycans

In order to ascertain that substituting a fluorine at the 2′ or 2″ position for the hydroxyl group does not alter the properties of the sugars significantly and that the two 2′-¹⁹F-substituted dimannoside- and the 2″-¹⁹F-substituted trimannoside are capable to act as specific and potent ligands, we tested their activity in blocking lectin binding to cells. As shown in Figure 1, 2′-¹⁹F-substituted glycans strongly inhibited lectin binding to human carcinoma cells, when compared to the natural monosaccharide. Similar data were also obtained for a T lymphoma (Jurkat) line (not shown).

Figure 1.

Cell-binding of monofluorinated di- and trimannosides. Human colon adenocarcinoma cells (SW480) were stained with fluorescein-labelled CV-N. Controls representing staining with CV-N (100%-value at 1 μg/ml; green line) and without (0%-value; shaded area) are included in each panel (quantitative data for the percentage of positive cells/mean fluorescence intensity are listed). Top panel: control experiment without added glycan (green) and in the presence of 100 mM mannoside (left), and in the presence of 4 mM or 8 mM ¹⁹F-Man2 (right). Bottom panel: control experiment without added glycan (green) and in the presence of 100 mM mannoside (left), in the presence of 1 mM or 2 mM ¹⁹F-Man3 (right).

Therefore, incorporation of the ¹⁹F sensor did not impair bioactivity, confirming that the 2′ group5 are not critical for cell surface receptor binding on the two cell types tested, in line with expectations based on results with the 2′-deoxy derivative of the dimannoside.^[34] These results prompted us to proceed to studying the interaction with the lectin by ¹⁹F-NMR spectroscopy.

Ligand-detected binding of fluorinated sugars to CV-N

For ligand-detected NMR experiments we recorded 1D-¹⁹F NMR spectra of ¹⁹F-Man2 (200 μM) and ¹⁹F-Man3 (50 μM) at 280 K. Spectra of 1:1 mixtures of sugar:protein are provided in Figure 2A, and Figure 2B for [CVN^ΔA]_ssm, CVN^mDB and CV-N^P51G with ¹⁹F-Man2 and ¹⁹F-Man3, respectively.

Figure 2.

1D-¹⁹F NMR spectra of ¹⁹F-Man2/Man3 in the presence of [CVN^ΔA]_ssm, a variant that contains a single glycan binding site on domain B, CVN^mDB, a variant that contains a single glycan binding site on domain A, and CV-N^P51G that is wild-type-like with both glycan binding sites present, at 280 K. CV-N domains are represented by two elongated crescents, representing the two binding sites. Site 1 on domain B is colored pink and site 2 on domain A in green. A) ¹⁹F-1D NMR spectra of 200 μM ¹⁹F-Man2 in the presence of [CVN^ΔA]_ssm, CVN^mDB, and CV-N^P51G. B) ¹⁹F-1D NMR spectra of 50 μM ¹⁹F-Man3 in the presence of [CVN^ΔA]_ssm, CVN^mDB and CV-N^P51G. Bound-state and free-state signals are in slow exchange for [CVN^ΔA]_ssm in C) and CVN^mDB in D. In the titration curves (bottom panels) the bound fraction is derived from the signal intensity (1-I_free/I₀) during the titration, with I_free the intensity of the free ligand signal at each point in the titration, and I₀ the intensity of the free ligand signal at the beginning of the titration. In C) [CVN^ΔA]_ssm was titrated into 50 μM ¹⁹F-Man3 (top panel), with molar ratios of protein/glycan: 0:1, 0.5:1, 1:1, 1.5:1. In D) CVN^mDB was titrated into 50 μM ¹⁹F-Man3 (top panel), with protein/glycan molar ratios: 0:1, 0.5:1, 1:1, 2:1, and 3:1.

Two distinct resonances, separated by a Δδ >1 ppm, correspond to the protein-free and protein-bound conformations. The resonance corresponding to the protein-bound conformation of ¹⁹F-Man2 when bound to domain A is downfield shifted from the free resonance while that corresponding to the domain B protein-bound is upfield shifted. The two resonances of ¹⁹F-Man3 when bound to domain A and B both experience a downfield shift, compared to the free sugar signal. Unambiguous assignments to the domain A-bound or domain B-bound state was achieved by using [CVN^ΔA]_ssm and CVN^mDB, variants of CV-N in which the binding site in domains A and B were obliterated.^[35–37] Gratifyingly, the two resonances corresponding to bound ¹⁹F-Man2 in the CV-N^P51G wild-type variant of CV-N that possesses both binding sites, exhibit identical chemical shifts to those observed in the single site mutants, confirming that the site that remains in the mutant is identical to its wild-type counterpart. The same holds for the ¹⁹F-Man3 resonances, although as pointed out above, both bound resonances are downfield of the free sugar signal.

Much of the power of fluorine NMR studies of biological systems derives from the high sensitivity of the fluorine shielding parameter to changes in local environment. Shielding is observed when the fluorine is in the close contact to H bond donors of the protein or solvent molecules.^[38–40] On the contrary, a fluorine nucleus in the vicinity of an electronegative atom downfield shifts. Also, deshielded fluorines are found in the close contact with hydrophobic side chains.^[38–40] Inspection of the crystal structure of a mutant CV-N protein, in which the carbohydrate binding site in domain A was abolished, complexed with a dimannoside ligand, P51G-m4-CVN: Man2 (PDB accession code 2RDK),^[41] revealed that in the binding site on domain B, a 2′-¹⁹F substitution could act as a H-bond acceptor with a vicinal water molecule, causing an upfield shift. Unfortunately, structural data for dimannoside binding in domain A are not available, therefore we can only speculate that such H-bond is not present for 2′-¹⁹F-Man2 interacting with domain A. The small ¹⁹F downfield chemical shift observed in this case is most likely caused by either the anionic repulsion from an electronegative protein atom or the interaction of fluorine atom with the hydrophobic protein binding pocket. For Man3, the crystal structures of wild type CV-N bound to Man-9 (PDB accession code 3GXZ)^[32] allows to delineate residues in domain A that interact with the trimannoside on the D1 arm of Man-9. The negative Glu101 side chain carboxylate causes electrostatic repulsion of the anionic 2″-¹⁹F, causing a downfield shift. A similar scenario or possibly contacts with hydrophobic protein side chains are likely for ¹⁹F-Man3 binding to domain B, although direct supporting structural data are not available.

