Streptavidin-modified magnetic beads (200 nm) were purchased from Ademtech (France). Biotin-labeled CD45 antibody and biotin-labeled CD66b antibody was obtained from Biolegend (United States). FITC-labeled anti-EpCAM antibody, FITC-labeled anti-E-calmodulin antibody, APC-labeled anti-N-calmodulin antibody and Alexa Fluor 647-labeled anti-CD45 antibody were from Biolegend (United States). Alexa Fluor 488-labeled anti-CK antibody was purchased from Abcam (United States). Hoechst 33342 was from Beyotime (China). The human breast cell line (MDA-MB-468,MDA-MB-453,MDA-MB-231,MCF7) were from the Cell Resource Center, China Academy of Medical Sciences (China).

Biotin-labeled antibodies were ligated to streptavidin-modified magnetic beads via streptavidin-biotin interactions. The loading amount of the magnetic beads was 1 mg of streptavidin magnetic beads loaded with up to 10 µg of biotinylated antibody. Based on this, 20 μL and 10 µL of streptavidin-conjugated magnetic beads (5 mg/mL) were added to two 1.5 mL centrifuge tubes, respectively, and 2 µL and 1 µL of biotinylated CD45 antibody (0.5 mg/mL) and CD66b antibody (0.5 mg/mL) were added, respectively, and then PBS was added to the volume of the reaction system to be 100 µL and then incubated for 1 h.

The particle size distribution and zeta potential of CDmix@MNPs were tested by dynamic light scattering instrument (DLS,Nano-ZS-2019, Malvern, United Kingdom). The spectral properties of CDmix@MNPs were examined using a UV-Vis spectrophotometer (Lambda-950, Perkin Elmer Instruments, United States of America). The morphology of CDmix@MNPs was characterized by transmission electron microscopy (TEM, 120 kV, Hitachi-HT7700). The morphology of CDmix@MNPs captured leukocytes was characterized by scanning electron microscopy (SEM, WangZX-JSM-IT500). It was further verified by Prussian blue staining kit (Solarbio).

For SEM sample preparation, the wafers were first cleaned and dried with anhydrous ethanol, respectively. After coating the wafers with poly-lysine solution for 1 h, the magnetic beads and captured cells were added and settled for 1 h. The samples were dehydrated with 10%, 30%, 50%, 70%, 90%, and 100% ethanol for 10 min.

To induce the transformation of MCF7 cells from epithelial to mesenchymal phenotype MCF7 cells were starved with DMEM without fetal bovine serum for 24 h, and then 10 ng/mL of transforming growth factor-β (TGF-β) was added to DMEM containing 3% fetal bovine serum at 6 h, 24 h, and 36 h, respectively.

To facilitate cell imaging and counting, cultured cells were pre-stained with Hoechst33342 staining solution after digestion with 0.25% trypsin in a cell incubator at 37°C. CDmix@MNPs were incubated with 5,000 cells for spiking cell capture experiments. The final captured recovered cells were counted by fluorescence microscope scanning.

Blood samples from 20 cancer patients and 10 healthy volunteers were obtained from clinical hospital. Peripheral blood samples were collected and used for CTC analysis. Informed consent was obtained from patients and healthy volunteers before blood collection. The study was approved by the Ethics Committee. Peripheral blood was collected using EDTA vacuum tubes, and blood samples were stored at 4°C and processed within 72 h after blood collection. Mononuclear cells were enriched from 1 mL of blood by density gradient centrifugation, and the supernatant was recovered by enrichment in a magnetic field after incubation with a quantitative amount of mixed magnetic beads, and the final cell precipitate was obtained by centrifugation at 1500 rpm for 10 min. Cells were incubated with Alexa Fluor 488-labeled anti-CK antibody and Alexa Fluor 647-labeled anti-CD45 antibody for 1 h and settled on slides. The slides were sealed using DAPI-containing sealer, and cells that were CK-positive, CD45-negative and had intact nucleus were counted by microscopy.

