Structural characterisation of high affinity Siglec-2 (CD22) ligands in complex with whole Burkitt’s lymphoma (BL) Daudi cells by NMR spectroscopy

Siglec-2 undergoes constitutive endocytosis and is a drug target for autoimmune diseases and B cell-derived malignancies, including hairy cell leukaemia, marginal zone lymphoma, chronic lymphocytic leukaemia and non-Hodgkin’s lymphoma (NHL). An alternative to current antibody-based therapies is the use of liposomal nanoparticles loaded with cytotoxic drugs and decorated with Siglec-2 ligands. We have recently designed the first Siglec-2 ligands (9-biphenylcarboxamido-4-meta-nitrophenyl-carboxamido-Neu5Acα2Me, 9-BPC-4-mNPC-Neu5Acα2Me) with simultaneous modifications at C-4 and C-9 position. In the current study we have used Saturation Transfer Difference (STD) NMR spectroscopy to monitor the binding of 9-BPC-4-mNPC-Neu5Acα2Me to Siglec-2 present on intact Burkitt’s lymphoma Daudi cells. Pre-treatment of cells with periodate resulted in significantly higher STD NMR signal intensities for 9-BPC-4-mNPC-Neu5Acα2Me as the cells were more susceptible to ligand binding because cis-binding on the cell surface was removed. Quantification of STD NMR effects led to a cell-derived binding epitope of 9-BPC-4-mNPC-Neu5Acα2Me that facilitated the design and synthesis of C-2, C-3, C-4 and C-9 tetra-substituted Siglec-2 ligands showing an 88-fold higher affinity compared to 9-BPC-Neu5Acα2Me. This is the first time a NMR-based binding study of high affinity Siglec-2 (CD22) ligands in complex with whole Burkitt’s lymphoma Daudi cells has been described that might open new avenues in developing tailored therapeutics and personalised medicine.


Results
Binding of 9-BPC-4-mNPC-Neu5Acα2Me to Burkitt's lymphoma (BL) Daudi cells. In the first instance, we have monitored the binding of 9-BPC-4-mNPC-Neu5Acα 2Me (3) directly to Burkitt's lymphoma (BL) Daudi cells using Saturation Transfer Difference (STD) Nuclear Magnetic Resonance (NMR) spectroscopy. STD NMR spectroscopy is an ideal tool to study the interaction of larger-sized biomolecular targets with low-molecular-weight ligands, and has been previously used to investigate ligand interactions with bulky targets, for example whole virus particles 12,23 , cells 14,16 , sepharose immobilized recombinant protein 24 , virus-like particles 13,25 and intracellular organelles 20,21,26 . The viability of BL Daudi cells in 2.0 mM deuterated HEPES buffer was assessed by monitoring the cells with increasing sodium chloride concentration. The absence of sodium chloride in the buffer resulted in a 79% decrease in viable cells after 1 h. The addition of either 300 mM, 150 mM or 75 mM sodium chloride to 1.5 mM deuterated HEPES chloride prevented the BL Daudi undergoing osmotic shock with only 2.9%, 1.9% and 2.1% loss in viability, respectively. In order to keep the NMR pulse length short, a buffer with lower salt concentrations (70 and 140 mM) was chosen. The 1 H NMR spectra of 3 in the presence of 5.0 × 10 5 BL Daudi cells buffered (1.5 mM deuterated HEPES with 140 mM NaCl) are shown in Fig. 2a; the corresponding STD NMR spectrum (Fig. 2b) reveals strong and significant signals for 3 when bound to the Daudi cells. We have determined the viability of the BL Daudi cells before (95.5%) and after incubation (1 hour at 37 °C and 4 °C) in the NMR buffer (1.5 mM HEPES, 75 mM NaCl, D 2 O). These experiments revealed that 92.5% of BL Daudi cells were still viable after 1 h incubation at 4 °C and 91.2% at 37 °C demonstrating that only a very small amount of cells have died during NMR spectra acquisition. However, cell viability after a 3 h incubation period was 83% at 4 °C and 63% at 37 °C indicating that the cells considerably loose their viability over longer time periods and that only shorter acquisition times can be used.
