A Peptide Mimetic of 5-Acetylneuraminic Acid-Galactose Binds with High Avidity to Siglecs and NKG2D

We previously identified several peptide sequences that mimicked the terminal sugars of complex glycans. Using plant lectins as analogs of lectin-type cell-surface receptors, a tetravalent form of a peptide with the sequence NPSHPLSG, designated svH1C, bound with high avidity to lectins specific for glycans with terminal 5-acetylneuraminic acid (Neu5Ac)-galactose (Gal)/N-acetylgalactosamine (GalNAc) sequences. In this report, we show by circular dichroism and NMR spectra that svH1C lacks an ordered structure and thus interacts with binding sites from a flexible conformation. The peptide binds with high avidity to several recombinant human siglec receptors that bind preferentially to Neu5Ac(α2,3)Gal, Neu5Ac(α2,6)GalNAc or Neu5Ac(α2,8)Neu5Ac ligands. In addition, the peptide bound the receptor NKG2D, which contains a lectin-like domain that binds Neu5Ac(α2,3)Gal. The peptide bound to these receptors with a KD in the range of 0.6 to 1 μM. Binding to these receptors was inhibited by the glycoprotein fetuin, which contains multiple glycans that terminate in Neu5Ac(α2,3)Gal or Neu5Ac(α2,6)Gal, and by sialyllactose. Binding of svH1C was not detected with CLEC9a, CLEC10a or DC-SIGN, which are lectin-type receptors specific for other sugars. Incubation of neuraminidase-treated human peripheral blood mononuclear cells with svH1C resulted in binding of the peptide to a subset of the CD14+ monocyte population. Tyrosine phosphorylation of siglecs decreased dramatically when peripheral blood mononuclear cells were treated with 100 nM svH1C. Subcutaneous, alternate-day injections of svH1C into mice induced several-fold increases in populations of several types of immune cells in the peritoneal cavity. These results support the conclusion that svH1C mimics Neu5Ac-containing sequences and interacts with cell-surface receptors with avidities sufficient to induce biological responses at low concentrations. The attenuation of inhibitory receptors suggests that svH1C has characteristics of a checkpoint inhibitor.

Some receptors may have non-glycan ligands in vivo yet contain an extracellular lectin-like domain. The receptor NKG2D on natural killer (NK) cells, γδ T cells and CD8 + cytotoxic T cells is regulated by endogenous polypeptide ligands such as MICA/B, ULBP, Rae-1 or H60 [24][25][26], but NKG2D also contains a lectin domain adjacent to the polypeptide binding site that binds Neu5Ac(α2,3)Gal-sequences [27].
Because siglecs are important regulators of the immune system, ligands with high affinity should provide valuable tools to address therapeutic opportunities [11,12,28]. A question of interest is how to design ligands that bind to these regulatory receptors with sufficient avidity and specificity to achieve manipulation of the immune system. To explore this possibility, we asked whether short peptides can mimic the ligands of lectin receptors, including siglecs, for this purpose. Peptide mimetics of sugars have potential advantages over glycans and glycoproteins because of the ease of chemical synthesis, their flexibility in design, and favorable physical properties. Multivalent peptides can be constructed that have much higher avidities to lectins than monosaccharides and are similar in binding avidity to natural multivalent glycoproteins and glycoconjugates such as fetuin and mucin [29,30]. A number of peptides that mimic sugars have been identified, some of which closely resemble specific sugars and bind to oligosaccharide-binding antibodies [31][32][33][34][35][36]. Some peptides can functionally mimic a sugar, such as those with the consensus core sequence YPY that inhibit the mitogenic activity of the Man-specific lectin concanavalin A, yet bind at a site different from the saccharide-binding site [37,38]. Peptide mimetics of carbohydrate antigens have been studied as vaccines to elicit antibodies against sugar antigens, including those on the surface of HIV, and complex oligosaccharides [36,39].
