Samples collected and used consisted of peripheral blood samples of 50 patients with GC diagnosed by pathology and tumor tissues of 20 matched patients. Sampling was approved by the Ethics Committee of Zhabei District Central Hospital of Shanghai (approval no. ZBLL20180823) and patients signed informed consents. The HER2-positive GC cell line NCI-N87 was purchased from the Institute of Biochemistry and Cell Biology, Shanghai Institute for Biological Sciences, the Chinese Academy of Sciences.

Peripheral blood (7.5 ml) was collected from patients with GC in the hospital by medical anticoagulant blood collection tubes and the anticoagulant was EDTA-K2. The samples were stored at 4°C to avoid freezing during storage, processing and transportation and were used within 72 h.

Peripheral blood was collected for CTC count (magnetic separation immunofluorescence method was used for CTC count determination).

The clinical information of patients included their name (sample number), sex (32 male, 18 female), age (14 <65, 36 ≥65 years), clinical stage, pathology type and medication. The GC samples (peripheral blood and tumor tissue) were collected from Zhabei District Central Hospital from September 2019 to November 2020.

A key aim of the present study was to explore the dynamic monitoring of the CTC changes of the enrolled gastric cancer patients at least twice. The two time points set were preoperative (pre-treatment, when the patient was admitted to the hospital) and postoperative (post-treatment, one week after surgery).

RPMI medium, fetal bovine serum and trypsin were obtained from Gibco (Thermo Fisher Scientific, Inc.), anti-HER2 antibody (cat. no. ab134182) was from Abcam. PathVysion HER2 kit was from Abbott Pharmaceutical, magnetic liposomes materials were obtained from Huzhou Lieyuan Medical Laboratory Company Ltd.), the fluorescence microscope (BX61) was from Olympus Corporation, BI-90Plus laser particle size analyzer/Zeta potential analyzer and atomic force microscope (AFM) from Brookhaven Instruments Corporation and vibrating sample magnetometer (MPMS-XL-7) from Quantum Design, Inc.

The Cancer Cell Line Encyclopedia (CCLE) database was employed to obtain HER2 mRNA expression in 38 gastric cancer cell lines (Fig. S1).

HER2 short interfering (si)RNAs were designed by Ambion online software and three siRNAs (HER2-siRNA-1, HER2-siRNA-2 and HER2-siRNA-3) were synthesized. The specific sequences were shown in Table SI.

Negative control (NC)-siRNA was the negative control. After extracting total RNA, cDNA was obtained according to the manufacturer's protocols instructions of the QuantiTect reverse transcription kit (cat. no. 205311; Qiagen GmbH). qPCR amplification was performed by the method provided in the instructions of the PCR detection kit (SYBR Premix Ex Taq II; cat. no. .RR820A; Takara). The relative quantitative method is 2^{−ΔΔCq (28)} and three independent experiments were performed.

The expression level of HER2 mRNA was detected in NCI-N87 cells transfected with HER2-siRNA-1, HER2-siRNA-2, HER2-siRNA-3 and NC-siRNA. The purity of RNA is 1.7<OD260/OD280<2.0 by spectrophotometer. The reaction conditions were pre-denaturation at 95°C for 10 min and denaturation at 95°C, then annealing at 55°C for 15 sec and extending for 1 min for 32 cycles in total.

The preparation steps of HER2 LMB (H-LMB) and EpCAM LMB (Ep-LMB) were as follows: Cholesterol, DOPC, GHDC and HQCMC were dissolved in dichloromethane and then Fe₃O₄ particles were added. With a simultaneous addition of 0.1 mol/l PBS, the mixture was stirred and heated to 25°C for 30 min to realize complete emulsification after shaking and mixing, which was the LMB. Furthermore, 0.6 mg anti-HER2 (anti-EpCAM) antibody was dissolved in 10 ml of isopropanol, supplemented with coupling agent EDC and NHS to mix with the prepared LMB, followed by stirring at 4°C for 24 h at a constant speed, which was the anti-HER2 (anti-EpCAM) antibody-modified LMB, i.e., H-LMB.

