A Remote Secondary Binding Pocket Promotes Heteromultivalent Targeting of DC-SIGN

Dendritic cells (DC) are antigen-presenting cells coordinating the interplay of the innate and the adaptive immune response. The endocytic C-type lectin receptors DC-SIGN and Langerin display expression profiles restricted to distinct DC subtypes and have emerged as prime targets for next-generation immunotherapies and anti-infectives. Using heteromultivalent liposomes copresenting mannosides bearing aromatic aglycones with natural glycan ligands, we serendipitously discovered striking cooperativity effects for DC-SIGN+ but not for Langerin+ cell lines. Mechanistic investigations combining NMR spectroscopy with molecular docking and molecular dynamics simulations led to the identification of a secondary binding pocket for the glycomimetics. This pocket, located remotely of DC-SIGN’s carbohydrate bindings site, can be leveraged by heteromultivalent avidity enhancement. We further present preliminary evidence that the aglycone allosterically activates glycan recognition and thereby contributes to DC-SIGN-specific cell targeting. Our findings have important implications for both translational and basic glycoscience, showcasing heteromultivalent targeting of DCs to improve specificity and supporting potential allosteric regulation of DC-SIGN and CLRs in general.

interactions formed by the aromatic substituents e.g. via a potential cation-π bond with K313.
The estimated KI values for selected mannosides were subsequently reproduced in 19 F R2-filtered and 15 N HSQC NMR titration experiments ( Figure S2 and S3, Table S3). Overall, the determined affinities were consistent with the screening results and validate glycomimetics either biphenyl or phenyl-indolinyl substituents in C1 of the Man scaffold as potent Langerin ligands.

Note S2 -Structure activity relationship in C6 -Langerin
The 19 F NMR RDA also served to estimate KI values for mannosides bearing sulfonamide substituents in C6 (Table S2). Here, the SAR is dominated by the affinity decrease associated with mannoside formation as exemplified by 46 (KI = 13±3 mM) (Figures S2, Table S3). Compared to this reference molecule, all three screened derivatives displayed a 6.0-fold affinity increase. While sulfonamide groups appear to represent suitable linkers in C6, no affinity increase was observed upon the introduction of phenyl rings for 26 (KI,est = 2 mM) and 27 (KI,est = 2 mM). Finally, the estimated KI value for 25 was subsequently validated in 19 F R2-filtered and 15 N HSQC NMR titration experiments ( Figure S2 and S3, Table S3).

