Single-protein identification by simultaneous size and charge imaging using evanescent scattering microscopy

Abstract

Separation and identification of different proteins is one of the most fundamental tasks in biochemistry that is typically achieved by electrophoresis and western blot. Yet, it is challenging to perform such analysis with small sample size. Using a principle analogous to these conventional approaches, we present a label-free, single-molecule technique to identify different proteins based on the difference in their size, charge, and antibody binding. We tether single protein molecules to a sensor surface with a flexible polymer and drive them into oscillation by applying an alternating electric field. By tracking the nanometer-scale oscillation of each protein molecule via a high-resolution scattering microscopy, the size and charge of each protein molecule can be determined simultaneously. Changes induced by varying the buffer pH and antibody binding are also investigated, which allows us to further expand the separation ability and identify two different proteins in a mixture. We anticipate our technique will contribute to single protein analysis and biosensing.

Graphical Abstract

Proteins are ubiquitous in biological processes^{1, 2} and play key roles in disease diagnosis and treatment^3–5. Size and charge are the most fundamental physical characteristics of protein, which are closely related to many molecular activities including binding and conformation changes^6–9. For this reason, mainstream techniques such as mass spectrometry, electrophoresis, western blot and conventional enzyme-linked immunosorbent assay (ELISA) separate or identify proteins based on the size, charge and specific interactions. Although these techniques are widely employed in laboratory and industry, most of them are detrimental to the native protein structure due to fragmentation and denaturation, require labels, and lack the sensitivity for measuring low-abundant samples¹⁰. Several optical imaging based single molecule detection techniques have been developed to measure single proteins without using label, including dark field imaging¹¹, interferometric scattering (iSCAT) ^{12, 13}, mass photometry¹⁴, and photothermal microscopy^{15, 16}. These methods take advantage of light scattering or absorption and have great performance on protein size and mass characterization, but they are not able to measure protein charge. Anti-Brownian electrokinetic (ABEL) trap complements in charge detection, as it measures the molecular diffusion and electrokinetic mobility (related to size and charge) by manipulating the motion of single protein via an electric field^{17, 18}, however, it requires fluorescent labelling and has limited throughput with only one molecule trapped and measured at a time.

We have recently developed a technique that can measure single protein size and charge simultaneously, and detect ligand binding using total internal reflection (TIR) based imaging¹⁹. We tether protein molecules to a sensor surface via flexible polymer linkers and apply an alternating electric field to drive the proteins into oscillation^20–22. By imaging the oscillation in response to the field, the size and charge can be extracted. The binding of ligands changes the size and/or charge of the protein and thus can be detected as well. However, this technique has two intrinsic limitations. Firstly, the imaging relies on the interference between the scattered light from the proteins and the reflected light from the surface. The strong reflected light easily saturates the camera full well capacity and impedes further improvement of signal-to-noise ratio (SNR) by increasing the incident light intensity. Also, the interference pattern appears as a parabolic shape with a size (5 μm) much larger than the diffraction limit (1 μm). As a result, only ~10 single proteins can be imaged simultaneously by a microscope with 60X amplification, making it difficult to study single molecules in a multiplexed fashion.

To address the above problems, we introduce an evanescent scattering microscopy (ESM) configuration to our imaging system, which only collects the scattered light from analyte and surface roughness rather than the reflected light^23,24. ESM has been reported to image sub-20 nm gold nanoparticles and lipid vesicles^{25, 26}, as well as the binding event of protein molecule to surface-bound lipid vesicles²⁷. Since the scattered light is much weaker than the reflected light, the SNR can be easily enhanced by increasing the incident light intensity without saturating the camera; a light source with shorter wavelength is also applied here to increase the scattering efficiency. The parabolic shape for single protein image is eliminated in the ESM configuration due to the absence of reflected light, allowing us to increase detection throughput with higher protein coverage.

