Rational Design of a DNA‐Scaffolded High‐Affinity Binder for Langerin

Abstract Binders of langerin could target vaccines to Langerhans cells for improved therapeutic effect. Since langerin has low affinity for monovalent glycan ligands, highly multivalent presentation has previously been key for targeting. Aiming to reduce the amount of ligand required, we rationally designed molecularly defined high‐affinity binders based on the precise display of glycomimetic ligands (Glc2NTs) on DNA‐PNA scaffolds. Rather than mimicking langerin's homotrimeric structure with a C3‐symmetric scaffold, we developed readily accessible, easy‐to‐design bivalent binders. The method considers the requirements for bridging sugar binding sites and statistical rebinding as a means to both strengthen the interactions at single binding sites and amplify the avidity enhancement provided by chelation. This gave a 1150‐fold net improvement over the affinity of the free ligand and provided a nanomolar binder (IC50=300 nM) for specific internalization by langerin‐expressing cells.


Synthesis of Glc2NTs-PNA conjugates
Automated solid-phase PNA synthesis: Linear solid-phase PNA synthesis was performed by using an Intavis ResPep parallel synthesizer and Intavis microscale columns. TentaGel R RAM resin (typical loading: 0.20 mmol/g, 2μmol scale) from Rapp Polymers (Tübingen, Germany) was allowed to swell in DMF for 30 min and then transferred to the synthesizer. Fmoc cleavage: 250 μL DMF/Piperidine (4:1, v/v) was added to the resin. After 2 min the resin was washed with 300 μL DMF (3x). The cleavage and washing step were repeated. Coupling of amino acid: 54 μL HCTU (5.4 eq, 0.2 M in NMP), 30 μL NMM (12.0 eq, 0.8 M in NMP) and 40 μL Boc-protected lysine (6.0 eq, 0.3 M in NMP) were mixed in a preactivation vessel. After 2 min, the preactivation solution was transferred onto the resin. After 30 min, the resin was washed with 200 μL DMF (3x) and the coupling was repeated.

Langerin ECD and CRD Receptor Expression and Purification
Langerin extracellular domain. Expression and purification were conducted as previously published. [9] Briefly, the trimeric Langerin extracellular domain (ECD) was expressed insolubly in E. coli BL21* (DE3) (Invitrogen). Following enzymatic cell lysis, inclusion bodies were harvested and subsequently solubilized. The sample was centrifuged and the Langerin ECD was refolded overnight via rapid dilution. Next, the sample was dialyzed overnight, centrifuged and purified via mannan-agarose affinity chromatography (Sigma Aldrich Langerin carbohydrate recognition domain. Expression and purification were conducted as previously published. [9] Briefly, the monomeric 15 N-labeled Langerin carbohydrate recognition domain (CRD) was expressed insolubly in E. coli BL21* (DE3) (Invitrogen). Following enzymatic cell lysis, inclusion bodies were harvested and subsequently solubilized. The sample was centrifuged and the Langerin CRDs were refolded overnight via rapid dilution. Next, the sample was dialyzed overnight, centrifuged and purified via StrepTactin affinity chromatography (Iba). After an additional dialysis step overnight, the sample was centrifuged and the buffer was exchanged to 25 mM HEPES, 150 mM NaCl , pH 7.0. The concentration of Langerin CRDs was determined via UV spectroscopy (ε280 = 56.170 mol -1 cm -1 ). Purity and monodispersity of Langerin CRD samples were analyzed via SDS PAGE and DLS.