Note S3 -Binding mode analysis for mannoside 43 -Langerin
We evaluated the binding mode of 43: 15N HSQC NMR confirmed interactions with the CBS and revealed unique chemical shift perturbations (CSPs) for D312 compared to the CSP pattern previously reported for Man reference 45, indicating favorable interactions by the biphenyl system ( Figure S1b). 4 These findings were corroborated by STD NMR experiments yielding uniformly strong STD effects for the substituent in C1 and suggesting a solvent exposed orientation of the sulfonamide linker (Figures S1c, S7 and S8). These observations are in accordance with binding modes obtained from tethered molecular docking, predicting the formation of a hydrogen bond between the carboxyl group of 43 and N292 ( Figure S1d). Additionally, the distal phenyl ring is located near to P310 and the trifluoromethyl group forms van der Waals interactions with A289. Both residues could not be assigned due the structural flexibility of the long loop but are found in proximity of D312. Finally, the sulfonamide linker is oriented towards G284, compatible with conjugation of 43 to liposomes. Figure S4: Analytical HPLC trace for mannoside 42. The purity of 42 was determined to be >95% via analytical reversed-phase HPLC utilizing an H2O: acetonitrile gradient (5% in acetonitrile for 10 min and 5% to 95% in acetonitrile in 15 min). Both solvents contained 0.01% TFA. The analysis was conducted on an Atlantis T3 column (Waters) at a flow rate of 0.5 mL·min -1 and the elution of the derivative was detected via ELSD or absorbance A254 measurements at 254 nm. Figure S5: Analytical HPLC trace for mannoside 43. The purity 43 was determined to be >95% via analytical reversed-phase HPLC utilizing an H2O: acetonitrile gradient (5% in acetonitrile for 10 min and 5% to 95% in acetonitrile in 15 min). Both solvents contained 0.01% TFA. The analysis was conducted on an Atlantis T3 column (Waters) at a flow rate of 0.5 mL·min -1 and the elution of the derivative was detected via ELSD or absorbance A254 measurements at 254 nm. 15 4 (c) Mapping the CSPs on the X-ray structure of the Langerin CRD (PDB code: 3P5F) validates a Ca 2+ -dependent binding mode as indicated by CSPs observed for E285 and K299. 5 Additionally, CSPs are observed for N297, K313 and A300, residues also affected upon recognition of Man and N-acetyl-mannosamine. 6 Two distinct features are observed for 43 compared to the recognition of natural glycan ligands. Here, prominent effects observed for D312 might be indicative for an interaction with the second phenyl ring or the carboxyl group while the CSP for G284 might be inducted by either the sulfonamide group or the acetylated ethylamino linker. These findings suggest that the binding mode of the Man scaffold is maintained. CSPs are also observed in remote regions of the CTL domain fold, particularly for K257 and G259 in the short loop region. This indicates a modulation of the previously reported allosteric network. 7     4 This observation is consistent with the comparable KI and KD values determined for these glycomimetics by NMR.                  Table S10.

Fuc-Lip 94%
LeX-Lip 41%    24 NMR spectra were processed in MestReNova (MestreLab Research). The specific optical rotation was determined using a Model 341 polarimeter (PerkinElmer). ESI-MS analysis was conducted using an 1100 Series LC/MS coupled to a Micromass ZQ spectrometer (Waters) or directly using an amaZon SL spectrometer (Bruker). HR ESI-MS analysis was conducted using a 6210 ESI-TOF spectrometer (Agilent). ATR-FTIR spectra were acquired using a Spectrum 100 FTIR spectrometer (PerkinElmer). Reversed-phase preparative HPLC was performed on a 1100 Series LC/MS (Thermo Scientific) using a preparative Nuleodur C18 column (Machery Nagel). Analytical HPLC was performed on an Aqcuity UPLC system using an analytical BEH C18 column (Waters) or on a 1200 Series LC/MS coupled to a 6130 ESI-Q spectrometer using an analytical Atlantis T3 column (Agilent).
Next, 2 M aqueous NaOH (340 µL, 665 µmol) was added and the reaction mixture was stirred at room temperature for 4 h. More 2 M NaOH (340 µL, 665 µmol) was added and reaction mixture was stirred at room temperature overnight. Product formation was analyzed via ESI-MS. Solvents were evaporated in vacuo and the residue was purified via reversed-phase MPLC (gradient: 100% H2O to 100% MeOH in 30 min) to yield 42 (12.3 mg, 22 µmol, 47% over 2 steps) as a white solid after lyophilization from H2O. 1
General procedure. DSPE-PEG2kDa-NHS (NOF Europe) was dissolved in anhydrous DMF (to 0.65 mM) and anhydrous N,N-diisopropylethylamine (1 µL per mL DMF) was added. 5.5 equivalents of the targeting ligand bearing a primary amino group (stock solution at 3.5 mM) were added and reaction mixtures was stirred for 18 h at room temperature. Solvents were removed in vacuo and the residue was dissolved in 0.1 M NaHCO3 in H2O and purified via dialysis (3 mL Slide-A-Lyzer cassette; twice against 2 L of 0.1 M NaHCO3 for 3 h and subsequently twice against H2O). Solvents were removed by lyophilization and the residue was dissolved in DMSO-d6 to determine coupling efficiencies using 1 H NMR spectroscopy by integration of characteristic resonances (Table S1_RW)