Besides the technical advance, we demonstrate the capability of ESM in measuring single-molecule size and charge with four different proteins. Towards protein identification, we measured the pH induced size and charge of single proteins, a strategy analogous to conventional two-dimensional (2D) electrophoresis. We also identify individual molecules in a two-protein mixture by the size, charge, and antibody binding, which resembles western blot. We hope this work will lay the groundwork for single-protein separation and identification and contribute to single-molecule biosensing.

Results

Detection principle

In ESM, a laser with 450 nm wavelength is directed to the surface of an indium tin oxide (ITO) coated cover glass at 65° of incidence, which is 3.5° above the total internal reflection angle of the glass/water interface, using a 60X high numerical aperture (NA) objective. An evanescent field is created on the ITO surface and propagate along the surface. The objects exist in the evanescent field, the tethered protein and surface roughness in this case, will scatter the light in all directions (Fig. 1a). The scattered light relates to the sixth power of the analyte size, which attenuates significantly when relates to the small molecules^{23, 24}, so the interference of scattered light from surface roughness and analyte dominates in the detection. This scattered light can be collected by another objective from top view and imaged via a CMOS camera, which can be denoted as

$$E_{cam}\sim 2\left|E_{sf}\right|\left|E_p\right|\cos (\theta ),$$ (1)

where E_sf and E_p are the light scattered by the ITO surface roughness and the protein molecules, respectively, and θ is the phase difference between the two scattered lights (Fig. 1b). For a given surface, E_sf is constant, while E_p is a function of protein size (hydrodynamic diameter, D_H) and the vertical distance from the surface (z) due to the exponential decay of evanescent field. Thus, E_cam is a function of both D_H and z. Since the proteins are charged in solution (with charge q), z can be modulated by applying an electric field in vertical direction. Therefore, both D_H and q can be extracted simultaneously by imaging the electrical response of the protein molecules. To continuously manipulate the same protein molecule and prevent diffusion, we tether the proteins to the ITO surface with a soft polymer, polyethylene glycol (PEG, with a molecular weight of 10 kDa). Then we drive the proteins into oscillation by applying an alternating electric field and record an image sequence at a frame rate several times higher than the frequency of the electrical field. By performing temporal fast Fourier transform (FFT) for every second of the recorded image sequence, an FFT image sequence of 1 frame per second (FPS) is obtained. The FFT amplitude images shows bright spots, which are the oscillating protein molecules. The oscillation amplitude in nanometer can be calculated based on the FFT amplitude image intensity of the bright spots (Fig. 1c, d). As the applied potential increases, the oscillation amplitude and FFT image intensity increases as well. But due to the restriction of the PEG tether, the FFT image intensity will reach a maximum when further increasing the potential cannot provide enough force to counter the entropic force of the stretched PEG. As a result, the FFT image intensity vs. applied potential (ΔI vs. U₀) plot presents a linear regime at low potential and a plateau regime at high potential (Fig. 1e). Using the linear regime and the plateau regime, the charge (q) and size of single molecules (D_H) can be calculated (see below).

Figure 1. Imaging the size and charge of single protein using evanescent scattering microscopy (ESM).

(a) Schematic of ESM and the electrochemical system. A vertical alternating electric field is applied to an ITO surface via a three-electrode configuration (see Fig. S3b and c for more details). A 450 nm laser is directed from the bottom objective to the ITO surface to excite the evanescent field. The scattered light from the oscillating molecules is collected by the top objective and imaged by a CMOS camera. (b) The protein molecule is tethered to the streptavidin functionalized (the capture layer) ITO surface by a PEG linker, and driven into oscillation by an alternating electric field in vertical direction. Lights scattered by protein (E_p) and surface roughness (E_sf) interfere with each other and eventually imaged by the camera. (c) An alternating potential with 0.5 V amplitude and 40 Hz frequency is applied to the ITO surface functionalized with tethered immunoglobulin G (IgG). An image sequence is recorded at 200 frames per second (FPS). (d) By performing temporal fast Fourier transform (FFT) to every 1 s images in the image sequence, an FFT image sequence of 1 FPS is obtained, in which oscillating proteins showing as bright spots. The arrow marks a single IgG protein (yellow spot). See Methods for details about how to identify protein oscillation from the ITO background. (e) The FFT image intensity of the marked protein is converted to oscillation amplitude (in nanometers) through a calibration (see Figure 2). By increasing the applied potential (U₀), the oscillation amplitude (Δz₀) shows a linear increase in the beginning and then reaches a plateau due to the restriction from the PEG tether. The hydrodynamic diameter and charge of the protein can be determined from the plateau regime and the linear regime, respectively. Scale bars in (c) and (d), and the image size in (e) are 5 μm.

