Time-Resolved and Label-Free Evanescent Light-Scattering Microscopy for Mass Quantification of Protein Binding to Single Lipid Vesicles

In-depth understanding of the intricate interactions between biomolecules and nanoparticles is hampered by a lack of analytical methods providing quantitative information about binding kinetics. Herein, we demonstrate how label-free evanescent light-scattering microscopy can be used to temporally resolve specific protein binding to individual surface-bound (∼100 nm) lipid vesicles. A theoretical model is proposed that translates protein-induced changes in light-scattering intensity into bound mass. Since the analysis is centered on individual lipid vesicles, the signal from nonspecific protein binding to the surrounding surface is completely avoided, offering a key advantage over conventional surface-based techniques. Further, by averaging the intensities from less than 2000 lipid vesicles, the sensitivity is shown to increase by orders of magnitude. Taken together, these features provide a new avenue in studies of protein-nanoparticle interaction, in general, and specifically in the context of nanoparticles in medical diagnostics and drug delivery.

was integrated for every particle. To produce Figure 2 in main text, time resolved single-vesicle and ensemble-average intensity plots were normalized to their respective maximum intensity values, while the background intensity plots were normalized to their respective non-background maxima. To calculate the bound protein mass based on the recorded intensity values, the temporally resolved scattering intensity data before protein addition and after saturated protein binding were noise-reduced using Gaussian and rolling window median smoothing (~50 to 200 data points), after which , and , were determined from the average or maximum intensity values, respectively. The bound mass attributed to each individual vesicle was then calculated using to the equations described in the main text and plotted versus the fourth root of the initial scattering intensity (Figure 3 in main text).

SPR measurements
The dual-wavelength SPR measurements were performed using an SPR Navi 220A instrument (BioNavis), utilizing wavelengths 670 nm and 785 nm, a scanning interval of 58-78 degrees and SiO2-coated sensors (SPR102-SIO2, BioNavis). The sensors were prepared by sonication in a bath sonicator for 15 min (ultrasonic frequency, 45 kHz; ultrasonic power level, 60%, VWR USC-THD) in 2%wt. sodium dodecyl sulfate (SDS, Sigma-Aldrich), rinsing with ultrapure H2O (Synergy systems, Merck Millipore Corporation), drying under N2 flow followed by O2 plasma treatment (Harrick Plasma cleaner, 30 W) for 3 minutes. The measurements were run at 20 μL/min at 25 °C using PBS (150 mM, pH 7.4, Sigma-Aldrich) as buffer. After establishing a baseline under the flow a PBS buffer, the sensor surface was sequentially exposed to the following 30 min functionalization steps: 50 µg/mL Poly(l-lysine)-graft-poly(ethylene glycol) (PLL(20)-g [3.5]-PEG(2), SuSoS AG), of which 5% was functionalized with biotin; 40 µg/mL of streptavidin, 39 nM biotinylated DNA-tethers; 100 µg/mL of vesicle solution and 20 µg/mL of CF488-labeled streptavidin (Biotium). Each of these steps was followed by rinsing with PBS. The system sensitivity was gauged by exchanging the regular H2O based PBS with a buffer identically prepared in D2O. The system decay lengths were determined in separate experiments where a DOPC supported lipid bilayer was formed on the sensor surface. Assuming a bilayer thickness of 4.5 nm, corrections to the theoretically calculated decay length values δ (see later section) were determined from / .

Nanoparticle Tracking Analysis
The vesicle size distribution was measured using a NanoSight LM10 NTA module (Malvern Instruments Ltd., United Kingdom; 488 nm laser) at room temperature under steady flow conditions using a NanoSight syringe pump. Figure S1: Temporal evolution of the normalized waveguide-microscopy scattering intensity s (ensemble average signal using around 600 vesicles) and the background scattering intensity s,bg (ensemble average signal from the area between the vesicles) for vesicles modified with 5 mole% biotin-lipids upon exposure to non-labeled streptavidin (18 nM). The fall in the signal during liquid injection is omitted (t ~45-47 s).

Dual wavelength SPR
Absolute mass quantification of protein binding to lipid vesicles immobilized using the same protocol as that used for the waveguide-obtained data presented in the main text was made using dual wavelength SPR [4]. The SPR response, , was monitored at two wavelengths in real time upon tethering of biotinylated vesicles (1.33 nM) to a streptavidin functionalized sensor surface followed by subsequent rinsing and exposure to either CF488-SA (360 nM) or anti-biotin-IgG (133 nM). The protein coverage per vesicle was then deduced from the response signal via standard means.
Dual wavelength SPR also allows the thickness, , of the lipid vesicle layer or, in other words, the vesicle diameter (provided vesicles are spherical) to be quantified without or with protein by presenting the ratio of the SPR responses measured at the two wavelengths [4], where are sensitivity factors and the decay lengths of the respective evanescent field intensities. were obtained by alternating between H2O and D2O based buffer solutions and measuring the responses while and / were derived theoretically [5] and from tabulated values [6], respectively. The values are summarized in Table S1.  The SPR data can be used to quantify both the vesicle coverage and the increase in mass by protein binding. Using a / of 0.148 cm 3 /g for vesicles and 0.16 cm 3 /g as a mean value for vesicles with bound proteins, the measured vesicle surface mass concentration corresponds to ~403 ng/cm 2 and the additional mass uptake upon SA binding is thus ~198 ng/cm 2 . However, these numbers refer to the coverage on the planar SPR sensor surface, and hence, to make a comparison with the protein coverage per vesicle, as obtained from the scattering data, one needs to normalize to the mass of a planar bilayer. Assuming a bilayer thickness of 4.5 nm and a specific density of 1.004 g/ml [7], one gets 452 ng/cm 2 as the mass for a planar bilayer, which corresponds to a protein mass concentration of ~220 ng per cm 2 membrane area, or ~850 streptavidin molecules per vesicle ( membrane area, or 470 anti-biotin-IgG molecules per vesicles. In the latter case, an increase in vesicle diameter obtained from Eq. S1 was also accounted for (see Figure S2d).

