Chapter 1: Carbohydrate–Protein Interactions: A 3D View by NMR Free

Nowadays it is well established that carbohydrates have exceptional properties for coding information.¹  This also holds true for carbohydrate receptors, lectins, antibodies and enzymes, which translate the sugar-based signals into cellular effects.²  In fact, one of the important roles of carbohydrates in Nature is to serve as recognition points for molecular receptors (which can be grouped into enzymes, lectins and antibodies), giving rise to a specific molecular recognition process that triggers a given biological response. The knowledge of the structural elements that govern such molecular recognition events is fundamental, and furthermore opens the possibilities to intervene in them with therapeutic purposes.

Oligosaccharides are involved in a plethora of regulatory processes, such as bacterial/viral infection, angiogenesis, inflammation, cell growth and development.³  Understanding the chemical basis of carbohydrate–receptor interactions not only gives a functional meaning to structures and changes occurring in diseases but also helps devise innovative therapeutic approaches. Therefore, the comprehension of the conformational, dynamics and spatial presentation features of saccharides is of paramount importance.

NMR spectroscopy has been demonstrated to be a robust tool for carbohydrate research. In fact, different NMR approaches are widely employed to study the interactions of carbohydrates⁴  and chemical analogues (glycomimetics) with their receptors up to the level of atomic resolution, both from the perspective of the carbohydrate ligand⁵  and from the receptor.⁶  The most accessible observables related to the structure are chemical shifts (δ), scalar couplings (J) and nuclear Overhauser effects (NOEs). However, in studies with complex oligosaccharides there are limits to the amount of relevant structural information provided by these observables, due to problems of signal overlapping, strong coupling and/or the scarcity of the key NOE information. In this sense, there is increasing use of additional parameters with structural information, such as residual dipolar couplings (RDCs), paramagnetic relaxation enhancements (PREs) or pseudo contact shifts (PCSs) induced by a paramagnetic ion. We will discuss all these parameters in this chapter.

Carbohydrates are rather flexible molecules. Therefore, NMR observables do not always correlate with a single conformer but with an ensemble of low free-energy conformers that can be accessed by thermal fluctuations. In this regard, NMR parameters should be complemented by computational methods in attempts to unravel the structural and conformational features of the molecular recognition process unambiguously.

Depending on the system under study, different NMR approaches can be followed to characterize protein–carbohydrate interactions; the standard methodologies can usually be classified as “ligand-based” or “receptor-based”. The selection of the proper methodology is usually determined by the size of the receptor, the dissociation constant of the complex (K_D), the availability of labelled protein (¹⁵N, ¹³C) and the access to soluble receptors at enough concentration for NMR measurements.

This is the most frequently employed methodology in NMR-based screening applied to drug discovery programs, both in academia and in industry. As the “ligand-observed detection” name suggests, detection takes place on the resonances of the free ligand. Ligand recognition can be identified thanks to the different motions of the receptor and ligand molecules (Figure 1.1): upon carbohydrate recognition the motional properties of the ligand are similar to the receptor and this change in mobility can be detected by different NMR experiments.

Ligand-based methods require complexes with relatively fast kinetics. This means dissociation constants in the range K_D≥100 μM. If k_on is well approximated by a diffusion-limited value (107–109 M⁻¹ s⁻¹), then the slowest exchange rate (k_off) values lie in the range 1000<k_off <100 000 s⁻¹. Even though, in the biological context, carbohydrates are generally attached to proteins or lipids (glycoproteins and glycolipids), most of the studies of the interactions between carbohydrates and their receptors by NMR in solution are carried out by using free sugars. Thus, in the absence of multivalent effects, their binding strengths to proteins are usually rather weak. Thereby, in neutral carbohydrates the dissociation constants are usually in the mM to low μM range and the ligand-based approach can be successfully used.

