Reversible capture and release of circulating tumor cells on a three‐dimensional conductive interface to improve cell purity for gene mutation analysis

Rare circulating tumor cells (CTCs) are always harvested with large numbers of white blood cells (WBCs) by current CTC isolation techniques, which influences the accuracy of CTC‐related gene and protein analysis. Therefore, it is very urgent to develop a method for efficient isolation of CTCs with high purity. In this work, we fabricated a reversibly assembled interface by layer‐by‐layer assembly of biotinylated poly‐(L‐lysine‐graft‐ethylene glycol) (PLL‐g‐PEG‐biotin), streptavidin, and CTC‐specific antibody on a three‐dimensional conductive scaffold, and further embedded it into a customized microchip for CTC capture. The assembled multilayers increased the roughness of the scaffold, which improved the capture efficiency. Importantly, the PLL‐g‐PEG‐biotin covalently coupled with CTCs could be detached from the scaffold by electrostatic repulsion, and the released CTCs could be recaptured on the conductive scaffold via electrostatic attraction while the WBCs were continuously drifted away. This reversible capture and release strategy significantly improved the purity of CTCs, and pure CTCs were obtained from cancer patients’ blood samples for gene mutation assay, which will be of great significance for assisting cancer diagnosis and treatment in a noninvasive way.


