Graphene Sensor Arrays for Rapid and Accurate Detection of Pancreatic Cancer Exosomes in Patients’ Blood Plasma Samples

Biosensors based on graphene field effect transistors (GFETs) have the potential to enable the development of point-of-care diagnostic tools for early stage disease detection. However, issues with reproducibility and manufacturing yields of graphene sensors, but also with Debye screening and unwanted detection of nonspecific species, have prevented the wider clinical use of graphene technology. Here, we demonstrate that our wafer-scalable GFETs array platform enables meaningful clinical results. As a case study of high clinical relevance, we demonstrate an accurate and robust portable GFET array biosensor platform for the detection of pancreatic ductal adenocarcinoma (PDAC) in patients’ plasma through specific exosomes (GPC-1 expression) within 45 min. In order to facilitate reproducible detection in blood plasma, we optimized the analytical performance of GFET biosensors via the application of an internal control channel and the development of an optimized test protocol. Based on samples from 18 PDAC patients and 8 healthy controls, the GFET biosensor arrays could accurately discriminate between the two groups while being able to detect early cancer stages including stages 1 and 2. Furthermore, we confirmed the higher expression of GPC-1 and found that the concentration in PDAC plasma was on average more than 1 order of magnitude higher than in healthy samples. We found that these characteristics of GPC-1 cancerous exosomes are responsible for an increase in the number of target exosomes on the surface of graphene, leading to an improved signal response of the GFET biosensors. This GFET biosensor platform holds great promise for the development of an accurate tool for the rapid diagnosis of pancreatic cancer.


S 1 Electronic readout system
Electrical characterization of the different graphene sensing units contained inside the GFET chip array was performed using a custom-designed and portable readout system with a dedicated graphical user interface (GUI) connected to Matlab (Mathworks Inc., USA) via a Universal Serial Bus (USB). The portable readout system employed time-multiplexed signal acquisitions (MUX) from the 12 channels connected to each graphene sensing unit, whereby the individual drain-source currents (iDS) were detected in a voltage-divider circuit topology, amplified, and digitized by a high-speed and high-resolution analogue-to-digital converter (ADC) connected to the central microcontroller (MCU) on the embedded electronics. The drain-source voltage level (vDS) on the graphene channels is imposed by the user during the iDS measurements through the GUI, which converts the digital value into an analogue equivalent level using digital-to-analogue conversion (DAC) circuitry within the range ± 8V and resolution of 250 μV. For measurement of the iDS -vGS curve, also known as Dirac's point, an additional DAC is employed to provide an automatic voltage level sweep across the floating gate and within the range specified by the user. iDS measurements in both stimulation modes (including extended temporal measurements) can be applied to a single graphene channel or to all twelve channels at once, with the true value for iDS being calculated by a Fast Fourier Transform (FFT) algorithm applied to the samples acquired by the ADC, thereby achieving a current resolution of 20 nA and bandwidth of 15 kHz.

S 2 Functionalization and characterisation of graphene surface
Various characterization techniques were performed to examine the functionalization of TCPP and antibody on the graphene surface.
The Raman spectra ( Figure S7) recorded for bare graphene showed peaks at 1580 cm -1 , which is commonly attributed to the G band of sp 2 carbon in graphene. In addition, the presence of peaks at 1350 cm -1 (D band), 2700 cm -1 (2D secondary D band) is also shown, indicating a high quality of monolayer graphene. Raman spectroscopy analysis of the graphene surface after functionalization with TCPP showed the characteristic peaks of TCPP, 1 which are the vibrations of C-N (1240 cm -1 ), C-H (1454 cm -1 ), C-C (1565 cm -1 ), and a weak shoulder peak (1496 cm -1 ). The Raman D-to-G peak of the graphene increased from 0.32 (bare graphene) to 0.89 (after functionalization with TCPP), indicating that there is some disorder in the graphene sheet. The intensity ratio I2D/IG = 1.89 indicates high-quality graphene after the functionalization process.
X-ray photoelectron spectroscopy (XPS) was used to confirm the existence of TCPP and antibody on the graphene surface. GPC-1 antibodies were selected as a pancreatic cancer exosome biomarker targeting GPC-1 proteins on the surface of cancer exosomes. Figure S7.shows the changes in principle C 1s and N 1s core levels after each functionalization step. The high-resolution N 1s spectra show a significant increase in the N 1s peak at 399.8 eV after GPC-1 antibody conjugation. The high-resolution C 1s spectrum is composed of 4 components at the surface, C-C at 284.9 eV, C-O at 285.9 eV, C-N at 286.4 eV, and O-C═O at 287.7 eV, corresponding to the large number of amine and amide bonds present in the antibodies.
Atomic force microscopy was used to reveal changes in the surface morphology and quality of graphene after each step of the functionalization process ( Figure S7). The surface roughness of pristine graphene is measured to be 0.3 nm. After surface treatment with 2 h of TCPP, the surface roughness increased to 0.41 nm. After antibody modification, the surface roughness further increased to 0.68 nm.
These results suggest that the graphene surface was successfully functionalized and can be used for sensing target exosomes.