Correction of the decay lengths for dual wavelength SPR
For thin films ≪ δ , equation S1 can be approximated as: Since and / are known, the theoretical determinations of and can be compared with the ratio obtained from the SPR response upon formation of films significantly thinner than the decay length. For pure gold sensors it was previously shown [4] that the ratio is in good agreement with theoretically determined decay lengths of 106 and 154 nm; the sensors in the current work were however coated with a 10-20 nm silica film, a change expected to slightly influence the absolute values of the decay lengths. Since SLB formation fulfill the thin film approximation ( ~5 nm and the corresponding ratio, i.e. 1.75, matches with the output from equation (S1)), each sensor was calibrated based on the ∆ and ∆ responses upon SLB formation resulting in an estimated reduction in decay lengths of 6 to 11 nm for different sensors.

Comparison of the scattering and fluorescence signal
Although quantitative interpretation of fluorescence intensity is in general a fairly complicated task that tends to require careful calibration measurements and/or extensive knowledge and consideration of experimental setup-parameters and fluorophore characteristics [8], the fluorescence intensity, , is expected to scale linearly with the number of bound proteins. First we note that increases with , , (Figure S3). Although both quenching and bleaching make quantitative interpretation of these plots complicated on the level of individual vesicles, these graphs indicate that there is a correlation between the amount of bound protein and vesicle size ( , ~ . This further supports that the seemingly higher protein coverage on small vesicles deduced from the scattering signal originates from overestimation of Δ / , for small , , although a higher protein coverage on small vesicles cannot be fully excluded. It is also worthwhile to note that is expected to scale linearly with the protein mass per membrane surface area, , calculated from the scattering intensity increase Δ / , as described in the Results and Discussion section of the main text, while a deviation from linearity is expected between Δ / , and . This is illustrated in Figure S4, which displays Δ / , and plotted as a function of for both SA and anti-biotin binding after correcting for photobleaching and/or quenching effects. In the plots, is determined using either a fixed protein refractive index n of 1.6 or a fixed thickness corresponding to the protein dimensions (5 and 15 nm for SA and antibiotin, respectively). As expected, clear deviations from linearity are observed for Δ / , versus , while a more linear relation is observed for for both film thicknesses. Note, however, that deviation from linearity for anti-biotin is greater than for the SA. Such deviations are partly related to the non-linear dependence of the scattering intensity on the particle mass and partly to other factors. For example, it may signal a protein coverage dependent variation in the homogeneity of the protein distribution and thus symmetry of the scattering object, which was previously observed for -synuclein induced vesicle deformation and rupture [7]. It might also be attributed to fluorescence quenching which is expected to increase with increasing surface coverage. In addition, the vesicle-size size distribution can here play a role, because the relative contributions of small and large vesicles to the signals change with increasing time and and for these vesicles are different. It should also be noted that although the actual protein mass concentration cannot be determined from the fluorescence signal alone, the quantitative nature of the scattering serves can serve to calibrate the fluorescence signal in terms of number of bound proteins.  Figure 2d in the main text) binding to the vesicles. The fluorescence signal was corrected for intensity decrease (see Figure 2 in main text) due to quenching and photobleaching.
was calculated from the temporal variation in the scattering intensity (rolling 50 s median value of raw intensity) for each individual vesicle, and converted to an equivalent increase in protein layer thickness Δ versus time using a fix of 1.6 (see Eq. 6 in the main text). The mean protein layer thickness of the particles (excluding top and bot 5% outlier values) for each time was then converted to mass using Eq. 6 (black). Essentially identical curves were obtained if the protein layer thickness was instead kept fix at 5 and 15 nm (green) for SA and anti-biotin, respectively, in which case was instead estimated from the increase in the effective refractive of the protein film (see Eq. 6 in the main text).

Diffusion-limited binding kinetics
The present experimental setup for the waveguide microscopy measurements involves pipetting a small volume of sample solution into a droplet of buffer solution located on the sensor surface, rapid mixing using the pipette and observing the response at the sensor surface as the sample permeates the droplet. The timescale characterizing relaxation of the solution motion after mixing is rather short (< 1 s as estimated according to hydrodynamics), much shorter that that characterizing the adsorption kinetics, and accordingly the adsorption can be considered to occur under stagnant conditions. To gain further insights regarding the kinetics of the protein adsorption process, we here discuss protein diffusion in this droplet environment. where is the vesicle radius, /6πηρ the protein diffusion coefficient (with ρ being the protein hydrodynamic radius, η the solution viscosity) and the local protein concentration (approximated as the global average concentration, 1 µg/mL). This results in a protein diffusion flux to a vesicle on the order of 900 proteins/second, a value considerably larger than what is experimentally observed. This excludes that is close to the average concentration and thus confirms a global diffusion limitation (i.e. on large distances compared to the vesicle size). Given a globally limited diffusion process, the protein flux towards the sensor surface can be approximated as: where is the bulk protein concentration (~1 µg/mL) and is the length scale characterizing the diffusion front, which in our case is estimated to be comparable with droplet height above the