19 F-NMR Assay Inhibition Experiments
The 19 F-NMR Assay has been previously described by Wamhoff et al. [10] Langerin ECD and CRD were obtained as described above.
Experiments with the Langerin ECD were performed at a receptor concentration of 50 μM or 25 μM in 25 mM Tris with 10 % DMSO, 10% D2O, 150 mM NaCl and 5 mM CaCl2 at pH 7.8 and 25° C. Experiments with the Langerin CRD were performed at a receptor concentration of 50 μM in 25 mM HEPES with 10% DMSO, 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 relaxation rates R2,obs for the reporter ligand were determined using the CPMG pulse sequence as previously published. [10] IC50 Determination Remarks: Measurements of IC50 values via 19F-NMR were considered as am initial screen for which hits were subsequently validated using an orthogonal assay (see chapter 12). Due to the substantial amount of ligand and protein required for the NMR assay, experiments were conducted as one independent 5-point titration per construct and standard errors of the mean (SEM) for IC50 and Hill factor were obtained directly from the fitting procedure. The following SPR experiments were conducted as duplicates We prioritised the value of two orthogonal methods, which we believe gives more reliable results than just one method with extensive S18 replicates. One could argue that this approach is more robust as it accounts, to some degree, for systematic errors and assay-dependent artefacts Further, the distance affinity screening measurements are not biological experiments, where a large statistical variance must be assumed, but rather physicochemical effects that display lower variance. IC50 values were determined in competitive binding experiments via the detection of binding of 0.1 mM 19 F-marked reporter ligand (propargyl-2-deoxy-2-trifluoroacetamido-α-mannoside) to either the Langerin ECD or CRD at six competitor concentrations as previously published. [10] Samples were prepared via serial dilution. Equation 1 served to derive IC50 values and Hills factors p from R2,obs values in a two parameter fit. R2,max represents the relaxation rate at 0.1 mM reporter ligand in presence of receptor and in absence of competitor. The measurements were done in single experiments (five point titration); standard errors were derived directly from the fitting procedures. The error-weighted, non-linear fitting procedure on experimentally determined transversal NMR relaxation constants (R2,obs, which themselves have associated statistical variance and SEMs) has been previously described and validated.1 Consequently, the magnitude of the SEM depends largely on the "goodness of fit" of the R2,obs values, i.e. how well the data points adhere to the assumed Hill equation. Variance stems from two factors: pipetting errors and technical errors from determining R2,obs both potentially resulting in "outliers". We assume the latter to be the main contributing factor as the SNR in some of the 19F NMR spectra when determining fast R2,obs values of more than 5 Hz will be low. This limitation is aggravated by inconsistent automated shimming when using 150 l NMR sample tubes. Both the low SNR and choice of sample tube are compromises to lower material requirements and increase throughput as previously described.1 In summary, the plotted SEMs reflect the technical error within each data set and the statistical variance in SEMs are due the presence of R2,obs outliers within some of these data sets. These SEMs only give limited insight into the variance when obtaining replicates from independent titration experiments. We fully expect error bars to be consistent under those conditions and have previously demonstrated that KI values can be reliably estimated (σ ≈ 20%) even from single-point experiments. 1 We further note that the relative errors obtain biv-Glc2NTs-13 are, in fact, comparable (ECD = 0.09 and CRD = 0.17). Comments: Constructs Biv-Glc2NTs-19 and 26 seem to display steeper curves. However, because they are weaker binders only the highest concentration results inhibition. Therefore, the curve shapes are poorly defined and the hill factors for these constructs were fixed to 1 to allow for a conservative IC50 estimation. We express caution and rather not over-interpret the curve shapes. The reason we refrained from using higher concentrations were concerns around solubility of the PNA-DNA constructs. Importantly, the solubility issue was addressed successfully by leveraging statistical rebinding effects (see chapter 12).
With the exception of constructs Biv-Glc2NTs-19 and 26, all other bivalent constructs display reduced p values around 0.5. It should be noted that the 19 F-NMR assay was performed at the assay limit, which will affect the hill slopes. The decrease in Hill factors could also be due to an autoinhibitory effect at higher concentrations where two or more individual bivalent constructs bind to a single ECD trimer and compete with chelate binding. Importantly, data obtained in the SPR assay (except for Biv-TriGlc2NTs-32 and Biv-Glc2NTs-07) displayed hill slopes >1 (chapter 12). This assay was not performed at the assay limit and high concentrations of proteins are avoided. The hill slopes > 1 are indicative of optimal chelate binding.