In such a situation, we profit from the fact that, in most cases, carbohydrate ligands dissociate relatively fast from the receptor's binding site. If the receptor is much larger than the ligand, its larger correlation time will dominate the relaxation properties of the ligand, provided that an interaction takes place. These are the basis of two of the most commonly used methods for obtaining structural information from the ligand's perspective in carbohydrate–protein interactions by NMR: saturation transfer difference (STD) and transferred NOESY (trNOESY). The main advantages of these methods are the small amount of receptor required (experiments are acquired with an excess of the ligand with respect to the receptor). In addition, there is no requirement of the labelled protein, no limitations in size for the target molecule, or any requirement of knowledge about the target structure.

STD is a powerful technique; it allows us not only to identify binders to a receptor, discriminating from non-binders, but also to define the epitope in the case of an intermolecular interaction. Opposite to the usually folded and packed structures of other biomolecules, such as proteins and nucleic acids, carbohydrates present extended and flexible structures with large hydrodynamic radii. Defining the ligand regions in closer contact with the protein provides an important piece of information in order to deduce the 3D shape of the ligand in the binding site, and to identify the structural elements that provide specificity to the recognition process. The simplest version of the methodology implies acquiring two sets of 1D ¹H-based spectra: in one of them, the protein's proton resonances are saturated (on-resonance spectrum) and in the other one they are not (off-resonance spectrum). In the first experiment (on-resonance) the saturation of magnetization (protein) is transferred to the protons of the bound ligand, whose resonances will thus suffer a decrease in intensity with respect to the second spectrum (off-resonance), in which no protein saturation occurs. By subtraction (on-resonance and off-resonance), the STD spectrum is obtained, in which only resonances of the protons close to the protein show up. Furthermore, their intensity will be dependent on the proximity to the protein, yielding epitope information.

Based on this simple experiment, different strategies have been developed with the aim of overcoming the problem of proton overlapping, typical of carbohydrates. One of them has been to extend the ¹H STD information into a second dimension, whether homonuclear (STD-TOCSY and STD-NOESY)⁷  or heteronuclear (STD-HSQC and STD-HMQC).⁸  These experiments have allowed us to define epitopes in larger glycans, such as undeca- and nonasaccharide N-glycans⁵  interacting with lectins or the hepta-xyloglucans recognised by a glycoside hydrolase.⁹  A different strategy has consisted of applying the STD strategy to heteronuclei, such as ¹⁹F.¹⁰  In this case we fully avoid the resonance overlapping, but fluorine atoms have to be introduced through chemical synthesis.

These strategies have been applied thoroughly to study the recognition of glycans and glycomimetics by different carbohydrate recognising elements of therapeutic interest, such as lectins, antibodies, enzymes, viruses or even cells.¹¹  Some of these examples are gathered together below.

The recognition of carbohydrates on the host cell surface is an important step in viral entry. Taking advantage of the large size of viruses and virus-like particles (VLPs), STD is a well-suited method for studding these recognition processes. Even though contradictory results had been reported about the relevance of sialic acid in certain human viral infections,¹²  much evidence is related to infectivity by rotavirus or influenza virus with the recognition of sialic acid on host cell surfaces. STD experiments demonstrated in 2009^13,14  that the previously thought sialidase-insensitive spike protein VP8* from the human rotavirus strain Wa is actually a sialic acid-dependent protein that recognises GD1 and GM1 and GM3-associated subterminal sialic acid residues. Later on, this recognition process was confirmed by performing STD experiments with native viruses, instead of the isolated viral protein. Furthermore, these experiments provided strong evidence that it is VP8* on the virus surface that is responsible for the recognition of sialic acids at the initial virus/host cell contacts.¹⁵  Indeed, the application of STD to study the interaction between ligands and native viruses had been previously proposed by Peters et al.,¹⁶  a strategy that requires very low amounts of virus and a high throughput of ligands. It was demonstrated to discriminate between binding (antiviral) from non-binding ligands to human rhinovirus serotype 2 (HRV2), providing, besides, complete binding epitope information. STD with viral particles has also been applied to gain insights into the infection process by avian influenza virus.¹⁷  Increasing evidence had proposed that human infectivity by the avian influenza virus H5N1 involved a switch in preference of the viral protein hemagglutinin (HA) from α-(2–3)-linked Neu5Ac (major form in avian intestinal cells in birds where the infection is enteric) to α-(2–6)-linked Neu5Ac (abundant in lung and airway epithelial cells in humans in which the infection is respiratory) that could be due to a single amino acid mutation of Asp190 to Glu190. STD experiments carried out with virus-like particles derived from the H5N1 avian influenza containing the hemagglutinin (HA) protein demonstrated that HA from non-humans is indeed able to discriminate between 3′SLN and 6′SLN, binding preferentially to the first one. A more recent work has used STD to relate HA of different subtypes with preference to bind either 3′SLN or 6′SLN.¹⁸  The interaction of noroviruses, the major cause of nonbacterial gastroenteritis, with host attachment factors, like human blood group antigens (HBGA)¹⁹  and inhibitors,²⁰  has also been studied through STD-NMR.