The FFT image intensity is converted into oscillation amplitude in nanometers by using ΔI = I₀ (e^−z₁/d − e^−z₂/d), where ΔI is FFT image intensity, Z₁and z₂ are the closest and farthest position of the oscillating tethered protein to the ITO surface, and d is the decay constant of evanescent field (~ 200 nm). The peak-to-peak height difference in oscillation can be denoted as Δz₀ = z₂ − z₁. I₀ is the intensity when the protein is at the ITO surface (z₁ = 0), which is determined by measuring the ΔI when the PEG is stretched to its maximal approachable length (Supplementary Notes 1–2). Assume the lowest oscillation point appears at the surface, the Δz₀ at different applied electric field can be determined, and the ΔI vs. U₀ curve is converted to Δz₀ vs. U₀ curve, and the protein charge can be obtained by (Supplementary Note 1),

$$\text{Δ}{z}_0=\frac{q}{k_{PEG}}{E}_0\left(\text{Δ}{z}_0\right)$$ (2)

where E₀(Δz₀) is the amplitude of the electric field at Δz₀ (proportional to U₀), q is the charge of the protein molecule and k_PEG is the entropic spring constant of the PEG tether. Details about how to calibrate the electric field can be found in Supplementary Note 2. By fitting the slope of Δz₀/U₀ in the linear regime of Fig. 1e, we can calculate the charge of protein molecule (q) by Eq. 2.

To determine the protein size, we establish a relationship between ΔI and D_H with polystyrene nanoparticles (PSNP). PSNPs are commonly used in biosensor calibration^{19, 24, 28–30} because their diverse size distribution and a close refractive index to the proteins^31–33. 3 kinds of PSNPs with different D_H (20, 50 and 100 nm) were tethered with PEG 10 kDa linker and then driven into oscillation via the same electric field series. The D_H of PSNPs are verified by the dynamic light scattering (DLS), which is a widely used technique to determine the bulk size distribution of small particles in a solution by measure the scattering light intensity change caused by Brownian motion³⁴. Then we applied the FFT filter and extract the oscillation signals at the plateau regime (Fig. 2a). The FFT intensity change at the plateau regime shows a proportional relationship with D_H in log-log scale, which also agrees with our calculation³⁵ (Supplementary Note 3).

Figure 2. Measuring the size and charge of single protein molecules.

(a) Relationship between particle size and FFT image intensity. Streptavidin coated polystyrene nanoparticles (PSNP) are tethered to the ITO surface by PEG and driven into oscillation. The average temporal FFT image intensity over the plateau regime is plotted vs. particle size in log scale and fitted with a linear model. The error bar represents the PSNP size distribution (determined by DLS) in horizontal axis and FFT intensity distribution (obtained from ~300 individual particles for each size) in vertical axis, respectively. (b) FFT image intensity of immunoglobulin G (IgG) measured at 0.5V, 40Hz. 5 individual IgG molecules and their intensity are selected as examples. FFT image intensity. Scale bar: 2 μm (c) and the corresponding oscillation amplitude (d) of the 5 IgG molecules in b at different potentials, from which the size (D_H) and charge (q) can be derived, (e) Measurement results of D_H and q from the data in c. Histogram of D_H (f) and q (g) for 293 IgG molecules are fitted with Gaussian distribution. The mean value and standard deviation are 10.4 ± 1.4 nm in size and −6.5 ± 1.8 e in charge, respectively. (h) Measuring the size of IgM, IgA, IgG and BSA. 256 IgM, 278 IgA, 292 IgG, and 282 BSA molecules are measured to generate the error bar. The solid curve describes the relationship between FFT image intensity and D_H as obtained in (a). (i) Comparison of D_H,q for different proteins determined by our oscillation approach with those measured by DLS.