We also carried out 1D-¹⁹F NMR experiments at 298 K. However, except for the resonance of the free fluorinated glycan, no signal for the ligand-bound to protein conformation was detected in the case of ¹⁹F-Man2, complexed with all CV-N variants. This is probably due to severe line broadening in the intermediate exchange regime at room temperature. For ¹⁹F-Man3, on the other hand, both ligand-free and ligand-bound resonances were detected at 298 K when [CVN^ΔA]_ssm, CVN^mDB or CV-N^P51G was added, although ligand-bound resonances were significantly broader than the ones at 280 K (data not shown). The 1D ¹⁹F-NMR titration data, monitoring the ¹⁹F-Man3 resonance intensities upon addition of [CVN^ΔA]_ssm and CVN^mDB at 280 K, are provided in Figure 2C, and Figure 2D respectively. The spectra for ¹⁹F-Man3 (50 μM) are displayed in the top panels and the resulting binding isotherm in the bottom panel. For both mannosides, free and bound resonances are in slow exchange. The K_D^F values were extracted from the binding isotherms, monitoring the intensity changes of the protein-bound ligand signal versus the protein:ligand molar ratio. The bound fraction (f_b) was estimated from the intensity changes of the free ligand resonance during the titration using f_b=1-I_free/I₀. The extracted K_D^F values were 10.8±0.3 μM and for [CVN^ΔA]_ssm and 33.5±3.5 μM for CVN^mutDB binding to ¹⁹F-Man3, respectively.

Exchange kinetics between free and protein-bound fluorinated glycans

The ¹⁹F-NMR resonances for free and bound states at 280 K are in slow exchange on the chemical shift scale, as evidenced by exchange peaks in the 2D ¹⁹F-¹⁹F NOESY spectra.

In order to extract kinetic parameters for the exchange process, a series of 2D ¹⁹F-¹⁹F homonuclear NOESY spectra with different mixing times were recorded. Representative spectra of ¹⁹F-Man2 (200 μM) in the presence of [CVN^ΔA]_ssm and CVN^mDB (1:1 molar ratio) at 280 K are provided in Figure 3A and Figure 3B. The exchange curves obtained from the intensity ratios of the bound, diagonal peak, to the exchange cross-peak (I_BB/I_BF) as a function of mixing time are shown next to the NOESY spectra. Analysis of the exchange curve for ¹⁹F-Man2 when bound to the site on domain B yields a free-to-bound exchange rate (k_ex) of 56.8±1.4 s⁻¹, while that for binding to domain A yields a k_ex value of 99.2±2.1 s⁻¹.

Figure 3.

2D ¹⁹F-¹⁹F NOESY exchange spectra of ¹⁹F-Man2/Man3 in the presence of [CVN^ΔA]_ssm and CVN^mDB protein. A) ¹⁹F-Man2 (200μM) bound to [CVN^ΔA]_ssm(left panel; t_mix=0.1 s), and the exchange curve of bound/bound over free/bound intensity ratio (I_BB/I_BF) versus mixing time (right panel) t_mix: 0.05, 0.1, 0.3, 0.5 s. B) ¹⁹F-Man2 bound to CVN^mDB (left panel; t_mix=0.05 s) and the exchange curve of bound/bound over free/bound intensity ratio (I_BB/I_BF) versus mixing time (right panel) t_mix: 0.025, 0.05, 0.1, 0.2 s. C) ¹⁹F-Man3 (50 μM) bound to [CVN^ΔA]_ssm (left panel; t_mix=0.1 s) and the exchange curve of bound/bound over free/bound intensity ratio (I_BB/I_BF) versus mixing time (right panel) t_mix: 0.1, 0.3, 0.5, 0.7, 0.9 s. D) ¹⁹F-Man3 (50 μM) bound to CVN^mDB (left panel; t_mix=0.05 s) and the exchange curve of bound/bound over free/bound intensity ratio (I_BB/I_BF) versus mixing time (right panel) t_mix: 0.05, 0.1, 0.15, 0.3 s. All spectra were recorded at 280 K for 1:1 molar ratios. The auto-correlation (diagonal) peaks I_FF, and I_BB arise from the free and bound states, while cross-correlation (off-diagonal) peaks I_FB=I_BF report on the exchange between free and bound states.

The equivalent data of ¹⁹F-Man3 (50 μM) in the presence of [CVN^ΔA]_ssm and CVN^mDB (1:1 molar ratio) are shown in Figure 3C and Figure 3D and resulted in k_ex values of 2.29±0.1 s⁻¹ and 10.5±2.9 s⁻¹ for the binding to domains B and A, respectively. Thus, the exchange between free and bound state of the di-and trimannoside is somewhat faster from domain A than domain B. Although easily observed at 280 K, no exchange peak was observed at 298 K for ¹⁹F-Man2 due to intermediate exchange that caused line broadening beyond detection. Those for ¹⁹F-Man3, on the other hand, are still present at 298 K.

