The past decades have witnessed a fast growth of chemical glycobiology, driven by an increased understanding of glycans ability to encode a large variety of biochemical information in physiological and pathological conditions.^1–3  One of the main objectives of the field consists in manipulating chemical information encoded in sugar structures (the sugar code⁴ ) to control or alter the flow of information read-out by sugar binding proteins, called lectins. Glycomimetic molecules that can disrupt the formation of sugar–protein complexes have been used in this context as probes of biological processes and are providing ideas for medicinal applications.^5–7  A large part of this work has been directed towards inhibitors of enzymes involved in glycan metabolism, i.e. glycosidases and glycosyltransferases. Inhibition of glycosidases by azasugar mimics of monosaccharides^8–11  and structure-based discovery of influenza sialidase inhibitors^5,12  have been milestones in the field and have both led to marketed drugs. Inhibition of glycosyltransferases has proven harder, but important steps forwards have been made with the discovery of potent inhibitors of O-GlcNAc transferase (OGT).^13,14  Seminal advances in the development of small molecule probes of glycan-processing enzymes have been recently reviewed.³ 

Antagonists of lectin-mediated sugar recognition events have also been discovered. As opposed to sugar-processing enzymes, lectins are proteins that recognize glycans with high specificity, but lack enzymatic activity. They are main mediators of sugar encoded information and are implicated in all processes involving cell–cell communication and pathogen recognition events. In the past, carbohydrates have been mostly disregarded as a class of molecules for drug development and lectins have rarely been exploited for the discovery of novel therapeutic opportunities.^6,11,15,16  The high density of functional groups, the large variety of glycans and their structural and synthetic complexity represent a formidable challenge for the development of antagonists. Additionally, lectins binding sites are shallow and exposed to the solvent, which makes recognition of oligosaccharides an intrinsically low-affinity process, to the point that questions have been raised concerning their druggability.¹⁷  Nonetheless, recent breakthroughs on pan-selectin antagonists¹⁸  and the involvement of major big-pharma players in the field^19,20  have highlighted the potential of glycomimetics as drugs and are attracting increasing attention to lectin antagonists as tools for basic research in glycobiology and for the development of drugs.

Progresses in the development of glycomimetics targeted against sugar binding proteins (lectins) have been comprehensively reviewed very recently.²¹  In this account, rather than once more providing a complete coverage of glycomimetic structures that have successfully been designed as lectin antagonists, we will try to highlight trends and novel research lines that have been developing over the past 5 years in the field.

Specifically, we have identified three topics that have attracted attention and provided new angles to glycomimetic research in recent years:

1. Multivalent enzyme inhibitors

2. Synthesis of oligosaccharide and glycoconjugate mimics with unnatural glycosidic bonds/linkers

3. Use of second site interactions in monosaccharide-anchored antagonists

The multivalent interactions of proteins with cell-surface glycoconjugates have stimulated intense interest in the carbohydrate community towards understanding and exploiting multivalency to antagonize the formation of lectin:sugar complexes.^22,23  In contrast to lectins, that are often polymeric, glycosidases are generally monomeric proteins and, as a consequence, multivalency was largely ignored as a tool to improve the activity of glycosidase inhibitors. However, early successful examples of multivalent enhancement obtained with different sialidases^24–26  stimulated additional research, suggesting that glycosidases and even glycosyltransferases can profit of multivalent effects, with a rapid growth of cases reported in the last 5 years (reviewed in ref. 27).

Clusters of iminosugars were synthesized using deoxynojirimycin (DNJ, 1, Fig. 1) on a variety of scaffolds,¹¹  from trivalent polyethers (6)²⁸  to fullerenes (7),²⁹  calixarenes (10 and 11) and porphyrins (12).³⁰  Inhibition studies performed on panels of glycosidases showed quantifiable inhibitory multivalent effects on various enzymes. Most importantly, these studies showed that multivalency can significantly modify the selectivity of DNJ. This is considered an exciting discovery, since the lack of glycosidase selectivity of iminosugars is a major cause of severe side-effects in clinical applications and a major drawback to the development of therapies based on glycosidase inhibitors.