To investigate if the binding of 3 to the BL Daudi cells is specific we added an equimolar concentration of a non-binding spy molecule (sucrose) and acquired a 1 H NMR spectrum as shown in Fig. 2c. The STD NMR spectrum of this equimolar 3:sucrose-mixture is shown in Fig. 2d. Most striking is the observation that no STD NMR signals can be detected for the spy molecule sucrose, demonstrating that it has no binding affinity to the Daudi cells. This result clearly supports the conclusion that the observed STD NMR signals of 3 in the presence of BL Daudi cells are a direct consequence of 3 binding to the cell surface protein.

9-BPC-4-mNPC-Neu5Acα2Me (3) bound to periodate pretreated BL Daudi cells reveals an increase in relative STD NMR effects.
In order to investigate if unmasking of siglecs present the BL Daudi cell surface that are engaged in cis interaction would result in more efficient binding and hence stronger STD NMR signals of 3, BL Daudi cells were pre-treated with periodate that specifically truncates the glycerol side chain of sialic acid of the glycosylated Siglec-2 27 . STD NMR experiment of 3 in complex with pretreated BL Daudi cells has revealed a significant increase in STD NMR signal intensities (Supplementary Figure 1) of 3 presumably due to the disruption of cis-binding resulting in an increase in Siglec ligand binding sites on the cell surface 28 .

9-BPC-4-mNPC-Neu5Acα2Me (3) bound to Siglec high BL Daudi cells reveals higher STD NMR effects compared to Siglec low BL Daudi cell fractions.
To further obtain evidence that 3 binds specifically to Siglec-2 expressed on BL Daudi cells we have performed fluorescence activated cell sorting of these cells into Siglec-2 high and Siglec-2 low fractions. The cells were stained using an anti-human Siglec-2 mAb (clone RFB4) as primary antibody and anti-antibody Brilliant Violet 421 conjugated secondary antibody. Cells were sorted into Siglec-2 low and Siglec-2 high fractions based on their median fluorescence intensities (Fig. 3a). Siglec-2 low and Siglec-2 high cell fractions were collected and counted to obtain an identical number of cells (5.0 × 10 5 BL Daudi cells), transferred into the cell-specific NMR buffer, complexed with 3, and subjected to STD NMR experiments. Figure 3b shows the absolute STD NMR signals of the methyl protons of the N-acetyl group of 3 in the presence of 5.0 × 10 5 Siglec-2 low (blue) and Siglec-2 high (red) BL Daudi cells. The spectra reveal that the Siglec-2 high cells have a 46% increased STD NMR signal intensity compared to the Siglec-2 low cells, correlating with the increased number of available Siglec-2 copies present on the cell surface. This analysis demonstrates that the binding of 3 to BL Daudi cells occurs via Siglec-2 expressed on the cell surface. To ensure that the binding of 3 occurred to Siglec-2 and not to other Siglecs expressed on the cells 29 , BL Daudi cells used in these experiments were stained with a panel of anti-human siglec antibodies and relevant controls. The quantitative analysis (Table 1 and Supplementary Figure 2a) revealed that a low percentage of cells express other Siglecs, eg. 5.9%, 6.7% and 4.5% for Siglec-10, Siglec-3 and Siglec-5/-14, respectively. Only Siglec-11 and Siglec-16 are present in 13.4% and 11.5% of the cells respectively, whereas less than 2% of cells express Siglec-1, -4, -6, -7, -8 and -9. This analysis suggests that the observed binding of 3 to BL Daudi cells is a direct consequence of its interactions with Siglec-2 and cross-reactivity with other Siglecs recognizing α (2,3)-/α (2,6)-Neu5Ac glycosides is negligible.