INTRODUCTION
It is estimated that the incidence of cancer will increase annually and is expected to increase further by nearly 50% by 2040. 2 Therefore, more efforts are required to expand early detection of cancer and to develop more effective cancer prevention and treatment methods. It is worth noting that various oncogenic mutations are commonly found in cancer patients and generally used as both biomarkers and rational targets for treatment of cancer. 3,4 For example, about 50% of Asian patients with lung adenocarcinoma are associated with epidermal growth factor receptor (EGFR) gene mutations and have good responses to the treatment with tyrosine kinase inhibitors. [5][6][7][8] In addition, pancreatic and colon cancer patients with Kirsten rat sarcoma viral oncogene (KRAS) mutations are quite common, present resistance to cetuximab therapy, and have a poor prognosis. [9][10][11][12] Therefore, detection of gene mutations has essential roles in assisting cancer diagnosis, prognostic assessment, and the selection of personalized treatment protocols for patients.
Currently, tissue biopsy is the standard method in the clinic for gene mutation detection. 13,14 However, tissue biopsy is invasive and very traumatic for patients, and it is difficult to obtain enough tissue through surgery for some patients, which leads to limitations in disease and treatment monitoring. Alternatively, for the past few decades, liquid biopsy has attracted extensive attention in the early diagnosis and dynamic monitoring of cancer due to its noninvasiveness and safety. [15][16][17][18] Circulating tumor cells (CTCs), important biomarkers for liquid biopsy, are shed from living tumors and carry complete genomic DNA and RNA information of cancer patients. [19][20][21] CTCs can not only be used for early diagnosis of cancer, but also provide biological fidelity about the tumors for gene and protein analysis. However, in patients' blood, extremely rare CTCs are always present together with large number of blood cells, and the genome and proteome of blood cells affect the accuracy of CTC downstream analysis. 22,23 Hence, it is vital and challenging to efficiently isolate CTCs with high purity for further analysis.
Nowadays, a number of methods have been developed to efficiently capture CTCs based on the distinct physical or biological properties of CTCs from blood cells. 20,[24][25][26] Among them, microfluidic devices are more widely used because of their variable designs, easy multifunctional integration, highly efficient processing of complex fluids, and so on. 22 Generally, microstructure or nanomaterials are often integrated into microfluidic devices to increase the contact frequency or enhance topographic interaction between substrates and cells, improving capture efficiency of CTCs, such as CTC-chip, 27 herringbone-chip (HBchip), 28 size dictated immunocapture chip (SDI-chip), [29][30][31] NanoVelcro chip, 32,33 etc. Meanwhile, in order to realize further downstream analysis of CTCs, it is eagerly necessary to build release systems in microchips for recovering highly pure and viable CTCs. 34 As a consequence, a range of release methods have been developed, classified as thermosensitive release, 35 light-controlled release, 36 ligandcompetition release, 37,38 electrochemical release, [39][40][41][42] chemical-responsive release, 43 etc. However, the cellular integrity and viability may be damaged by release process, and a large number of nonspecifically adsorbed white blood cells (WBCs) are usually co-collected with CTCs by current CTC isolation and release techniques, restricting downstream analysis of CTCs. Therefore, it is still desirable to develop a versatile platform to obtain highly pure CTCs in a simple way for downstream gene analysis.
In our previous work, we have developed a threedimensional (3D) scaffold gelatin-microchip 44,45 and gold nanotubes (Au NTs)-coated 3D conductive scaffold microchip 46 to efficiently capture and release CTCs, respectively. In order to further improve the purity of CTCs for gene mutation analysis and simplify the fabrication of microchips, herein, we described a new type of 3D conductive scaffold microchip with superior performance in isolating high-quality CTCs (Figure 1). The 3D scaffold was sequentially assembled with poly(3,4ethylenedioxythiophene) (PEDOT), carboxyl graphene (CG), biotinylated poly-(L-lysine-graft-ethylene glycol) (PLL-g-PEG-biotin), streptavidin (SA), and biotin-labeled anti-EpCAM antibody, and then embedded into a customized indium tin oxide (ITO)-based single channel manifold for a microchip. The microchip had flexible porous scaffold, chaotic migration mode, and increased surface roughness, which cooperated to improve capture efficiency of CTCs. When multiple cyclic positive potentials were applied, the scaffold was highly positively charged, inducing the desorption of positively charged PLL-g-PEG-biotin from the scaffold to release CTCs. Furthermore, the released CTCs coupled with PLL-g-PEG-biotin could be reabsorbed on the 3D conductive scaffold microchip via electrostatic attraction, while free WBCs released together with CTCs flowed through the microchip after the second washing process. Therefore, the purity of CTCs was significantly improved. This new type of 3D conductive scaffold microchip could simply and efficiently achieve reversible capture and release of CTCs, which was further applied to cancer patients' blood samples for gene mutation analysis of CTCs.   Scanning electron microscopy (SEM) images were taken by a Zeiss Sigma field-emission scanning electron microscope (Zeiss, Sigma). Electrochemical measurements were conducted using a CHI 660E electrochemical workstation (CHI Instruments) at room temperature. Fluorescence microscopy images were obtained using a Zeiss microscope (AxioObserver Z1, Zeiss, Germany). Laser scanning confocal microscopy images were taken by ZEISS LSM 900.

2.2
Assembly of conductive interface on 3D porous scaffold A 3D porous PEDOT/CMC/PDMS scaffold (denoted as PCP scaffold) was prepared according to previous work. 44,45,47 Briefly, Ni foam slice of 30 mm × 4 mm ×1 mm with an interconnected network was used as a template to fabricate 3D porous PDMS scaffold. After the 3D PDMS scaffold was treated with a plasma cleaner for 5 min, it was immersed in 10 mM CMC solution for 1 day and then dried at 80 • C. Subsequently, the CMC/PDMS scaffold was immersed in 0.8% PEDOT:PSS solution (with 5% DMSO) and centrifuged at 8000 rpm for 4 min. Then, excess of PEDOT:PSS solution was removed by centrifugation at 1500 rpm for 2 min in another empty centrifuge tube. After being dried at 80 • C for 30 min, the 3D porous conductive PCP scaffold was obtained.
Next, after being treated by plasma cleaner for 3 min, the PCP scaffold was immersed in 2 mg/ml CG solution. Then, it was incubated for 1 h on a shaker and dried at 70 • C for 30 min. After that, the above steps were repeated once. Thus, the CG/PEDOT/CMC/PDMS scaffold (denoted as CG/PCP scaffold) was obtained.