A rationalization of the multivalent effect in enzyme inhibition is not complete, yet. Increase of local concentration of the inhibitor should clearly be beneficial in most situations.³¹  However, selectivity studies suggest that specific effects must be operative. In the case of Jack-bean α-mannosidase (JBα-Man), multivalent iminosugars were shown to promote the formation of aggregates, allowed by the dimeric nature of the enzyme.³⁰  This mechanism is unavailable to other glycosidases, such as baker's yeast α-glucosidase. This monomeric enzyme has a similar affinity as JBα-Man for DNJ, but it does not respond to multivalency of the inhibitors, possibly because it misses a cross-linking mechanism.

Another puzzling observation concerning selectivity came from the group of Ortiz Mellet and co-workers working with JBα-Man and sp²-iminosugars 4 (1N-oxomethylnojirimycin, 1N-ONJ) and 5 (1N-oxomethylmannojirimycin, 1N-OMJ), the latter being ca. 200 times more potent than the former, at the monovalent level.³²  These authors reported that some specific combinations of multivalent support and linker length (Fig. 1) were able to abrogate the intrinsic difference in inhibition potency of the ligand, yielding multivalent inhibitors of identical activity (compare K_i values for 8 and 9, Fig. 1).

Multivalent iminosugars have also been investigated for pharmacological chaperone therapy of lysosomal storage disorders, a set of diseases that depend on defects of lysosomal glycosidases leading to accumulation of glycosphingolipids.³³  The strategy is based on the conformational stabilization of the defective glycosidase by sub-inhibitory concentrations of reversible competitive inhibitors. The inhibitor stabilizes the protein enough to prevent its premature degradation and thus rescues its catalytic activity. Multivalent DNJ derivatives were found to improve the affinity of DNJ for β-glucocerebrosidase (GCase) in vitro, with relative potency enhancements up to 21 (valency-corrected) with the heptavalent cyclodextrin-based cluster 14 (Fig. 2). However, in cellular studies the monovalent ligand (13) and the heptavalent cluster had very similar chaperoning activity,³³  possibly because of poor membrane permeation by 14. Recently, a more systematic exploration of multivalent DNJ derivatives led to similar conclusions and highlighted a lack of correlation between β-glucocerebrosidase inhibition and chaperoning activity.³⁴ 

Another important field of application of glycosidase inhibitors consists in their use as “correctors” of the endoplasmic reticulum (ER) glycosidases. Reducing the activity of ER glycosidases allows misfolded proteins to escape quality control and degradation and helps to protect defective proteins that are nonetheless functional, once they are trafficked to their correct cellular location. This approach is actively studied to restore the function of misfolded cystic fibrosis transmembrane regulator (CFTR), the main cause of cystic fibrosis.³⁵  Multivalent DNJ derivatives were evaluated as correctors, measuring the ability to restore mature CFTR to its correct location. The best effects were obtained for a trivalent cluster 15³⁶  that gave relative potency enhancements of ca. 300 fold (Fig. 2). However, the effect did not correlate with inhibition of the ER glycosidase.

Summarizing, multivalent presentation of glycosidase inhibiting iminosugars has raised big hopes for the development of novel approaches to the cure of severe diseases. The effect has been observed for a limited number of enzymes, so far, which provides a good opportunity for the development of selective inhibitors. However, more research is required to identify potential targets and to understand the basis of the operating effects.

Many modifications have been introduced in the structure of carbohydrates to generate glycomimetics with improved drug-like characteristics and stability to enzymatic degradation.