Binding of 9-BPC-4-mNPC-Neu5Acα2Me (3) to transiently expressed Siglec-2 on HEK293T cells. To further demonstrate the binding of 3 to Siglec-2 expressed on cells an additional control NMR experiment was performed using HEK293T cells before and after transfection with the full length Siglec-2 coding sequence (Uniprot P20273-1). The quantitative analysis of Siglec expression (Fig. 4a) in untransformed HEK293T cells revealed that a very low percentage of cells express Siglec-2 (1.2% of the population) making this cell line an ideal negative control to study Siglec-2 ligand interactions. Only minimal expression of Siglec-4, Siglec-10, Siglec-11 and Siglec-16 can be detected on HEK293T cells hence, minimizing the contributions coming from the interaction of the ligand of interest with other Siglecs. (Table 1 and Supplementary Figure 2b). The flow cytometric analysis of Siglec-2 transfected HEK293T cells (Fig. 4b) revealed the highest Siglec-2 expression (91.7% positive, median fluorescence intensity 26.9) after 24 hours post transfection using 3.75 μ L of Lipofectamine 3000 reagents to prepared DNA-lipid complexes. Non-transfected and Siglec-2 transfected HEK293T cells were then complexed with 3 and STD NMR experiments were undertaken. Very low signal intensity was observed for 3 in complex with non-transfected HEK293T cells (Fig. 4c). An identical NMR experiment performed with Siglec-2 transfected HEK293T cells in complex with 3 showed a 75-fold increase in STD NMR signal intensity correlating with the significantly increased number of available Siglec-2 binding sites present on the HEK293T cell surface (Fig. 4d). This result does not only support that 3 binds specifically to Siglec-2 expressed on the cell surface, but also that a higher protein expression level leads to an increased STD NMR signal intensity due to the availability of an increased number of binding sites.  Binding epitope of 3 bound to BL Daudi cells. Our structural approach comprises of the quantitative evaluation of STD NMR effects of 3 when bound to BL Daudi cells and to generate a binding epitope that consequently facilitate the design of second-generation Siglec-2 ligands. First, we were interested to compare the relative STD NMR values of 3 in complex with BL Daudi cells to our recently determined STD NMR values obtained for 3 bound to recombinantly-expressed Siglec-2 11 , The binding epitope of 3 in complex with BL Daudi cells was determined by double difference STD (STDD) NMR spectroscopy (Fig. 5c) and compared to published relative STD NMR values obtained for a 3-Siglec-2 complex (red and blue STD NMR values in Fig. 5, respectively). Remarkably, the overall relative STD NMR effects of the 3-Siglec-2 and 3-BL Daudi cells complexes are comparable suggesting an identical binding mode. The BPC moiety at C-9 and the mNPC ring at C-4 show in both complexes the strongest saturation transfer. The STD NMR spectra of 3 in the presence of BL Daudi cell further revealed that the methyl protons of the C-2 aglycon methoxy group in 3 receive the same level of saturation as the C-5 N-acetyl functionality indicating significant interaction of the C-2 aglycon moiety with Siglec-2 expressed on BL Daudi cells. This cell-derived binding epitope is therefore in excellent agreement with the relative STD NMR values obtained for 3 in complex with recombinantly-expressed Siglec-2 suggesting that incorporation of bulkier groups of C-2 aglycon moiety may increase protein contacts and hence further enhance binding affinity. This is also consistent with previous studies that have shown that aromatic moieties at the anomeric (C-2) position, such as a α -benzyl glycoside 30 or α -2′ ,3′ -dichlorobenzyl 10 glycosides, of Neu5Ac-based ligands are well tolerated by the binding site and result in enhanced binding affinities. Our cell-derived epitope maps prompted us to introduce a longer alkyl chain. We have also sought to place a carboxylate group directed away from the anomeric position of the Neu5Ac-based ligand, whilst maintaining an aromatic functionality which mimics the interaction of the galactose moiety as present in the natural ligand Neu5Acα (2,6)Galβ (1,4)GlcNAc. Due to the presence of aromatic groups at both C-9 and C-4 of the Neu5Ac-based ligand scaffold, an increase in hydrophilicity is required while maintaining the pharmacophore aromatic character. Therefore, a functionalized    Synthesis of second-generation Siglec-2 binding