Protein-detected binding between ¹⁹F-Man2 and ¹⁹F-Man3 with CV-N

Direct protein-observed NMR titration experiments were carried out with ¹⁹F-Man2 at 298 K by following chemical shift perturbations in the ¹H-¹⁵N HSQC spectrum of ¹⁵N uniformly labeled [CVN^ΔA]_ssm (75 μM, Figure 4) and CVN^mDB (400 μM, Figure S1) as a function of added sugar. Spectra in the absence (black contours) and presence (green contours) of 25 molar equivalents of ¹⁹F-Man2 are provided in Figure 4A and Figure S1A, respectively. In the ¹H-¹⁵N HSQC NMR spectra of [CVN^ΔA]_ssm and CVN^mDB titrations at 298 K, resonances are in fast exchange, in contrast to the situation in the ¹⁹F-NMR spectra, where resonances at 298 K were in intermediate exchange and broadened beyond detection. Following the chemical shift changes throughout the titration yielded the binding isotherm displayed in the right panels of Figure 4A and Figure S1A, respectively.

Figure 4.

NMR titration of [CVN^ΔA]_ssm with ¹⁹F-Man2 recorded at 298 K and ¹⁹F-Man3 recorded at 280 K, respectively. A) Perturbed amide resonances in the [CVN^ΔA]_ssm spectrum clearly reside in domain B and are labeled by amino acid name and number. The left panel display a superposition of ¹H-¹⁵N HSQC NMR spectra of [CVN^ΔA]_ssm in the absence (black) and presence of 25 molar equivalents of ¹⁹F-Man2 (green). The right panel displays the titration curve (chemical shift difference versus ligand/protein molar ratio), with the chemical shift difference defined as: Δδ = [(Δδ_H)² + (Δδ_N × 0.17)²]^1/2, where Δδ_H and Δδ_N are the observed chemical shift changes for ¹H and ¹⁵N, respectively. The system is in fast exchange, as evidenced by a single signal corresponding to glycan free and glycan bound [CVN^ΔA]_ssm. B) Superposition of ¹H-¹⁵N HSQC spectra of 50 μM [CVN^ΔA]_ssm in the absence (black) and presence (green) of 3 molar equivalents of ¹⁹F-Man3 (left panel). The right panel displays the titration curve at glycan/protein molar ratios: 0:1, 0.5:1, 1:1, 1.5:1, 2:1, and 3:1. The system is in slow exchange, as evidenced by separate signals for glycan free and glycan bound [CVN^ΔA]_ssm. The binding isotherm is derived from the relative signal intensities (1-I_free/I₀) (right panel).

The equilibrium dissociation constant (K_D^F) for ¹⁹F-Man2 binding was determined from the titration shifts (chemical shift Δδ (ppm) versus the ratio of ligand to protein concentration) using several resonances. ¹⁹F-Man2 binding to the single sites on domains B and A yielded K_D^F values of 960±91 μM and 1.37±0.075 mM, respectively. The latter value is about a factor of two larger than the one determined for Man2 binding to site A on CVN^mutDB (757±80 μM),^[36] suggesting that the substitution of fluorine at the 2′ position reduces the binding affinity only by a very small amount. As noted previously, Man2 exhibits a slight preference for Domain B over Domain A, respectively.^[28,31,32]

Equivalent NMR titration experiments, monitoring chemical shift perturbations in the ¹H-¹⁵N HSQC spectra of [CVN^ΔA]_ssm and CVN^mDB were carried out at 280 K with ¹⁹F-Man3. A superposition of [CVN^ΔA]_ssm (50 μM) spectra in the absence (black contours) and presence (green contours) of 3 molar equivalents of ¹⁹F-Man3 is provided in Figure 4B. The binding isotherm derived from the intensity changes of free resonances is depicted in the right panel. The equivalent titration experiment was performed for CVN^mDB (50 μM) and the data are provided in Figure S1B. The dissociation constants were extracted as described for ¹⁹F-Man2 (above) and binding to a single site on domain B (in [CVN^ΔA]_ssm) yielded a K_D^F value of 63.5±5 μM, while the one for binding to a single site on domain A (in CVN^mDB) was 150.3±7 μM.

ITC binding

A direct comparison of binding parameters that were extracted from the different NMR titrations with those from ITC measurements was performed.

In particular, reverse titrations in which ¹⁹F-Man3 (50 μM) was placed into the cell and protein into the injector were carried out. In this manner, an equivalent situation to the one in ¹⁹F-glycan-observed NMR titrations at 280 K is created. The measured titration heat as well as the derived binding isotherms are provided in Figure 5A and Figure S2A for [CVN^ΔA]_ssm and CVN^mDB, respectively. The extracted K_D^F values were 12.6±1.2 μM for [CVN^ΔA]_ssm and 28.5±3.5 μM for CVN^mutDB, in excellent agreement with the values extracted from the ¹⁹F-glycan-observed NMR titrations (see above).

Figure 5.

ITC titration experiments for the interaction between ¹⁹F-Man3 and [CVN^ΔA]_ssm at 280 K. In A) the trace for titrating [CVN^ΔA]_ssm into 50 μM ¹⁹F-Man3 (reverse titrations) for 28 automated injections is shown (top panel), and the derived binding isotherm is displayed in the bottom panel. The nonlinear least-squares best fit to the experimental data using a one-site model is shown by the line connecting the data points. In B) the ITC trace of titrating ¹⁹F-Man3 into 30 μM [CVN^ΔA]_ssm (direct titrations) for 28 automated injections (top panel) with the derived binding isotherm (bottom panel) is displayed. The nonlinear least-squares best fit to the experimental data using a one-site model is shown by the line connecting the data points.

