A Bifunctional Spin Label for Ligand Recognition on Surfaces

Abstract In situ monitoring of biomolecular recognition, especially at surfaces, still presents a significant technical challenge. Electron paramagnetic resonance (EPR) of biomolecules spin‐labeled with nitroxides can offer uniquely sensitive and selective insights into these processes, but new spin‐labeling strategies are needed. The synthesis and study of a bromoacrylaldehyde spin label (BASL), which features two attachment points with orthogonal reactivity is reported. The first examples of mannose and biotin ligands coupled to aqueous carboxy‐functionalized gold nanoparticles through a spin label are presented. EPR spectra were obtained for the spin‐labeled ligands both free in solution and attached to nanoparticles. The labels were recognized by the mannose‐binding lectin, Con A, and the biotin‐binding protein avidin‐peroxidase. Binding gave quantifiable changes in the EPR spectra from which binding profiles could be obtained that reflect the strength of binding in each case.

Reverse phase HPLC was carried out using a 250 x 21.2 mm Phenomedia C18 5-micron column at a flow rate of 15 mL/min. All solvents were obtained from Sigma Aldrich and used as received. Perdeuterated solvents for NMR spectroscopy were obtained from Sigma Aldrich. Alkyl thiolates were obtained from Prochimia Surfaces (Poland). All other commercially available chemicals were obtained from either Aldrich or Fisher, unless indicated otherwise. Water used for nanoparticle synthesis was HPLC grade obtained from Sigma Aldrich.
Citrate capped nanoparticles were passivated with excess alkyl thiols (a mixture of polyethylene glycol 18 and carboxylic acid 19 terminated) and pH adjusted to > 11 with NH 4 OH. 6 After 24 h stirring the nanoparticles were sonicated for ~5 min to break up any aggregated particles and dialyzed 3 x with 14K MWCO tubing to remove any residual akyl thiols. The nanoparticles were then dialyzed once more in 0.01 M potassium borate buffer, pH 9.
General procedure for peptide coupling on gold nanoparticles 1.6.2 To a stirring solution of thiol passivated nanoparticles (10 mL Abs (a.u.)
To investigate the avidity (agglomeration) of the mannose spin labelled nanoparticles with Con A, 1 mL samples of the nanoparticles were placed in Eppendorf micro centrifuge tubes, and varying amounts of a 0.1 mg/mL Con A solution was added, to give a range of concentrations (0-9.1 µg/mL). The nanoparticles were incubated at room temperature, and their UV-Vis spectra obtained at 3, 8, and 24 hours. Addition of Con A led to an observable colour change and agglomeration of the nanoparticles (Fig.   S16 S2), but only for those treated with > 5 µg/mL Con A. The UV-Vis spectrum of the untreated nanoparticles showed a maximum absorbance at 530 nm, corresponding to the surface plasmon band. 8 On addition of Con A, there was a shift of this maximum and a reduction in intensity. Using the change in the surface plasmon band intensity at different Con A concentrations Fig. S3 was obtained, and the data for 3, 8, and 24 hours fitted with Hill functions. This yielded the dissociation rate constants displayed in Table S1. After 24 hours, the nanoparticle solution with the highest Con A concentration was subjected to an excess of Dmannose, and then sonicated, briefly. This resulted in the nanoparticles no longer agglomerating, and a regeneration of the surface plasmon resonance band in the UV-Vis spectra ( Figure S4).  Table S1: Rate constants obtained from the Hill functions obtained from the data in Figure S3.
The changes in the UV-visible spectra upon titration with Con A could be fitted, giving an average apparent IC 50 ≈ 45 nM ( Figure S3) Figure S3: The change in the surface plasmon band (530 nm) intensity with Con A concentration at different times after lectin addition. A 0 is the surface plasmon resonance band intensity in the absence of Con A. 400 500 600 700 800 With Con A -24h With Con A -24 + Mannose Without Con A Normalised abs (a.u.) Wavelength (nm) Figure S4: Normalised (at 400 nm) UV-Vis spectra of mannose spin labelled Au nanoparticles treated with Con A (9.1 µg/mL) for 24 hours (Black), treated with excess D-mannose (Red), and compared against Au nanoparticles with no Con A (Grey). S18

