Triggerable Multivalent Glyconanoparticles for Probing Carbohydrate–Carbohydrate Interactions

Carbohydrate–carbohydrate interactions are proposed to be biologically significant but have lower affinities than the well-studied carbohydrate–protein interactions. Here we introduce multivalent glyconanostructures where the surface expression of lactose can be triggered by an external stimulus, and a gold nanoparticle core enables colorimetric signal outputs to probe binding. Macromolecular engineering of a responsive polymer “gate” enables the lactose moieties to be presented only when an external stimulus is present, mimicking how nature uses enzymes to dynamically regulate glycan expression. Two different carbohydrate–carbohydrate interactions are investigated using this tool.

* S Supporting Information ABSTRACT: Carbohydrate−carbohydrate interactions are proposed to be biologically significant but have lower affinities than the well-studied carbohydrate−protein interactions. Here we introduce multivalent glyconanostructures where the surface expression of lactose can be triggered by an external stimulus, and a gold nanoparticle core enables colorimetric signal outputs to probe binding. Macromolecular engineering of a responsive polymer "gate" enables the lactose moieties to be presented only when an external stimulus is present, mimicking how nature uses enzymes to dynamically regulate glycan expression. Two different carbohydrate−carbohydrate interactions are investigated using this tool. C ells from across all kingdoms of life are coated with glycans (polysaccharides, glycolipids, glycoproteins) which provide internal and external cues for recognition, disease state, signal transduction, and more. 1 The protein "readers" of glycans are termed lectins, which engage in relatively weak interactions with individual glycans with mM affinity. 2,3 To overcome this, glycans are presented on cells in a multivalent format, enabling a nonlinear increase in binding affinity (often < nM), termed the "cluster glycoside effect", 4 leading to dramatic increases in affinity. 4 Inspired by this, synthetic materials bearing multiple carbohydrates have been developed which also have high affinity 5 and can compete for native ligands in antiadhesion therapy 6−8 or as biosensors. 9,10 While carbohydrate−lectin interactions have been extensively studied and a huge variety of tools exist to probe them, carbohydrate−carbohydrate interactions (CCIs) where the sugars directly engage with each other are much less studied, and their biological importance is not entirely clear. 11 Glycosphingolipid CCIs have been implicated in the initial stages of cell−cell recognition, 12 and the aggregation of marine sponges is linked to sulfated glycan CCIs. 13,14 The affinity of CCIs is extremely weak, even compared to carbohydrate−lectin interactions (∼mM), and require divalent metal ions to bridge the glycans; multivalent presentation is essential to enable significant interaction forces to be generated. 11,15,16 Murthy et al. have synthesized lactose-functional β-cyclodextrin which engages with GM3 glycolipid via a CCI to enable successful delivery of doxorubicin and may contribute to the lactose "trojan horse" effect for gaining entry into mammalian cells. 17,18 Dendritic-and nanoparticle-based systems have been synthesized and evaluated for CCIs. 11,16,19 These extremely weak interactions have been measured using solution NMR, 20 interfacial techniques such as Langmuir−Blodget 19 and scanning probe microscopy, 21 and also surface plasmon resonance (SPR). 14 Seeberger and co-workers developed multivalent 4 nm silica glyconanoparticles to study GM3− Gg3 interactions. 16 Penades and co-workers have used glycanfunctional gold nanoparticles with SPR to demonstrate strong CCI multivalent effects. 15 Russel and co-workers used PEGlactose functional gold particles for the plasmonic detection (red-blue color change) of CCIs, 22 enabling simple and accessible read-outs, which have also found application in biosensing. 10 A challenge in studying CCIs is that many of these are selfinteractions, e.g., lactose−lactose. Therefore, it is desirable to prevent spontaneous interactions between synthetic multivalent systems, by introducing an inducible "trigger" such that the glycan is presented with temporal or spacial control to underpin their study. In nature this is achieved by expression of glycosyltransferases which installs the glycan when required, depending on internal/external cues such as disease state. 23,24 There are many examples of stimuli-responsive polymers, 25 capable of changing their form (e.g., coli−globule transition) in response to redox, 26 light, 27 and metal ions. 28 Mastrotto et al. have used the reversible collapse of poly(N-isopropylacrylamide) to present folate on particle surfaces for cellular delivery using thermal 29 and pH stimuli. 