We also carried out direct ITC titrations at 280 K with ¹⁹F-Man3 in the injector and [CVN^ΔA]_ssm(30 μM) and CVN^mDB(50μM) in the cell. The observed binding isotherms are provided in Figure 5B and Figure S2B yielding K_D^F values of 62.5±2 μM and K_D^F of 153.3±5 μM for [CVN^ΔA]_ssm and CVN^mDB, respectively. These values are in good agreement with those obtained from the NMR direct titration experiment for both [CVN^ΔA]_ssm and CVN^mDB variants, that were 63.5±5 μM and 150.3±7 μM, respectively. Interestingly, it appears that ¹⁹F-Man3 no longer preferentially binds to domain A, as was observed for the non-fluorinated Man3,^[28,31,32] possibly due to the ¹⁹F substitution. The extracted dissociation constants for ¹⁹F-Man3 binding are provided in Table 1.

Table 1.

Dissociation constants for the interaction of ¹⁹F-Man3 with single binding site CV-N variants determined by ligand-observed ¹⁹F-NMR, reverse ITC, protein-observed NMR, and direct ITC experiments.

Protein	KDF (μM) 19F-NMR	KDF (μM) reverse ITC	KDF (μM) 1H-15N HSQC	KDF (μM) direct ITC
[CVNΔA]ssm	10.8 ± 0.3	12.6 ± 1.2	63.5 ± 5	62.5 ± 2
CVNmDB	33.5± 3.5	28.5 ± 3.5	150.3 ± 7	153.3 ± 5

Calorimetry is the only technique that directly permits to evaluate the basic physical forces in sufficient detail between and within molecules by measuring heat quantities or heat effects.⁴² Therefore, a deeper understanding of the molecular basis of protein–ligand interactions can be gained by thoroughly characterizating and quantifying the energetics that govern complex formation.^24,42 Analysis of the binding isotherm for ¹⁹F-Man3 with [CVN^ΔA]_ssm reverse ITC titrations yielded a ΔG value of −5.1 kcal mol⁻¹, ΔH= −2.9 kcal mol⁻¹, and a binding entropy TΔS of 2.2 kcal mol⁻¹, while for the direct ITC titration a ΔG value of −5.3 kcal mol⁻¹, ΔH= −5.1 kcal mol⁻¹, and a binding entropy TΔS of 0.2 kcal mol⁻¹ was obtained. Similarly, a ΔG value of −5.8 kcal mol⁻¹, ΔH= −1.5 kcal mol⁻¹ and TΔS of 4.3 kcal mol⁻¹ was measured in the reverse ITC titration of ¹⁹F-Man3 with CVN^mDB, while a ΔG value of −4.88 kcal mol⁻¹, ΔH= −4.08 kcal mol⁻¹ and TΔS of 0.8 kcal mol⁻¹ was obtained for the direct titration. A summary of all thermodynamic data determined for ¹⁹F-Man3 binding to [CVN^ΔA]_ssm and CVN^mDB in reverse and direct ITC titrations is provided in Table 2. It should be mentioned that standard deviation values arise from the fitting procedure.

Table 2.

Thermodynamic parameters extracted from ITC binding data

Protein	Enthalpy ΔH (kcal mol−1)	Entropy TΔS (kcal mol−1)	Free energy ΔG (kcal mol−1)	Stoichiometry n
[CVNΔA]ssm reverse ITC	−2.9 ± 0.05	2.2 ± 0.04	−5.1 ± 0.06	1.01 ± 0.03
[CVNΔA]ssm direct ITC	−5.1 ± 0.1	0.2 ± 0.08	−5.3 ± 0.1	1.0 ± 0.0
CVNmDB reverse ITC	−1.5 ± 0.03	4.3 ± 0.02	−5.8 ± 0.04	1.02 ± 0.04
CVNmDB direct ITC	−4.08 ± 0.06	0.8 ± 0.03	−4.88 ± 0.05	1.0 ± 0.0

For both mutants, binding to ¹⁹F-Man3 was driven by enthalpic contributions (with negative ΔH values of −2.9 kcal mol⁻¹ and −5.1 kcal mol⁻¹ in the case of [CVN^ΔA]_ssm, and −1.5 kcal mol⁻¹ and −4.08 kcal mol⁻¹ for CVN^mDB, for the reverse and direct ITC titrations, respectively), with favorable entropic contributions. Somewhat larger positive TΔS values (2.2 kcal mol⁻¹ for [CVN^ΔA]_ssm and 4.3 kcal mol⁻¹ for CVN^mDB) were obtained for the reverse ITC titrations, compared to the direct ones (0.2 kcal mol⁻¹ for [CVN^ΔA]_ssm and 0.8 kcal mol⁻¹ for CVN^mDB), suggesting that water molecules are more eficiently released from the complex surface in the case of reverse titrations.⁴² The overall ΔG values, however, are not very different.

ITC also allows determining the stoichiometry of binding, independent of the binding affinity.²⁴ Since a 1:1 stoichiometry was obtained from the experimental (sigmoid) curves in the cases of reciprocal ITC titrations, a one binding site model was used to fit the data and binding affinity parameters were extracted. In the case of direct ITC titrations, since ~1:1 stoichiometry was observed, a stoichiometry value of n=1 was fixed in order to fit the experimental (parabolic) curves, based on the fact that a single binding site was detected for the equivalent, protein-observed ¹H-¹⁵N HSQC titrations (Figure 4B and Figure S1B). Furthermore, the fully saturated samples resulted either at the end of protein-observed ¹H-¹⁵N HSQC or direct ITC titrations, were tested on 1D-¹⁹F NMR as well, and indeed, only one protein-bound peak corresponding to a single binding site was observed. Therefore, the reverse titration allowed us not only to check the stoichiometry, but also to ensure that a suitable binding model was used to fit the ITC data. For a 1:1 binding stoichiometry, the measured thermodynamic parameters are expected to be invariant when changing the orientation of the experiment.²⁴ However, this is rarely the case since one of the two species may display greater aggregation or less solubility when concentrated.^24b Indeed, a comparison of the data in Table 1 revealed that titrating ¹⁹F-Man3 into solutions of [CVN^ΔA]_ssm and CVN^mDB using protein-observed ¹H-¹⁵N HSQC spectra or direct ITC experiments yielded K_D^F values approximately five times higher than the ones obtained from ligand-observed ¹⁹F-NMR and reciprocal ITC titrations. Naturally, if aggregation occurs, this will happen at high concentrations of the ligand. Therefore, it is advantageous to measure binding on the ligand, either by NMR or reverse ITC, since lower ligand concentrations are needed.