General EPR
CW X-band EPR spectra were acquired on a Bruker EMX micro instrument, equipped with a Bruker high Q X-band resonator. Bruker strong pitch (g = 2.0028) was used as a g-value reference. The nitroxide radicals and spin labelled nanoparticles were measured in 0.01 M potassium borate buffer (pH 9) at room temperature (295 K) using a flat cell (unless specified otherwise). 4-oxo-TEMPO was used as an intensity reference. Typical parameters for spin labelled gold nanoparticles were: modulation amplitude, 1.5 G; sweep width, 100 G; Scans, 250; attenuation, 20 mW.

Mannose/Con A EPR
Spin labelled mannose 4 2.2.1 Con A exists as a homotetramer above pH 6, and each subunit contains both a Mn 2+ and Ca 2+ ion.
However, both ions are not directly involved with glycoside binding in the active site, but are structurally essential for lectin to function. Despite the paramagnetic nature of Mn 2+ , no observable EPR signal was observed at room temperature solution, which is consistent with previous literature, and is attributed to the flexible nature of the lectin Mn 2+ site and possibly a fast relaxation time in solution. 10 The crystal structure of the methyl-mannose glycoside bound Con A complex is known in the literature; 11 using this crystal structure it was observed that the Mn 2+ ion is >15 Å away from the mannose binding site, and even further from the predicted nitroxide location; attached via a linker, increasing the distance to over 20 Å.
Therefore, it is unlikely that there will be significant broadening observed from dipolar coupling. These samples were placed into capillary tubes and run using a Bruker EMX Micro spectrometer with an ER 4102ST X-band resonator.
The CW EPR spectra obtained were normalised to the center field line and compared ( Figure S5).
With increasing concentration of Con A an increased broadening of the resonances was observed, especially of the high field peak, along with a small increase in hyperfine splitting. This broadening has been attributed to the spin label entering a reduced motion environment on binding.
Simulations were created using EasySpin, a MATLAB toolbox, created by Stefan Stoll and Arthur Schweiger, allowing the simulation and fitting of a wide range of EPR spectra. 12 Simulations were created S20 using the data shown in Figure S5 for the mannose spin label 4 with and without Con A ( Figure S6). A mixture of two species could be fitted to this data; a fast and a slow-motion species, corresponding to the free spin label, and bound label-Con A complex, respectively. reduced both the broadening and hyperfine changes. This was ascribed to displacement of the spin labelled mannose by free mannose ( Figure S7) and indicated that the observed changes in lineshape were due to Con A binding rather than changes in viscosity or polarity. On excess mannose addition, the broadening due to the binding of the spin label to Con A was significantly reduced, and a recovery of the centre resonance height was observed. This, again, indicated that the broadening is unlikely to be caused by changes in viscosity on the addition of Con A, as adding mannose to the solution is likely to increase the viscosity further, yet the broadening is reduced.
Mannose spin labelled nanoparticles 2.2.2 Con A was purchased from Sigma Aldrich as a lyophilised powder, and dissolved in a 0.01 M potassium borate buffer to make a stock solution. Mannose functionalised spin labelled nanoparticles (1 S22 mL) were placed into 2 mL Eppendorf tubes, and varying amounts of a stock solution of Con A added, creating Con A concentrations between 10-200 µM. The nanoparticle solutions were vigorously stirred and left to equilibrate at room temperature for 10 minutes, before placing into an aqueous flat cell for EPR analysis.
When the normalised free spin label 4 CW EPR spectra was compared to the spin labelled nanoparticles ( Figure S8) there was significant broadening of all three nitroxide resonances, indicating a reduced molecular motion of the spin label when on the surface of a nanoparticle. Addition of Con A (100 µM) to the mannose functionalised nanoparticles led to a visible colour change, and agglomeration of the nanoparticles. A change in the EPR line height ratios between the high and centre field resonances was also observed, similar to that seen for the free spin label, but was also seen for the low and centre field resonances, indicating that the rotational motion of the spin labelled mannose was perturbed by Con A, demonstrating binding.