STD has also been applied to living cells, by overexpressing the corresponding receptor on the cell surface. It was first applied to demonstrate the recognition of the S. cerevisiae mannan by cells containing the lectin DC-SIGN (dendritic cell-specific ICAM-3 grabbing nonintegrin).^11a  One of the drawbacks of this approach is to maintain the living cells suspended in solution. More recently, the combination of this strategy with high-resolution magic angle spinning (HR-MAS) was successfully applied to the recognition of the glycoside natural product phlorizin by the Na⁺/glucose co-transporter hSGLT1, for which it is known to be a potent inhibitor.²¹ 

In all the above-mentioned examples, the carbohydrate binding protein attached to the surface of either cells, viruses or virus-like particles is a lectin. Lectins are proteins of non-enzymatic and non-immune origin in charge of recognising carbohydrates, playing a wide variety of roles in different biological events. Indeed, lectins can be potential targets for drug development in different pathologies.²²  In particular, the role of certain lectins in cancer development and apoptosis has focused the attention on them as possible therapeutic anti-tumour targets.²³ 

A typical strategy in drug discovery is the development of enzyme inhibitors with the aim of blocking their activity. Many of them are involved in carbohydrate processing. STD has also been applied in order to understand how substrates and products are recognised by their processing enzymes, an important piece of information for designing inhibitors. Interesting targets are the enzymes involved somehow in the synthesis of the precursors of the glycoconjugates on the parasite's cell surface, like UDP-glucose pyrophosphorylase from Leishmania major²⁴  or trans-sialidase from Trypanosoma cruzi²⁵  or in bacterial cell walls, like UDP-galactopyranose mutase.²⁶  In this last case, STD not only provided epitopic information but STD competition experiments were used in order to compare binding affinities of fluorinated inhibitors.²⁷ 

STD was initially proposed for screening ligand libraries, permitting the use of mixtures of compounds to speed up the identification of ligands for a given target.⁷  An interesting proposal has been to apply STD to dynamic libraries. The method was applied to the identification of the best β-galactosidase inhibitors among a pool of hemithioacetals, formed through the fast and reversible reaction between thiomonosaccharides with small chemical fragments containing aldehyde functionalities. By using eight different monosaccharides and three different fragments, a dynamic library of 18 possible inhibitors was created, from which the best one was identified by giving the strongest STD response.²⁸ 

Carbohydrates act also as antigens, being recognised by antibodies and generating an immune response. Even though the details that confer immunogenicity to carbohydrates are not yet fully understood, their therapeutic use is promising and STD is being used as an important tool for describing epitope selectivity in these recognition processes. Some recent examples involve the use of antibodies as targets in autoimmune diseases like Guillain–Barre syndrome,²⁹  in cancer,³⁰  in infectious diseases like HIV³¹  or in bacterial infections like those produced by Y. pestis,³²  group A streptococcus (GAS),³³  Shigella flexneri³⁴  or Bacillus anthracis.³⁵ 