During characterization, corresponding particle size and potential were measured following the dilution of 10 µl H-LMB (Ep-LMB) with 1 ml of distilled water. The prepared H-LMB (Ep-LMB) was observed using an atomic force microscope (AFM). An amount of 10 µl H-LMB (Ep-LMB) was taken and diluted with 1 ml of distilled water and then 50 µl of the diluted sample was smeared on the mica sheet for observation. The absorbance was measured at a wavelength of 280 nm with ultraviolet (UV) spectrophotometer with the dilution of another 10 µl H-LMB (Ep-LMB) with 1 ml of distilled water. The magnetic properties of H-LMB (Ep-LMB) and LMB were measured by a vibrating sample magnetometer. The method of obtaining Ep-LMB was the same as H-LMB.

After counting, cultured GC cells with high expression of HER2 were divided into five groups at different cell count gradients of 10, 20, 50, 100 and 200 cells/ml. These were divided into H-LMB and EpCAM LMB (Ep-LMB) groups after dilution with 7.5 ml PBS solution, with three replicates in each group. These groups were used to test the ability of the prepared H-LMB to capture NCI-N87 GC cells and to compare with Ep-LMB. An amount of 20 µl H-LMB was added to 7.5 ml of sample, which was then incubated at room temperature for 20 min and mixed once after every 5 min. The centrifuge tube was then inserted into the Magnetic Separation Rack for 10 min (room temperature) and the PBS solution was added to wash twice after discarding the supernatant. Staining was performed in the dark for 20 min (room temperature) following the addition of 20 µl FITC-labeled CK19 antibody (CK19-FITC), 20 µl DAPI staining solution and 10 µl PE-labeled CD45 antibody (CD45-PE). Following staining another two times of washing with double distilled water were performed on the Magnetic Separation Rack to remove completely the unbound antibody and staining solution. In the final step, 20 µl double distilled water was added into the centrifuge tube to suspend again and evenly distributed on the polylysine-treated slide. Observation and counting were performed under a fluorescence microscope (10× magnification) after the liquid dried.

All specimens were processed by marking, sectioning, fixation and other steps within 30-40 min following collection. The tissues were fixed using 10% neutral formalin (room temperature for 2-12 h). All the samples were confirmed to be GC by medullary staining of a qualified pathological examination institution. The present study used the most representative specimens from gastroscope biopsy and the primary tumor specimens were resected surgically (postoperative surgical specimens). The specimens were paraffin-embedded, serially sectioned (5 µm), dewaxed with xylene and dehydrated using 100, 100, 95, 90, 80 and 70% ethanol (12–15 min each). The mouse anti-human HER2 monoclonal antibody (OriGene Technologies, Inc.) was used as the primary antibody. Immunohistochemical SP staining kit was purchased from OriGene Technologies, Inc. IHC was used to detect the protein expression of HER2 in three tissues. The positive control was provided by the Pathology Department and PBS was taken as the negative control for staining instead of the first antibody.

The peripheral blood of 50 patients with GC and the tissues of 20 matched patients with GC were detected first by HER2-FISH test to calculate cases with HER2 abnormal expression. H-LMB and Ep-LMB were used to separate and screen CTCs from peripheral blood samples of 50 patients. Immunofluorescence identification and FISH were performed on the separated CTCs before and after treatment. Positive-CTCs were stained with -DAPI andCK19-FITC but not CD45-PE and exhibited typical tumor cell morphology (inconsistent size, shape and staining of nuclei). H-LMB and cells were stained with Prussian Blue Staining kit [nuclear fast red; cat. no. 60533ES20, San Yi Biotechnology (Shanghai) Co., Ltd.] for 10 min at room temperature.

The centromere of chromosome 17 (CEP17) was labeled green and the HER2 gene was labeled red. Following heat and high-pressure treatment for 3 min, pepsin (0.2 mg/ml) digestion for 20–40 min, dehydration and drying, 10 µl probe mixture was added in the target area of tissue slice followed by slice covering and molding. The tissue slices were then placed into an in situ hybridization system, denatured at 83°C for 5 min and hybridized at 45°C overnight (14–18 h). With a rapid washing, the tissue sections were dried at room temperature, 15 µl DAPI solution added to the target area and sealed. The number of 100-cell colonies was counted after reaction in the dark for 20 min and the results were observed under the fluorescence microscope.

The clinical characteristics of 50 patients with GC, such as age, tumor size, TNM stage and metastasis, were collected and analyzed in terms of their relationship with HER2 abnormal expression.