The endocyclic oxygen was replaced by a carbon atom (cyclitols or carbasugars),³⁷  a nitrogen atom (imino sugars),⁶  or a sulfur atom (thio sugars).^38,39  A PO group was used to replace the anomeric carbon (phospha sugars).^40,41  Alternatively, the exocyclic oxygen was replaced by the same set of atoms, giving, respectively, C-glycosyl compounds, N-glycosyl compounds, S-glycosides and phostines (Fig. 3). A recent review by Werz et al. covers advances in the field from 2000 to 2010.⁴² 

Fluorosugars have also attracted a great deal of attention as inhibitors of carbohydrate processing enzymes.^43–47  Replacement of the anomeric oxygen by a CHF^43,45  or CF₂^44,46  provides molecular entities with different mimicking abilities than C-glycosyl compounds. Indeed, fluoro-C-pyranosyl rings appear to adopt consistently the classic ⁴C₁ conformation of natural O-glycosides. Non-exo conformations of the anomeric bond are favored by CHF glycosides, while the exo-anomeric effect is restored when the ring oxygen is replaced by a CF₂ group (gem-difluorocarbadisaccharide) (Fig. 3). Development of appropriate synthetic methodology is still needed for full exploration of this class of compounds.

Sulfonamide-bridged oligosaccharides have been recently introduced as glycomimetics by Poulsen and co-workers.⁴⁸  These molecules incorporate a sulfonamide linker in place of the glycosidic bond and are prepared from S-glycosyl thioacetates and amino sugars, as exemplified in Fig. 4. The synthesis is straightforward and structures up to the octasaccharide have been prepared.

Carbohydrate-based sulfonamides have been developed as a new class of carbonic anhydrase inhibitors with good isoform specificity dictated by a cellular permeability profile controlled by the sugar moiety.⁴⁹ 

Among unnatural glycoconjugates, we have explored α-linked glycosyl amides. Natural glycopeptides are invariably β-linked, thus the unnatural, α-linked isomers should be stable to enzymatic hydrolysis. In particular, we have used an α-fucosylamide anchor to design antagonists of P. aeruginosa PA-IIL⁵⁰  and of DC-SIGN (16–19, Fig. 5).^51,52  The synthesis of these mimics relies on a one-pot Staudinger reduction/aza-Wittig sequence described by De Shong.⁵³  The structure of the DC-SIGN antagonist 19 (Fig. 5) was initially designed as a mimic of Lewis^X, a natural ligand of DC-SIGN, and further simplified, following suggestions by NMR and computational studies,⁵⁴  to get ligand 16.⁵¹  Both 19 and 16 were found to have very similar affinity to the natural ligand and to inhibit DC-SIGN binding to highly mannosylated surfaces (IC₅₀ values of 0.35 mM for 19, 0.5 mM for 16, and 0.8 mM for Lewis^X). A radical reduction of the structure complexity, as in the fucosyl-β-alanylamide 17, still yielded a molecule with a measurable affinity for the protein (IC₅₀ 0.9 mM). Remarkably, gold nanoparticles 20 functionalized with 17 bound to cellular DC-SIGN and induced internalization as effectively as similar particles coated with comparable amounts of the Lewis^X oligosaccharide.⁵⁵  They were found to be neutral towards dendritic cell maturation and IL-10 production, thus they may have interesting applications as tools for targeted imaging and delivery.

A limited number of approaches are available for the selective synthesis of α-glycosyl amides. We have recently described a new method based on the traceless Staudinger ligation of glycosyl azides with functionalized phosphines bearing an acylating agent (21⁵⁶  and 22,⁵⁷  Fig. 6). The method is particularly effective for furanoses, as it offers a stereodivergent way of synthesizing glycofuranosyl amides of either anomeric configuration starting from the same azide. The stereochemical outcome of the process is controlled through the absence/presence of acetyl protecting groups on the sugar (Fig. 6a).⁵⁸  This approach was applied to an easy synthesis of α-N-ribosyl-asparagine (24) and α-N-ribosyl-glutamine (25) building blocks starting from 5-tert-butyldiphenylsilyl-β-d-ribofuranosyl azide 23 (Fig. 6b) and phosphine 22.⁵⁹  The N-glycosyl aminoacids (24 and 25) were produced in good yields as pure α-anomers, suitably protected on the sugar ring for peptide synthesis.