ligands 7 and 8. The synthetic approach towards 7 and 8 commenced with the preparation of 2,3-β -epoxy 4-azido-4-deoxy-Neu5Ac derivative 5 31 that is readily accessible from the corresponding 2,3-unsaturated 4-azido-4-deoxy-Neu5Ac2en derivative 4. Following our recently developed method for accessing 3-hydroxy-Neu5Ac α -glycosides 32 , the key synthetic intermediate 3-hydroxy-2-α-propargyl-Neu5Ac 6 was obtained through an acid catalysed α -stereoselective opening of epoxide 5 (Fig. 6). To our knowledge, this is the first report of a high yielding reaction generating α -glycosides from 2,3-β -epoxy 4-azido-4-deoxy-Neu5Ac (5). This method offers great potential for accessing 4-azido-4-deoxy-3-hydroxy-Neu5Ac α -glycosides and could be used to introduce a range of functionalities at the anomeric position to explore interactions with biologically important sialic acid-recognizing proteins. The presence of a C-3-hydroxyl group in N-acetylneuraminic acid derivatives has been shown to increase resistance to enzymatic hydrolysis by sialidases 33 and, in some cases, may yield slight improvements in binding affinity compared to the unsubstituted parent compound 9 . As shown in Fig. 6, intermediate 6 offers promising opportunities for selective derivatisation at the C-2, C-4 and C-9 positions. Employing our established chemistry 11 , the biphenylcarboxamido group at C-9 and aromatic amide at C-4 were introduced. The aromatic-like moiety at C-2 was then added via a 1,3 dipolar cycloaddition of methyl 2-azidoacetate [CuSO 4 • 5H 2 O, sodium ascorbate, MeOH, rt, 2 h] to produce the desired derivatives, which upon base-catalysed deprotection [aq. NaOH, MeOH, rt, 2 h] led to the final compounds 7 and 8.

Binding of 7 and 8 to Daudi Burkitt's lymphoma cells.
We then investigated the binding of 7 and 8 directly on B lymphoblast cells. Figure 7a shows the 1 H NMR spectrum of 7 and the corresponding STD NMR spectrum of 7 in complex with BL Daudi cells is shown in Fig. 7c. The corresponding STD NMR spectrum of 7 in complex with recombinantly-expressed Siglec-2 is shown in Fig. 7b,d   On-resonance frequency was set to − 1 ppm and the off-resonance to − 300 ppm. The relative STD NMR effects of 7 in the presence of cells (red values) and recombinantly expressed Siglec-2 (blue values) are depicted at the molecular structure. The binding epitope was calculated using a double difference (STDD) NMR spectrum by subtracting the control spectrum obtained in the absence of cells b) from the spectrum acquired for the 7-cell complex c). The control experiment of cells in the absence of ligand did not show any significant background signals and was therefore not used in the quantitative analysis.
Scientific RepoRts | 6:36012 | DOI: 10.1038/srep36012 recombinantly-expressed Siglec-2 ( Fig. 7d) indicates similar signal intensities. We have quantified the relative STD NMR signals of 7 when bound to BL Daudi cells (red) and pure Siglec-2 (blue) as shown on the molecular structure of 7. Remarkably, the binding epitope of 7 is very similar, if not, identical when comparing protein-based and cell-based NMR spectra. This result demonstrates that 7 presumably binds to Siglec-2 expressed on the BL Daudi cell surface. We have performed a similar analysis of the binding epitope of 8 in complex with Siglec-2 and BL Daudi cells (Fig. 8). Figure 8c,d show the STD NMR spectra of 8 in complex with BL Daudi cells or recombinantly-expressed Siglec-2, respectively. The experimentally determined relative STD NMR values and binding epitope revealed very similar values indicating that 8 binds almost exclusively to Siglec-2 expressed on BL Daudi cells. Figure 9 outlines the inhibition potencies of compounds 2, 3, 7 and 8 as determined by hapten inhibition assays. The addition of a meta-nitrophenylcarboxamido (mNPC) moiety at C-4 with a biphenylcarboxamido (9-BPC) at C-9 (3) shows enhanced binding to Siglec-2 by a factor of 15.9 compared to the benchmark compound 2 (Fig. 9). The hapten inhibition assay for 7 that bears an additional triazole-carboxy moiety at C-2 and the hydroxyl group at C-3 revealed a relative IC 50 (rIP) of 86 compared to the benchmark compound 2. The rIP of compound 8 was 58 compared to 2. Absolute binding affinities were also determined using Surface Plasmon Resonance (SPR) measurements. Dissociation constants (K D ) were obtained for 7 and 8 (K D = 182 nM and K D = 175 nM, respectively) ( Fig. 9 and Supplementary Figure 3).