Liposome preparation
PEGylated liposomes were prepared via thin film hydration and subsequent pore extrusion as previously published. 26 Liposomes used in this study were comprised of DSPC (57 mol%), cholesterol (38 mol%) and DSPE-PEG2kDa (5 mol% total), with the latter also containing the glycolipids for targeting and Alexa 647-lipids (0.25 mol%).
Notably, since efficiencies of the coupling to DSPE-PEG2kDa-NHS differed between ligands, effective mol ratios deviated from calculated quantities. Effective mol ratios are given in the results section. Alexa 647-lipids were prepared as previously described. 4 Briefly, The DSPE-PEG components were dissolved in dimethyl sulfoxide, added to test tube and lyophilized. Next, DSPC (NOF Europe) and cholesterol (Sigma Aldrich) were dissolved in chloroform, added to the test tube and the solvents removed initially using an N2 gas stream and subsequently in vacuo. The residue was dissolved in PBS (pH = 7.4) and the mixture was vortexed and sonicated repeatedly to obtain a homogeneous suspension. The resulting unilamellar liposomes further treated using a pore extruder (Avanti Polar Lipids) with polycarbonate membranes of 800, 400, 200 and finally 100 nm pore size (Avanti Polar Lipids). Liposomes were stored at 4° C. Characterization by dynamic light scattering (DLS, Malvern Zetasizer Nano ZS) and electrophoresis experiments was conducted to characterize liposome dispersity, size and Ζ potential (Scheme S4).

Cell culture
If not stated otherwise, all media and supplements for cell culture experiments were purchased from Thermo Fisher Scientific. Raji, THP-1 and U937 cell lines (ATCC) were grown in complete growth medium containing RPMI1640 medium, 10% FCS, 100 U mL -1 Penicillin-Streptomycin and GlutaMax at 37°C and 5% CO2. Cells were monitored with a light microscope (IT40 5PH, VWR) and subcultured every 2-3 days to maintain cell densities ranging from of 0.5 -3 x 10 6 cells per mL.
CLR-expressing cell lines were generated as described before. 4,27 Briefly, DC-SIGN, Langerin, murine Langerin and murine Dectin-1 cDNAs (Sinobiologicals) were cloned into a lentiviral BIC-PGK-Zeo-T2a-mAmetrine:EF1A construct by Gibson assembly (NEB) according to the manufacturer's protocol. HEK293 cells were transfected with the lentiviral vector together with third-generation packaging vectors and viral particles were then used for transduction of Raji cells. In analogy, DC-SIGN and Langerinexpressing THP-1 and U937 cell lines were generated by lentiviral transduction.

Liposome binding assayflow cytometry
Liposome binding to CLR-expressing cells was assayed as previously described. 27 Briefly, 0.05 x 10 6 cells were plated in transparent conical-bottom 96 well microtiter plates (Nunc) in a volume of 100 µL medium. Plates were centrifuged at 500 x g for 3 minutes, supernatant was aspirated, and cells were resuspended in 100 µL medium, containing 16 µM liposomes. In control experiments, cells were incubated with medium containing 10 mM EDTA or 50 µg mL -1 mannan for 15 min at 4 °C, prior to liposome application. After 1 h incubation at 4 °C in the dark, cells were centrifuged at 500 x g for 3 min and the supernatant was discarded. Cells were resuspended in 200 µL icecold medium and analyzed by detecting the co-formulated Alexa 647 dye via flow cytometry with a 654 nm laser and a 670/14 nm filter (Attune Nxt, life technologies). Flow cytometry data was processed using FlowJo (BD Bioscience). Mean fluorescence intensity (MFI) was further analyzed and plotted using GraphPad Prism (GraphPad Software).
To monitor allosteric activation of DC-SIGN by glycomimetic 48, the described protocol was adjusted as follows. DC-SIGN + Raji cells were plated in transparent conicalbottom 96 well microtiter plates (Nunc) in a volume of 100 µL medium. Plates were centrifuged at 500 x g for 3 minutes, supernatant was aspirated, and cell were resuspended in 25 µL medium supplemented with varying concentrations of monovalent carbohydrates or 0.5 mM 48. Homomultivalent liposomes carrying either natural carbohydrates or 48 were added at a final concentration of 16 µM. Liposome binding was subsequently quantified by measuring Alexa 647 fluorescence in flow cytometry experiments. Statistical significance by means of a student's t-test was assessed based on averaged normalized MFIs from four biological replicates each conducted as technical duplicates. A p-value < 0.05 was set as significance cut-off.