Functionalization and characterization of the 3D conductive CG/PCP scaffold
Firstly, the 3D CG/PCP scaffold was functionalized with PLL-g-PEG-biotin. The scaffold was treated with 200 μg/ml PLL-g-PEG-biotin in 10 mM N-2-hydroxyethylpiperazine-N-ethane-sulphonicacid (HEPES) buffer (pH 7.4) for 1 h at room temperature. Then, the scaffold was washed with phosphate-buffered saline (PBS) three times. Secondly, the scaffold was immersed in 50 μg/ml SA solution to incubate for 1 h and then washed with PBS three times. Thirdly, 10 μg/ml biotin-labeled mouse antihuman anti-EpCAM monoclonal antibody was incubated with the scaffold for 1 h and then washed with PBS. The scaffold functionalized with antibody can be stored at 4 • C until being used. Additionally, SA and biotin-FITC were used to identify the successful modification of PLLg-PEG-biotin on the surface of the CG/PCP scaffold, while the scaffold without modifying PLL-g-PEG-biotin was used as a control. DyLight 488-labeled goat anti-mouse immunoglobulin G (IgG) was employed to identify the successful conjugation of antibody on the surface of the scaffold, while the scaffold without conjugation of antibody was used as a control.
The colorimetric sandwich enzyme-linked immunosorbent assay (ELISA) was employed to compare the amount of anti-EpCAM antibody immobilized on the PDMS scaffold and CG/PCP scaffold. The two kinds of scaffolds were functionalized with anti-EpCAM antibody, which were then integrated into customized microchips with a single channel manifold. Then, HRP-labeled goat anti-mouse IgG was pumped into the chips, incubated for 1 h, and washed with PBS at a flow rate of 50 μl/min for 3 min. TMB (100 μl) was introduced into the chips to react with HRP for 15 min in the dark. Subsequently, 2 M H 2 SO 4 was introduced into the chips to stop the reaction. The reaction solution was collected in a 96well plate, and the signal was recorded at 450 nm using a microplate reader. All the experiments were repeated three times.
Since the 3D scaffold had a continuously porous structure, it was difficult to measure the contact angles. Therefore, the complanate PDMS film was obtained by spin coating and assembled with CMC, PEDOT, and PLL-g-PEG-biotin to characterize the water contact angles by a contact angle measurement system (DSA 100S, KRUSS, Germany), proving the assembly processes of multiple layers on the 3D scaffold.
To characterize the stability of assembly layers on the scaffold, the scaffolds functionalized with anti-EpCAM antibody and DyLight 488-labeled goat anti-mouse IgG were integrated into customized microfluidic chips. Then, PBS was pumped into the microchips at different flow rates (10, 50, 100 μl/min) for 60 min. A microchip without PBS washing was used as a control. The fluorescence images were taken every 10 min during this process. All these experiments were repeated three times. In addition, the fluorescence intensity was calculated by ImageJ software and normalized.

Optimizing the release condition of electrical stimulation
After being functionalized with anti-EpCAM antibody and DyLight 488-labeled goat anti-mouse IgG, the scaffolds were operated as a working electrode with a standard three-electrode setup. The cyclic potential of electrical stimulation with four different kinds of voltages swept from -0.8 to +0.5 V, from -0.8 to 0 V, from 0 to +0.5 V, and from 0 to +1.0 V were applied to the scaffolds at a scan rate of 100 mV/s for 20 cycles in PBS. Then, the scaffolds were washed with PBS for 10 min. All these experiments were repeated three times. Fluorescence microscopy images of the scaffolds were taken before and after electrical stimulation. Then, fluorescence intensity was calculated by ImageJ software. Thus, the release efficiencies under different conditions were obtained.