In the same year, van de Marel, Filippov and co-workers described the synthesis of the same building blocks by direct glycosylation of the carboxamide side chains of asparagine and glutamine with ribofuranosyl N-phenyltrifluoroacetimidates 26 (Fig. 7).⁶⁰ 

N-glycosyl triazoles were introduced in 2006 by Dondoni and coworkers as isosteric replacements of glycosidic bonds.⁶¹  More recently, they have been shown to possess in vitro metabolic stability and plasma stability.^62,63  Direct synthesis of triazole-linked glycoconjugates from unprotected sugars was investigated by the groups of A. Fairbanks and M. Brimble.⁶⁴  These authors developed a one-pot sequence that generates glycosyl azides from unprotected monosaccharides in D₂O/CH₃CN mixtures and used it in situ for the synthesis of glycoconjugates by copper catalyzed azide–alkyne cycloaddition (CuAAC) reaction (Fig. 8). 2-azido-1,3-dimethylimidazolinium hexafluorophosphate 27 was used as both the activating agent of the anomeric group and the azide source. The configuration of the resulting glycosyl azide is β, except for mannose, which gives exclusively the α anomer. Click reaction occurs smoothly with a variety of alkynes, including oligosaccharides and peptides. Thus the method allows the easy synthesis of mimics of oligosaccharides and glycopeptides by direct conjugation of reducing sugars to alkynes under aqueous conditions in a two-step one-pot transformation, and it is totally steroselective.

Sugar azides have also been used by Volonterio et al. in a multicomponent domino process⁶⁵  that allows the combinatorial synthesis of multivalent glycoconjugates (Fig. 9). Staudinger reduction of glycosyl azide 28 in the presence of isocyanates or isothiocyanates 29 yields sugar-carbodiimides 30 that react with fumarates 31 in a domino process affording glyco-idantoins 32 or glyco-peptides (33–35), in the presence of appropriate nucleophiles. Using a combination of glycosylated reagents (sugar azides, sugar isothiocyanates and sugar amines), the authors prepared a group of 34 divalent and trivalent glycomimetics, including aminoglycoside conjugates. The approach appears to be very general, to work efficiently with a number of different substrates and, coupled with appropriate screening procedures, should be promising for the discovery of lectin antagonists.

A similar approach was adopted by Overkleeft and co-workers to obtain a library of monovalent C-glycosyl derivatives of imino sugars (IUPAC not recommended term: aza-C-glycosides) through a domino Staudinger/aza-Witting/Ugi three-component reaction (Fig. 10).⁶⁶  Sugar-derived azido-pentanals and azido-hexanals 36 were treated with PMe₃ to afford imine 37, as the aza-Wittig product. A 3-component Ugi reaction of 37 with isonitriles and 4-pentenoic acid yields 38, as a mixture of separable diastereoisomers. Upon removal of the pentenyl group, further manipulation and diversification of the aza sugar allowed the synthesis of 62 compounds from seven sub-libraries.

Elimination or substitution of hydroxy groups by other functional groups, introduction of aliphatic or aromatic substituents in the structure of the glycomimetic or use of different ring sizes (e.g. polyhydroxylated azepanes,⁶⁷  Fig. 11) are other possibilities that have been explored to mimic carbohydrates. Sialic acids with single or multiple and combined modifications of positions 2, 3, 4, 5 and 9 of the sialic acid core have been studied as ligands of human Siglec-2 (CD22) for immunoglycotherapy (Fig. 11).^68,69 