Thus, ligand-detected binding methods are preferable compared to methods that measure protein signals, since any errors associated with the solubility issue, which therefore might introduce incorrect ligand concentrations, are avoided.

Detection of a second binding mode in the interaction of ¹⁹F-Man3 with CV-N

The concentration of ¹⁹F-Man3 in the ligand-detected ¹⁹F-NMR experiments was 50 μM, while for those with ¹⁹F-Man2 was 200 μM. To directly compare the results of [CVN^ΔA]_ssm, CVN^mDB and CV-N^P51G binding to the two fluorinated glycans, we also recorded a ligand-detected ¹⁹F-spectrum of ¹⁹F-Man3 (200 μM) with CV-N variants (1:1 molar ratio). As can be observed from the comparison of the spectra illustrated in Figure 6, at higher ¹⁹F-Man3 concentrations ( >150 μM) additional resonances emerge. These new resonances exhibit exactly the same chemical shifts (Figure 6B) that were seen for the signals of ¹⁹F-Man2 when bound to domains A and B of CV-N, respectively (Figure 6A).

Figure 6.

1D-¹⁹F spectra of 200 μM ¹⁹F-Man2/Man3 in the presence of different CV-N variants, illustrating the power ¹⁹F NMR for discriminating between different binding sites and modes. The presence of a second binding mode for CV-N’s interaction with ¹⁹F-Man3 is clearly observed. All spectra were recorded at 280 K. A) 1D-¹⁹F spectra of 200 μM ¹⁹F-Man2 in the presence of [CVN^ΔA]_ssm, CVN^mDB and CV-N^P51G. B) 1D-¹⁹F spectra of 200 μM ¹⁹F-Man3 in the presence of [CVN^ΔA]_ssm, CVN^mDB and CV-N^P51G. Schematic depiction of ¹⁹F-Man2 and the two binding modes of ¹⁹F-Man3 to CV-N variants are represented on the right panels of A) and B). The individual sugar units on the oligosaccharide are color coded according to their linkage pattern. C) 2D ¹⁹F-¹⁹F NOESY spectra ¹⁹F-Man3 in the presence of 200 μM [CVN^ΔA]_ssm and CVN^mDB (1:1 molar ratio of glycan:protein). The auto-correlation (diagonal) peak intensities I_FF, I_BB in the 2D ¹⁹F-¹⁹F NOESY spectra arise from the free and bound states, while cross-correlation (off-diagonal) resonances I_FB=I_BF and I_FB’=I_BF’ arise from exchange between the two binding modes for CV-N and ¹⁹F-Man3. D) Competitive binding of Man2 to the complex of ¹⁹F-Man3:[CVN^ΔA]_ssm (200 μM;¹⁹F-Man3:[CVN^ΔA]_ssm:Man2 molar ratios: 1:1:0; 1:1:1; 1:1:5; 1:1:10; and 1:1:20). Schematic representation of Man2 competing with ¹⁹F-Man3 binding to [CVN^ΔA]_ssm is represented on the bottom panel. Man2 competitor displace first ¹⁹F-Man3 bound to protein with non-reducing ¹⁹F-Manα(1–2)Manα unit (the low affinity binding mode: solid arrow), and later the one bound with the unit closer to O-methyl group (high affinity binding mode: dashed arrow).

The presence of the second binding mode can also be observed in the 2D ¹⁹F-¹⁹F exchange NOESY experiment recorded with a 50 ms mixing time on the sample of 200 μM ¹⁹F-Man3 with [CVN^ΔA]_ssm and CVN^mDB (Figure 6C). Given that the additional peaks exactly match the chemical shifts of the bound conformations of ¹⁹F-Man2 when sitting in domains A and B of CV-N and are only observed at relatively high concentrations, it appears that each domain of CV-N binds first to the Manα(1–2)Manα unit that is closer to O-methyl group in ¹⁹F-Man3 (high affinity binding mode: 1) and subsequently to next non-reducing end ¹⁹F-Manα (1–2)Manα unit (low affinity binding mode: 1′).

To further substantiate our findings we also carried out a competition binding experiment, adding non-fluorinated Man2 to a solution of ¹⁹F-Man3 (200 μM) and [CVN^ΔA]_ssm (1:1 molar ratio) and monitored the change in 1D-¹⁹F NMR ligand signal intensity as a function of Man2 addition. The resulting data up to a 20 fold molar excess of Man2, are provided in Figure 6D. As can be noted, the resonance that corresponds to the bound conformation of the non-reducing end ¹⁹F-Manα(1–2)Manα unit is the one that is affected first by the addition of non-fluorinated Man2, while the resonance corresponding to the bound conformation of the second Manα(1–2)Manα unit, the one closer to O-methyl group, is affected only at higher Man2 addition. Such differentiation was not observed when ¹⁹F-Man3 or non-fluorinated Man3 was used in protein-detected titration experiments.