We initially identified several sequences of amino acids that bound to specific plant lectins by screening phage display libraries [40]. Sequences were further refined by in silico modeling of docking of a peptide into the sugar binding site of the crystal structures of lectins. These studies predicted that several short peptide sequences, 5 to 8 amino acids in length, would have significant affinity to lectins chosen as analogs of cell-surface receptors. The sequences identified by this approach were incorporated into tetravalent structures, a design based on the concept of avidity as a function of ligand density and entropic factors [41][42][43] and also to accommodate the possibility that cross-linking of receptors may be required for modulation of signal transduction pathways [44,45]. In binding assays, the peptide with the sequence HPSLK (designated sv6B) had characteristics of a general sugar mimetic and bound to several lectins that bind monosaccharides such as Neu5Ac, Gal, GalNAc or fucose with avidities several orders of magnitude greater than their natural ligands [29,30]. Surprisingly, when the HPSL core was retained but the peptide lengthened to NPSHPSLG (designated svH1D), and in particular the SL!LS inversion to generate NPSHPLSG (designated svH1C), the peptides did not bind to lectins that bound monosaccharides but bound strongly to those specific for di-or trisaccharides that terminate with Neu5Ac-Gal. When svH1C was added to cultures of human peripheral blood mononuclear cells (PBMCs), endocytosis of opsonized microspheres by adherent cells was activated [29,30]. Additionally, svH1C stimulated phosphorylation of STAT6, an indication that the peptide triggered a receptor-mediated signal transduction pathway [29]. These observations suggested that the peptide has biological activity and may bind productively to specific receptors. As a step to further understand the mechanism of action of svH1C, we hypothesized that, as a mimetic of Neu5Ac-Gal, the peptide binds to receptors such as siglecs and NKG2D that are specific for these sugars.
The amino acid residues in the binding sites of many lectins and receptors have been identified and, although the protein structures obtained from data bases are rigid, the validity of the in silico approach to predictions of binding energies for peptides is supported by the concept that flexible ligands select specific conformations of a binding site according to the Monod-Wyman-Changeux hypothesis [46]. To confirm that the peptide interacted with a glycan binding site from a flexible, non-structured conformation, CD and NMR spectra of svH1C were examined. Although an unusual CD signal was observed at elevated concentrations, a specific structure for the peptide was not observed by NMR spectroscopy. We designed solid-phase assays in which binding of biotin-tagged svH1C to recombinant receptors occurred under conditions in which the peptide sequences retained full flexibility. We also observed binding of the peptide to a subset of the monocyte population in human PBMCs after treatment of cells with neuraminidase. Furthermore, svH1C induced activation and proliferation of immune cells in the peritoneal cavity of mice.

Ethics Statement
Studies of peritoneal cells were performed by Biomodels, LLC, Watertown, MA, which is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. The protocol was approved by Biomodels' IACUC and has IACUC approval number 13-0611-01. C57BL/6 mice were obtained from Charles River Laboratories (Wilmington, MA). The studies were performed in animal rooms provided with HEPA filtered air at a temperature of 70°F +/-5°F and 50% +/-20% relative humidity. Animal rooms were set to maintain a minimum of 12 to 15 air changes per hour. The room was on an automatic timer for a light/dark cycle of 12 hours on and 12 hours off with no twilight. A cage card or label with the appropriate information necessary to identify the study, dose, animal number, and treatment group marked all cages. Animals were identified by an ear punch corresponding to an individual number. Animals were fed with a sterile Purina Labdiet 5053 rodent diet and water was provided ad libitum.
Human peripheral blood mononuclear cells (PBMCs) were prepared from purchased discarded cellular residues of a plasma preparation procedure (TRIMA) performed on blood of anonymous donors at the Blood Centers of the Pacific, San Francisco, CA, and sold commercially for research use.

Synthesis of svH1C
The amino acid sequence identified previously [29,30] as a Neu5Ac-Gal-mimetic was incorporated into a tetravalent peptide that has the structure [(NPSHPLSGGGGS) 2 K] 2 K-NH 2 . The tetravalent peptide was synthesized by CBL Biopharma (Patras, Greece) on a tri-lysine core by standard procedures [47] utilizing Fmoc (9-fluorenylmethoxycarbonyl)-protected amino acids. The sequence-GGGS-is a linker that extends the mimetic from the tri-lysine core. Modifications at the C-terminus made during synthesis consisted of an amide group (no tag) or an extension of the peptide by addition of ε-biotinyl-lysine-amide for the binding assays. The 'arms' were synthesized separately by solid-phase synthesis by standard chemistry and subsequently condensed in solution with the four amino groups generated by the tri-lysine core [48]. Quality of the synthetic product was assessed by HPLC and mass spectroscopy. Purity of the peptide was 97.7%, with net peptide content 73% without counter-ion and 87% when the counter-ion was included. The remaining material was largely water of hydration of the polar peptide. Solutions of the peptide in water were neutralized and passed through a 1 x 10 cm column of DEAE-Sephadex A-25, Clform, at pH 5 to 6 to remove trifluoroacetic acid.