Measuring the size and charge of single IgG molecule

After establishing the detection method, we demonstrate the capability of ESM for measuring size and charge using immunoglobulin G (IgG, with molecular weight 150 kDa). A series of potentials from low to high were applied to the tethered IgG and the images were recorded. Fig. 2b shows the FFT image at 0.5 V, 40 Hz, where the bright spots are single IgG molecules. The total image intensity (in gray scale) for each molecule is integrated to represent the scattered light intensity of the molecule (ΔI). After plotting the intensity (ΔI) vs. different potentials (U₀) (Fig. 2c), the average intensity at the plateau was calculated and plugged into the calibration curve (Fig. 2a) to calculate D_H. Then the intensity in gray scale was converted into oscillation amplitude (Δz₀) in nanometers (Fig. 2d). By fitting the slope of the linear regime of Δz₀ vs. U₀ plot prior to the plateau, q for each IgG molecule was obtained. The D_H and q for the 5 representative IgG molecules were shown in Fig. 2e. A total of 293 IgG molecules were measured for statistical analysis. By fitting the histogram with Gaussian distribution (Fig. 2f and g), D_H and q are determined to be 10.4 ± 1.4 nm and −6.5 ± 1.8 e (e, the elementary charge), respectively.

Using the above method, the D_H of another three kinds of proteins: immunoglobulin M (IgM), immunoglobulin A (IgA), and bovine serum albumin (BSA) were also measured, and the results are shown in Fig. 2h. The charge of the proteins was measured as well (Fig. 2i). To evaluate the accuracy of our detection, we measured the size and charge of the same proteins with DLS (Fig. 2i). The results from our approach show good consistency with those determined by DLS and agree with the value reported in literatures^36–40.

Dependence of charge on pH

2D electrophoresis creates a pH gradient in one dimension to separate proteins by tuning the mobility/charge of proteins at different pH. As a preliminary test towards this goal, here we show that ESM can measure the charge change of single IgG molecules at different pH. To alter the charge of IgG, we tuned the solution pH from below to above the isoelectric point (pI) of IgG (pI ranges from 6.6 to 8.2)⁴¹, which changed the polarity and amount of charge.

In the experiment, the same IgG molecules were oscillated and imaged at different pH (5.0, 7.0 and 9.7) (Fig. 3a). Fig. 3b shows the Δz₀ vs U₀ curve of an example IgG molecule at pH 5.0 and 9.7, where the IgG charge polarity were positive and negative, respectively. The oscillation amplitude Δz₀ were similar at different U₀, indicating the charge in the IgG is similar, while the polarity was opposite by examining the phase of the oscillation. We plotted the oscillation profile of the representative IgG molecule at pH 9.7 and found that the oscillation was out-of-phase with the applied potential (Fig. 3c), implying the molecule was negatively charged. After switching the pH to 5.0, the oscillation phase was inversed, which reflected the fact that charge polarity was positive when pH < pI.

Figure 3. Dependence of charge on pH.

(a) FFT images showing the same tethered IgG (white circle) in acidic (left panel) and basic solutions (right panel). Applied potential, 0.3V at 40 Hz. The pH for the solutions were tuned to 5.0 and 9.7 respectively while maintaining the ionic strength the same as the 100 times diluted 1X PBS. Scale bar: 2 μm (b) Oscillation amplitude of the same IgG molecule at different potentials in acidic and basic solutions. The oscillation profiles at 0.3 V (circled) were plotted in c. (c) Oscillation of the IgG molecule under pH 5.0 (red curve) and 9.7 (blue curve). The potential applied was 0.3 V, 40 Hz. The oscillation under pH 5.0 was in phase with the applied potential, indicating positive charge of the molecule. The phase was flipped ~180° at pH 9.7 because the charge polarity became negative. (d) and (e) Histograms of D_H and q at different pH. 312, 300 and 314 IgG molecules were measured under pH 5.0, 7.0 and 9.7 respectively. The data was fitted with Gaussian distribution (solid curves). The size remains constant at different pH, but the charge amount and polarity were changed by pH.