¹⁹F-NMR competition assay

To further highlight the advantages of using ¹⁹F-NMR, we also carried out a quantitative competition binding experiment, titrating non-fluorinated Man3 into a complex of ¹⁹F-Man3 (250 μM) with relatively low (25 μM) of CVN^mDB (10:1 molar ratio), that ensure that only the high affinity binding mode is involved. Indeed the inspection of 1D-¹⁹F NMR spectra show only the presence of the peak corresponding to high affinity protein-ligand binding mode, while no additional low affinity binding peak is observed.

The intensity change in the 1D-¹⁹F NMR free ligand signal was monitored as a function of Man3 addition. A schematic representation of competitive binding scenario is illustrated in Figure 7A. The resulting data are provided in Figure 7B. The intensity increase in the 1D-¹⁹F NMR spectrum of the free ¹⁹F-Man3 signal in complex with CVN^mDB, upon addition of the non-fluorinated competitor (Man3), corresponds to the amount of non-labeled ligand bound to CVN^mDB. The dissociation constant of Man3 binding to CVN^mDB was obtained from the relevant competition titration data for the ¹⁹F-Man3:CVN^mDB:Man3 complex (Figure 7B). In order to extract the binding constant k_d of Man3 interacting with CVN^mDB, it is necessary to know the K_D^F value of ¹⁹F-Man3 binding to CVN^mDB (Figure 7A). We determined the value by ¹⁹F-NMR and reciprocal ITC titrations (see above). The extracted k_d value of 4.65 ± 0.6 μM is consistent with our previous affinity data of the non-labeled Man3 with CVN^mDB (3.4±0.05 μM), obtained from ITC titration experiment.^[36]

Figure 7.

Competition of Man3 and ¹⁹F-Man3 for 250 μM ¹⁹F-Man3 in complex with 25 μM CVN^mDB. A) Schematic representation of the competitive binding scenario. Cheng–Prusoff equation based on a competitive site binding model, where [¹⁹F-Man3] and [Man3] are the concentrations of the fluorinated and non-fluorinated ligand, and K_DF the dissociation constants of ¹⁹F-Man3 for the glycan-binding site on domain A of CVN^mDB. The data were fitted using the maximum increase in free ¹⁹F-Man3 resonance intensity upon Man3 addition, ΔI_max, and k_d the dissociation constant for Man3 binding as adjustable parameters. B) ¹⁹F-NMR spectra were recorded for ¹⁹F-Man3:CVN^mDB:Man3 molar ratios of 10:1:0, 10:1:0.25, 10:1:0.5, 10:1:0.75, and 10:1:1. In the titration curve the bound fraction of Man3 is derived from the difference in signal intensity of the ¹⁹F-Man3 free signal ΔI=I_free-I₀ during the titration, with I_free the intensity of the ¹⁹F-Man3 free signal at each point in the titration, and I₀ the intensity of ¹⁹F-Man3 free signal at the beginning of the titration.

Conclusion

We presented an approach to quantitatively assess glycan-lectin interactions in which the changes in the ¹⁹F-NMR resonances of fluorinated glycans are monitored upon protein addition. Our method exploits the relatively large ¹⁹F-NMR chemical shift range that results in well-resolved resonances in 1D spectrum. Employing fluorinated ligands permits studying protein-ligand interaction by NMR even for high molecular mass proteins (> 100kDa) and requires only minimal amounts of ligand. Therefore, novel avenues for screening in a facile fashion and relatively short time can be envisaged.

In particular we show that:

1. Quantitative binding affinity data can be obtained using ¹⁹F-NMR experiments with fluorinated glycans.

2. Information about protein-ligand binding exchange can be extracted from ¹⁹F-¹⁹F NOESY experiments and we determined chemical exchange rate constant, k_ex, for each of the two binding sites of CV-N.

3. Different binding modes can be discerned for the two neighboring Manα(1–2)Manα units in ¹⁹F-Man3 when interacting with CV-N.

4. ¹⁹F-NMR competition experiments can be employed for quantitatively extracting affinity constants of non-fluorinated glycans with target lectins.

In summary, utilizing ¹⁹F as sensor for studying protein-ligand interactions represents a promising perspective for ¹⁹F-NMR spectroscopy as a universal tool in drug discovery for any target receptors for which fluorinated ligands are available or can be synthesized.

Experimental Section

Lectins

The Cyanovirin-N wild-type stabilized variant CV-N^P51G and the two mutants [CVN^ΔA]_ssm and CVN^mDB that possess only one sugar-binding site, either in domain B ([CVN^ΔA]_ssm) or domain A (CVN^mDB) were prepared as described previously.^[35–37] The two fluoro-deoxy sugar derivatives, methyl 2-deoxy-2-fluoro-α-D-mannopyranosyl-(1→2)-α-D-mannopyranoside and methyl 2-deoxy-2-fluoro-α-D-mannopyranosyl-(1→2)-α-D-mannopyranosyl-(1→2)-α-D-mannopyranoside were obtained using synthetic procedures outlined in Supplementary Scheme S1.

Cell assays

CV-N^P51G was fluorescently labeled by fluorescein isothiocyanate under activity-preserving conditions, separated from reagents by gel filtration and applied in cytofluorometry as described for other lectins previously.^[43]

NMR Spectroscopy

¹⁹F-NMR spectra were recorded on a Bruker 600 MHz AVANCE spectrometer, equipped with a Bruker CP TXO triple-resonance, X-nuclei observe, z-axis gradient cryoprobe (Bruker Biospin, Billerica, MA). ¹H-¹⁵N HSQC experiments were recorded on Bruker 600 MHz AVANCE spectrometer equipped with CP TCI triple-resonance, z-axis gradient ¹H detect cryoprobe (Bruker Biospin, Billerica, MA). Spectra were processed with NMRPipe^[44] and analyzed with NMRview.^[45]

NMR Titrations –monitoring resonances of the fluorinated ligand

Ligand-observed ¹⁹F-NMR experiments were performed using fluorinated deoxy sugars, ¹⁹F-Man2 (200 μM) and ¹⁹F-Man3 (50 μM), and single binding site variants, [CVN^ΔA]_ssm, CVN^mDB as well as the stabilized CV-N^P51G wild-type protein at 1:1 protein:ligand molar ratios, in 20 mM sodium phosphate buffer, pH 6.0, 0.01% NaN₃, 90% H₂O/10% D₂O.