2.5
Evaluating capture and release performance of the 3D conductive scaffold microchip In this work, a customized ITO-based microchip with a single channel manifold 46 was used and H1975 cells were selected as the model cells to evaluate the capture and release performance of the 3D conductive scaffold chip. The H1975 cells were stained with DiI and washed with PBS three times. Then, the cells were resuspended at 10 6 cells/ml and further diluted to the desired concentration. Different numbers of H1975 cells (20,50,100,200, 500 cells) resuspended in 500 μl PBS were pumped into the microchips at a flow rate of 25 μl/min, followed by PBS washing at a flow rate of 50 μl/min for 3 min. The H1975 cells captured on chips were counted under a fluorescence microscope to calculate capture efficiency. Subsequently, a cyclic potential of electrical stimulation with a certain voltage was applied to the microchips at a scan rate of 100 mV/s for 20 cycles in PBS. Then, the chips were washed with PBS at a flow rate of 200 μl/min for 10 min to release the captured cells. After that, the released cells were counted to calculate release efficiency.
Hela cells and WBCs were used as controls to test the specificity of anti-EpCAM antibody-functionalized 3D conductive scaffold microchip. WBCs were obtained from lysed human whole blood. They were stained with DiI and resuspended at 2 × 10 6 cells/ml in PBS. After that, 500 μl PBS containing 10 6 Hela cells or WBCs was pumped into microchips at a flow rate of 25 μl/min, followed by PBS washing at a flow rate of 50 μl/min for 3 min. The cells captured on the chips were counted to calculate capture efficiency. In addition, 500 μl PBS containing 200 DiI-prestained H1975 cells was pumped into 3D CG/PCP scaffold chip without functionalization with anti-EpCAM antibody, which was also used as a control.

Viability and culture of the released cells
The released H1975 cells by this method were stained with Calcein-AM and PI at 37 • C for 30 min to analyze their viability, and the unprocessed cells were employed as controls. Then, the cells were observed under a fluores-cence microscope. Furthermore, the released H1975 cells were cultured for several days in 24-well plates under a CO 2 incubator to detect their proliferation ability, and the unprocessed cells were used as controls. The number of cells in the 24-well plates was recorded every day.

2.7
Isolation of CTCs from mimic cancer patients' whole blood Five hundred microliters of healthy people's whole blood spiked with 20-400 H1975 cells (20,50,100,200, 400 cells) was used to mimic cancer patients' blood samples, which were injected into microchips at a flow rate of 25 μl/min. After that, PBS was used to wash the hematologic cells at a flow rate of 50 μl/min. Then, 4% paraformaldehyde was injected into the chips to incubate for 30 min and washed with PBS at a flow rate of 50 μl/min for 3 min. Subsequently, Triton X-100 was introduced into the chips to incubate for 10 min and washed with PBS. Then, 5% BSA was injected into the chips to incubate for 1 h to reduce the nonspecific binding of fluorescent dyes, followed by washing with PBS. Finally, FITC-CK, PE-CD45, and Hoechst 33258 were pumped into the chips, incubated for 2 h at 4 • C, and then washed with PBS at a flow rate of 50 μl/min for 10 min to remove the extra fluorescent dyes. At last, they were observed under fluorescent microscope to count cancer cells and obtain the capture efficiency. Captured cancer cells were classified as Hoechst+/FITC-CK+/PE-CD45-, and WBCs were defined as Hoechst+/FITC-CK-/PE-CD45+.
After capture, a cyclic potential of electrical stimulation with a certain voltage was applied to the microchips at a scan rate of 100 mV/s for 20 cycles in PBS. Then, the chips were washed with PBS at a flow rate of 200 μl/min for 10 min to release the captured cells. The released cells were collected and counted under fluorescence microscope.