S23
Agglomeration and subsequent precipitation led to a decay of the nitroxide spin label over time. This precipitation process had little effect on the shape of the EPR spectrum, as only a very small proportion of the nitroxide spin labels on the surface of the nanoparticles are bound by Con A, and most of the remaining spin labels are free to rotate/flex on the surface of the nanoparticle. EPR spectra were obtained for mannose functionalised nanoparticles with Con A (0-100 µM) which showed that there was a relationship between Con A concentration and the centre/high field ratio ( Figure S9), although it was not possible to fit the data to a binding model. CW EPR were also obtained for mannose spin labelled nanoparticles passivated with only 10 % carboxylic acid terminated thiols, treated with the same peptide coupling procedure as the 1:1 nanoparticles above ( Figure S10). The spectra obtained showed less broadening than the 1:1 S24 nanoparticles, with much better signal to noise, but also a smaller difference between the free spin label and the spin labelled nanoparticle. This reduction in broadening could be due to the close proximity of the spin labels to each other (in the 1:1 nanoparticle) causing exchange coupling, however, this would lead to the observation of a half-field transition at low temperature, which was not detected. Alternatively, the broadening could be attributed to the reduced molecular motion of the spin label in a more crowded environment Magnetic Field (mT) Figure S10: EPR spectra for spin labelled mannose functionalised nanoparticles. Normalised CW EPR spectra for spin labelled Au nanoparticles (1:9 functionalised alkylthiols/PEG terminated alkyl thiols) with and without Con A (100 µM).
Addition of Con A (100 µM) to the mannose functionalised nanoparticles once again led to a visible colour change, agglomeration of the nanoparticles, and a change in the ratio between the high and centre field line height, which appeared to be more pronounced than that seen for the 1:1 nanoparticles.
Therefore, it appeared that the surface coverage of these nanoparticles was particularly important to control not just their bio-chemical properties, but also their physical/spectroscopic properties too.

S25
The 10 % spin labelled nanoparticles could be simulated by one component, with a correlation time of log 10 -9.3 ( Figure S11). Magnetic Field (mT) b) Figure S11: Simulations of the mannose functionalized nanoparticles a) without Con A and b) with Con A, compared with the obtained EPR spectra. The spectra are normalised to have the same mid field resonance height.

Spin labelled biotin 5 2.3.1
The biotin spin label 5 (2 µM) was treated with streptavidin (1.8 µM), and the CW EPR spectra obtained both with and without streptavidin. Using this data, simulations were created for the biotin spin Unfortunately, the average molecular weight was unknown for the commercially available avidinperoxidase, however, the molecular weights of both avidin and peroxidase are known, as well as the extinction coefficient for horseradish peroxidase. For the EPR titration studies, a 2 mg/mL solution of avidin-peroxidase was created, and small amounts added to 1 mL solutions of the biotinylated gold nanoparticles. In order to calculate the concentration of avidin-peroxidase binding sites, this 2 mg/mL solution was diluted to 0.2 mg/mL, and the UV-Vis spectrum obtained ( Figure S13).
Using the absorbance at 403 nm (0.30647) and the extinction coefficient for horseradish peroxidase at 403 nm (1.02 × 10 5 M -1 cm -1 ) 13 it was possible to obtain the HRP concentration (3.13 × 10 -6 M). 14 Using the molecular weight of HRP (44,000 g/mol, obtained from the Sigma-Aldrich website) the concentration of HPR in mg/mL was 0.138 mg/mL. Therefore, assuming the rest of the mass is avidin, the concentration of avidin was 0.062 mg/mL (0.2 -0.138). Given the molecular weight of the avidin tetramer is 68,000 (from Sigma-Aldrich website), the concentration of avidin in a 0.2 mg/mL solution of the avidin peroxidase S28 solution provided from Sigma-Aldrich, was 9.1 × 10 -7 M. Therefore, the avidin binding site concentration of the 2 mg/mL avidin-peroxidase complex solution used in the titrations, was 3.64 × 10 -5 M. Using this concentration, the concentration of the avidin-peroxidase/nanoparticle solutions were calculated, and are shown below in Table S2.
This UV-Vis data indicated that the Conjugate is probably a mixture with an average composition (avidin)(HRP) 3.44 . This is higher than stated on the Sigma-Aldrich website: "extent of labelling: 0.  On addition of avidin-peroxidase, there was no observable change in the colour of the nanoparticle solution, and no observable agglomeration, even after 24+ hours with excess avidin peroxidase. Initially, this was prescribed to the addition of peroxidase to the avidin protein (as discussed above), however, on closer inspection of the literature, 15 it was revealed that agglomeration with avidin is slow and will only create small clusters, having no perceivable impact on the colour of the nanoparticle solution and no precipitation.