For titration experiments with ¹⁹F-Man3 sugars (50μM), ¹⁹F-spectra were monitored upon addition of increasing amounts of protein, [CVN^ΔA]_ssm and CVN^mDB, (5 mM stock solution) in 20 mM sodium phosphate buffer, pH 6.0, 0.01% NaN₃, 90% H₂O/10% D₂O. A series of 1D-¹⁹F NMR spectra were recorded at 280 K at protein:sugar molar ratios of 0:1, 0.5:1, 1:1, and 1.5:1 for [CVN^ΔA]_ssm and 0:1, 0.5:1, 1:1, 2:1 and 3:1 for CVN^mDB. Free and bound ¹⁹F ligand resonances throughout the titration are in slow exchange on the chemical shift scale. The observed signal intensity change during the titrations is directly proportional to the fraction bound (f_b), given by: f_b=1-f_f=1-I_free/I₀= [PL]/[L], where [L] is the total concentration of ligand and [PL] the concentration of the resulting ligand-protein complex. I_free is NMR peak intensity of the free ligand signal at each point in the titration, and I₀ is the peak intensity of the free ligand signal at the beginning of titration. Binding curves were derived from the relative ratios of ligand resonance intensities (1-I_free/I₀) versus molar ratio M of protein:sugar, and apparent K_D^F values were obtained by non-linear best fitting of the titration curves using KaleidaGraph (Synergy Software, Reading, PA), using the following equation:

$$f_b=1-\frac{I_{\mathit{free}}}{I_0}=0.5\ast \left(M+1+\frac{{{\text{K}}_{\text{D}}}^{\text{F}}}{[L]}-\sqrt{{\left(M+1+\frac{{{\text{K}}_{\text{D}}}^{\text{F}}}{[L]}\right)}^2-4M}\right)$$

NMR Titrations – monitoring the protein resonances

Protein-observed NMR titration experiments were performed on Bruker 600 MHz AVANCE spectrometers, equipped with 5-mm, triple resonance, three-axis gradient probes or z-axis gradient cryoprobes, using uniformly ¹⁵N-labeled [CVN^ΔA]_ssm(75 μM) or CVN^mDB(400 μM) with increasing amounts of ¹⁹F-Man2, in 20 mM sodium phosphate buffer, pH 6.0, 0.01% NaN₃, 90% H₂O/10% D₂O. A series of ¹H-¹⁵N HSQC spectra at 298 K were recorded after the addition of sugar aliquots from a stock solution (10 mM) at sugar:protein molar ratios of: 1:0, 2:1, 4:1, 6:1, 8:1, 10:1, 14:1, 18:1, 22:1, and 25:1 for [CVN^ΔA]_ssm and 0:1, 0.5:1, 1:1, 1.5:1, 2:1, 2.5:1, 3:1, 3.5:1, 5:1, 6:1, 7:1, 8:1, 9:1, 11:1, 12:1, 15:1, 16:1, 21:1, and 25:1 for CVN^mDB.

Free and sugar-bound protein resonances throughout the titration are in fast exchange on the chemical shift scale. The observed chemical shift changes during the titration is given by: Δδ= [PL]/[P] (δ_b − δ_f), with [P] and [PL] the concentrations of protein and ligand-protein complex and δ_b and δ_f the chemical shifts of protein resonances in the free and fully bound state. The chemical shift difference was calculated as: Δδ= [(Δδ_H)² + (Δδ_N × 0.17)²]^1/2, with Δδ_H and Δδ_N the observed chemical shift changes for ¹H and ¹⁵N, respectively. The dissociation constant K_D^F, was obtained by best-fitting the titration curve (chemical shift change Δδ vs. molar ratio M of sugar:protein) using KaleidaGraph software and the following equation:

$$\mathrm{\Delta }\delta =0.5\ast \mathrm{\Delta }{\delta }_{max}\left(M+1+\frac{{{\text{K}}_{\text{D}}}^{\text{F}}}{[P]}-\sqrt{{\left(M+1+\frac{{{\text{K}}_{\text{D}}}^{\text{F}}}{[P]}\right)}^2-4M}\right)$$

Analogous NMR titration experiments were performed using uniformly ¹⁵N-labeled [CVN^ΔA]_ssm(50 μM) and CVN^mDB(50 μM) with ¹⁹F-Man3, in 20 mM sodium phosphate buffer, pH 6.0, 0.01% NaN₃, 90% H₂O/10% D₂O. A series of ¹H-¹⁵N HSQC spectra were recorded at 280 K after addition of sugar aliquots from a stock solution of 10 mM at sugar:protein molar ratios of: 0:1, 0.5:1, 1:1, 1.5:1, 2:1, and 3:1 for [CVN^ΔA]_ssm and 0:1, 0.5:1, 1:1, 1.5:1, 2:1, 2.5:1, 3.5:1, and 4:1 for CVN^mDB. In this case, free and sugar-bound protein resonances throughout the titration were in slow exchange on the chemical shift scale. The observed signal intensity change during the titration is direct proportional to the fraction bound (f_b), given by: f_b=1-f_f=1-I_free/I₀= [PL]/[P], with [P] the total concentration of protein and [PL] the concentrations of the protein-ligand complex. I_free is the resonance intensity of the ligand-free protein signal at each point in the titration, and I₀ is the resonance intensity of the ligand-free protein signal at the beginning of titration. Binding curves were derived from the relative ratios of ligand resonance intensities (1-I_free/I₀) versus molar ratio M of protein:sugar, and apparent K_D^F values were obtained by nonlinear best fitting of the titration curves using KaleidaGraph (Synergy Software, Reading, PA), averaging over the eight titration curves using the following equation:

$$f_b=1-\frac{I_{\mathit{free}}}{I_0}=0.5\ast \left(M+1+\frac{{{\text{K}}_{\text{D}}}^{\text{F}}}{[P]}-\sqrt{{\left(M+1+\frac{{{\text{K}}_{\text{D}}}^{\text{F}}}{[P]}\right)}^2-4M}\right)$$