2.8
Repeat the capture and release process to improve the purity of CTCs According to the above description, these mimic blood samples were processed by anti-EpCAM antibodyfunctionalized 3D conductive scaffold microchips to first capture and release CTCs, which were stained by the three-color immunocytochemistry (ICC) with FITC-CK, PE-CD45, and Hoechst 33258. Then, the first released cells were collected and re-pumped into the 3D conductive CG/PCP scaffold microchips without PLL-g-PEG-biotin and antibody modification at a flow rate of 25 μl/min for the second capture. After that, the recaptured cells were released again by applying a cyclic positive potential at a scan rate of 100 mV/s for 20 cycles and washing with PBS. The numbers of CTCs and WBCs that were captured and released twice were both recorded.
In order to verify the necessity for CTC purification, three groups of cells were prepared, namely, 50 H1975 cells were added to whole blood samples without any treatment, and 14 H1975 cells were added to 1000 or 100 WBCs. They were then lysed for DNA extraction using the TIANamp Genomic DNA Kit and analyzed by amplification refractory mutation system polymerase chain reaction (AMRS-PCR) method using a human EGFR gene mutations detection kit to detect EGFR gene mutations.

Isolation of CTCs from cancer patients' whole blood and their gene mutation analysis
The study protocol was reviewed and approved by the Institutional Review Board (IRB) of Renmin Hospital of Wuhan University. This research was conducted in line with the guidelines established by the Declaration of Helsinki. Ethylenediamine tetraacetic acid anticoagulated whole blood samples from cancer patients (including eight lung cancer patients and one colon cancer patient) and healthy volunteers were obtained from Renmin Hospital of Wuhan University. One milliliter of whole blood samples were injected into the anti-EpCAM antibodyfunctionalized microchips, and PBS was used to remove the hematologic cells. Following, captured cells in the chips were fixed with 4% paraformaldehyde for 30 min, permeabilized with Triton-X 100 for 10 min, and then treated with 5% BSA for 1 h. Then, FITC-CK, PE-CD45, and Hoechst 33258 were used to stain for 2 h at 4 • C. After washing, they were observed under fluorescence microscope to record the number of cancer cells. Captured cancer cells were classified as Hoechst+/FITC-CK+/PE-CD45-, and WBCs were defined as Hoechst+/FITC-CK-/PE-CD45+. Next, the captured cells were released by applying cyclic positive voltage to the chips and then counted under fluorescence microscope.
Three of these patients were shown to have mutations in either the EGFR gene or the KRAS gene by tissue biopsy. Therefore, the released cells from these samples were re-pumped into the 3D conductive CG/PCP scaffold microchips for a second capture and release to purify the obtained CTCs. The number of CTCs was observed and recorded after the second capture or release. Next, the obtained cells were lysed for DNA extraction using the TIANamp Genomic DNA Kit and analyzed by AMRS-PCR method using a human gene mutation detection kit to detect gene mutations.