STD can also provide an estimation of the binding affinity. The traditional strategy has been based on following the ligand STD changes upon the presence of a competitor, whose K_D is known.³⁶  Alternatively, single ligand titration experiments have been proposed to directly extract K_D from STD measurements.³⁷ 

In spite of the wide possibilities of STD, this method is applicable to a limited range of binding affinities, between μM and mM for K_D, due to the requirements for fast dissociation rate complexes. In cases where the affinity is too high (in fact, when k_off is too slow), no STD signal will be detected, because proton ligand relaxation will occur before detection. In such cases the absence of STD signal might be misinterpreted as a nonbinding event. On the other hand, nonspecific interactions can give rise to a STD response, especially when dealing with aromatic patches prone to establish hydrophobic interactions, particularly in water, and whose long relaxation rates can bias the STD intensities.³⁸ 

The transferred NOESY³⁹  is in this sense a more robust and reliable technique. It is based on the different sign of the NOE depending on molecular correlation time, and thus on molecular size: positive for small molecules, but negative for large ones.⁴⁰  As stated above, if a small ligand is in the presence of an interacting large protein, and in fast exchange between the free and the protein-bound states, the large correlation time of the protein–ligand complex will dominate, even if only a small fraction of the ligand is in the bound state. As a result, while the NOEs of the free ligand are positive, the NOEs of the ligand in the presence of the protein will become negative. This NOE sign change is known as transferred NOE. This method permits us to prove that one small ligand binds to a large receptor, although the reverse is not necessarily true.

The first applications of trNOE permitted us to identify different scenarios that might take place in the recognition of carbohydrates. One of them is a conformational selection process, in which a single conformation of the ligand, among the ones present when it is free, is recognised, as in the recognition of the disaccharide melibiose [Gal-α-(1→6)-Glc] by the lectin domain of ricin, where only the major ligand's conformation found in the free state is recognised by the lectin.⁴¹  In a few cases the conformational selection proceeds in such a way that a ligand conformation with minor contributions in the free state is the one recognised, like the case of the GlcpNAc-β-(1→6)-α-Manp disaccharide bound to the lectin WGA.⁴² 

The study of the recognition of sialyl Lewis X tetrasaccharide by E-selectin, a key recognition event in inflammatory processes, allowed identifying a conformational selection process in which the major conformer in solution is the one being recognised.⁴³  In parallel, the binding mode of a family of E-selectin antagonists with higher affinity than the natural ligand was also studied by trNOE. Interestingly, the comparison between the natural ligand's and antagonist's binding modes allowed the researchers to relate conformational preorganization with increased binding affinity.⁴⁴  Many other examples of lectin/carbohydrate recognition processes have been studied by trNOE.⁴⁵ 

Glycosaminoglycans (GAGs) have well-known therapeutic applications, such as anticoagulants or antithrombotic drugs. The advances in structure–function relationship knowledge are providing them with new possible therapeutic applications.⁴⁶  trNOESY has contributed to gain insights into the recognition of GAGs by their receptors, where glycan flexibility involves not only dihedral angles⁴⁷  but also pyranose ring conformation, like in the case of the synthetic 6-nonsulfated tetrasaccharide ANS-I2S-ANS-I2s, where the IdoA residues adopt a ¹C₄ conformation when bound to FGF2.⁴⁸ 

One the main difficulties in trNOESY is to achieve the proper experimental conditions to detect such an effect, the protein/ligand ratio being a critical parameter. The most important bias in this experiment comes from spin-diffusion in the ligand mediated by the protein.⁴⁹  This can give rise to the misinterpretation of spin-diffusion crosspeaks as trNOESY crosspeaks, which would lead to a virtual bound conformation. The way to identify such effects is through the co-acquisition of trROESY experiments.⁵⁰ 