We have studied 312, 300 and 314 IgG molecules at pH = 5.0, 7.0 and 9.7, respectively, and summarized the size and charge data in Fig. 3d and e. The results show the charge are +7.6 ± 1.7 e, −5.9 ± 2.2 e and −7.4 ± 2.1 e at pH 5.0, 7.0 and 9.7, consistent with the charge change trend with pH. In contrast, the size does not alter much at different pH (10.4 ± 1.8 nm, 10.4 ± 1.4 nm and 10.5 ± 2.2 nm for pH 5.0, 7.0 and 9.7), which suggests IgG size is not responsive to pH. Together, these results demonstrate that our method is capable of differentiating proteins at different pH.

Identification of single proteins in a two-protein mixture

In western blot, proteins are firstly separated by electrophoresis and then identified using antibody. Analogously, we differentiated individual IgA and BSA by size and charge in an IgA/BSA mixture and identified IgA molecules via antibody binding (Fig. 4a). Prior to the measurement, we mixed the samples at 1:1 ratio and then tethered them on the ITO surface. Next, we measured the size and charge of individual proteins in the mixture and plotted the data in a 2D graph with axes showing size and charge. Two separated domains are found in the graph, which are assigned to BSA and IgA (Fig. 4d). Projection of the 2D plot on each axis shows the size and charge distribution of the mixture, where the size distribution has two distinct peaks corresponding to IgA and BSA (Fig. 4d, top panel); However, only one peak is shown in the charge projection (Fig. 4d, right panel) because the charges of these proteins are similar.

Figure 4. Identification of single protein in a two-protein mixture.

(a) IgA and BSA were mixed and tethered to the surface. The specific binding of anti-IgA changes both the size and charge of IgA, while those of BSA are not changed. (b) and (c) Real-time monitoring of the size and charge change in single IgA molecules upon antibody binding, 0.4V, 40 Hz potential was applied for size detection (where oscillation was within the plateau regime) and 0.2V, 40 Hz potential was applied for charge detection (where oscillation was within the linear regime). Anti-IgA was introduced at t = 0 s (marked by an arrow). (d) and (e) 2D size and charge distribution of the IgA/BSA mixture before and after adding anti-IgA. 425 and 424 individual molecules were measured before (blue) and after (red) adding anti-IgA, respectively. Before introducing the antibody. Two domains can be found which were identified as BSA and IgA. After anti-IgA binding, a portion of the IgA domain shows a shift due to size and charge changes. The top and side panels are projection of the 2D plot showing size or charge distribution. The size distribution is fitted with two Gaussian distributions, and the full width at half maximum (FWHM) is used to identify individual BSA and IgA in the sample. (f) An example showing several IgA and BSA molecules were identified based on the size and charge, where the BSA and IgA are marked in red and blue. The rest of the spots in yellow color were molecules that were not identified or backgrounds that did not show clear response to the electric field.