Assembly and characterization of conductive interface on 3D scaffold microchip
In this work, 3D scaffold was assembled with multiple functional layers to be endowed with both conductivity and biological specificity. Simply, 3D PDMS scaffold was coated with CMC and PEDOT to form a uniform and coherent conductive coating on the scaffold. 47 Herein, PCP scaffold was further assembled with CG via electrostatic attraction. Compared with bare PCP scaffold, CG/PCP scaffold was rougher and had more wrinkles, as shown by SEM images (Figures 2A and S1A). Furthermore, it was characterized by cyclic voltammetry in K 3 Fe(CN) 6 solution that the CG/PCP scaffold had improved conductivity and excellent electrochemical performance compared with PCP scaffold ( Figure 2B, red and black lines).
The CG/PCP scaffold was then assembled with positively charged PLL-g-PEG-biotin (denoted as PPB/CG/PCP scaffold) and characterized by sequential incubation with SA and biotin-FITC. As shown in Figure (S2), the fluorescence of the PPB/CG/PCP scaffold was strong and uniform, while the CG/PCP scaffold without assembly of PLL-g-PEG-biotin showed very weak fluorescence, proving the successful assembly of PLL-g-PEG-biotin. Besides, as shown in Figure (2C), the water contact angle decreased with the assembly of CMC, PEDOT, and CG on the PDMS and then increased to 75.8 • after modification with PLL-g-PEG-biotin because of its long hydrophobic chain.
For CTC recognition property, the PPB/CG/PCP scaffold was further conjugated with SA and biotin-labeled anti-EpCAM antibody. As shown in Figures (2D and  S3), the anti-EpCAM antibody-functionalized CG/PCP scaffold showed strong and uniform fluorescence characterized by DyLight 488-labeled goat anti-mouse IgG, while the fluorescence of CG/PCP scaffold without modification with anti-EpCAM antibody was almost negligible. Since antibody blocked the electron transfer, the current of the antibody-modified CG/PCP scaffold was slightly decreased compared with that of the pure CG/PCP scaffold ( Figure 2B, blue and red lines). Both the fluorescent and electrochemical characterization proved the successful conjugation of anti-EpCAM antibody on the scaffold. Furthermore, ELISA assay showed that CG/PCP scaffold could immobilize more anti-EpCAM antibody than that of PDMS scaffold since the rougher surface of CG/PCP scaffold could provide more surface area to effectively bind with PLL-g-PEG-biotin, offering more biotin sites to interact with SA as well as anti-EpCAM antibody ( Figure S4). As-prepared flexible and elastic anti-EpCAM antibody biofunctionalized CG/PCP scaffold was embedded into a customized single channel manifold for a microchip 46 and the stability of anti-EpCAM antibody-functionalized 3D conductive scaffold microchip was determined. As shown in Figure (2E), when DyLight 488-labeled goat anti-mouse IgG-functionalized CG/PCP scaffold was flushed by PBS at different flow rates (10, 50, 100 μl/min), the green fluorescence of 3D scaffold barely decreased, confirming that each layer was tightly assembled on the scaffold and stable enough for CTC capture.
Studies have shown that PEDOT could be used to fabricate conducting polymer bioelectronic substrate for releasing biotargets from the substrates by electrostatic repulsion under electrostimulation. 41,[48][49][50] Here, DyLight 488-labeled goat anti-mouse IgG-functionalized PPB/CG/PCP scaffold was applied with cyclic positive voltage for electrostimulation ( Figure S5) and the green fluorescence was decreased ( Figure S6). Quantitative results showed that the decrease in fluorescent intensity was the greatest when swept from 0 to 1.0 V ( Figure 2F). This phenomenon could be explained by the conductive interface on 3D PPB/CG/PCP scaffold accumulating abundant positive charges under cyclic positive potential, resulting in the desorption of positively charged PLL-g-PEG-biotin monolayer from CG/PCP scaffold. Therefore, this cyclic potential (0 to 1.0 V) was chosen as the optimized condition for releasing process.

3.2
Performance of 3D conductive scaffold microchip for capture and release of H1975 cells H1975 cells overexpressing EpCAM were used as model cells to evaluate the performance of the 3D conductive scaffold microchip (Figure 3A), and Hela cells and WBCs were used as negative controls. As shown in Figure (3B), about 97.1% of H1975 cells could be successfully captured. Fluorescence images showed that H1975 cells could be captured  Figure  S7). In contrast, the capture efficiencies of Hela cells and WBCs were only 0.033% and 1.5%, respectively. Meanwhile, the bare conductive interface without functionalization of anti-EpCAM antibody could only nonspecifically trap 9.0% of H1975 cells. These results indicated that 3D conductive scaffold microchip could specifically capture CTCs with high efficiency. This may be attributed to the rough surface of the scaffold enhancing topographic interaction between cells and microchip ( Figure 3C) and the increased amount of anti-EpCAM antibody immobilized on the scaffold providing more binding sites for specific recognition. The captured H1975 cells could be released by electrochemical stimulation swept from 0 to 1 V at a scan rate of 100 mV/s for 20 cycles in PBS ( Figures 3D, S8 and S9). Quantitative results showed that when 20-500 DiI-prestained H1975 cells were spiked in PBS and introduced into the microchip, 95.9% of H1975 cells were captured successfully, and 90.8% of captured H1975 cells were released after cyclic scan. All these data supported that this new type of 3D conductive scaffold microchip can capture and release EpCAM-positive cells specifically and efficiently.
The viability of cells was very important for downstream analysis, which was evaluated by Calcein-AM and PI stain-ing here. As shown in Figures (S10A and 3E), the majority of released H1975 cells were stained by Calcein-AM with a viability of 99.8%, which was similar to the unprocessed cells of 99.9% viability, indicating their good viability after capture and release process. The good viability of released H1975 cells endowed them with perfect proliferative ability. As shown in Figures (3F and S10B), the released H1975 cells could stick well on the cell culture plate and proliferate normally during 4 days of culture, which was concordant with the performance of control cells. Actually, the released H1975 cells could be used for long-term cell passages ( Figure S10C). These results demonstrated that this electrically stimulated release method was moderate and had no significant negative effect on the viability and proliferation of cells.