30 Sequential revealing of ligands on liposomes has enabled improved delivery of liposome formulations. 31 Spain and Alexander have used DNA strand displacement for oligonucleotide-specific ligand expression. 32 Won et al. demonstrated that responsive polymers can provide a reversible steric "shield" to present glycans at gold nanoparticle surfaces, ensuring lectin binding only occurs when the system is activated. 33 Herein we describe dynamic nanomaterials where the surface expression of glycans can be externally triggered, to induce a carbohydrate−carbohydrate interaction. We show that polymeric gates can effectively block a (shorter) polymer-bound glycan and present it at the surface of a gold nanoparticle on demand and exploit the coupling of gold particle SPRs to monitor the interaction colorimetrically under the control of a small temperature change. By engineering an inducible trigger, these systems will facilitate the study and exploitation of CCIs in biotechnological fields.
Our design principle requires a nonresponsive polymer bearing a glycan and a responsive polymer to act as the reversible gate: poly(N-hydroxyethyl acrylamide) (pHEA) and poly(N-isopropylacrylamide) (pNIPAM) were, respectively, chosen. 33 pNIPAM was polymerized using CTA-2 (2-(dodecylthiocarbonothioylthio)-2-methylpropionic acid). pHEA was polymerized using CTA-1 (pentafluorophenyl 2-(dodecylthiocarbonothioylthio)-2-methylpropionic acid) which enables installation of amino-glycans onto the chain end by displacement of the PFP group ( Figure 1). 34 PFP-pHEA 15 was functionalized by addition of 2-amino-2-deoxy-D-galactose (GalNH 2 ) or 1-aminopropyl-D-lactose (LacNH 2 ), which was synthesized in 4 steps from lactose (SI). Addition of the glycans also results in cleavage of the RAFT agent, producing a terminal thiol suitable for immobilization onto citrate-stabilized gold nanoparticles. To ensure that the glycan-bearing polymer (pHEA) can be "hidden" by the responsive polymer (pNIPAM) the targeted degrees of polymerization were 15 and 50, respectively. 33 SEC analysis showed slightly higher molecular weights than NMR for the pHEA and that polymers were well-defined, with Đ < 1.1. 1 H NMR and IR spectroscopies confirmed installation of the glycan. pNIPAM 50 had a cloud point of 38°C at 1 mg mL −1 (Table 1). Once immobilized on the nanoparticle, the actual transition temperature will vary, 35 and hence a full concentration-dependent study on this well-known polymer was not undertaken. 26 60 nm gold nanoparticles (Au 60 ) were coated with the indicated (Table 2) ratios of each polymer to generate particles bearing either pure glycan coatings or mixed glycan/responsive polymer coatings, which have been found to give grafting densities ∼0.3 chains/nm −2 . 36 Polymer coating was confirmed by the shift of the SPR MAX peak to longer wavelength, and dynamic light scattering showed an increase in diameter from 58 to ∼70 nm. There was also a corresponding increase in the zeta potential. The particles had a net-negative charge as is widely reported for coating by this method, 37 but the particles (see below) remain stable in buffered saline solutions. pNIPAM-containing polymers had cloud points above 50°C, but it is crucial to note that the actual chain collapse occurs below this temperature, which is essential for the present work. 35 With this panel of nanoparticles bearing both responsive/ nonresponsive units and two different glycans in hand, control experiments were undertaken to demonstrate glycan-driven interactions at the particle surface. Lac-pHEA 15 @Au 60 and GalNac-pHEA 15 @Au 60 (i.e., nonresponsive coatings) were exposed to 0.1−100 mM of Ca 2+ . Ca 2+ is the primary divalent ion which bridges appropriate glycans to initiate the CCI. Following 30 min incubation Lac-pHEA 15 @Au 60 showed distinct shifts in its UV−vis with the SPR MAX decreasing and Abs 700 increasing, consistent with cross-linking ( Figure 2). 36,38 Conversely, addition of Ca 2+ to GalNAc-pHEA 15 @Au 60 showed very minor spectral changes with essentially zero aggregation (isotherms in SI). TEM analysis confirmed these observations ( Figure 2E). GalNAc does not undergo self−self CCIs and hence excluded the possibility of nonspecific interactions due to changes in the ionic strength of the buffers. Similar results were obtained using Mn 2+ , also a divalent cation which can drive CCI (SI). To further corroborate the link between glycan presentation and aggregation, a lectin-mediated (carbohydrate−protein) interaction was also studied. SBA (soy bean agglutinin) was chosen as it has high affinity for GalNAc but lower affinity for lactose (i.e., opposite response to CCI expected). 39 A serial dilution of SBA was added to both nanoparticles from 10 to 100 nM and UV−vis spectra recorded ( Figure 2C/D). SBA requires calcium to be present in the