For in vivo studies, svH1C was prepared by absorption of 1 g peptide on a 2.5 X 10 cm column of CM-Sephadex C-25 in 100 mM NaCl. The column was washed with 200 mM NaCl and svH1C was then eluted with 500 mM NaCl that was prepared in endotoxin-free water (<0.005 EU/mL, HyClone Laboratories, Logan, UT). Endotoxin was undetectable (<0.01 EU/ 10 mg peptide) in the eluted peptide as assayed by a quantitative colorimetric LAL assay (Lonza, Walkersville, MD). Peptide concentration was determined by the bicinchoninic acid assay (Pierce Biotechnology, Rockland, IL) using a dansylated peptide (extinction coefficient, ε mM = 5.7 cm -1 at 336 nm) as standard or with a calculated [49,50] svH1C extinction coefficient of 25 OD 210 nm /mg/mL.

CD and NMR
CD spectra were obtained with 100 μM svH1C dissolved in 50 mM sodium borate, pH 9.0, using a JASCO J-170 CD spectropolarimeter operated in continuous scan mode and 200 nm/ min scan speed. Spectra were generated from five serial accumulations. Samples were either diluted with water or adjusted to pH 2 for additional spectra.
Samples of the tetravalent peptide in H 2 O/D 2 O (90%/10% v/v) were analyzed by NMR at 298K with a Bruker Avance DPX 400 MHz spectrometer. 1 H 1D NMR spectra were recorded using spectral width of 12-17 ppm with presaturation of the H 2 O signal. 1 H-1 H 2D TOCSY [51,52] were recorded using the MLEV-17 spin lock sequence using τ m = 80 ms, and 1 H-1 H TPPI NOESY [53,54] spectra were acquired using mixing time τ m = 250 or 300 ms applying water suppression during the relaxation delay and mixing time. All 2D spectra were acquired with 16.0204 ppm spectral width, consisting of 2K data points in the F2 dimension, 64 transients and 512-1024 complex increments in the F1 dimension. Raw data were multiplied in both dimensions by a pure cosine-squared bell window function and Fourier-transformed to obtain 2048×2048 real data points. A polynomial base-line correction was applied in both directions. For data processing and spectral analysis, the standard Bruker software (Topspin 1.3) and XEASY program [55] (ETH, Zurich) were used.

Binding assays
Isothermal microcalorimetry was performed by TA Instruments Professional Services (New Castle, DE) by titrating 10 μM NKG2D (monomer concentration) (R&D Systems, Inc., Minneapolis, MN) with 100 μM svH1C in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1 mM EDTA. Recombinant human receptors were obtained from R&D Systems as chimeric proteins. NKG2D is a type II transmembrane protein and the extracellular lectin-type domain was fused at the N-terminus with a human IgG 1 Fc domain. Recombinant siglecs contained a C-terminal IgG 1 Fc domain or a His 6-10 tag. The lyophilized proteins were reconstituted in phosphatebuffered saline, pH 7.2 (PBS), to 100 μg/mL. The Fc-chimeric proteins were bound to protein A/G-coated wells of microtiter plates (Pierce Biotechnology). His-tagged proteins were bound in Nickel-coated plates (Pierce Biotechnology). Sufficient proteins were added to wells to saturate the binding capacity of the wells. After a wash to remove unbound receptor, wells were blocked with 1% gelatin in wash buffer. Wash buffers were either 10 mM phosphate, pH 7.2; 10 mM HEPES, pH 7.4; or 50 mM Tris-HCl, pH.7.5, each in 150 mM NaCl containing 0.05% Tween-20. Biotinylated peptides were prepared as 2 μM solutions in the wash buffer and 50 μL (100 pmoles) were added to each well and incubated at room temperature for one hour. After four washes with 350 μL buffer containing 0.05% Tween-20, each assay was incubated with 50 μL of 1 μg/mL streptavidin-peroxidase (Sigma-Aldrich, St. Louis, MO). The wells were again washed four times with buffer-Tween-20 and then 50 μL peroxidase substrate (1-Step ULTRA TMB, Pierce Biotechnology) were added. The reaction was allowed to proceed at room temperature for 5 min, and the reaction was stopped with 50 μL of 2 N H 2 SO 4 or 1 M H 3 PO 4 . Absorbance was measured at 450 nm and converted to the amount of streptavidin-peroxidase bound with the specific activity of the conjugate determined under the conditions of the assay. Absorbance of assays with svH1C was corrected for background levels of binding of streptavidin to the receptors in wells without a peptide or with an inactive peptide, which were less than 10% of the value with bound peptide, except for Siglec-5 and Siglec-14. Binding of streptavidin to these siglecs caused a background level of nearly 50% of binding in the presence of svH1C. Interestingly, the sequences of the first 200 amino acids of these two siglecs are identical [18], which suggests that streptavidin bound directly to these receptors. For inhibition studies, fetal bovine fetuin (purity, 88%; cat. no. 341506, Calbiochem, EMD Chemicals, Philadelphia, PA) and sialyllactose (cat. no. A0828, Sigma-Aldrich) were added to the assay along with svH1C.
