2.3 Multivalent Lectin Ligands
2.3.4 Ligands Based on Nanomaterials
Multivalent Lectin Ligands
that facilitate detection of molecular interactions. The nature of the material can be manifold and examples reach from metals, polymers, dendrimers to carbon and silica nanostructures.[105] An early report by Penadés and coworkers introduced gold nanoparticles which are decorated with different mono- and oligosaccharides.[106] The saccharides were tethered to oligoethylene glycol spacers with a terminal thiol that dimerizes the molecules upon oxidation to the corresponding disulfides. Treatment of these with tetrachloroauric acid under reducing conditions yielded in the corresponding nanoparticles.
They were shown to have potential as antiadhesives in tumor metastasis.[107] Carbon nanoparticles are often made of graphene derivatives as fullerenes[108] (Figure 7, 11) or carbon nanotubes[109]. A ligand for WGA was based on CdSe/ZnS quantum dots which were decorated with GlcNAc residues via a linker with a terminal thiol (Figure 7, 12).[30] Quantum dots are usually made of semi-conductive materials and their electrons can only adapt discrete energy levels. The applications of such glycosylated nanomaterials are manifold, especially for biosensing of lectins in vitro and in vivo.[105]
Figure 7 A ligand based on fullerene 11[108] and carbohydrate modified quantum dots 12[30]. 2.3.5 Ligands Based on Self-Assembly or Combinatorial Chemistry
If no structural information about a target lectin is available, it can be necessary to screen many ligands in order to find hits that fit the geometry of the lectin and bind with high affinity. Combinatorial chemistry is an ideal tool to create big, structurally diverse libraries of related compounds. One approach is the use of dynamic combinatorial chemistry.[110] Here, a library is created by reacting several different compounds with each other via a reversible reaction. All products that are formed are in equilibrium.
Then an external force is applied to the system. This can be for example a physical parameter as temperature or pressure, or a binding partner for the compounds in the library. The library adapts to the new conditions and certain library members are amplified while the content of others decreases depending on the conditions. In case of lectin binding as a selection criterion, the best binders can be amplified. Requirements that have to be met are reversible reactions for the creation of the library and the possibility to detect the amplified species, for example by “freezing” of the library at a certain state, transforming the members into stable derivatives, stopping the exchange by changing the conditions[111], or deconvolution techniques[112]. Ramström and Lehn used disulfide exchange as reversible reaction to create dimeric ligands for Con A using different monosaccharides (Scheme 1).[111]
Scheme 1 Schematic representation of a dynamic combinatorial library based on disulfide exchange. Divalent Gal species 13 and divalent Man species 14 react via disulfide exchange to the mixed compound 15.[111]
The library solution was incubated with sepharose beads conjugated to Con A. The bound species from the sepharose beads could be eluted at acidic pH which also stopped the disulfide exchange and the species containing two mannose residues 14 were identified as the most dominant which is in line with the sugar specificity of Con A.
Another way of using combinatorial chemistry is the creation of “static” libraries. They can be created by various synthetic strategies. Wittmann and Seeberger used split-mix synthesis[113] for the creation of a diverse library of cyclic peptides by solid phase peptide synthesis (SPPS). The peptides served as scaffolds for the attachment of carbohydrates.[15] The idea behind the cyclic structure of the peptides was to create a conformationally restricted and preorganized ligand, leading to a small enthalpic penalty upon binding to the protein. The optimal peptide sequence and position of the carbohydrates on the peptide should be found by exploring a big number of different ligands. For the creation of the library in each coupling step a set of different amino acids was used. The resin for the SPPS was distributed on several reaction vessels according to the number of different amino acids used. After the coupling, the resin portions were combined, mixed and split again for the next coupling cycle. That way a library of about 20 000 members could be created (Figure 8).[20] In the synthesis amino acids with protected side chain amino groups were used that were conjugated to GlcNAc residues after the peptide synthesis.
Figure 8 Design of the glycopeptide library by Wittmann and Seeberger.[20]
The resin beads were then incubated with biotin labeled WGA and after washing, the beads which bound most WGA were identified by a color reaction catalyzed by an alkaline phosphatase conjugated to an anti-biotin antibody. The glycopeptides were released from the resin beads and identified. Compound
Multivalent Lectin Ligands
(depending on the assay, see Chapter 2.4.2)[114] affinity enhancement over the monovalent ligand GlcNAc.