Discussion
In the current study, we have demonstrated the binding of high-affinity Siglec-2 ligands directly to BL Daudi cells using NMR spectroscopy. Our NMR-derived results suggest that ligand binding occurs exclusively to Siglec-2 present on BL Daudi cells. Control NMR experiments using HEK293T cells that naturally express Siglec-2 at a very low level revealed very weak ligand STD NMR signals, whereas Siglec-2 transfected HEK293T cells showed a significant increase due to the availability of an increased number of Siglec-2 binding site. In an additional control experiment, by spiking the ligand-BL Daudi cell complex with a non-binding spy molecule sucrose, we have shown that sucrose does not bind to the cells and that therefore the binding of the ligand is specific to Siglec-2 expressed on the BL Daudi cells. The likelihood that the synthesised N-acetylneuraminic acid-based ligands also bind to other Siglecs (e.g. Siglec-1, -3, -4, -5, -6, -7, -8, -9, -10, -11, -14 and -16) present on BL Daudi cells is remote as they are present in low populations (Table 1). In addition, STD NMR data for 9-BPC-4-mNPC-Neu5Acα 2Me (3) in the presence of the Siglec-2 high cell fraction clearly revealed significantly higher STD NMR signal intensity due to the presence of an increased number of Siglec-2 ligand binding sites. Interestingly, our previously reported analysis of N-acetylneuraminic acid derivatives in complex with various Siglecs revealed that binding to different Siglecs resulted in distinct binding epitopes 11 . The cell-derived binding epitope of 3 presented in the current study was similar to the ligand in complex with recombinant protein and was used to guide the synthesis of 7 that showed a 88-fold affinity in binding affinity. The close proximity of the equatorial proton at C-3 to the protein surface led to the introduction of a C-3 hydroxyl group. Strong relative STD NMR effects of H o and H p of 3 adjacent to the meta-nitro group in ring A have led to the introduction of a meta-methoxy group. Finally, STD NMR signals revealed the importance of the C-2 methoxy aglycon moiety of 3 and a more bulkier functionalized [1,2,3]-triazole ring was introduced. Overall, we have shown that the use of whole BL Daudi cells in NMR binding experiments is a straightforward approach and can add invaluable structural information to the design process of second-generation high-affinity ligands. We have also demonstrated for the first time disruption of cis-binding directly at cell-level by NMR spectroscopy.
In conclusion, our whole-cell NMR approach does not require cloning, expression and purification of the target protein and NMR experiments can be acquired in less than an hour. This is the first time a STD NMR-based study of high affinity Siglec-2 ligands directly at cell level has been described and the approach, in general, provides new avenues in the development of tailored therapeutics and personalised medicines. Specifically, our approach could enable the direct assessment of ligand and inhibitor affinity using patient-derived cells in a fast and efficient manner. Moreover, the obtained structural information could be utilised to design novel inhibitors and potential drugs with higher affinity and efficacy. A few challenges in the preparation of samples for cell-based NMR experiments remain. Samples of high homogeneity are essential and cells need to survive the acquisition time, often in non-optimal buffer conditions suitable for NMR acquisition. However, with further development of cryogenic probe technology, reduced NMR acquisition time will offer an excellent platform for structure-guided inhibitor design at cell level.  NMR experiments using BL Daudi and HEK293T cells. STD NMR spectra were performed using a Bruker Advance 600 MHz spectrometer, equipped with a 5-mm TXI probe with triple axis gradients at 283 K. 5 × 10 5 BL Daudi and 1 × 10 6 HEK293T cells were washed three times with the optimised deuterated NMR buffer by centrifugation at 1,300 rpm, at room temperature for 5 min and resuspending in 200 μ L. The ligand concentration was adjusted to 0.5 mM for all experiments. To determine the binding epitope of the ligands when bound to cells STDD (Saturation Transfer Double Difference) spectra were generated by subtracting STD NMR signals from the ligand-cell complex from the spectrum obtained in the absence of cells or protein.