Molecular modeling was performed of docking of a single (monovalent) sequence into the glycan-binding site of cell-surface receptors. The crystal structures of the receptors were downloaded from the Protein Data Bank into the docking mode of ArgusLab 4.0.1 software (Mark A Thompson, Planaria Software LLC, Seattle, WA, http://www.arguslab.com), which was used for modeling. Amino acid residues that comprise the binding site of a receptor were selected from the literature that describes each protein.

Phospho-immunoreceptor assay
Human PBMCs were incubated at a density of 2 x 10 6 cells/mL overnight in RPMI-1640 medium supplemented with 0.1% ovalbumin, 1:100 dilution of penicillin-streptomycin solution (Mediatech, Inc., Herndon, VA), and 1 mM glutamine at 37°C and 5% CO 2 . svH1C was added to a final concentration of 100 nM, and after 5 min the cells were collected by centrifugation at 500g for 2 min. The cells were lysed and spread on an array of capture antibodies against cell-surface receptors on immune cells (Human Phospho-Immunoreceptor Array, R&D Systems). The protocol for detection of phosphorylated receptors with an anti-phospho-tyrosine antibody conjugated to horseradish peroxidase and detection by chemiluminescence with xray film was followed as described by the manufacturer. Spots on the film were analyzed by ImageJ software.

In vivo studies
C57Bl/6 mice, 22-24 g body weight, were injected subcutaneously with 1 nmole svH1C per g on days 0, 2 and 4. On days 1, 3 and 5, primary peritoneal cells from three animals were isolated and pooled, and the populations of macrophages, dendritic cells (DCs), CD3 + , CD4 + and CD8 + T cells, natural killer (NK) cells and B cells were evaluated by flow cytometry with specific biomarkers. FACS was performed using a MACSQuant Analyzer (Miltenyi Biotec, Teterow, Germany). Data were analyzed with FlowJo software, version 10.0.6.

Insight into the structure of svH1C
As an initial approach to structural analysis, CD spectra were performed with svH1C in borate buffer, pH 9. The spectrum for a 100 μM solution of the peptide had a strong negative signal at 234 nm (Fig 1, curve A). The minimum broadened and shifted to 227 nm when diluted to 25 μM (Fig 1, curve B), and disappeared when the peptide was diluted to 10 μM, with a new minimum at 203 (Fig 1, curve C). The unusual spectra at higher concentrations may be a function of the tetravalent structure of the peptide, because similar negative peaks were observed between 225 nm and 235 nm with 100 μM solutions of several other tetravalent peptides with similarly short but different sequences. When the sample that provided spectrum A in Fig 1 was acidified to pH 2, the minimum at 234 nm was replaced with a minimum at 211 nm and a shoulder at about 230 nm (Fig 1, curve D). This spectrum suggests a weak α-helical structure [56], but formation of helical structures is restrained by the two proline residues in each arm. These spectra suggest that the peptide aggregates at the higher concentrations. However, because the receptor binding studies were performed with 2 μM peptide, svH1C was likely in a random, flexible structure under those conditions. NMR spectra were determined with solutions of 1 mM svH1C in PBS (S1 Fig). Attempts to perform sequence specific assignment for the peptide faced the drawback of resonance degeneration and lack of a defined tertiary conformation, indicative of flexibility of the svH1C arms. Analysis of the TOCSY spectra was based on the hypothesis that the identical residues existing in each unit of the tetramer resonate at the same (or extremely similar) field. Consequently, the individual spin patterns were identified according to (i) the intra-residue connectivities, and (ii) the characteristic spin-patterns of each of the amino acids present in the peptide sequence. A single set of peaks was assigned to each of the residues of the svH1C segment/unit: [NPSHPLSGGGGS]K. Because of the different chemical structures of the three lysine residues, which are linked not only through the typical peptide bonds but also through iso-peptide bonds (connectivities between the iso-peptide HN groups and the side-chain protons were identified for two of the three lysine residues), we managed to attribute three spin-patterns to the three lysine residues of the svH1C peptide (S1 Table). Standard sequence specific proton assignment was also prevented by the fact that NOESY spectra (250-300 ms mixing time) lacked any meaningful inter-residue connectivity (i.e., Hα(i)-HN(i+1)). Data from the ΗN regions of the 2D 1 H-1 H TOCSY spectra are shown in Fig 2 (the 1 H 1D spectrum is shown in S1 Fig), which suggest the absence of a well-ordered structure. These analyses indicate that binding to lectin-type receptors occurred with flexible peptide arms capable of forming conformations that accommodate glycan binding sites.
