Purification of HCC-specific extracellular vesicles on nanosubstrates for early HCC detection by digital scoring

We report a covalent chemistry-based hepatocellular carcinoma (HCC)-specific extracellular vesicle (EV) purification system for early detection of HCC by performing digital scoring on the purified EVs. Earlier detection of HCC creates more opportunities for curative therapeutic interventions. EVs are present in circulation at relatively early stages of disease, providing potential opportunities for HCC early detection. We develop an HCC EV purification system (i.e., EV Click Chips) by synergistically integrating covalent chemistry-mediated EV capture/release, multimarker antibody cocktails, nanostructured substrates, and microfluidic chaotic mixers. We then explore the translational potential of EV Click Chips using 158 plasma samples of HCC patients and control cohorts. The purified HCC EVs are subjected to reverse-transcription droplet digital PCR for quantification of 10 HCC-specific mRNA markers and computation of digital scoring. The HCC EV-derived molecular signatures exhibit great potential for noninvasive early detection of HCC from at-risk cirrhotic patients with an area under receiver operator characteristic curve of 0.93 (95% CI, 0.86 to 1.00; sensitivity = 94.4%, specificity = 88.5%).


Fabrication of polydimethylsiloxane (PDMS) chaotic mixers
PDMS chaotic mixers were fabricated from a photolithographically-prepared master wafer prepared by inductively coupled plasma-reactive ion etching (ICP-RIE). To fabricate the master wafer, a protective layer of chromium and positive photoresist was first cast on top of a 4-inch silicon wafer; the silicon wafer was then exposed to UV light through a photomask of three rectangular microfluidic channels (60 mm in total length and 2 mm in width). The structure of the three microfluidic channels is shown in Supplementary Figure 1b. Next, the exposed regions of the photoresist were dissolved away and the exposed chromium was removed with an acid wash.
ICP-RIE was then applied to etch the silicon to a depth of 115 µm. A similar procedure was then used to etch 40 µm deep herringbone ridges on top of the silicon wafer. The structure of the herringbone ridges is shown in Supplementary Figure 1c. Prior to making replicas via injection moulding, the Si master was pre-treated by exposure to trimethylchlorosilane vapor for 1 min. Wellmixed PDMS precursor (RTV 615 A and B in a 10 to 1 ratio, GE Silicones) was injected to the mold and then incubated in an oven at 80 °C for 48 h to make a 5 mm-thick slab. The resulting PDMS chaotic mixers were peeled off from the silicon master wafer/molds. During injection moulding, two holes were also fabricated at the ends of the channel for insertion of tubing.

Computational simulation and scanning electron microscopy (SEM) for extracellular vesicle (EV) distribution analysis
The PDMS chaotic mixer of the EV Click Chip (Supplementary Figure 3a) has three parallel rectangular microfluidic channels (60 mm in total length and 2 mm width) connected head-to-tail.
The herringbone structure is capable of passively inducing a microvortex that stirs the flow, facilitating repeated physical contact between Si nanowire substrate (SiNWS) and the flow-through hepatocellular carcinoma (HCC) EVs, further enhancing EV capture performance. Computational fluid dynamics (CFD) was used to study the trajectories of the flow-through HCC EVs in the chaotic mixer. Supplementary Figure 3c and d show the trajectories of EVs under the assumption that i) the inlet flowrate is 1 mL h -1 , ii) the fluid density is 1060 kg m -3 , and iii) the viscosity is 0.0036 kg m -1 s -1 . Near the tops of the nanowires, the velocity of the flow approximates to zero because of the boundary layer and the no-slip condition, allowing the EVs to touch the SiNWS. Figure 3e shows a cross-sectional SEM image of SiNWS with HepG2 EVs captured onto both the tops of the Si nanowires and at different depths (0-1 μm, 1-2 μm, 4-5 μm, and 7-10 μm) along the sidewalls of the Si nanowires. Most HepG2 EVs were captured on the tops of the Si nanowires. Because the flow velocity approximates to zero near the SiNWS in the 1.3 μm thick boundary layer, EVs in this area can diffuse into the SiNWS for click chemistry-mediated capture via Brownian motion. The dissipative particle dynamics (DPD) simulation can predict the balance of forces on the bead (the simulated EV) quickly using Newton's momentum equation shown in equation (1).