In many cases, STD and trNOESY experiments are combined together in order to have a more complete view of the interaction from the ligand perspective: the epitope mapping and bound conformation. This strategy has significantly contributed to gain insights into biological processes with therapeutic applications. For instance, it provided confirmation that the different binding modes found by X-ray⁵¹  for the recognition of the terminal Man-α-(1–2)-Man fragment by DC-SIGN also takes place in solution.⁵²  DC-SIGN is a C-type lectin found on the surface of dendritic cells and macrophages, known to recognise mannose and Lewis X-based pathogen-associated molecular patterns (PAMPs) to subsequently activate phagocytosis. It has attracted much attention since it was found to be the target to which some viruses (e.g. VIH,⁵³  Ebola,⁵⁴  hepatitis C⁵⁵ ), bacteria (e.g. H. pylori, K. pneumoniae, M. tuberculosis⁵⁶ ), parasites (e.g. L. pifanoi) and fungi (Candida albicans)⁵⁷  bind, being essential for the infection. Thus, the development of inhibitors with the aim of blocking its carbohydrate binding properties has been proposed as a potential therapeutic strategy.⁵⁸  In certain cases the combination of trNOESY and STD has allowed the identification of their different binding modes.⁵⁹ 

Despite the advantages of the STD and trNOESY experiments, there are systems in which this methodology cannot be applied due to strong binding. In the context of carbohydrates, this is usually the case in charged systems such as glycosaminoglycans that establish electrostatic interactions with their protein receptors.³  In these systems the carbohydrate conformation can be inferred by using filtered experiments with ¹³C-labelled protein samples. The protons of the receptor (attached to ¹³C) can be selectively removed from the spectrum, allowing the analysis of the carbohydrate signals (protons attached to ¹²C) without interference from the protein signals. The ¹³C double-filtered 2D NOESY method has been successfully applied to obtain the conformation of a hexasaccharide bound to acidic fibroblast growth factor.⁶⁰ 

These experiments can also be acquired in the 3D version. For instance, the 3D ¹³C F1-edited F3-filtered HSQC-NOESY experiment has been applied in the structure calculation of the complex between Coprinopsis cinerea lectin 2 (CCL2) and a fucosylated chitobiose. In this case, not only intramolecular NOEs but also 82 intermolecular NOEs could be identified.⁶¹ 

Residual dipolar couplings (RDCs) originate from the dipolar interaction occurring between a pair of nuclei (e.g. ¹H–¹³C, ¹H–¹H). RDCs are orientational in nature since they depend on the angle θ that forms the vector joining two nuclei with the spectrometer magnetic field and are independent of the distance between protons. This feature makes RDCs highly complementary to the conventional distance constraints derived from NOE data and are especially useful in systems with small numbers of NOEs due to carbohydrate flexibility or extended shapes.

The dipolar coupling between a pair of nuclei (e.g. ¹H–¹³C) displays the angular dependence shown in eqn (1.1), where θ is the angle that forms the vector joining the two nuclei with the direction of the B₀ spectrometer magnetic field and r_HC is the distance between interacting spins:

The terms enclosed within brackets in eqn (1.1) correspond to the time average distance 〈r_HC⁻³〉 and time angular averaging 〈cos²θ〉. In liquid-state NMR the function goes to zero due to the isotropic distribution of the ¹H–¹³C vectors and no dipolar couplings are observed. Therefore, RDC measurement requires a weak alignment of the carbohydrate molecules in order to obtain some orientations of the interatomic vector connecting the two coupled nuclei slightly more favourably than others.⁶² 

The calculation of the dipolar interaction between a pair of dipolar-coupled nuclei requires determination of 〈cos²θ〉 in eqn (1.1). This term can be determined by using the alignment tensor methodology.⁶³  In the so-called principal axis frame, in which the three axes of the alignment tensor are diagonal, eqn (1.2) provides the theoretical value of the RDC for any ¹H–¹³C vector of the carbohydrate, where A_ax and A_rh are the axial and rhombic components, respectively, θ is the angle that forms vector H–C with the z-axis in the molecular frame and ϕ is the angle that describes the projection of vector H–C over the plane xy of the molecular frame (see Figure 1.2):

For a given signal in the spectrum, the magnitude of the RDC is determined from the difference in splitting between the sample oriented in the aligned media (RDC + J) and the isotropic non-aligned sample (J).