S 3 Selection of antibody for PDAC cancer exosome detection
Exosomes are extracellular vesicles that carry proteins, nucleic acids, and lipids. Many exosomeenriched proteins have been reported. Specific exosomal enriched proteins are expressed more on pancreatic cancer cell-derived exosomes compared to healthy ones, which offers the possibility to diagnose and distinguish patients with PC. 2 -4 Protein biomarkers enriched on pancreatic cancercell-derived-exosomes include glypican-1 (GPC-1), 2 RHOB, 5 CD63 and Prom1. 6 All of these proteins can be expressed on both cancerous and healthy exosomes. 4 In order to choose an appropriate biomarker for detection using GFETs, the protein should be highly expressed on cancerous exosomes with low expression on healthy exosomes. Therefore, the selectivity of each protein biomarker to cancer cell-derived compared to healthy exosomes was tested. Four types of antibodies (CD63, GPC-1, Prom1, and RhoB) were immobilized on the graphene surface to investigate their sensitivity and selectivity to PDAC cancer exosomes in patient plasma samples (see section S2 for details of the functionalization protocol for the graphene surface). Figure S8 summarizes the performance of these antibodies in differentiating PDAC cancer exosomes from healthy exosomes in blood plasma. Using AFM, it was found that all protein biomarkers are abundant on the PDAC cancerous exosomes. All the selected protein biomarkers were found to have increased expression in plasma exosomes in PDAC patients compared to the healthy controls. Furthermore, CD63 antibodies exhibited the highest capture rate for both cancerous and healthy exosomes. These results were expected, because CD63 has been observed to be enriched on all exosomal membranes. 7,8 The high concentration of CD63+ exosomes confirmed the presence of exosomes in the tested samples. 7,9,10 We found that Prom1 antibodies have low selectivity, while that of RHOB was higher. However, we also found that the GPC-1 antibodies had the highest selectivity to the PDAC cancer exosomes and the lowest capture number of healthy exosomes ( Figure S8), showing significantly higher specificity to the PDAC cancer exosomes. The specificity of each protein for cancer and healthy exosomes were 1.4:1, 3.2:1, 1.6:1, and 1.8:1 for CD63, GPC-1, Prom1 and ROHB respectively.

S 4 Dynamic light scattering (DLS)
Dynamic light-scattering measurements were performed using a Zetasizer Nano ZS (Malvern Instruments Ltd, Worcestershire, UK). The intensity, volume, and distribution data for each sample were collected on a continuous basis for 4 minutes in sets of three. Three different measurements from the isolated exosome samples were performed. Figure S1. Illustration of the optical image of on-chip integrated GFET devices.

B)
Supplementary Table   Table S1. Summary of the patient cohort.   Figure S9. Device exclusion criteria for GFET reproducibility and validity. Results from n = 281 clinically tested devices that satisfied the exclusion criteria are included in the study. Devices (n=25) with < 0.47 or > 0.95 were excluded in this study. Devices (n=12) that exhibited values of > 1700 Ω < 700 Ω were also excluded.