Supplementary
In our DPD simulation, the diameter of EVs was 50 nm, the length, diameter, and spacing of Si nanowires were 10 μm, 100 nm, and 150 nm, respectively. There were a total of 48 EVs in this simulation model. The initial condition placed all EVs above the SiNWS in 2 μm. Figure 3f and h show the distribution probability profiles along the depths of the Si nanowires in SiNWS. The results show that the trend from the DPD simulation (n = 48) is similar to the experimental data (n = 108, counted in the SEM micrographs shown in Supplementary Fig. 3e), with the distribution probability profile of the DPD simulation showing 37.5%, 15% and 2% of EVs located at depths of the 0 μm (top), 1-2 μm, and 5-6 μm, respectively. Figure 3g shows the DPD-simulated distribution of the EVs (blue) captured along Si nanowires (brown, depth = 0-2 μm). Most EVs are immobilized on the tops of the Si nanowires.

Supplementary
In our study, the TCO-conjugated antibody agents were incubated with EVs in the artificial plasma samples before being subjected to the EV click chips. The EV recovery yield is flow-rate dependent as shown in Figure 2g

EVs and background EVs from artificial samples
Artificial plasma samples were prepared by spiking a) 10-µL aliquoted HepG2 cell-derived EVs into 90-µL plasma from a female healthy donor or female cirrhotic patient, b) 10-µL aliquoted SNU387 cell-derived EVs into 90-µL plasma from a male healthy donor or male cirrhotic patient and c) 10-µL aliquoted Hep 3B cell-derived EVs into 90-µL plasma from a male healthy donor or male cirrhotic patient. The EV recovery yield of the male HCC cell line (HepG2) observed for EV Click Chips can be obtained from the following equation (2) (the copy numbers of SRY transcripts in the original 10-µL aliquoted HepG2 EVs and the EV Click Chip-recovered HepG2 EVs were denoted as SRY transcriptsori-EV and SRY transcriptsrec-EV, respectively): The purities of the male HCC cell line (HepG2) EVs harvested from EV Click Chips were calculated as the ratio of recovered SRY transcripts (contributed by recovered HepG2 EVs only) to C1orf101 transcripts (contributed by both recovered HepG2 EVs and the non-specifically captured background female plasma-derived EVs, denoted as C1orf101transcriptsrec-EV) using the following equation (3): For HCC cell lines without SRY transcripts (SNU 387, Hep3B), cancer cell-derived EVs were spiked into the plasma from a male donor or male cirrhotic patient, and the EV recovery yields and purities can be calculated using the following equations (4) and (5)

Characterization of EVs by fluorescence microscopy and dynamic light scattering (DLS)
To track the capture and release processes of HCC EVs spiked in blood plasma samples in EV Click Chips, we conducted two parallel characterization studies. In the first study, RNase-treated HepG2 EVs were first labeled with PKH26 dye and then spiked into healthy donors' plasma samples. The is very similar to that observed for HepG2 EVs in PBS (Figure 3). In the second study, RNasetreated HepG2 EVs were first exposed to protease K, followed by PKH26 dye labeling. The Dynamic light scattering (DLS) was adopted to characterize the size distribution of HepG2 EVs in solution. For these studies HepG2 EVs were placed into a disposable microcuvette and analyzed using a Zetasizer Nano instrument (Malvern Instruments Ltd., UK) at room temperature.

Validation of primers and probes for the 10 HCC-specific mRNA markers using ddPCR
To demonstrate the presence of SRY, C1orf101 and the 10 HCC-specific genes, i.e. To ensure the reproducibility of the ddPCR assay, we validated the PCR primers and probes using cDNA obtained from HepG2 cells, HepG2 EVs, and HCC EVs purified from 5 HCC patients' plasma samples. Each HCC patient's plasma was split into 3 samples for independent analysis for HCC EV purification and HCC-specific mRNA profiling, as shown in Supplementary Figure 8e,   The reproducibility of EV Click Chips was evaluated by calculating the percent coefficient of variation (%CV) for recovery yields. Intra-assay variability was measured for one operator who performed three tests on one day, whereas inter-assay variability was measured across three operators who performed five assay runs total (one run per day), with each run consisting of three tests (15 chips total). Source data are provided as a Source Data file. The reproducibility of EV Click Chips was evaluated by calculating the percent coefficient of variation (%CV) for recovery yields and recovery purities. HD, healthy donor; LCD, liver cirrhotic donor. Source data are provided as a Source Data file.