RDCs have been efficiently used to derive the conformation of carbohydrates in solution.⁶⁴  However, fewer examples have been described on the use of this parameter to study the carbohydrate conformation in the bound state with their receptors. In contrast to transferred NOEs that are dominated by NOE bound contributions even with high ligand/receptor ratios, RDCs of exchanging ligands can be dominated by free-state contributions. In order to overcome this drawback, it is possible to enhance the alignment of the protein under study by addition of a hydrophobic alkyl tail that anchors to the bicelles of the ordering medium.⁶⁵  However, this approach is not generally applicable because of the nature of the required protein modifications. The use of a His-tagged protein with a nickel chelate-carrying lipid inserted into the lipid bilayer-like alignment media has been proposed as an alternative approach.⁶⁶ 

Pseudo contact shifts (PCSs) also arise from a dipolar interaction, but in this case the effect is due to the dipolar interaction between the unpaired electrons of a paramagnetic ion and the nuclei in the vicinity of this ion. PCSs give rise to large changes in chemical shifts that decrease with the distance between the metal ion and the nuclear spin with a 1/r³ dependence. Therefore, PCSs provide long-range structural information. The equation that governs PCSs is given in eqn (1.3), where Δδ_PCS is the difference in chemical shifts measured between diamagnetic and paramagnetic samples, r is the distance between the metal ion and the nuclear spin and θ and ϕ are the polar coordinates describing the position of the nuclear spin with respect to the principal axes of the Δχ tensor:

The effect is proportional to the anisotropy of the magnetic susceptibility tensor (Δχ) and independent of the isotropic component of the χ tensor. It is therefore sufficient to describe the PCS as a function of the axial and rhombic components of the Δχ tensor, Δχ_ax and Δχ_rh.

In addition, all lanthanide ions with non-vanishing susceptibility anisotropy tensors (Δχ) generate weak molecular alignment with the magnetic field and therefore RDCs.^67a  Conformational studies of diamagnetic carbohydrates have been carried out by converting them to paramagnetic compounds through attachment of a lanthanide binding tag by chemical synthesis. There is an increasing number of examples of the use of PCS methodology in the carbohydrates field, from disaccharides to nonasaccharide structures.^67,68  The standard protocol is based on the generation of an ensemble of conformations by molecular dynamics simulations and to back-calculate the expected PCS values for each conformation. The fitting experimental versus calculated PCSs allow defining the carbohydrate conformation.

PCSs have been recently used to derive the conformation of a complex type N-glycan. The characterization of this molecule is challenging since the molecular pseudo-symmetry leads to isochronous NMR in both branches (see Figure 1.3).⁶⁸  N-Glycans are common modifications of membrane/secreted proteins and immunoglobulin G (IgG) antibodies. For instance, the Fc domain of IgG from healthy humans has a complex-type biantennary N-glycan attached to Asn297. While the Fc glycans are quite heterogeneous, a high percentage is normally terminated with galactose. Deviations from this pattern correlate well with certain diseases, as well as differential effects on inflammatory response.⁶⁹  Despite their importance, there is a need for development of new methods to characterize these flexible carbohydrates. The conformational features of the different classes of N-glycans (from complex to high-mannose type) have been investigated using different techniques, especially NMR spectroscopy,⁷⁰  since the flexibility of the glycosidic linkages has usually hampered their study by X-ray crystallography. However, typical NMR parameters (J couplings and NOEs) provide only short-range structural information. Therefore, it is not easy to define the global shape of these nonglobular molecules.