Isolation of CTCs from mimic cancer patients' whole blood
The mimic cancer patients' whole blood samples were used to evaluate the performance of 3D conductive scaffold microchip, and three-color ICC through FITC-CK (green), PE-CD45 (red), and Hoechst 33258 (blue) was employed to identify the captured cells. When 20-400 H1975 cells were spiked into healthy human blood and flowed through the 3D conductive scaffold microchip, 87.7% of H1975 cells were successfully captured and marked by Hoechst+/FITC-CK+/PE-CD45-with a size larger than 10 μm, while WBCs were labeled by Hoechst+/FITC-CK-/PE-CD45+ with a size less than 10 μm (Figures 4A,C and S11A). Then, when a cyclic potential of electrical stimulation swept from 0 to 1 V was applied to the chip, 79.0% of captured H1975 cells were released successfully ( Figures 4B,C and S11B). These results indicated that the 3D conductive scaffold microchip could perform excellently to obtain free CTCs from clinical cancer patients' blood.
Beyond the enumeration of CTCs, molecular analysis of CTCs would provide more information to support diagnosis and therapy of cancer. H1975 cells, as a nonsmall cell lung cancer (NSCLC) cell line, carry two EGFR gene mutations, T790M and L858R, 18,35,51 which were harvested for downstream gene mutation analysis. As shown in Figures (4D and S12), when 14 H1975 cells were cocollected with 1000 WBCs, no gene mutation signal was detected ( Figure 4D, column B), which was consistent with the results of 50 H1975 cells spiked in whole blood without purification process. In contrast, the T790M and L858R mutations were both detected when 14 H1975 cells were mixed with 100 WBCs. These results indicated that if rare CTCs were immerged in whole blood or isolated with less efficiency and purity, the concomitant WBCs may induce false negative results.
In this work, released CTCs were covalently bound with positively charged PLL-g-PEG-biotin monolayer, which could react with CG/PCP scaffold again via electrostatic attraction. Therefore, the released CTCs could be re-pumped into the 3D conductive CG/PCP scaffold microchip for the second capture. Free WBCs that were nonspecifically attached to the substrate and released concomitantly with CTCs would not bind to the scaffold again and then drifted away from microchip due to the F I G U R E 5 (A) Representative fluorescence images of single circulating tumor cell (CTC) captured from cancer patients' blood samples by three-dimensional (3D) conductive scaffold microchip. (B) Representative images of single CTC released from cancer patients' blood samples. (C) Quantitative data of single CTC captured and released from patients' blood samples (the cancer patients were numbered 1-9 from left to right). (D) Gene mutation analysis for CTCs isolated from the blood sample of the seventh cancer patient by amplification refractory mutation system polymerase chain reaction (ARMS-PCR) method. IC, internal control; Qc, quality control second fluid flow process. After the cyclic positive potential was applied to the scaffold again, the CTCs with improved purity were obtained. As shown in Figure (4E), 197-2022 (average number: 1018) WBCs and 29-195 (average number: 95) WBCs were recorded after the first and second release processes, respectively. Clearly, the number of WBCs decreased more than 10 times after the second capture and release process. Meanwhile, 76.4% of capture efficiency and 79.8% of release efficiency of CTCs were maintained during the second process ( Figure 4F), meaning that these processes would not obviously ruin the recovery efficiency of CTCs but would dramatically purify CTCs.