Figure 9 One of the most potent ligands for WGA found in the screening of a 20 000-membered library.[20]
Self-assembling systems are also a valuable tool for the creation of multivalent ligands. Monomers that interact in a noncovalent, defined way can build up large supramolecular structures of high valency comparable to ligands based on polymers. Brunsveld and co-workers developed disc-shaped building blocks 17 with an aromatic core (Figure 10 A).[92] The periphery of the disc was decorated with mannose.
The discs were able to build columnar structures by stacking mediated through -interactions (Figure 10 B).
Figure 10 (A) Monomeric building block 17 for the (B) self-assembly of columnar multivalent ligands.[92]
The binding affinities were measured by an ELLA and proved to be three times higher for the polymer formed from the mono-mannose disc compared to the -methyl mannoside.
2.3.6 Multivalent Ligands for WGA
The lectin WGA possesses eight carbohydrate binding sites making it a target often used to study multivalent carbohydrate lectin interactions. Early multivalent ligands for WGA by Zanini and Roy were based on a peptidic dendrimer with a valency of up to eight (Figure 11). Compound 18 with highest valency showed a 20-fold enhancement of binding compared to the monovalent analogue as determined by an enzyme-linked lectin assay (ELLA).[115] Calixarene 19 had a 312-fold inhibitory potency over monovalent GlcNAc, derived from a hemagglutination assay. In the study, ligands of varying linker lengths between the calixarenes and the saccharides were tested. Ligands with shorter linkers showed lower affinities.[21] As described above, Wittmann and Seeberger employed combinatorial chemistry to synthesize a library of 20 000 cyclic glycopeptides[15] and screened it for affinity to WGA.[20] One of the hits (16, Figure 9) was further optimized to give 20 (Figure 11), which had an affinity 25 500-fold higher as GlcNAc.[11] Lactotriaose dendrimers 21 were also effective binders for WGA.[28] Masaka et al.
prepared tetravalent glycoclusters with 22 being the most potent one. They also performed precipitation studies showing that their ligands were able to precipitate WGA. At high concentrations of ligand, the aggregates dissolved again.[116] Beckmann et al. used click chemistry to tether N,N-diacetylchitobiose to di- and trivalent core structures with compound 23 being among the best ligands for WGA at that time.[117] Soon after, Fiore et al. presented the first example of a WGA inhibitor with nanomolar affinity in terms of IC50 and also Kd value which was based on a cyclic peptide scaffold.[12] They also observed precipitation and attributed the high affinity partly to aggregation. The same group developed ligands on the base of octasilsesquioxanes (24, Figure 11)[118] and dendrimers with a carbohydrate number of up to 48 carbohydrate residues[29]. Currently compound 24 and the 48-valent dendrimer belong together with compound 20 to the best ligands for WGA.
Schwefel et al. performed a thorough structural study of a series of divalent ligands and cyclic tetravalent glycopeptides binding to WGA.[11] They could show that divalent ligand 25 (Figure 12 A) was able to bridge two adjacent binding sites and four ligands were able to occupy all eight binding sites in the crystal structure of 25 with WGA (Figure 12 B). Ligand 25 showed a 400 times higher affinity in an ELLA assay compared to GlcNAc. Structurally related compound 26 (Figure 12 A) bound even 2350 times better than the monosaccharide. Also cyclic peptide 20 (Figure 13 A) could be crystallized together with WGA (Figure 13 B). Here the ligand also bridges two adjacent binding sites but the structure of the ligand is only partly resolved in the crystal structure (Figure 13 A, shaded gray) and it cannot be seen if the carbohydrates at D-Dab2 (D-diamino butanoic acid) and D-Dab7 are participating in the binding.
Multivalent Lectin Ligands
Figure 11 Multivalent ligands for WGA with scaffolds based on peptides (18, 20), calixarenes (19) ethylene glycol (22), ammonia (23) and silsesquioxanes (24).
Figure 12 (A) Divalent ligands 25 and 26 for WGA. (B) WGA in complex with ligand 25 (PDB ID: 2X52). The ligand is represented by a stick model (black) and the protein is shown as surface representation (gray). Secondary binding sites are located at the back of the protein and are not shown here.