In this context, the use of the lanthanide tag has permitted breakage of the inherent pseudo-symmetry of the N-glycan as shown in Figure 1.3, revealing that the T-shaped gg rotamer at the Man-α-(1–6)-Man junction is the major one in solution, with minor contributions of other backfolded geometries. In addition, the recognition of this nonasaccharide by human galectin-3 has been studied. In this line, the novel methodology employed has permitted the characterization of the binding epitopes of the symmetrical N-glycan, showing that both arms are involved in the recognition of human galectin-3.⁶⁸ 

The monitoring of molecular recognition processes using these NMR methods is based on the comparison of different NMR parameters of the receptor molecule resonances in the presence and absence of the ligand (or mixtures of putative ligands). Owing to the direct observation of receptor signals, this method cannot be applied to large proteins (it is usually employed in receptors below 40 kDa). The advantage of the receptor-based approach is that it allows defining the location of the protein binding site. This is especially important in the understanding of the action mechanism of a drug, since it is possible to find additional binding sites (as happens in allosteric modulators). The main drawback of receptor-based techniques is the requirement of a previous assignment of the protein NMR resonances. This assignment should be ideally combined with the a priori knowledge of the receptor's 3D structure (either from X-ray or NMR) to drive lead generation.

One of the most frequently used receptor-based methods in drug discovery programs relies on the perturbation of the receptor's chemical shifts upon binding of the ligand. The chemical shift is an extremely sensitive parameter to the environmental changes and therefore can be used to detect interaction processes. Chemical shift mapping obtained via NMR titration experiments is a straightforward NMR technique to define those protein residues that are involved in the binding to a partner molecule. Usually, a series of NMR experiments (generally ¹H–¹⁵N HSQC or ¹H–¹³C HSQC spectra) are recorded with increasing amounts of the ligand (Figure 1.4).

This approach has been used to unravel the role of heparin in the activation of fibroblast growth factors (FGFs) and fibroblast growth factor receptors (FGFRs).^6,71  FGFs are key pharmacological targets since they are involved in cell replication, angiogenesis, differentiation, cell adhesion, migration and wound healing. Down- and up-regulation of FGFs are associated with many pathologies.⁷² 

The interactions of heparin fragments with both acidic FGFs and FGFRs were characterized using NMR.^69,71  As an example, in Figure 1.4 is shown the superimposition of the ¹H–¹⁵N HSQC spectra of the immunoglobulin-like 2 domain (Ig2) of FGFR in the presence of increasing amount of a heparin pentasaccharide.⁷³  As a first step, assignment of the backbone HN correlations of the protein was carried out by combination of the 3D triple resonance HNCO, HN(CA)CO, HNCACB and CBCA(CO)NH experiments. ¹H–¹⁵N HSQC experiments were collected with different ligand/protein ratios from 1 : 0.25 to 1 : 2. The residues with the largest chemical shift perturbations are clustered on a well-defined surface of the protein, indicating that there is a specific interaction between the protein and the oligosaccharide at this particular region.

In the heparin field, NMR receptor approaches also have been proved to be useful in determining the role of heparin in the cytotoxicity of eosinophil cationic protein (ECP).⁷⁴  ECP is secreted by activated eosinophil granulocytes during infections, acting as a mediator in human immune host defence. ECP's cytotoxicity is the subject of intense research due to the tissue damage associated with eosinophil degranulation at the inflammation site.⁷⁵  Novel eosinophil targeting therapies are under development and the anti-inflammatory properties of heparin derivatives are a promising research field.⁷⁶  In recent work, the structure of the ECP in complex with a heparin trisaccharide was obtained using NMR. ¹H–¹⁵N HSQC experiments of ECP were acquired in the presence of increasing trisaccharide concentrations to obtain both the location of the carbohydrate binding site and the estimation of the dissociation constant.⁷⁴ 