Binding of svH1C to siglecs Tetravalent svH1C, with the mimetic sequence NPSHPLSG, bound selectively to lectins from Sambucus nigra (SNA1) and Maackia amurensis (MAA) [29,30]. MAA is a mixture of M. amurensis leukoagglutinin (MAL) and M. amurensis hemagglutinin (MAH), both of which along with SNA1 bind preferentially to terminal Neu5Ac-Gal-sequences on complex glycans [57]. Binding of the peptide to several lectins specific for monosaccharides was not detected [30]. We then assayed binding of svH1C to several siglecs, which comprise a family of cell- surface receptors that are specific for Neu5Ac-Gal/GalNAc-sequences. Within this family of receptors are proteins that bind to Neu5Ac(α2,3 or α2,6)Gal, Neu5Ac(α2,6)GalNAc or Neu5Ac(α2,8)Neu5Ac [8][9][10][11][12]. For binding assays, sufficient recombinant homodimeric, chimeric proteins containing a siglec domain and a Fc domain per subunit were added to protein A/G-coated microtiter wells to saturate the binding sites. Recombinant Siglec-1 contained a Cterminal His tag rather than a Fc domain and was bound in Nickel-coated wells. Non-specific sites in the wells were blocked with 1% gelatin-0.05% Tween-20 in the wash buffer.
The composition of the buffer had a significant effect on binding of peptide to individual siglecs. Interestingly, Siglec-2 and Siglec-3 bound svH1C as strongly as the other siglecs in assays with Tris buffer. In contrast, the most discrimination of binding between the various siglecs was observed when HEPES buffer was used, with little or no binding detected to Siglec-2 or Siglec-3, which was similar to results in PBS (Fig 3). Binding of svH1C to Siglec-7, -9, and -11 was significantly less with HEPES buffer as compared with PBS. These results suggest that the peptide binds to all the siglecs but with differing avidities, with only the strongest interactions surviving the extensive washes. Binding of the peptide was inhibited by the Neu5Ac-rich, multivalent glycoprotein, fetuin.
Consistent with results from binding of peptides to lectins [29,30], biotinylated tetravalent peptides with the sequence HPSLK (sv6B) and NPSHPSLG (svH1D) also bound to siglecs, although with avidity patterns different from that of svH1C. In contrast, no binding occurred with a peptide with the structure [(VGGGSGGGS) 2 K] 2 K-ε-biotinyl-K-NH 2 , which was used as a negative control. No binding was detected with the lectin receptors CLEC9a (ligand unknown), CLEC10a (specific for GalNAc) or DC-SIGN (specific for Man). Binding of svH1C to these receptors was not detected regardless of the buffer used. Addition of CaCl 2 and MgCl 2 (1 mM each) did not increase binding of svH1C.

Binding of svH1C to NKG2D
NKG2D is a homodimeric, type II receptor whose recognized ligands are cell-surface proteins MICA and MICB, which are major histocompatibility complex class I homologs. Other polypeptide ligands include ULBP, Rae-1 and H60 [26]. Extensive studies of binding of MICA and MICB [24][25][26] have shown that each homodimer binds a single ligand at the interface of the subunits. Each subunit of NKG2D also contains a lectin domain that binds glycans with specificity for Neu5Ac(α2,3)Gal [27]. With the assay as used for siglecs, we found that svH1C bound strongly to NKG2D. Binding was inhibited by fetuin (Fig 4A), with an IC 50 of approximately 3 μM protein, when added with 2 μM svH1C (Fig 4B). Fetuin was as effective an inhibitor of binding of svH1C to NKG2D as it was of the binding of the peptide to siglecs. Fetuin contains three N-linked and three O-linked glycans per molecule [58][59][60][61][62], with trisialylated or tetrasialylated termini that are linked nearly equally as Neu5Ac(α2,3)Gal or Neu5Ac(α2,6)Gal. We previously found fetuin effectively competed with svH1C for binding to plant lectins, and that digestion of fetuin with neuraminidase nearly eliminated its ability to inhibit binding of svH1C [29]. From the number of multivalent glycans on each fetuin molecule, we concluded that svH1C effectively competed with natural multivalent glycan ligands for the same binding site on these receptors.