Glycomimetics have also been generated by modifications of glycoconjugates, for example by truncation of the native structure, or by replacement of oligosaccharide fragments with appropriate linkers. We used this approach in the search of non-hydrolyzable mimics of the GM1 oligosaccharide (GM1os 46, Fig. 12a).⁷⁰  The mimics were selected from a library of C-galactosyl groups linked to a N-sialyltriazole residue (47, Fig. 12a) using weak affinity chromatography with immobilized cholera toxin B, a known high-affinity receptor of GM1. Affinity could be enhanced by up to one or two orders of magnitude over the individual sugar residues. A further simplification was reported recently by Robina, Moreno-Vargas and co-workers⁷¹  who used S-galactosides and replaced the sialic acid residue of GM1 with a polyhydroxyalkylfuran moiety (48, Fig. 12b). This project actually led to the discovery of a non glycosylated polyhydroxyalkylfuran ligand 49 which shows interesting activity as a small molecule cholera toxin antagonist.

So far, the most successful linkers for oligosaccharide replacement strategy have been conformationally restricted scaffolds, designed to reproduce the 3D features of the ligands in their bioactive conformation. This approach was pioneered by Ernst in his work on mimics of sialyl Lewis^X (sLe^X) as selectin antagonists in the late ‘90s⁷²  and culminated in 2010 in the discovery of GMI-1070 (Rivipansel, 50, Fig. 13),¹⁸  a pan-selectin antagonists that is currently undergoing clinical trials for treatment of vaso-occlusive crisis in people with sickle cell disease. In these structures, a trans diequatorial cyclohexanediol replaces sLe^X native GlcNAc, while maintaining a stacked conformation of the Gal and Fuc rings, which is required for lectin interaction. Replacement of sLe^X sialic acid residue with an alkyl (S)-lactic ester completes the structural simplification of the natural ligand.

In a recent work, the Ernst group dissected the role of the cyclohexane core structure pre-organization in a series of E-selectin antagonists using Surface Plasmon Resonance (SPR) and Saturation Transfer Difference NMR experiments.⁷³  They showed that addition of hydrophobic substituents on the cycle as in 51 (Fig. 13) improved the affinity not by adding interactions with the receptor, but by restricting the conformational freedom of the structures in ways that increased their similarity with the native ligand.

Recently, tartaric esters have been introduced as conformationally constrained acyclic scaffolds for the design of selectin antagonists.^74,75  The conformational bias is dictated by minimization of dipole–dipole interactions between the carboxy groups and by the gauche effect of the tartrate diol. Intriguingly, the configuration of the lactic acid side-chain of the most active compounds (e.g. 52, Fig. 13) is the opposite one, relative to GMI-1070. The antagonists generated with this approach were found to be highly potent both by in vitro and in vivo assays.

Following Ernst's lead, in 2000 we introduced two enantiomerically pure, conformationally stable cyclohexanediols 53 and 54 designed to replicate carbohydrate branching motifs that incorporate one or more axial substituents and used them to synthesize mimics of the GM1os and of linear oligomannosides⁷  (Fig. 14).

The pseudo-dimannoside 55 and pseudo-trimannoside 56 (Fig. 15) generated from 54 as mimics of Manα(1,2)Man and Manα(1,2)Manα(1,6)Man, respectively, where used in multimeric format as antagonists of DC-SIGN mediated viral infections by HIV and Ebola virus. DC-SIGN (dendritic cell-specific ICAM-3 grabbing non-integrin) is a tetrameric calcium dependent (C-type) lectin, expressed by immature dendritic cells (DCs), which specifically recognizes highly-glycosylated structures displayed at the surface of several pathogens.^76–79  Recognition by DC-SIGN is known to play a key role in HIV transmission and is involved in a number of other viral infections, so that it is currently regarded as an interesting target for the design of anti-viral agents.²²  A tetravalent presentation of the pseudo-trisaccharide 56 in dendron 58, based on bis-hydroxymethylpropionic acid monomers (Fig. 15), was shown to inhibit trans infection of T lymphocytes by DC-SIGN expressing B-cells, which had been pre-incubated with HIV in the presence of 58. Infection was abrogated almost totally at 100 μM concentration of the dendron, and an IC₅₀ ca. 10 µM could be estimated.⁸⁰  Dendron 58 was also found to block HIV-1 infection of human cervical tissue, under conditions mimicking compromised epithelial integrity.⁸¹ 