NMR receptor-based methods have also been used to study DC-SIGN inhibitors. DC-SIGN is a C-type (Ca²⁺ dependent) lectin present in the surface of dendritic cells. DC-SIGN binds to mannosides and fucosides to mediate interactions with other cells or pathogens. DC-SIGN is involved in immune responses since it mediates pathogen recognition and cellular interactions that lead to pathogen neutralization. As a pathogen receptor, DC-SIGN recognizes several viruses (HIV,⁵³  ebola,⁵⁴  hepatitis C⁵⁵ ), bacteria (Mycobacterium tuberculosis⁵⁶ ) and fungi (Candida albicans⁵⁷ ). However, numerous studies have demonstrated that some of these pathogens subvert DC's function to escape immune surveillance by targeting DC-SIGN, such as HIV-1 that exploits the DC-SIGN internalization pathway to transinfect T cells.⁵³  DC-SIGN binds to the mannosylated surface glycoprotein gp120 on HIV, and this recognition event is the initial entry port of the human immunodeficiency virus to the host.

In this context, it is important to design inhibitors of the carbohydrate–lectin interaction. Oligomannose and oligofucose derivatives could be envisioned as potential inhibitors of DC-SIGN interactions. However, the presence of additional C-type lectins with similar specificities precludes their application. In order to overcome this problem, glycomimetics with higher affinity for DC-SIGN have been designed. Addition of the glycomimetic to DC-SIGN carbohydrate recognition domain resulted in ¹H–¹⁵N HSQC chemical shift perturbations similar to those obtained with N-acetylmannosamine or fucose.⁷⁷  These data demonstrate that the glycomimetic occupies the same carbohydrate binding site and interacts with the same side-chain residues on DC-SIGN as the natural ligands.

¹H–¹⁵N HSQC experiments were also used to study bifunctional inhibitors, designed for targeting both matrix metalloproteinases (MMPs) and galectins.⁷⁸  There is strong evidence supporting the fact that MMPs are involved in tumour progression.⁷⁹  In addition, galectins are overexpressed in several tumours.⁸⁰  The presence of galectins in the zone of tumour invasion and the role in tumour progression of up-regulating MMPs reveal a functional connection between these two effector proteins. In this context, a dual inhibitor was synthesized and its ability to bind both receptors was explored using ¹⁵N-labelled galectin-3 and ¹⁵N-labelled metalloproteinase 12.⁷⁸  The simultaneous binding of the dual inhibitor to the two proteins was demonstrated using a wise combination of ligand-based and receptor-based NMR experiments, employing the proper combination of labelled and non-labelled proteins to monitor the interaction process.

In the context of carbohydrate–protein interactions, it is also worth mentioning the application of nitroxide spin-labelled carbohydrates for mapping the carbohydrate binding sites on the protein surface. Nitroxide spin-labelled molecules, such as 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO), give rise to intensity losses in the NMR peaks due to the paramagnetic relaxation enhancement effect (PRE) of the nitroxide.

Nitroxide spin labels have been incorporated by chemical synthesis in N-acetyllactosamine,⁸¹  cellotriose and cellotetraose to carry out binding studies.⁸² 

The presence of the spin label allows defining the interaction surface in the receptor by analysing the perturbation of the intensities of cross-peaks in the protein ¹⁵N HSQC spectrum. In this context, the perturbation of the intensities of cross-peaks in the ¹⁵N HSQC spectrum of full-length galectin-3 has been monitored and used to identify protein residues proximate to the binding site for N-acetyllactosamine.⁸¹  In addition, it is possible to convert intensity measurements into distances between discrete protein amide protons and the bound spin label.⁸¹ 

Studying carbohydrate interactions by using NMR with the drug design perspective is a growing field. We foresee many developments in the coming years. The standardization of novel expression systems for biomedically relevant glycoproteins will allow us to generate these entities in the proper amounts for the required detailed generation and analysis of NMR parameters. Current NMR methodologies, based on those described herein, will also find their application in the derivation of the key factors influencing the existence of induced fitting or conformational selection processes. The combination with advanced modelling procedures will, with no doubt, generate novel methodologies, novel scientific achievements and also molecules that can be employed as drugs or probes.