To confirm the domain with larger D_H was IgA, we added anti-IgA to the surface and recorded the size and charge change of each single molecule. Binding of anti-IgA to IgA changes its D_H and q such that the IgA molecules can be identified from the mixture. In contrast, BSA should not show changes in D_H and q because it does not bind with anti-IgA. Fig. 4b and c show the real-time size and charge change of 3 representative IgA molecules upon antibody binding, where a sudden change was observed when the antibody bound to IgA. We measured ~60 anti-IgA binding events in real-time, and the average size and charge change were 1.7 nm and −1.8 e, respectively (Fig. S6). After the antibody binding, we switched the imaging area and measured the size and charge for more molecules (~420 molecules) in an end-point fashion (without capturing the binding moment) to construct the 2D plot (Fig. 4e). The 2D plots before and after antibody binding are compared. Before antibody binding (Fig. 4d), the plot shows two separate domains for the two proteins. In the size projection, two peaks with centers at 8.5 ± 1.0 nm and 12.1 ± 1.2 nm could be observed. For the charge projection, only one peak is shown, indicating that the proteins are indistinguishable by merely looking at the charge. After antibody binding (Fig. 4e), the right domain (IgA) shifts to the lower-right with a more dispersed contour. In this case, the two peaks in size projection have centers at 8.6 ± 1.3 nm and 13.6 ± 2.1 nm (Fig. 4e, top panel), where the second peak is ~1.5 nm larger compared to the histogram before the antibody binding. Simultaneously, a peak at ~ −10 e could be observed in the charge distribution (Fig. 4e, right panel), reflecting the contour change in 2D results. Compared to the IgA domain, the BSA domain is unaffected by the antibody. To confirm the assignment of two domains, we also measured IgA and BSA independently (Fig. S7). For BSA, distribution center in size and charge are at 8.2 ± 1.3 nm and −6.6 ± 1.3 e respectively; For IgA, the values are 12.2 ± 2.2 nm and −7.6 ± 1.8 e in size and charge respectively. They both agree with the mixture measurement.

The above analysis allows us to derive a criterion for distinguishing IgA and BSA on the same sensor chip. We used the width of distribution as a threshold to identify IgA or BSA, where the full width at half maximum (FWHM) for BSA and IgA are 2.2 nm and 2.8 nm (Fig. 4d, top panel). Together with the distribution center, we took the size of 7.4 nm - 9.6 nm for BSA and 10.7 nm - 13.5 nm for IgA. Using the size threshold and antibody binding induced size and charge change, we were able to identify the single proteins in the IgA/BSA mixture (Fig 4f).

Discussion

ESM improves both the SNR and the image resolution when compared with the tethered protein detection by reflection-based imaging¹⁹. ESM avoids the strong reflected light in reflection-based detection by collecting the scattered light from top of the sample, and thus dramatically improves the camera photon collection efficiency. ESM allows the use of much stronger incident light power to improve SNR in shot noise limited detection condition, where the SNR scales with the square root of scattered photons. Theoretical calculation shows ESM can use 5000 times higher incident power (5 kW·cm⁻² vs 1 W·cm⁻²) and has five times SNR improvement over reflection-based imaging on single BSA molecule detection (SNR_ESM ~ 16 and SNR_TIR ~ 3, based on received photon number calculation) (Supplementary Note 8). In our current ESM setup, limited by the camera quantum efficiency and loss of light in the optical path, the shot noise limited SNR is ~ 8.5. The actual measured SNR_ESM for single BSA molecule is ~ 7.2, indicating our measurement is shot noise dominant with minor contributions from other types of noises (Supplementary Note 8). Nevertheless, ESM still has a more than doubled SNR than TIR detection, which could be further improved by reducing photon loss and other type of system noises in the optical system.

ESM also improves the spatial resolution of the protein image and the detection throughput compared to reflection-based imaging. By eliminating the parabolic interference pattern (~5 μm in length) in TIR image, the resolution of ESM image is decided by the wavelength of the detection light and the numerical aperture (NA) of the objective, which has a spatial resolution of ~1 μm in both X and Y directions. Therefore, ESM can detect individual protein signals as long as their distance is larger than 1 μm, which allows several times higher protein oscillator packing density on the sensor surface compared to reflection-based imaging (Supplementary Note 9). In addition, the Gaussian distributed point spread function of ESM also makes the automatic imaging processing possible, enhancing the efficiency of detection.