S29
On binding, there was a significant broadening of the EPR spectrum; so significant, that it was almost impossible to detect with the particularly weak signal/noise that was obtained on measurement of the nanoparticles ( Figure S13). With 1000 scans, it was possible to observe the remarkably broad component of the avidin-peroxidase bound nanoparticles, which showed that the observed EPR resonances barely overlap with the unbound nanoparticle resonances. This leds to the decreasing intensity of all three resonances of the spectra, which can be quantified to yield binding profiles ( Figure S14). Using the avidin binding site concentrations shown in Table S2, and the centre line heights obtained from two separate experiments ( Figure 4 in the main text corresponds to the data obtained in Figure   S14), binding profiles were created. These profiles can be fitted by a 1:1 binding model using Dynafit, (Figure 4 and S15) to yield binding constants of 7.6 and 10.7 nM, respectively, with a spin label concentration ([L]) of 210 nM (which was calculated from the double integral of the EPR spectrum of the spin labelled gold nanoparticles). This indicated that the average binding constant for the biotinylated nanoparticles was somewhere in the region of 9.2 nM with a standard deviation of ± 1.55 nM.

S31
Simulations of biotin spin labelled nanoparticles with streptavidin peroxidase 2.3.3 The biotin functionalised spin labelled nanoparticles could be simulated using the data shown in  Figure S16a). The nanoparticles with avidin-peroxidase could roughly be fitted with a particularly broad component with a rotational correlation time of 10^-8.0 s, indicating that the label was in a hindered motion environment ( Figure S16b). However, due to the significant broadening, it was difficult to obtain an exact fit, as the smaller peaks are lost in the noise. In contrast to the Con A binding, there was a significant broadening of the peaks, most likely due to the tight binding of the avidin-peroxidase leading to significantly reduced molecular motion. Using these two components, it is possible to create a simulated titration, to demonstrate how the line shape would be influenced by an increasing population of the protein-label complex ( Figure S17). When the simulations were not normalised, we observed that the line heights were significantly smaller than those of the free spin label. It is likely that it would be difficult to detect the changes in the line height for  To further rule out viscosity as a source of the binding profile observed in Figure S7, a viscosity study was conducted with 100 µM spin labelled mannose 4 with varying amounts of glycerol, the viscosity of the solution increases, there is a reduction in tumbling rates of the nitroxide spin label, and a broadening of the resonances. As before, the ratio between the high field and central field resonance heights may be plotted against the % glycerol to give a binding profile ( Figure S18). Figure S18: The change in relative line height between the high field and mid field resonances against % glycerol for 100 µM solutions of the nitroxide radical 4 in 0.01 M potassium borate buffer with various amounts of glycerol.