Methods
Transfection of HEK293T cells with Sigle-2 coding sequence. The full-length codon-optimized coding sequence of Siglec-2 (Uniprot entry P20273-1) cloned in plasmid pD2517 was purchased from DNA2.0 (Menlo Park, CA, USA). Plasmid amplification was performed by transformation into chemically competent E. coli NEB 5-alpha (Ipswich, MA, USA) followed by purification using a NucleoBond XTRA EF Plasmid Purification Kit (Machery Nagel, Duren, Germany, EU). HEK293T transfection was performed using Lipofectamine 3000 Transfection Reagent Kit (Thermo Fisher Scientific, Carlsbad, CA, USA) according to the supplier's standard two-steps optimization protocol in Tissue Culture treated 6-wells plates (Corning, Corning, NY, USA) using 500 ng of purified plasmid per well. Transfection efficiency was monitored by flow cytometry analysis 24 and 48 h post-transfection. Briefly, the cells were detached with PBS supplemented with EDTA at a final concentration of 5 mM and subsequently stained with mouse anti-human Siglec-2 primary antibody (clone RFB4) and anti-mouse secondary antibody BV421 conjugated.

Periodate treatment of BL Daudi cells.
Mild periodate treatment to disrupt cis-binding was performed following published procedures 28 . BL Daudi cells were resuspended in phosphate buffer (pH 7.4) containing freshly dissolved 2 mM NaIO 4 and incubated at 4 °C in the dark. After 30 min, the excess of periodate was quenched by adding 10 μ L of 20% deuterated glycerol followed by immediate washing with deuterated buffer optimised for cell NMR experiments. The STD NMR on-resonance frequency was set to − 1.00 ppm and the off-resonance to 300 ppm. To ensure the viability of the cells during NMR acquisition 256 scans were chosen resulting in a total acquisition time of 53 min.

Synthesis of Siglec-2 ligands.
The N-acetylneuraminic acid derivatives 7 and 8 and intermediates were prepared according to Fig. 6. Experimental protocols, a detailed scheme, 1 H and 13 C NMR spectra for all compounds are provided in the Supplementary Material. Production of recombinant Siglecs. Recombinant Siglec-2 (CD22) Fc-chimeras were expressed in CHO-Lec1 and protein purification was performed as described 34 . The recombinant Siglec-2-Fc chimera, in which the three N-terminal Ig-domains of Siglec-2 were fused to the C-terminal human IgG1 Fc domain, was expressed in CHO-Lec1 cells.
Scientific RepoRts | 6:36012 | DOI: 10.1038/srep36012 Hapten inhibitions assays. Hapten inhibition assays were performed following published procedures 34 . In brief: Fetuin was used as target in hapten inhibition assays, since Siglec-2 binds to this glycoprotein containing α (2,3)-and α (2,6)-linked N-acetylneuraminic acids. The assays were performed in 384 wells microtitre plates determining siglec-Fc binding to immobilized fetuin in the presence of increasing concentrations of compounds to be tested as competitive inhibitors. Specific binding was obtained by subtracting unspecific binding to immobilized desialylated fetuin as target. The half maximal inhibitory concentration of siglec binding (IC 50 ) was determined from corresponding binding curves. Average IC 50 values and corresponding standard deviations were calculated from these experiments. Each sample was measured in triplicates and each experiment was repeated at least three times. Standard deviations for all compounds were below 15% of the average IC 50 values. For a better comparison between different assays, for each sample relative IC 50 (rIP) values were calculated.
Surface plasmon resonance (SPR) assays. Siglec-Fc chimeras were either captured to anti-Fc antibody-derivatised dextran surface of a Reichert SR7500DC flow cell or directly immobilized by amide coupling. N-acetylneuraminic acid derivatives at various concentrations in PBS were passed over the flow cell at a flow rate of 20 μ L/min followed by buffer for dissociation. Specific binding was determined by subtracting the signal obtained from the reference cell (ethanolamine derivatised dextran surface) from the sample cell derivatised with Siglec-2 Fc chimeras. The data were analysed and fitted using the software Scrubber (BioLogic). Corresponding sensorgrams for 7 and 8 are shown in Supplementary Figure 2a,b, respectively.