Inhibition of binding of svH1C to NKG2D by sialyllactose further supported binding of the peptide in the glycan-binding site. Sialyllactose was a relatively weak inhibitor (IC 50 18 μM), which may have resulted because the Neu5Ac(α2,3)Gal linkage, which is the preferred ligand for NKG2D, was a minor component of the mixture of sialyllactose isomers, among which the isomer with the Neu5Ac(α2,6)Gal linkage was the major form (Fig 4A and 4B). We then assayed binding of svH1C to NKG2D directly by isothermal microcalorimetry. svH1C (100 μM) was titrated into a solution of NKG2D (10 μM, subunit concentration) and changes in heat content of the system were measured. The concentration of NKG2D was introduced into the data analysis as the total number of lectin-like binding sites. The titration curve (S2 Fig) yielded a K A = 1.7 x 10 6 M -1 (K D = 0.58 μM). The dissociation constant corresponds to a ΔG' = -38.4 kJ/mol. The ΔH of the binding reaction determined by microcalorimetry was positive, 16.4 kJ/mol, which suggested that entropy was a major factor in the energy of binding. The affinity of svH1C for NKG2D was similar to that of the endogenous polypeptide ligands, which have K D values in the range of 0.6 to 1.0 μM [24,25].
In the analysis of the microcalorimetry data, the stoichiometry, n, yielded a value of 0.71, which suggests that each binding site on approximately half of the dimeric protein was filled with one peptide molecule while the binding sites on the remaining dimers were either crosslinked by one peptide or that only one of the subunits was bound with a peptide. If each arm of the tetravalent structure was in the extended β-conformation, as the CD and NMR data in Figs 1 and 2 suggest, the distance from each mimetic domain to another is approximately 70Å. This distance is sufficient to span the approximately 34Å between the lectin domains in the structure of the dimeric NKG2D [24]. Thus it is feasible that the two binding sites on the dimeric protein are accessible to a single peptide molecule. As an alternate interpretation of the stoichiometry, a fraction of the receptor may not have achieved the native conformation during reconstitution.
The inhibition by fetuin of binding to siglecs and NKG2D indicated that the dissociation constants for svH1C were similar with the two types of receptors. Molecular modeling was used to compare the predicted energies of binding of the peptide to Siglec-5 and NKG2D. The glycan binding site for Siglec-5 was described by Zhuravleva et al. [23]. The predicted binding energy calculated by the docking software was -47 kJ/mol (Fig 5A). A model of NKG2D, with the binding site as described by Li et al. [24] and McFarland et al. [25], provided a predicted binding energy of -40 kJ/mol (Fig 5B). This value is similar to the binding energy calculated from the dissociation constant determined by microcalorimetry.
To determine whether binding of svH1C to receptors could be observed with whole cells, human PBMCs were incubated on ice with biotinylated svH1C in PBS for 30 min, washed with PBS, stained with labeled antibodies against cell-surface markers plus streptavidin-PerCP-Cy5.5, washed and fixed. By flow cytometry, streptavidin was detected on small subsets of NK, CD4 + and CD8 + T cells. However, no binding was detected with monocytes. Collins et al. [63] and Zhang and Varki [64] found that binding of multivalent Neu5Ac-containing glycans to B cells was detected only after digestion of the cells with neuraminidase to remove cellsurface Neu5Ac. We repeated the experiment to include a treatment of PBMCs with neuraminidase at 37°C before cooling the cells on ice and incubating with biotin-tagged svH1V on ice. The binding step was performed on ice to prevent endocytosis of the peptide that could be mediated by siglecs [11,12]. As shown in Fig 6B, significant binding was found with a subpopulation among CD14 + monocytes after cells were treated with neuraminidase. Treatment with neuraminidase did not have a significant effect on binding of svH1C to NK, CD4 + or CD8 + T cells. Histograms of samples treated with neuraminidase that was inactivated by heating in boiling water (Fig 6A) were similar to the control without peptide.

Effect of svH1C on receptor phosphorylation
We then asked whether svH1C would alter the phosphorylation status of cell-surface receptors. Human PBMCs were cultured overnight in RPMI-1640 medium supplemented with 0.1% ovalbumin, which was added in place of serum to minimize effects of possible regulatory factors in serum. svH1C was added to 100 nM and cells were collected after 5 min of further incubation. Cells were lysed and spread on an array of cell-surface receptor antibodies (see Materials and Methods). After washes, an anti-phospho-tyrosine antibody conjugated to horseradish peroxidase was added and phosphorylated tyrosine was detected by chemiluminescence.
Sections of films corresponding to a selection of receptors are shown in Fig 7. Phosphorylated siglecs included on the array were decreased by more than half, with the greatest reduction on Siglecs-3, -5 and -7 in cells treated with svH1C. The amounts of phosphorylated CD229/SLAMF3, LAIR-1 and FcγRIIA were also dramatically reduced. However, phosphorylation of FcRH4/IRTA1, NKp46 and SHP-1 was not changed, while that of BLAME/SLAMF8 increased. The major phosphorylated protein, the ε subunit of the CD3 T cell receptor complex, was slightly reduced. These effects indicated that the peptide interacted with cells and, by inducing dephosphorylation, attenuated the activity of inhibitory receptors.