ACS Macro Letters
Letter binding mix, but at concentrations which did not cause CCI aggregation. Lac-pHEA 15 @Au 60 showed no aggregation, but GalNAc-pHEA 15 @Au 60 underwent strong dose-dependent binding to SBA. These control experiments confirmed the differential presentation of glycans on the nanoparticle surface that the end-group glycosylations provide sufficient density on the particle surface to induce a cluster-glycoside enhancement.
The above data demonstrated the challenge of self−self CCI study; when lactose-functional materials are placed in the biological buffers, spontaneous self-aggregation is always going to occur. This complicates detailed studies, especially in cellbased assays as the expression of lactose is static, not controllable, but has been reported to drive cell uptake by unknown mechanisms. 17 To evaluate steric gating Lac-pHEA 15 :pNIPAM 50 ratio 9:1@Au 60 was monitored for calcium-mediated aggregation at 20°C. At this temperature, the pNIPAM is fully extended and provides a steric block to CCIs, preventing aggregation even at 100 mM Ca 2+ ( Figure  3A). Time-dependent studies by monitoring Abs 700 showed only a very slow change, and TEM revealed no evidence of aggregation ( Figure 3C/D). Hence it can be seen that just 10 mol % of pNIPAM on the particle surface can sterically block CCIs and is "gate closed".
Identical experiments as above were conducted but this time at 40°C which is above the chain collapse temperature of pNIPAM but below the threshold for them to aggregate (avoiding false positives, see Table 2) ( Figure 3E/F). Addition of Ca 2+ leads to a clear increase in Abs 700 , and time-dependent monitoring of these showed rapid aggregation significantly faster and to a greater extent than that at 20°C. TEM confirmed the formation of aggregates in the presence of Ca 2+ but that in the absence of Ca 2+ (even at 40°C) there were no nonspecific particle−particle interactions. This is the first example of using an external trigger to control a CCI interaction and provides a dynamic tool for studying this area of glycoscience.
To further demonstrate the potential of these responsive nanoparticles, an assay was devised to mimic cell-surface hetero-CCI interactions, namely, lactose-GM-3. 18 A polystyrene plate was modified by physiosorption of GM-3 glycolipid to create a cellular membrane mimic substrate. This was then incubated with both static and dynamic nanoparticles and binding evaluated by UV−vis and high-resolution scanning electron microscopy (SEM) after washing ( Figure 4A). Due to the low concentration (a monolayer) low UV−vis signals were obtained, but the presence of the SPR peak at ∼550 nm could be seen when Lac-pHEA@Au 60 was incubated with calcium at 20°C; however, no SPR peak was seen for the other static particles ( Figure 4B). Figure 4C shows the dynamic particles, Lac-pHEA 15 (9)/pNIPAM 50 (1)@Au 60 , at both 20 and 40°C, showing nanoparticles were only bound at the higher temperature, when the gate is "open". SEM of the GM-3 surfaces corroborated the observations from UV−vis. Figure  4D/E shows that static nanoparticles are captured by GM-3 only for lactose not GalNAc functionality. Figure 4F shows the dynamic nanoparticles with "gate closed" at 20°C (no binding) and Figure 4G with "gate open" at 40°C with significant Lac-GM-3 interactions occurring.

ACS Macro Letters
Letter Herein, we have demonstrated stimuli-responsive, "triggerable" multivalent glyconanoparticles for the study of carbohydrate−carbohydrate interactions. At the surface of gold nanoparticles nonresponsive polymers were used to anchor lactose and a responsive polymer, which can collapse upon application of a thermal stimulus used as a steric "gate". By exploiting this trigger, and the unique optical properties of the gold nanoparticle core, it was possible to monitor aggregation due to the lactose−lactose interactions in a label-free manner and upon application of the external stimulus. The hetero Lac-GM-3 interaction was also probed by this method using a binding assay which mimics a cell−surface interaction, demonstrating selective binding only upon application of the stimulus. This method enables temporal control of when the glycans are expressed on the nanoparticle surface, comparable to enzymatic control in nature preventing unwanted spontaneous interactions, entirely under the operator's control. This will facilitate the study and exploitation of multivalent carbohydrate interactions.

Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version.

Notes
The authors declare no competing financial interest.