A B
Figure 13 (A) Cyclic peptide 20 and (B) crystal structure of its complex with WGA (PDB ID: 2X3T). Only the gray shaded substructure of 20 is resolved in the crystal structure and shown as stick model (black). The protein is shown as surface representation (gray). D-Dab = D-diaminobutanoic acid
A B
Methods for the Investigation of Multivalent Interactions.
2.4 Methods for the Investigation of Multivalent Interactions.
For the investigation of multivalent interactions many different methods have been applied.[70] A lot of them yield in binding affinities in terms of association constants or IC50 values (defined as the ligand concentration that reduces the initially detected signal of the binding assay to 50 %). Examples are:
enzyme-linked lectin assay (ELLA)[20, 114-115, 119], fluorescence spectroscopy[30, 120], isothermal titration calorimetry (ITC)[121], microarrays[27], microscale thermophoresis[122], quartz crystal microbalance (QCM)[123], surface plasmon resonance (SPR)[124], or total internal reflection spectroscopy[125]. Information about the structure of complexes formed by multivalent interactions can be gained from analytical ultracentrifugation[83, 120], crystallography[11, 19, 82, 126-130], dynamic light scattering (DLS)[21, 30,
131-132], transmission electron microscopy[30] or electron paramagnetic resonance (EPR) spectroscopy[133]. Often combinations of the above techniques are needed to elucidate the binding mechanisms that lead to the observed multivalent effects. In the following the methods used in this work are discussed in more detail.
2.4.1 Isothermal Titration Calorimetry
Isothermal titration calorimetry is a technique mainly used for the analysis of binding events between two binding partners (that may interact also in stoichiometries other than one to one). It allows the determination of the association constant Ka, the binding enthalpy H, and the stoichiometry n of the binding event in one single experiment. Neither of the binding partners has to be labeled and the measurement takes place in solution. This make ITC an ideal method to study interactions of biomolecules.[134-135] The instrument consists of two cells: a reference cell and the sample cell which both are heated with a constant power supply to maintain a steady, constant temperature (Figure 14 A).
The reference cell is filled with the solvent that is used for the interaction and the sample cell is filled with a solution of one binding partner. The syringe is filled with the second binding partner dissolved in the same solvent and this solution is gradually titrated into the sample cell. Upon each addition, heat is released into or absorbed from the solution depending on whether the interaction is exothermic or endothermic. The sample cell warms up or cools down and the heater has to supply more or less power to maintain the temperature of the sample cell constant to the reference cell. This power provided to the sample cell is plotted as µcal s–1 against the time (Figure 14 B). The area under the resulting peaks for each injection is plotted against the ratio of the binding partners (Figure 14 C). This plot is used for the fitting of the binding curve with Ka, H and n as adjustable parameters.
Figure 14 (A) Schematic representation of isothermal titration calorimeter. (B) Raw data, power used to heat sample cell plotted against time. (C) Integrated peaks and fitting curve.
The fitting equation can be derived as follows.[135] For an interaction between ligand (L) and receptor (R) that form a complex LR
the equilibrium constant is
𝐾𝑎= [𝐿𝑅]
[𝐿][𝑅]
and the total concentrations of ligand and receptor are
[𝐿]𝑡𝑜𝑡 = [𝐿] + [𝐿𝑅] and [𝑅]𝑡𝑜𝑡= [𝑅] + [𝐿𝑅] = [𝐿𝑅] + [𝐿𝑅]
𝐾𝑎[𝐿]. After resolving for [L] and substitution, the quadratic equation
[𝐿𝑅]2+ [𝐿𝑅] (−[𝑅]𝑡𝑜𝑡− [𝐿]𝑡𝑜𝑡− 1
𝑘𝑎) + [𝑅]𝑡𝑜𝑡[𝐿]𝑡𝑜𝑡= 0 results that can be resolved to
[𝐿𝑅] =−𝑏−(𝑏2−4𝑐)
1 2
2 with 𝑏 = −[𝑅]𝑡𝑜𝑡− [𝐿]𝑡𝑜𝑡− 1
𝐾𝑎 and 𝑐 = [𝑅]𝑡𝑜𝑡[𝐿]𝑡𝑜𝑡.