Effect of svH1C on immune cells in vivo
To determine whether a decrease in activity of inhibitory receptors is reflected by proliferation of immune cells in vivo, svH1C was injected subcutaneously every other day into C57Bl/6 mice at a dose of 1 nmole per g body weight. Injections were administered on day 0, 2 and 4, and peritoneal lavage was performed to obtain primary cells 24 h after each injection from different groups of mice. Cells from three animals at each time point were pooled and populations of immune cells were measured by flow cytometry. We found in preliminary experiments that significant changes in cell populations did not occur in the 24 h after the first injection. However, as illustrated in Fig 8, populations of most cells types increased over the period of treatment. In particular, populations of DCs (CD11c + ), NK cells (NK1.1 + ), T cells (CD3 + ), T H cells (CD4 + ), cytotoxic T cells (CD8 + ), and B cells (CD19 + ) increased several-fold, including those that expressed the activation marker CD69. Memory B cells that expressed CD73, CD80 and CD273 increased significantly.

Discussion
Our data support peptide svH1C as a mimetic of Neu5Ac-Gal/GalNAc or Neu5Ac-Neu5Ac. The CD spectra of svH1C suggest that the arms of the peptide aggregate at a sufficiently high concentration. The aggregation may reflect the PxxP sequence in the peptide, which is similar to collagen. Indeed, the CD spectrum of gelatin obtained under the same conditions had a minimum at 238 nm. Interestingly, the negative signal at 234 nm in the spectrum at 100 μM is fortuitously similar to that of sialic acid lactones, including the sequence Neu5Ac-Gal [65], but this minimum disappeared as the peptide was diluted. The tetravalent structure of the peptide provides a high local concentration within the sphere of a single molecule, but this feature is not sufficient to produce a CD signal. Analysis of solutions of the peptide by NMR spectroscopy did not reveal a definite structure, which suggested that the 'arms' are extended and have a high degree of flexibility. At the low concentrations used for the binding studies, the peptide apparently achieved a random structure. Evidence for a high degree of flexibility does not provide information on a specific binding mode. It simply provides an idea for the plasticity of the peptide and its tendency to explore conformational space and adopt the preferred conformation for partner recognition and interaction. The flexibility of the peptide allows it to interact with multiple members of the siglec family of receptors.
The flexibility of the arms of the tetravalent structure was retained in binding assays. When added to streptavidin-coated wells, the biotinylated peptide was anchored at the C-terminus, which allowed full flexibility of the arms [29]. Likewise, when the receptors were first bound to wells via a Fc domain or a His extension, the interaction with peptide occurred in solution and the peptide that remained bound after extensive washes was then detected by streptavidin binding to the C-terminal biotin tag.
Because svH1C bound to plant lectins specific for Neu5Ac-Gal sequences, we tested the ability of the peptide to bind to recombinant human lectin-type receptors that bind ligands containing these glycans. The siglec family is the most prominent among the receptors for which sialylated glycans serve as ligands. Differences in binding of svH1C were found among those tested, with strong binding to Siglec-1, -5, -7, -11, -9 and -14 but less with Siglecs-2, and -3 in PBS. The low binding detected with several siglecs was sensitive to the conditions of the assay. It is likely that svH1C binds to Siglecs-2 and -3 but that the interaction does not survive the extensive washes with buffer. However, binding to other lectin-type receptors such as CLEC9a, CLEC10a and DC-SIGN was not promoted by use of other buffers or inclusion of Ca 2+ . These observations suggest that the standard assay was capable of discriminating within a range of avidity and specificity. A common feature between Siglec-5 and -14 is their preference for a terminal Neu5Ac(α2,8)Neu5Ac or Neu5Ac(α2,6)GalNAc sequence [17]. Siglec-1 is specific for terminal Neu5Ac(α2,3)Gal, which is also a ligand for NKG2D [27]. The inhibition of binding by the multivalent fetuin suggests that the peptide interacts with the glycan binding sites of these receptors.