This dendron and higher valency constructs based on similar scaffolds (such as 60) were also tested in a DC-SIGN dependent Ebola infection model based on pseudotiped viral particles.⁸²  In these experiments, the tetravalent systems 57 and 58 were active in the low micromolar range, and the multivalent systems G₃(pseudosugar)₃₂ 59 and 60 showed a very strong inhibition effect with IC₅₀ in the nanomolar range. Surprisingly, relatively small differences were observed between equivalent constructs obtained from the pseudo-di 55 and the pseudo-trisaccharide 56, even if, at the monovalent level, 56 was found to be an order of magnitude more active than 55 by SPR competition experiments.⁸⁰  Recently, we have been able to explain this puzzling observation by showing, with a number of biophysical techniques, that 56 as a monovalent ligand is able to cluster DC-SIGN tetramers, leading to an artificially overestimated inhibitory potency in the SPR assay.⁸³  Additional optimization of the pseudo-disaccharide 55 as DC-SIGN ligand is described in the following section.

The identification of unnatural inhibitors of lectin–sugar recognition events has been approached most often by reducing the complexity of oligosaccharide ligands to mono- or disaccharide binding determinants working as the lectin anchor. These units are linked to aglycones that can be designed to reproduce the 3D structure of the native oligosaccharide as discussed above, and/or provide additional functional units for interaction with the lectin in the proximity of the primary binding site. In both cases, the aglycone is also used to impart to the molecule some pharmacologically favorable properties, such as improved lipophilicity or resistance to hydrolytic enzymes.

The pioneering study of Sharon on the use of mannosides as antagonists of bacterial adhesion mediated by FimH,⁸⁴  together with structural information obtained from X-ray studies⁸⁵  led to the rational design and optimization of high affinity antagonists directed against uropathogenic E. coli.^86–90  All these compounds are targeted to FimH binding site by their mannose component and carry extended aromatic aglycones that establish high affinity interaction with lipophilic residues lining the rim of the sugar-binding pocket of the protein (61–63, Fig. 16). These lipophilic aglycones have also been shown to tune the selectivity of the antagonists against a panel of human mannose-binding lectins, a key element for in vivo application development.⁹¹ 

We recently used a similar approach to optimize the activity and selectivity of the pseudo-dimannoside antagonist of DC-SIGN 55. A library of amide derivatives was prepared following the approach shown in Fig. 17.⁹²  SPR inhibition tests showed that tertiary amides are no longer recognized by the lectin, but a group of bis-benzylamido derivatives 64 was found to be more active by a factor of 3–4 than the parent pseudo-mannobioside 55 and displayed good selectivity against Langerin, a C-type lectin similar to DC-SIGN, but protective against HIV infection. NMR studies showed relatively high saturation transfer for the aromatic amide moieties in the STD spectrum, suggesting a close proximity between the aromatic groups of the ligand and the binding site of DC-SIGN.

The small affinity improvement observed can be amplified by multivalent presentation,⁹³  as clearly shown by the IC₅₀ values of similar dendrimers of equal valency carrying different monovalent ligands (Fig. 18). IC₅₀ in the low micromolar range are rapidly achieved using the most powerful ligand 64a; other mannosylated and pseudo-mannosylated materials are less effective, even when the sugar is presented in higher valency on the dendrimers. This is one of the few cases reported in the literature where the relationship between the intrinsic affinity of monovalent ligands for a lectin target and the level of affinity enhancements observed in their polyvalent presentation has been studied. The results contrast sharply with the observations of Ortiz Mellet described above for clustered glycosidase inhibitors.³²  Her work clearly shows that clustering a more potent inhibitor is not automatically rewarded, but the results will depend on the operating binding mode. Whether these contrasting data point to a fundamental difference between lectins and glycosidases will deserve further attention in the future.