We next compare the performance of ESM with other interference imaging techniques, and specifically, with the state-of-the-art iSCAT. The resolution of iSCAT can reach 0.3 μm, which is 3 times better than our current ESM setup, due to the wavelength of detection light (405 nm) and NA (1.46) of the objective. It is possible to improve the spatial resolution of ESM by collecting the scattered light from bottom using the same high NA oil immersion objective, by blocking the reflected light with a partial beam blocker, similar to iSCAT⁴². Another important difference between ESM and iSCAT is the incident light intensity. ESM has a ~5 times enhancement by the evanescent field, which means the incident power is ~5 times stronger than iSCAT if they are using the same light source. In addition, the oscillation approach removes the noise other than at the oscillation frequency, allowing longer measurement time for the same protein to further reduce shot noise. Theoretically these unique features will offer ESM a smaller molecular mass detection limit, or a larger imaging area (higher single molecule detection throughput than iSCAT). In practice, the background signals from double layer charging effects is a noise source of the oscillator signal. . Our ESM can measure BSA in a 35 μm × 35 μm area with SNR of 7.2, while iSCAT can achieve imaging BSA in 4.5 μm × 4.5 μm area with SNR of 10. Therefore, ESM can reach near two order of magnitude larger detection area with similar SNR than iSCAT. Note that ESM image is processed with FFT for each second videos, which reduces the real frame rate to 1 fps and thus reduced the time resolution of detection compared to iSCAT that can have tens of frames per second. This may limit the application of ESM in fast biological process sensing, but in general, 1 s time resolution is sufficient for most biological process including most binding kinetics studies.

Conclusion

We have developed a detection method that can quantify the size and charge of single proteins by evanescent scattering microscopy, which has improved SNR and spatial resolution over the total internal reflection imaging method we developed earlier. Two different protein molecules in mixture can be distinguished based on measuring the size, charge, and antibody binding, analogous to 2D electrophoresis and western blot, but at the single-molecule level. We anticipate that this method could contribute to single-molecule studies including measuring small volume samples (e.g. single cells), revealing molecular heterogeneity, and understanding the functions of various biological macro molecules in diseases and drugs.

Methods

Materials

ITO coated cover slips with resistance of 70-100 Ω were purchased from SPI Supplies. Streptavidin and hydrogen peroxide were purchased from VWR. (3-Glycidyloxypropyl) trimethoxysilane and BSA were purchased from Sigma-Aldrich. IgG (goat anti human IgA, 150 kDa) was purchased from Bio-Rad. Secretory IgA (from human colostrum, MW = 385 kDa) and IgM (from human plasma, MW = 950 kDa) were purchased from Athens Research and Technology. Streptavidin coated polystyrene nanoparticles (20 nm, 50 nm and 100 nm) and biotin-PEG-NHS (MW = 3.4, 5 and 10 kDa) were purchased from Nanocs. 1X phosphate buffered saline (PBS) was purchased from Corning. Ammonium hydroxide was purchased from Mallinckrodt Chemicals. Deionized (DI) water with resistivity of 18.2 MΩ·cm was used in all experiments.

Experimental Setup

An 80 mW laser diode (PL450B, Thorlabs) with central wavelength of 450 nm was used as the light source. 450 nm light was selected for a balanced consideration between the larger scattering cross section and the potential photo damage to biological samples of the short wavelength light (Eq. S10). The light was collimated by a lens group and then focused to the back focal plane of a 60× oil immersion objective (NA = 1.49, Olympus). The light spot on the ITO surface was 30 μm by 30 μm with an intensity of 5 kW·cm⁻². The reflected light was collected by a camera (Pike F-032B, Allied Vision) which was used to find the total internal reflection angle. Light scattered by the protein molecules was collected from top by a 50× objective (NA = 0.42) and imaged by a second camera (MQ003MG-CM, XIMEA) at 200 FPS. The camera exposure time and gain were optimized for different experiments (Supplementary Note 6). A function generator (33521A, Agilent) and a potentiostat (AFRDE5, Pine Instrument Company) were used to apply potential to the ITO surface. A sinusoidal potential (f = 40 Hz) was applied to the ITO surface through a three-electrode configuration, where the ITO, an Ag/AgCl wire and a Pt sheet served as the working, reference, and counter electrodes, respectively. A USB data acquisition card (USB-6251, National Instrument) was used to control and synchronize the applied potential, current and the camera.