2D ¹⁹F-¹⁹F NOESY exchange experiment

Two-dimensional exchange spectroscopy (NOESY) represents a powerful tool for probing ligand-receptor exchange.^[46] NOESY experiments are particularly suitable to evaluate chemical exchange processes that occur on a slow time scale, where the exchange rate has little effect on the lineshape. A series of 2D homonuclear exchange ¹⁹F-¹⁹F NOESY experiments were recorded at 280 K on a sample containing 200 μM ¹⁹F-Man2, complexed with [CVN^ΔA]_ssm and CVN^mDB at 1:1 molar ratios and mixing times (t_mix) of 0.05, 0.1, 0.3, 0.5 s and 0.025 0.05, 0.1, 0.2 s, respectively. An equivalent series was recorded for 50 μM ¹⁹F-Man3, complexed with [CVN^ΔA]_ssm and CVN^mDB at mixing times (t_mix) of 0.1, 0.3, 0.5, 0.7, 0.9 s and 0.05, 0.1, 0.15, 0.3 s, respectively. The exchange rate constant was extracted by fitting the experimental intensity ratios (diagonal peak (I_BB) of bound conformation, to the bound to free exchange cross-peak (I_BF)) versus mixing time (t_mix), using the following equation:

$$\frac{I_{BB}}{I_{BF}}=\frac{k_{eq}\left[exp(k_{ex}\times t_{\mathit{mix}})+1\right]}{\left[exp(k_{ex}\times t_{\mathit{mix}})-1\right]},$$

where k_eq is the equilibrium constant, t_mix the mixing time, and k_ex the free-bound exchange rate.^[47]

Isothermal Titration Calorimetry

Calorimetric titrations were performed using a VP-ITC isothermal titration calorimeter (MicroCal, LLC; Northampton, MA). Titrations were carried out at 280 K and all solutions contained 50 mM sodium phosphate buffer, pH 7.4, 0.2 M NaCl, 0.02 % NaN₃. ITC measurements were carried out with [CVN^ΔA]_ssm and CVN^mDB solution in the cell (~1.44 ml active volume) at 30 μM and 50 μM concentrations, respectively, stirred at 310 rpm. 28 injections of 10 μl aliquots of 2.5 mM ¹⁹F-Man3 in the case of [CVN^ΔA]_ssm, and 35 injections of 8 μl aliquots of 4 mM ¹⁹F-Man3 in the case of CVN^mDB, were performed at 2 min intervals from a 285 μl stirring syringe. Reverse ITC measurements were performed as follows: 50 μM ¹⁹F-Man3 solution was placed in the calorimeter cell (~1.44 ml active volume), stirred at 310 rpm, while [CVN^ΔA]_ssm or CVN^mDB solution were placed in the injector. 10 μl aliquots of 0.6 mM [CVN^ΔA]_ssm and 11 μl aliquots of 1 mM CVN^mDB were added at 2 min intervals from the 285 μl stirring syringe. A total of 28 injections were performed for [CVN^ΔA]_ssm and 25 injections for CVN^mDB. Binding isotherms were fit using the Origin 7.0 software using a standard one-site model, to extract the apparent number of binding sites and affinity parameters. Values for the binding enthalpy, the apparent number of binding sites and affinities were obtained from the fit to the experimental data. Other thermodynamic parameters were calculated using the standard expressions: ΔG = −RT ln Ka; ΔG = ΔH −TΔS.

¹⁹F-NMR competition binding assay

The binding affinities of Man3 for the CVN^mDB variant were quantified using a ¹⁹F-NMR-based competition assay. In this experiment, the non-labeled competitor ligand Man3 was titrated into a solution of 250 μM ¹⁹F-Man3 and 25 μM CVN^mDB in 20 mM sodium phosphate buffer, pH 6.0, 0.01% NaN₃, 90% H₂O/10% D₂O. A series of 1D-¹⁹F NMR spectra were recorded at 280 K with ¹⁹F-Man3:CVN^mDB:Man3 molar ratios of: 10:1:0; 10:1:0.25; 10:1:0.5; 10:1:0.75; and 10:1:1. The intensity increase of the free 1D NMR signal of ¹⁹F-Man3 complexed with CVN^mDB, upon addition of the inhibitor (Man3), is proportional with the amount of non-labeled ligand (Man3) bound to CVN^mDB. The apparent dissociation constants k_d for Man3 binding to the Domain A of CVN^mDB, was obtained from best fitting the response curves with KaleidaGraph (Synergy Software, Reading, PA), using a corresponding competitive binding Cheng–Prusoff equation,^[48] where K_D^F is the dissociation constant of ¹⁹F-Man3 binding to the Domain A of CVN^mDB (see above).

$$\frac{\mathrm{\Delta }I}{\mathrm{\Delta }{I}_{max}}=1-\frac{1}{1+\frac{{K_D}^F}{[{}^{19}F-\mathit{Man}3]}\left(1+\frac{[\mathit{Man}3]}{k_d}\right)}$$