DC-SIGN carbohydrate recognition domain (CRD).
His-tagged DC-SIGN CRD wildtype and the M270F mutant-encoding pET28a plasmids were expressed and purified from inclusion bodies as described previously, with minor changes. 28 Briefly, transformed E.coli BL21(DE3) were grown in M9 minimal medium containing 15 Nlabeled NH4Cl (Silantes), supplemented with 35 mg L -1 ampicillin at 37°C. Protein expression was induced with 1 mM IPTG at OD600 ~ 0.9 for 4 h at 37°C. Bacteria were harvested by centrifugation (4.000 x g, 30 min, 4°C), resuspended in lysis buffer and lysed by sonication on ice. IBs were harvested by centrifugation (15.000 x g, 90 min, 4°C) and washed thrice with lysis buffer and ultrapure water to remove soluble proteins. Washed IBs were solubilized in 20 mL denaturation buffer for 1 h at 37°C. After centrifugation (15.000 x g, 90 min, 4°C), solubilized IBs where rapidly diluted into 180 mL refolding buffer and stirred overnight at 4°C. The protein solution was then dialyzed overnight at 4°C against 5 L TBS (100 mM Tris-HCl,150 mM NaCl, pH 7.8). After another dialysis step against 5 L TBS, precipitated protein was removed by centrifugation (15.000 x g, 15 min, 4°C) and the DC-SIGN CRD was purified using Ni 2+ -NTA affinity chromatography according to manufacturer's instructions (Qiagen). Purified receptor was dialyzed against 5 L MES low salt buffer (20 mM MES, 40 mM NaCl, pH 6.0) supplemented with 5 mM CaCl2 overnight at 4°C. DC-SIGN CRD samples were concentrated using centrifugal filtration and concentration was quantified via UV spectroscopy (with A280, 0.1% = 2.966). Sample purity was analyzed via SDS PAGE. The protein solution was aliquoted, snap frozen in liquid N2 and stored at -80°C until further usage.

F R2-filtered NMR
General remarks. 19 F R2-filtered NMR experiments were conducted on a PremiumCompact 600 MHz spectrometer (Agilent). Spectra were processed in MestreNova (Mestrelab Research) and data analysis was performed with OriginPro (OriginLab). Experiments with the Langerin ECD were performed at a receptor concentration of 50 μM in 25 mM Tris with 10% D2O, 150 mM NaCl and 5 mM CaCl2 at pH 7.8 and 25°C. Experiments with the DC-SIGN CRD were performed at a receptor concentration of 50 μM in 25 mM HEPES with 10% D2O, 150 mM NaCl and 5 mM CaCl2 at pH 7.0 and 25°C. TFA served as an internal reference at a concentration of 50 μM. Apparent transverse relaxation rates R2,obs for the reporter ligand were determined using the CPMG pulse sequence as previously published. 6,30,31 Screening of mannoside library. Estimated affinities KI,est were determined in competitive binding experiments at 0.1 mM of reporter ligand 49 at 0.1 mM using the Langerin ECD at a single competitor concentration as previously described. 6 33 Receptor resonances were suppressed using a T1,rho filter at a relaxation time τ of 35 ms.
Epitope mapping. The binding epitope for 43 was determined at a concentration of 500 µM. For each spectrum 512 scans were recorded. The relaxation delay d1 was set to 6 s and spectra were recorded at 5 different saturation times tsat varying from 0.25 to 6.00 s. Equation 1 served to derive the STD effect STD for each analyzed resonance from the corresponding on-and off-resonance spectra. 34 I0 represents the integral of a resonance in the off-resonance spectrum and Isat represents the integral of a resonance in the on-resonance spectrum.