The hexavalent presentation of 64a (65, Fig. 19) was shown to block both DC-SIGN mediated uptake of Dengue Virus by Raji cells and HIV trans-infection of T cells at low μM concentration.⁹³  Inclusion of a rigid spacer at the dendrimer core yielded elongated structures that displayed significantly higher activity in infection studies, reaching IC₅₀ in the nM range with 66 (Fig. 19), which was found to be 40 times more active than 65 in HIV trans-infection studies.⁹⁴  The activity of 66 appears to stem from a combination of three elements: an effective monovalent ligand, a rigid core of appropriate length, allowing to bridge two adjacent binding sites of the lectin, and two trivalent dendrons, providing increased local density at each site.

Other research programs making use of second site interactions have been developed for the discovery of ligands of the human asialoglycoprotein receptor (ASGPR),⁹⁵  P. aeruginosa LecB⁹⁶  and Galectin 3.^97,98 

ASGPR is a galactose binding receptor expressed by hepatocytes and is actively studied for hepatic delivery of drugs and diagnostic probes. Early work for the discovery of active ligands focused on GalNAc and reported various replacements of the amide group showing little effect on the activity. Recently, combinatorial modification of all the positions of galactosamine except the ASGPR anchoring C3/C4-diol was performed, yielding one-order of magnitude improvements over the starting scaffold (Fig. 20).⁹⁵ 

Titz and co-workers, instead, applied a structure-based approach to the optimization of ligands for LecB, a crucial factor for maintenance of P. aeruginosa biofilm.^96,99  The authors addressed a cleft on the lectin surface that is adjacent to the C6 hydroxy group of O-methyl-mannoside in the crystal structure of its LecB complex. They synthesized a library of 6-modified mannosides, identifying two low micromolar ligands, 67 and 68, with different binding modes (Fig. 21). Despite the affinity improvement over methyl-mannoside (as measured by ITC), neither of these molecules displayed improved ability to disrupt biofilm and reduce bacterial adhesion.

Nilsson and co-workers have used second site interactions for the structure-based rational design of potent Galectin 3 (Gal3) antagonists.^97,98  The X-ray structure of N-acetyllactosamine-Gal3¹⁰⁰  shows an extended binding groove that could be probed with different aromatic fragments connected to galactosides or thiodigalactosides. This project led to the discovery of TD139 (69, Fig. 22) a modulator of the VEGF/VEGFR-2 signaling pathway in angiogenesis, currently in clinical trial.¹⁰¹ 

Optimization of E-selectin antagonists was recently carried out by Ernst using fragment-based discovery techniques to select ligands able to bind in a second site near the sLe^X mimic binding site.¹⁰²  Fragments were screened for E-selectin binding by spin-lock filtered NMR experiments. The hits were re-tested in the presence of the first site ligand, modified with a spin-label probe. The fragments binding in the vicinity of the spin-label were identified by paramagnetic relaxation enhancement spectroscopy and then connected to the sLe^X mimic through flexible linkers of variable length. The extended antagonists so obtained (e.g. 70, Fig. 23) were tested by SPR for interaction with E-selectin and exhibited strong interaction (in the nanomolar range) and improved binding kinetics, as expected from a two-point ligand.

The quest for glycomimetic molecules capable of antagonizing the native ligands of lectins and sugar-processing enzymes is actively pursued by a number of groups and some exciting advances have occurred during the past five years. Some new molecules have reached the market, others are currently in clinical trial. Fundamental improvements have also been achieved in our ability to understand and use the design principles leading to efficient mimicry of complex carbohydrates. Much work remains to be done in this area to improve both synthetic procedures and design tools to finally reach the goal of obtaining mimetics optimal in size, shape and valency and finely tuned to the supramolecular architecture of individual lectins.