Surface functionalization of ITO

ITO coated coverslips were cleaned by sonication in acetone, ethanol and DI water serially, each for 10 min. Then the ITO coverslips were soaked in H₂O₂/ NH₃·H₂O/ H₂O (1:1:5) and at 80 °C for 30 min, followed by rinsing with DI water and drying with nitrogen flow. Next, the ITO coverslips were incubated with 2% (3-glycidyloxypropyl) trimethoxysilane in isopropanol at 80 °C for 2 hours, rinsed with isopropanol and DI water, and dried with nitrogen flow. The epoxy-functionalized ITO coverslips were incubated with 100 μL of 1 mg·mL⁻¹ streptavidin or 1 mg·mL⁻¹ BSA in 1X PBS for 1 hour to immobilize streptavidin or BSA on the surface. 1 mg·mL⁻¹ BSA was flowed to the streptavidin coated surface for 10 minutes to block the non-specific binding sites.

Tethering nanoparticles to ITO surface

The BSA-functionalized ITO surface was incubated with 100 uL of 0.5 nM biotin-PEG-NHS (3.4 kDa, 5 kDa or 10 kDa) at 4 °C overnight to allow the covalent bonding between NHS and the primary amine groups on BSA. Then the surface was slowly washed with 1X PBS to remove unbound molecules. Streptavidin coated polystyrene nanoparticles (20 nm, 50 nm or 100 nm) were attached to the PEG linkers via streptavidin-biotin conjugation by incubating the surface with nanoparticle solution (10¹⁰ particles/mL) for 30 min. Finally, the ITO surface was slowly flushed with 100 times diluted PBS to remove the free particles.

Tethering proteins to ITO surface

Biotin-PEG-NHS with MW = 10 kDa was used to tether protein to the streptavidin-functionalized ITO surface. The protein (IgM, IgA, IgG or BSA) was first incubated with biotin-PEG-NHS tether at 5:1 ratio to form a protein-PEG complex in 1X PBS at 4 °C overnight. Excess amount of protein reduced the chance that multiple tethers were linked to the same protein molecule. The complex solution was added to the streptavidin functionalized ITO surface and incubated for 1 hour to allow the biotin end of PEG coupling to the streptavidin. The ITO surface was slowly washed with 100 times diluted PBS to remove free protein molecules in the solution before measurement.

Oscillation detection and data processing

A 40 Hz alternating electrical potential sequence was applied to the ITO surface with amplitude ramped from 0.1 V to 0.5 V in 0.05 V step and 1 s duration for each step. At the same time, oscillation images were captured at a frame rate that satisfies Nyquist theorem by the top camera. The exposure time was optimized for different measurements to achieve sufficient SNR while not saturating the camera (see Supplementary Note 6 for frame rate and exposure time). To determine oscillation amplitude (in gray scale) from the captured images, temporal FFT was performed by MATLAB (R2019b, MathWorks) for all pixels in every 1 s images in the image sequence, and the 40 Hz amplitude component was extracted. The FFT image showed bright spots that were assigned to single molecules. We integrated the total gray scale intensity of each bright spot within a circular region of interest (ROI, diameter of 7 pixels or 1 μm, the size of Airy disk in this case) as oscillation signal, and then used them to plot the I vs. U₀ curve. Each bright spot in the FFT image was checked, and only the ones with plateau appeared at high potential were selected as protein signals. From this curve, the molecular size was determined from the average plateau value and the calibration curve in Fig. 2a. Next, we took the average plateau intensity as where the PEG tether reached its maximum extension, and converted into nanometer (Supplementary Notes 1–2). At the same time, the U₀ could be converted to E₀(Δz₀) after the electric field was calibrated (Supplementary Note 2). At last, the charge of protein was determined by fitting the slope of the linear regime in Δz₀ vs. E₀(Δz₀) plot. Because of the existence of charge screening effect (Supplementary Note 4), the charge determined here should be regarded as the effective charge in 100 times diluted 1X PBS (1.5 mM).

For oscillation phase measurement, a ROI containing the protein molecule was selected. The average image intensity within the ROI was calculated for each frame in the image sequence. Then we used FFT to extract the phase information of the protein spot.

Data Availability:

The data that support the findings of this study are available from the corresponding author upon reasonable request.