SPR Assay Inhibition Experiments
Experiments were performed on a Biacore X100 instrument (GE Healthcare Europe GmbH Uppsala, Sweden) at 25 °C. Langerin ECD was obtained as described above. [1] The procedure was based on a previously described method. [7] For immobilization of biotinylated α-Dmannose-PAA, the HBS-EP buffer (10 mM  ). For testing the experimental set up, single cycle kinetics were performed at five concentrations of Langerin, ranging from 62 nM to 5000 nM, and finally a KD value of 1.5 µM was determined. This study also confirmed the used protein contact times of 120 s, the dissociation time of 300 s and a flow rate of 20 µl/min. The chip surfaces were regenerated at 30 µl/min with 10 mM EDTA pH 8 and a contact time of 60 s. In the dose response experiment before injection, each protein sample ( [Langerin]=500 nM) and a serial dilution of the complexes (dissolved in 20 mM HEPES, 150 mM NaCl, 1mM CaCl2, pH 7.4) were incubated for a minimum of 5 min at rt. The samples were injected over the reference and measuring channel. For evaluation the reference channel data were subtracted from the measuring channel data. The Langerin control was measured before and after every dose response series. By calculating the regression between both values an individual Langerin baseline drift was calculated for every dose response measurement. Corrected response values were calculated by dividing the RU of the dose response experiment by the individual Langerin baseline RU. The corrected response values were used for curve creation and IC50 fitting procedure. Responses of the sample injections were extracted between report points set at the start of the injection (0 s) and at the end of the dissociation phase (250 s). Each data point represents the mean value (SEM) of 2 measurements (duplicates).
In the single concentration experiment the protein sample [Langerin]=500 nM) and the [complexes] = 10 µM (dissolved in 20 mM HEPES, 150 mM NaCl, 1mM CaCl2, pH 7.4) were incubated for a minimum of 5 min at rt. The samples were injected over the reference and measuring channel. For evaluation the reference channel data were subtracted from the measuring channel data.

Establishment of C-Type Lectin+ Model Cells
The production of the model cells has been previously described. [4]

Flow Cytometry Assay
Raji cells, suspended in cell culture medium (RPMI1640 (Sigma Aldrich)), were counted, centrifuged at 500 g for 3 min, aspirated, incubated with blocking buffer (0.2 mg/mL Salmon Sperm DNA, 0.2 % BSA in PBS) centrifuged at 500 g for 3 min, aspirated resuspended in culture medium at 37° C and 5% CO2. 50,000 cells were added to the 96 well microtiter plates (Nunc) to obtain a volume of 100 μl. To monitor internalization and binding, Biv-Glc2NTs-12 or Biv-TriGlc2NTs-13 in HBS (20 mM HEPES, 150 mM NaCl, 1mM CaCl2, pH 7.4) were added to the cells at a final concentration of either 660 nM (Method I) or 66 nM (Method II).
In case of control experiment 250 μg•mL -1 mannan in PBS was added to the cells and constructs. The cells and constructs were incubated for 45 min at 4° C and subsequently centrifuged at 500 g for 3 min. Cells were aspirated and resuspended in cell media for 60 min at 37° C and 5% CO2. Afterwards, the cells were resuspended in fresh cell culture medium before analyzing (Method I) or directly analysed without fresh medium (Method II). Internalization and binding of fluorophore marked complexes was evaluated by flow cytometry on an BA Accuri C6 Plus Flow Cytometer equipped with an autosampler by detecting the conjugated dye Cyanine5 with a 640 nm laser and > 670 filter set. 5000 events were measured for every well. The data was analysed with CFlow Plus. For normalization the Mean Fluorescent Intensity (MFI) of the Raji cells (autofluorescence) was subtracted from the measured MFI of cells with ligands.