Equation 1
The apparent saturation rate ksat and the maximal STD effect STDmax were derived from Equation 2 in a two-parameter fit. 35 Standard errors were derived directly from the fitting procedures. These parameters were used to calculate the initial slope of the STD build-up curves STD'0 via Equation 3. STD'0 values were normalized and mapped on the corresponding ligand structure. Only resonances for which at least part of a multiplet was isolated were considered for the epitope mapping.

Equation 3
15 N HSQC NMR General remarks. 15 N HSQC NMR experiments were conducted on an Ascend 700 MHz spectrometer (Bruker). 36 Spectra were processed in NMRPipe. 37 Data analysis was performed using CCPN Analysis, MatLab (MathWorks) and OriginPro. 38 Experiments with the Langerin CRD were performed at a receptor concentration of 100 μM in 25 mM HEPES with 10% D2O, 150 mM NaCl and 5 mM CaCl2 at pH 7.8 and 25° C. DSS-d6 served as an internal reference at a concentration of 100 μM. Spectra were referenced via the internal spectrometer reference. Spectra were acquired with 128 increments and 32 scans per increments for 150 μl samples in 3 mm sample tubes (Norrell). The relaxation delay d1 was set to 1.4 s and the acquisition time tacq was set to 100 ms. The W5 Watergate pulse sequence was used for solvent suppression. 39 The used resonance assignment for the Langerin CRD has been published previously. 7 Titration experiments with 9, 25, 43 and 46 were conducted in presence of 10% DMSO. Here, assignments were transferred from a reference spectrum in absence of DMSO to the nearest neighbor in the reference spectrum in presence of DMSO. In case this approach was ambiguous, the corresponding resonances were flagged during data processing and analysis.

Equation 4
A standard deviation σ of 0.02 ppm was previously determined for the measurement of chemical shifts in 15  KI values were determined as previously described for Langerin and the DC-SIGN CRD. 6

Equation 9
The KD value was calculated from CSP19F, T values in a two-parameter fit using Equation 10, using a fixed protein concentration [P]T and a CSP19F of 0 ppm as a lower bound. CSP19F,b corresponds to the top asymptote and was kept variable. As in Equation 9, pb represents the fraction of bound protein. 19