Differentiation for [L]tot gives
Reference cell
Methods for the Investigation of Multivalent Interactions.
The heat change that results from the titration is proportional to the change in concentration of the complex [LR]:
𝑑𝑄 = 𝑑[𝐿𝑅] ∙ Δ𝐻𝑜∙ 𝑉𝑂
Together with the expression above, the following term is obtained:
𝑑𝑄
The data gained from the experiment is the differential heat 𝑑[𝐿]𝑑𝑄
𝑡𝑜𝑡 which is connected to the binding enthalpy Ho and the association constant Ka as described by the equation before. All other variables are known. These parameters are determined by least-squares fitting that means the sum of squared residuals is minimized by variation of the parameters Ka and Ho. A residual is defined as “the difference between a measured data point and its calculated counterpart”.[136]
ITC has often been used to characterize lectin carbohydrate interactions[18-19, 24, 102, 104, 121, 137-153]. Bains et al. were first to investigate the binding properties of WGA to its different ligands.[64] The monosaccharide GlcNAc had a Kd value of 2.5 mM and the affinity rose up to 50 µM for the pentasaccharide. The binding enthalpy for GlcNAc was stated as 7 kcal mol–1. In general, the investigation of monosaccharides with ITC is problematic because the binding affinities are generally low. That would either call for samples with very high concentrations of protein and ligand or, because high concentrations can be difficult to achieve, the evaluation has to be performed with binding curves that lack a sigmoidal shape. This may result in inaccurate values for H and n.[154]
Brewer and coworkers investigated the binding of multivalent ligands to Con A and Dioclea grandiflora lectin (DGL) thoroughly.[121, 151-152] They reported increased affinities compared to the monosaccharide as it is usually found for multivalent ligands and the stoichiometries showed dependency on the valency.
The binding enthalpy H was proportional to the number of carbohydrate residues on the ligand while the enthalpy TS did not behave proportionally. The values were much lower than expected for a proportional behavior leading to only moderate binding affinities. This was attributed to the big distances between the binding sites on Con A and DGL that did not allow a chelating binding mode for the ligands used in the study. Because of that behavior G and thus the affinity was low for the ligands that could not bridge adjacent binding sites. The same group reported negative cooperativity for their crosslinking ligands[151] and decreasing affinity constants for individual carbohydrate residues of the ligands upon sequential binding using reverse titrations with the macromolecule being titrated to the ligand.[152] In studies by Rao et al. with systems that allow chelating binding modes also TS was proportional to the number of individual epitopes.[155]
ITC data for multivalent ligands binding to WGA is rare. Fiore et al. investigated a series of cyclic peptides using ELLA and ITC and identified neoglycopeptide 27 (Figure 15) as the best binder with a dissociation constant of 9 nM.[12] No other examples of ITC data for multivalent WGA ligands are available until now.
Figure 15 Tetravalent cyclic peptide 27 with nanomolar affinity for WGA.[12]
2.4.2 Enzyme-Linked Lectin Assay
The enzyme-linked lectin assay is a method to determine relative affinities of carbohydrate ligands binding to lectins and was introduced by Goldstein and coworkers.[119] The principle is similar to the well-known enzyme-linked immunosorbent assay (ELISA). A microtiter plate is coated with a saccharide that binds to the lectin of interest. For the coating different methodologies exist. The first examples used glycoproteins[119, 156], others used porcine stomach mucin (PSM)[20, 115], or glycopolymers[12] that bind non-covalently to the surface of the microtiter plate. Maierhofer et al.
developed an assay which uses amino functionalized microtiter plates that are coated with monosaccharides using covalent immobilization.[114] The amino groups on the microtiter plate are coupled to 1,4-phenylene diisothiocyanate 28 to form a thiourea derivative (Scheme 2). The second isothiocyanate group is then reacted with amine 29 that is connected via a linker to the reference ligand, in this case GlcNAc.
Scheme 2 Covalent functionalization of microtiter plate for ELLA.