Statistical analyses were performed using GraphPad Prism (GraphPad Software, San Diego, United States). Differences with a p-value <0.05 were statistically significant.

The results from dynamic light scattering revealed the average sizes of MNPs, CD45@MNPs, CD66b@MNPs, and CDmix@MNPs to be 213.1 nm, 231.2 nm, 222.8 nm, and 240.1 nm, respectively (Figure 1A). Correspondingly, the zeta potentials were measured as −26.07 ± 1.02 mV, −29.77 ± 1.05 mV, −28.03 ± 0.81 mV, and −30.7 ± 0.44 mV, respectively (Figure 1B). UV-vis spectrophotometer graphs demonstrated that MNPs lacked characteristic absorption peaks within the studied wavelength range of 200–800 nm. In contrast, CD45@MNPs and CD66b@MNPs exhibited distinctive absorption peaks around 260 nm, while CDmix@MNPs displayed a more pronounced absorption peak at the same position (Figure 1C). Transmission electron microscopy was employed to investigate the topographic characteristics of MNPs and CDmix@MNPs. The results illustrated that the size of CDmix@MNPs was larger compared with that of bare MNPs, which confirmed the effective attachment of antibodies to the bare magnetic beads (Figure 1D). Further investigation of the morphology of CDmix@MNPs capturing leukocytes, conducted through scanning electron microscopy, revealed a substantial number of CDmix@MNPs adhering to the surface of leukocytes, underscoring the strong affinity between CDmix@MNPs and leukocytes (Figure 1E). Prussian blue staining provided additional evidence that CDmix@MNPs specifically recognized and targeted leukocytes, predominantly distributing on the cell surface (Figure 1F).

Upon capturing 1 mL of peripheral blood using single CD45 antibody magnetic beads, the clearance of leukocytes exceeded 90%, yet the clearance of neutrophils was approximately 45%. Subsequent magnetic bead capture using CD66b antibody led to an increased neutrophil clearance of 80%, but the clearance of leukocytes dropped to less than 40%. Thus, a combination of CD45 and CD66b antibody beads became imperative. Dual-antibody hybrid bead capture achieved a leukocyte clearance exceeding 95%, while the neutrophil clearance rate approached 80% (Figure 2A). After establishing the initial capture method, further optimizations were undertaken. The effectiveness of direct and indirect methods for cancer cell recovery was compared. In the indirect method, cells were incubated with antibodies before the addition of magnetic beads, whereas in the direct method, antibodies were coupled with magnetic beads before incubation with cells. The direct method exhibited an approximate 20% increase in the recovery of cancer cells compared to the indirect method (p < 0.01) (Figure 2B). Subsequently, two methods of connecting the two antibodies to the magnetic beads were compared: mixing the antibodies first and then coupling them to the magnetic beads, or coupling the antibodies separately to the magnetic beads and then mixing. The latter method of attachment proved to be more effective (p < 0.01) (Figure 2C). The concentration of hybrid magnetic beads and incubation time were further studied to optimize capture conditions. The concentration of hybrid magnetic beads demonstrated a dual effect, with higher concentrations resulting in increased antibody attachment and faster separation. However, excessive concentrations led to self-aggregation of beads and increased non-specific binding. An optimal concentration of 0.75 mg/mL for hybrid magnetic beads was identified, as further increases did not significantly improve the cancer cells recovery rate (Figure 2D). The optimization of incubation time revealed that cancer cell recovery increased with longer incubation times, with 30 min identified as the optimal incubation time to meet the demand for rapid analysis (Figure 2E).

The results indicate the successful optimization of capture conditions for dual antibody hybrid magnetic beads, demonstrating their efficacy in recovering spiked cancer cells (MDA-MB-468) at varying concentrations (50–5000/mL). The precision of spiked cancer cell counts was ensured through meticulous enumeration using fluorescence microscopy before their introduction into healthy blood samples. Notably, the dual antibody magnetic beads exhibited the capability to recover cancer cells even at a low concentration of 50 spiked cancer cells per 1 mL of blood sample, with a remarkable recovery rate exceeding 80% at a spiked concentration of 5000 cells/mL (Figure 3A). Linear analysis revealed a robust and positive linear relationship between the recovery rate of cancer cells and the number of spiked cancer cells, where recovery increased linearly with the number of spiked cells (Figure 3B).