The abundance of sialylated glycans on the cell surface, estimated to provide a local Neu5Ac concentration of 110 mM on B cells [13,63], and the many receptors that bind Neu5Ac-containing glycans call into question whether a drug can bind with sufficient avidity and specificity to achieve a biological response. Cell-surface proteins are heavily glycosylated, which allows siglecs to bind sialylated proteins on the same cell (cis) or on neighboring cells (trans) and serve as adhesion proteins [63,64]. Siglecs tend to be specific for the linkage between terminal Neu5Ac and the penultimate sugar, usually Gal, which suggests a certain rigidity of the binding site and the ligand. A flexible peptide can therefore adjust its confirmation to fit into a rigid binding site, consistent with the Monod-Wyman-Changeux hypothesis [46]. The siglecs bind their natural glycan ligands with relatively low affinity, with K D values from the 100 to 400 μM range [9,66] into the mM range [23]. Siglec-2 (CD22) binds to CD45, a heavily glycosylated and abundant cell surface protein, with a K D of 117 μM [66]. The kinetics of binding, with a rapid off rate, suggest the probability of easy exchange of ligands on siglecs [63,64,66].
Most siglecs are inhibitory receptors containing an ITIM [9][10][11][12] and express full activity when a ligand is bound [16]. As documented with Siglec-2 on B cells, a functional ligand binding site is required for tyrosine phosphorylation, recruitment of the phosphatase SHP-1, and inhibition of B cell activation [67]. However, Kelm et al. [68] showed that binding of a sialoside, methyl-α-9-N-(biphenyl)-4-carbonyl)-amino-9-deoxy-Neu5Ac (BPC-Neu5Ac), which bound Siglec-2 with an IC 50 = 4 μM after cells were treated with neuraminidase, reduced tyrosine phosphorylation and attenuated the receptor's inhibitory activity. This work was followed with synthesis of additional sialoside derivatives, which bound human Siglec-2 with an IC 50 as low as 2 nM and also attenuated its inhibition of activation of B cells by anti-IgM [69].
As shown in Fig 7, a rapid decrease of phosphorylated tyrosine in siglecs was detected within a few minutes of treatment with svH1C, with the greatest reduction on Siglec-3, -5 and -7. Phosphorylated Siglec-2 and -9 were abundant in the untreated PBMC cultures, but significant dephosphorylation also occurred after addition of svH1C. These siglecs are expressed by several different immune cell types [11,12]. FcRH4/IRTA1 and NKp46, which had less reduction in phosphorylation, are expressed on memory B cells and NKT cells, respectively [12,63,70]. The decrease in the active forms of siglecs, and the receptor LAIR-1, which is expressed on most mononuclear leukocytes and contains two ITIMs [71], suggest a reduction of inhibitory activities within monocytes. CD229, member 3 of the SLAM (signaling lymphocyte activation molecule) family of immune receptors, contains two cytoplasmic ITSM motifs and can initiate positive or negative signals [72]. CD229 is expressed most strongly by CD4 + and CD8 + T cells, to a lesser extent by B cells and slightly by NK cells. Ligation of CD229 or the T cell receptor (TCR) by specific antibodies leads to phosphorylation of CD229 and recruitment of Grb2, which favors internalization [73,74]. The dramatic loss of phosphorylated CD229 and other receptors may result from activation of phosphatases. The effects are clearly induced by interactions of svH1C on the cell surface, but whether changes in phosphorylated receptors other than siglecs are direct or indirect effects of svH1C is not known.
NKG2D lacks a signaling motif in its cytoplasmic domain and functions as an activating receptor in association with adaptor proteins, DAP10 and DAP12 [75]. In addition to NKG2D, NK cells express other activating cytotoxicity receptors such as NKp30, NKp44 and NKp46. These receptors bind to tandemly multivalent heparan sulfate/heparin sequences with K D values of 0.3 to 2 μM [76]. Similarly to NKG2D, NKp46 also binds multivalent sialyl-Lewis X with a K D of approximately 0.6 μM [77]. The native in vivo ligands for NKG2D are divergent proteins, MIC-A/B, ULBP, Rae-1 or H60, which bind the receptor with K D values of approximately 1 μM [24][25][26]. In comparison with these data, svH1C binds to NKG2D with a similar K D value, 0.6 μM, as determined by microcalorimetry. The avidity of svH1C to NKG2D was also estimated by the inhibition of binding by fetuin, a glycoprotein with multiple glycans with terminal Neu5Ac-Gal sequences, which yielded a binding constant of near 1 μM (Fig 4).
Multivalent structures achieve several orders of magnitude greater avidity than single glycans [21][22][23]41]. Thus, these results lead to the conclusion that svH1C is an effective glycan mimetic and binds with avidities equal to or greater than natural multivalent glycan structures. svH1C or similar glycan mimetics are readily synthesized, non-toxic in vivo, and may find practical application as immunomodulatory therapeutic agents. Because the data suggest that the activity of svH1C leads to attenuation of inhibitory receptors and apparent stimulation of immune cells in vivo, this peptide expresses characteristics of checkpoint inhibitors that are important in immunotherapy [78,79].
Supporting Information S1