The lectin that is to be investigated bears a reporter group which is needed for detection. This can be an enzyme which catalyzes for example a color reaction. Here, it is a horseradish peroxidase which oxidizes 2,2'-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) to a green product. The lectin-reporter
Methods for the Investigation of Multivalent Interactions.
the surface and the remaining protein is washed away. The amount of surface bound lectin is then determined by the intensity of the color produced by the reporter group (Scheme 3 C). This results in inhibition curves that yield in IC50 values. The IC50 value is the concentration of ligand at which 50 % of the binding of the lectin to the microtiter plate is inhibited by the ligand. The IC50 values are relative values and strongly depend on the assay. For example, the material of the microtiter plate, the density of carbohydrates on the surface, and the method used for immobilization influences the values.
Wittmann and coworkers determined the IC50 values for tetravalent cyclic glycopeptide 16 (Figure 9) using an ELLA with either a PSM coated surface[20] or a covalently coated surface[114]. The PSM assay gave an IC50 of 380 µM and the covalent assay yielded in 16 µM. Thus, it is difficult to compare the results derived from different ELLA assays.
Scheme 3 ELLA assay: (A) Preincubation of labeled lectin with ligand; (B) incubation of lectin-ligand mixture with saccharide coated microtiter plate; (C) visualization of protein bound to surface with enzyme catalyzed color reaction.
2.4.3 Dynamic Light Scattering
Dynamic light scattering (DLS) is used to determine the hydrodynamic radii of species in solution. These can be biologic macromolecules as DNA or proteins, polymers, nanoparticles, or other supramolecular assemblies. The sample containing the particles of interest is irradiated with laser light and the scattered light is detected at a certain angle (Figure 16 A). The light is scattered by the particles in solution which are all at different positions which leads to interference. The Brownian motion of the particles changes the positions of the particles constantly which leads to fluctuations in the intensity of the scattered light.
Figure 16 (A) Basic experimental setup for a DLS experiment, (B) correlation function, (C) intensity distribution.
ABTS ABTS∙+ ligand
The signal is analyzed using an autocorrelation function (Figure 16 B). The speed of the particles is dependent on the diffusion coefficient D. This is related to the hydrodynamic radius r via the Stokes-Einstein equation
𝐷 = 𝑘𝐵𝑇 6𝜋𝜂𝑟
with Boltzmann’s constant kB and the dynamic viscosity η.[157] The analysis of the correlation function yields in an intensity distribution of the radii of the species in solution (Figure 16 C). The particles are usually treated as spheres and the obtained radii refer to spherical particles that have the same diffusion properties as the original particles. Depending on the shape of the particle, the obtained radii may deviate from the actual size of the species. The method has often been used in the context of lectins and multivalency to detect the aggregation of multivalent ligands with lectins[25, 82-83, 130, 153, 158-159] or to study the interactions of nanomaterials with lectins[160]. The advantage of the method is, that it is non-invasive, fast, and uses only a small amount of sample.
2.4.4 Electron Paramagnetic Resonance Spectroscopy
Electron paramagnetic resonance (EPR) spectroscopy is related to nuclear magnetic resonance spectroscopy but observes unpaired electron spins instead of nuclear spins. Unpaired electrons are introduced into the system of interest in the form of spin labels, in organic molecules often as stable nitroxide radicals. The double electron-electron resonance (DEER) or pulsed electron double resonance (PELDOR) technique allows the measurement of distances between electron spins in the range of 2 to 10 nm utilizing the magnetic dipolar coupling between two unpaired electron spins.[161-162] It has been especially useful in a biological context for the structural investigation of proteins[163-168]. The currently most commonly used technique is the four-pulse DEER experiment (Figure 17).
Figure 17 Four pulse sequence of DEER measurement.[161]
The DEER technique separates the magnetic dipole-dipole interaction of pairs of electron spins from other contributions to the EPR signal. An example for a resulting EPR spectrum is shown in Figure 18
Methods for the Investigation of Multivalent Interactions.
denominated spins A and B as shown in Figure 18 B. Due to their different Larmor frequencies (corresponding to the different spectral positions), spins A and B can be addressed individually by two different microwave frequencies 1 and 2 as shown in Figure 17. The first /2 pulse at the resonance frequency 1 of A turns the magnetization into the xy-plane and evolves over 1under the influence of the magnetic dipole-dipole interaction with spin B and the resonance offset of the A spins.[169] Then a
pulse creates a spin echo (dotted) at 21. A second pulse after 21+2 refocuses the A spins. At the
pulse creates a spin echo (dotted) at 21. A second pulse after 21+2 refocuses the A spins. At the