Exploiting Electrostatic Interaction for Highly Sensitive Detection of Tumor-Derived Extracellular Vesicles by an Electrokinetic Sensor

We present an approach to improve the detection sensitivity of a streaming current-based biosensor for membrane protein profiling of small extracellular vesicles (sEVs). The experimental approach, supported by theoretical investigation, exploits electrostatic charge contrast between the sensor surface and target analytes to enhance the detection sensitivity. We first demonstrate the feasibility of the approach using different chemical functionalization schemes to modulate the zeta potential of the sensor surface in a range −16.0 to −32.8 mV. Thereafter, we examine the sensitivity of the sensor surface across this range of zeta potential to determine the optimal functionalization scheme. The limit of detection (LOD) varied by 2 orders of magnitude across this range, reaching a value of 4.9 × 106 particles/mL for the best performing surface for CD9. We then used the optimized surface to profile CD9, EGFR, and PD-L1 surface proteins of sEVs derived from non-small cell lung cancer (NSCLC) cell-line H1975, before and after treatment with EGFR tyrosine kinase inhibitors, as well as sEVs derived from pleural effusion fluid of NSCLC adenocarcinoma patients. Our results show the feasibility to monitor CD9, EGFR, and PD-L1 expression on the sEV surface, illustrating a good prospect of the method for clinical application.

fluorescence measurements on the GFP tagged sEVs were performed with the 100x oil immersion lens of a Zeiss inverted microscope Colibri 5, equipped with a Hamamatsu CCD Camera (Orca Flash 4). The 475 nm wavelength LED was used for exciting the GFP tagged sEVs, and their images were recorded with 2s acquisition time across an area of 133.2 µm × 133.2 µm. The capture of sEVs was done following the functionalization scheme of APTES-GA and PPB-avidin, as mentioned in the materials and methods section. The resulting images are shown in the figure S1. On an average, a total of 567 sEVs were captured by the APTES-GA surface in a microscopic area, whereas 511 sEVs were captured by the PPB-avidin coated surface. The surface coverage is hence about 10% lower in case of the PPB-avidin surface.
This shows that despite the lower extent of sEVs captured by the PPB-avidin functionalization scheme, the net signal when measured by the streaming current method is still higher. Figure S1: Fluorescence images of the sEVs from H1975 cells stably expressing GFP-CD9 captured by (a) GA-APTES and (b) PPB-avidin functionalization schemes reveal that the number of sEVs captured is about 10% less in the PPB-avidin method.

S2. Effective surface charge density
The derivation of surface charge density of a flat surface in contact with an electrolyte is possible by solving the Poisson-Boltzmann equation. The resulting solution, known as Gouy-Chapman equation, is given by 1 , Where c is the ion concentration, N is Avogadro's number, 0 is the permittivity of the medium, is the Boltzmann constant, T is the temperature and 0 is the surface electrostatic potential. This solution to the Poisson-Boltzmann equation however is valid only for an ideally flat surface. In reality, as the surface is not flat, this equation only provides a lower bound for the estimation of , rather than an accurate expression 1 . The quantity "effective charge density" can be defined as the sum of the charge density of the flat surface, as well as the surrounding ions enclosed by the slip plane. This term is related with the zeta potential, * , and the relation, which has been derived elsewhere, is given by, Using the above equation, the value of could be determined from * obtained experimentally for the surfaces at various stages of the functionalization, and are given in

S4. EGFR-TKI Erlotinib and Osimertinib responses in H1975 cells
The response of H1975 cells to erlotinib (1µM) or osimertinib (0.1µM) was analysed at 48h post treatment when also the sEVs were harvested from cell culture media (figure S3a). As seen erlotinib at this dose had minor effect on cell morphology relative to untreated cells while osimertinib caused clear growth inhibition. The selection of these doses for erlotinib or

S5. Limit of detection (LOD)
The LOD was estimated as the concentration of the target corresponding to the MDS level of the sensor. This was estimated from the calibration curve in a semi-logarithmic scale as shown in figure S4, by considering the point of intersection between the calibration plot and the MDS level (0.1 mV) of the sensor. The linear regime of the sensor response was considered for this purpose. The LOD was obtained to be 4.9 × 10 6 particles/mL. This is an improvement of about two orders of magnitude over the LOD obtained by us previously 3 .
The LOD estimation was done using CD9 surface protein in the present study and EGFR surface protein in the previous study. As the signals in both cases are in the same order of magnitude, this does not affect our claim of improvement in LOD.

S6. Simulation parameters
The Adamczyk model 4 was used for the simulations, according to equation 1. Given the fact that the Debye length is 2.3 nm in our case, and the average radius of the sEVs is much higher (∼50 nm), the values of the parameters Ci and Cp lie in the saturation range according to the figure 8 of the reference 4 , and were taken to be 6.5 and 10.2 respectively. Moreover, according to the results of our previous work 5 that the roughness of the surface requires a reduction in Ci, we made it 2 times smaller. The of the EVs was assumed to be -30 mV. In reality, EVs are very heterogeneous in terms of their surface protein profiles, which can influence their . This was ignored for simplicity.

S7. sEV isolation and characterization
In this study sEVs were isolated from cell culture media of the NSCLC cell line H1975 (ATCC ® CRL-5908™,  For fluorescence experiments of sEVs, H1975 NSCLC cells were stably transfected with a CD9-GFP plasmid (Origene #RG202000). Briefly, cells at around 70% confluence were transfected using Lipofectamine 2000 (Thermo Fisher Scientific) and 0.7 ng plasmid.
Plasmid uptake in the cells were selected using Geneticin (G418, Thermo Fisher Scientific).
The isolation of sEVs from media from these cells were performed similarly as described for H1975 cells above.

S8. Cell morphology and cell viability analyses post EGFR-TKI treatment
The cell morphology of H1975 after 48h treatment with either erlotinib or osimertinib was analysed using a Nikon Eclips TS100 microscope using a 10x lens. The erlotinib or osimertinib induced cytotoxicity on H1975 cells were studied using 3-(4,5-dimethylthiazol-2yl)-2,5 diphenyltetrazolium bromide (MTT) assay. Thus H1975 cells were plated at a density of 5,000 cells/well 96-well plate using cell culture media and supplements as indicated above.

S9. Preparation of biotinylated EGFR-binding antibody
The EGFR-binding antibody cetuximab (Erbitux, Merck) was biotinylated with NHSactivated biotin (EZ-Link™ Sulfo-NHS-LC-Biotin, Thermo Fisher Scientific) that reacts with primary amino groups in the antibody. 0.5 mg/ml cetuximab was incubated with 6 times molar excess of biotin for 30 minutes at room temperature. After the biotinylation reaction, the mixture was purified and buffer exchanged to PBS with Amicon Ultra Centrifugal 100k