Molecular docking
Langerincarbohydrate binding site (CBS). Molecular modelling procedures were performed in MOE (Chemical Computing Group). Deviations from default options and parameters are noted. The AMBER10:EHT force field was selected for the refinement of docking poses and the hydrogen bond network while the MMFF94x force field was utilized for the generation conformers. [45][46][47] Receptor surfaces were visualized in Connolly representation. 41 A structural alignment of the Langerin CBS in complex with different Man-type oligosaccharides was performed (PDB codes: 3P5D, 3P5E and 3P5F). 48 Based on this visualization, a pharmacophore model was defined with features for O3, O4 and O5 of the Man scaffold. The spatial constraint on the O3 and O4 was defined by a sphere with a radius r of 0.5 Å while the position of O5 was constrained by a sphere with a radius r of 1.0 Å. Chain B of the Langerin CRD in complex with a Man-type disaccharide served as the structural basis for the docking of 43 (PDB code: 3P5F). Of the two binding modes included in this model, the orientation for targeting the identified pockets in axial direction of C2 was selected. Additionally, an alternative conformation for K313 observed for the Langerin complex with Gal-6-OS was modeled and included into the analysis. 48 Overall model quality and protein geometry were evaluated in MolProbity and maintained utilized MOE's Structure Preparation. 49 Next, protonation states and the hydrogen bond network of the complex were simulated with MOE's Protonate 3D followed by the removal of all solvent molecules.
Conformations for 43 were generated utilizing MOE's Conformation Import. A pharmacophore-based placement method was utilized to generate docking poses that we scored using the London ΔG function. Highly scored poses were refined utilizing molecular mechanics simulations, rescored via the GBIV/WSA ΔG function, filtered using the pharmacophore model and written into the output database. 50 Conformational flexibility of the CBS was accounted for by introducing B-factor-derived tethers to side chain atoms. Refined docking poses were ranked according to their the GBIV/WSA ΔG score and evaluated visually in the context of the conducted 15 N HSQC and STD NMR experiments.
DC-SIGNsecondary binding pocket. A model of ligand 48 was built and optimized using the VMD Molefacture plugin, and the X-ray structure of the DC-SIGN CRD (PDB code: 1SL4) was used as a receptor structure. 51,52 Only protein residues and Ca 2+ ions were kept while structural waters were removed. The receptor and ligand structures were prepared following the standard AutoDock protocol. 53 All non-polar hydrogens were merged, and Gasteiger charges and atom types were added. The grid size and position were chosen to include all the amino acids belonging to the secondary binding pocket (Q306, M270, Y268, T261, F302, F269, and I124) and the spacing between grid points was set at 0.375 Å. AutoDock Bias protocol was applied to perform a biased docking experiment, considering the information from NMR experiments (ΔCSPs). 54 Briefly, based on ideal interaction estimated using ideal_interaction_sites.py, a hydrogen bond donor and acceptor restraint were added by modifying their respective energy grids (HD and OA map, respectively) using prepare_bias.py script. 55,56 For each system, 100 different docking runs were performed and the results were clustered according to the ligand heavy atom RMSD using a cut-off of 2 Å. The Lamarckian Genetic Algorithm parameters for each conformational search run were kept at their default values (150 for initial population size, 1•10 7 as the maximum number of energy evaluations, and 2.7•10 4 as the maximum number of generations). The docking results for 48 were further analyzed by visual inspection.

MD simulations
The complex between 48 and DC-SIGN was further analyzed using the protocol described by Blanco et al. with the modifications described below. 57 Briefly, the system was prepared with the leap module from the AMBER package using ff14SB and TIP3P force field for amino acid and water molecules. 58 The BMP parameter was obtained using the Antechamber module from the AMBER package using the GAFF force field.
The system was first optimized using a conjugate gradient algorithm for 5000 steps, followed by 150 ps. Long constant volume MD equilibration, in which the first 100 ps were used to gradually raise the temperature of the system from 0 to 300 K (integration step = 0.0005 ps per step). The heating was followed by a 300 ps long constant temperature and constant pressure MD simulation to equilibrate the system density (integration step = 0.001 ps per step). During these temperature and density equilibration processes, the protein α-carbon atoms were constrained by 5 kcal•mol -1 •Å -1 force constant using a harmonic potential centered at each atom starting position. Next, a second equilibration MD of 5 ns was performed, in which the integration step was increased to 2 fs using the SHAKE algorithm, and the force constant for restrained α-carbons was decreased to 2 kcal•mol -1 •Å -1 followed by 5 ns long MD simulation with no constraints. Finally, 20 ns long production MD simulations were carried out using the 'Hydrogen Mass Repartition' method, which allows an integration step of 4 fs. 59 The trajectory processing and RMSD analysis were performed with the CPPTRAJ module of the AMBER package. 60

Webserver-based allosteric site prediction
The AllositePro server was used to predict potential allosteric binding sites in the DC-SIGN CRD X-ray structure (PDB code: 1SL4). 10,11 Binding sites were defined as allosteric based on an Allosite score of > 0.5 detected pockets, resulting from the weighted sum of a feature score, describing structural features physicochemical properties of the pocket, and a perturbation score derived from significant changes in normal mode analysis of apo and holo states of the protein. 10