Isolation of CTCs from clinical cancer patients' whole blood and their gene mutation analysis
The 3D conductive scaffold microchip was applied to clinical patients' blood samples, including nine cancer patients and three healthy donors. Figure (5A) shows that some cells were successfully stained by Hoechst+/FITC-CK+/PE-CD45-with a size larger than 10 μm, which could be identified as CTCs. After cyclic electrical stimulation, the cells could be successfully released ( Figure 5B). In total, 4-33 CTCs were captured from 1 ml of nine cancer patients' whole blood samples, and 3-25 CTCs were successfully released and collected with a release efficiency of 78.2% ( Figure 5C and Table S1). There were no CTCs found in healthy persons' blood samples. Hence, all of the above results proved the excellent performance of 3D conductive scaffold microchip in isolating CTCs from clinical cancer patients' blood samples.
It is worth noting that three of nine patients were confirmed by tissue biopsy to have EGFR or KRAS mutations, and this kind of blood sample underwent a two-round capture-release process for gene analysis. Figure (5D) indicates that EGFR gene exon 19 deletion was successfully detected in CTCs isolated from blood sample of the seventh NSCLC patient. The eighth and ninth patients, with gene mutations detected by tissue biopsy rather than CTCs, had received chemotherapy and targeted therapy a few days before blood collection, and the undetectable gene mutations by CTCs could reflect good cancer prognosis after cancer treatment, which is another advantage of CTC-based liquid biopsy. Our results suggested that the 3D conductive scaffold microchip was expected to be used for isolating free CTCs together with sensitive PCR techniques for CTC gene analysis in clinical studies, which might provide more information for precise cancer diagnosis, prognosis assessment and therapy guidance of cancer.

CONCLUSIONS
In summary, we successfully designed a new type of 3D scaffold microchip with biofunctionalized conductive interface. The 3D scaffold was simple in fabrication, and assembled CG and PLL-g-PEG-biotin increased the surface roughness and amount of antibody modification on the scaffold, improving the capture efficiency of CTCs. Meanwhile, owing to electrostatic repulsion and electrostatic attraction, PLL-g-PEG-biotin monolayer could be desorbed under multiple cycles of electrical stimulation and reabsorbed on the CG/PCP scaffold, mediating the reversible capture and release of CTCs, which could highly improve the purity of CTCs. This electrically stimulated release method was also moderate and did little harm to cell viability, facilitating downstream analysis of CTCs. Furthermore, this new type of 3D conductive scaffold microchip has been successfully used to capture and release CTCs from nine clinical cancer patients' blood, and EGFR gene exon 19 deletion was detected from isolated CTCs. Taken together, this work presents a promising multifunctional platform to efficiently capture and release highly pure and viable CTCs, which might be used for downstream molecular analysis and in vitro culture, assisting cancer diagnosis and treatment in a noninvasive way.

A U T H O R C O N T R I B U T I O N S
Yi-Ke Wang and Ming Wang contributed equally to this work. Shi-Bo Cheng, Yi-Jing Chen, and Cui-Wen Li helped to accomplish the experiments. Min Xie and Wei-Hua Huang supervised the project and revised the manuscript.

A C K N O W L E D G M E N T S
This work was supported by the National Natural Science Foundation of China (grants nos. 21974098, 22274120, 21725504, and 21721005).

C O N F L I C T O F I N T E R E S T
The authors declare no conflict of interest.