Limitation of EpCAM-based positive enrichment due to phenotypic heterogeneity of CTCs and variability in EpCAM expression, three breast cancer cell lines with differing EpCAM expressions—MDA-MB-231, MDA-MB-453, and MDA-MB-468—were selected. Flow cytometry confirmed high, moderate, and minimal EpCAM expressions in MDA-MB-468, MDA-MB-453, and MDA-MB-231, respectively (Figure 3C). Spiking these breast cancer cell lines into 1 mL of peripheral blood mononuclear cells (PBMC) in peripheral blood showed no significant differences in recovery rates, regardless of EpCAM expression levels (Figure 3D). The robustness of this approach was further validated by inducing epithelial-mesenchymal transition in cancer cells using TGF-β. MCF7, a breast cancer cell line, exhibited significant downregulation of EpCAM and E-cadherin expression and upregulation of N-cadherin expression following TGF-β induction (Figure 3E). When 5000 MCF7 cells and induced MCF7 cells were admixed with 1 mL of peripheral blood for capture, no significant difference in recovery between the two groups was observed (Figure 3F). These findings collectively reinforce the versatility and robust performance of the dual antibody hybrid magnetic bead approach in capturing CTCs, even in the context of phenotypic variability and EMT induction.

The impact of varying concentrations of hybrid magnetic beads on cell viability was investigated using the MTT method. The results demonstrated that as the concentration of hybrid magnetic beads increased from 0.0675 mg/mL to 1.00 mg/mL, there was no significant effect on cell viability during the 30 min incubation (Figure 4A). This indicates that the hybrid magnetic beads exhibit low toxicity to cells. To further assess the viability of recovered cells, Calcein-AM and propidium iodide (PI) dyes were employed for staining. Live and dead cell groups were included as controls, revealing that the vast majority of recovered cancer cells were viable (Figure 4B). This finding reinforces the notion that the hybrid magnetic beads exhibit good biocompatibility and are suitable for downstream analysis. Furthermore, the ability of recovered cancer cells to continue proliferating is a crucial aspect for subsequent functional analyses. Continuous culture of the recovered cancer cells for 7 days revealed that they maintained robust proliferative properties within this timeframe (Figure 4C). This observation underscores the suitability of the hybrid magnetic bead approach for downstream functional analyses, affirming the potential of this method in facilitating follow-up studies on recovered cancer cells.

Given the remarkable cancer cell capture efficacy demonstrated by the hybrid magnetic beads in vitro, samples from cancer patients (n = 20) with various tumor types, including liver (n = 10) and breast (n = 10), as well as samples from healthy volunteers (n = 10), were collected for circulating tumor cell capture and enumeration. Peripheral blood (1 mL) underwent pretreatment via density gradient centrifugation to aspirate the peripheral blood mononuclear cell, which was then incubated with the optimal concentration of hybrid magnetic beads under optimal conditions. Immunofluorescence staining was employed to identify recovered samples after capture, with nuclei stained using DAPI, tumor cells and leukocytes differentiated using anti-cytokeratin (CK) and anti-CD45 antibodies. Cells that stained positively for DAPI and CK and negatively for CD45 were identified as cancer cells. Representative CK-positive and CD45-negative CTCs were observed in blood samples from all 20 cancer patients (Figure 5A), with no CTCs detected in blood samples from any of the 10 healthy volunteers. In contrast, 2–14 CTCs were isolated from hepatocellular carcinoma patient’s blood samples, and 4–12 CTCs were isolated from breast cancer patient’s blood samples (Figure 5B). The ability to completely differentiate between blood samples from cancer patients and healthy volunteers indicates that the capture of CTCs by dual antibody magnetic beads exhibits excellent specificity.