A set of sequencing primers (NGS Adapter Primer) were used to design by Sangon Biotech (sangon.com/), to sequence the exons and ±10 bp side introns of several related genes. DNA samples were used to prepare the captured target gene library and the Illumina high-throughput sequencing method (Novaseq, Hybrid selection, paired) was used for sequencing (detection by Genowise, genowise.com/). In the present study, some NGS detection results have been used as statistics, as shown in Table I. The sequencing data are available in NCBI Sequence Read Archive (ncbi.nlm.nih.gov/sra) with the accession number PRJNA767212.

All data in the present study were analyzed by SPSS 20.0 statistical analysis software (IBM Corp.). Pairwise comparisons were performed using an un-paired Student's t-test. P=value of tumor stages (I/II, III and IV), T stage (T1/T2/T3/T4), N (N0/N1/N2/N3) and M (M0/M1) were calculated by Fisher's test. the analysis of CTC numbers in pre-treatment group and post-treatment group was performed by a paired Student's t-test. The difference in the expressed level of Her2 among groups were calculated by ANOVA. Bonferroni was used for the post-hoc test used following ANOVA. All data were the results of three independent experiments. P<0.05 was considered to indicate a statistically significant difference.

Further research was focused on the separation and identification, quantitative statistics and gene detection of CTCs from peripheral blood of clinical samples. The detection technology is shown in Fig. 1.

The preparation of LMB needed verification and the characterization of related indicators needed to be completed at the beginning of the experiment, as shown in Fig. 2. The charge of the synthesized magnetic lipid nanoparticles was +31.5 mV; the average particle size was 261.4±3.7 nm distributed between 206.2 and 327.4 nm, showing a relatively narrow and uniform particle size distribution (Fig. 2A). The UV absorption spectrum showed that there was no absorption peak at 280 nm for LMB and a wide absorption peak at 280 nm for H-LMB, which were the properties of the protein. This suggested that the surface of the LMB was conjugated with an anti-HER2 antibody that was connected by EDS-NHS coupling reaction, exhibiting good specificity (Fig. 2B). AFM images revealed that the nanoparticles were spherical in shape with a size of ~250 nm with regular shape and without agglomeration. This was consistent with the particle size detection results (Fig. 2C). Fig. 2D shows the magnetic property test results of H-LMB. The hysteresis loop of H-LMB passed through the origin; besides, the coercive force and remanence were close to zero. This suggested that the prepared LMBs were superparamagnetic particles. It was concluded that the prepared nanoparticles were superparamagnetic LMBs modified with HER2 antibody.

The relative expression levels of HER2 messenger RNA (mRNA) in 38 different gastric cancer cell lines are shown in Fig. S1A. HER2 overexpression in recombined plasmids (HER2 overexpression) and the empty vector (Vector) was detected by RT-qPCR (Fig. S1B). The results proved that the expression level of HER2 mRNA was downregulated in NCI-N87 cells following transfection with HER2-siRNA-1, HER2-siRNA-2 and HER2-siRNA-3, compared with that in NC-siRNA group. In addition, following stable transfer of the HER2 overexpression plasmid into NCI-N87 cells, the expression level of HER2 mRNA increased significantly; ~7 times higher than compared with the negative control group. With the increase of H-LMB concentration, the activity of gastric cancer cell line NCI-N87 decreased significantly (P<0.05).

The activity of cells below the cell concentration of 10 µM was still >90% after 6 h, which indicated good biocompatibility of H-LMB and Ep-LMB (Fig. 3A). Subsequent study focused on the experimental simulation of the capture ability of H-LMB for CTCs from NCI-N87 GC cells with HER2 abnormal expression to verify the separation effect on CTCs, as shown in Fig. 3B. The recovery rate of H-LMB to NCI-N87 GC cells was stable at different concentrations compared with Ep-LMB and the capture efficiency of H-LMB was >80%. Therefore, H-LMB could be used to capture CTCs in clinical samples of GC with HER2 abnormal expression. The experimental results are shown in Fig. 3C. There was no blue signal in the blank control group, but it was observed on the cell surface after the addition of H-LMB. This suggested that H-LMB adsorbed on the cell surface and the signal intensity was positively correlated with incubation time and H-LMB concentration. The results of IHC showed that the staining of tissue was superior compared with CTCs (Fig. 3D). More detection results are shown in Fig. S2.

H-LMB and Ep-LMB were used to detect CTCs in blood samples of 50 patients with GC. Fig. 4A shows the immunofluorescence identification of CTCs captured by H-LMB. The CTCs in patients with GC had obvious cell morphology under the white light microscope, the green fluorescence of CK19-FITC was strongly positive and the blue fluorescence of DAPI was strongly positive, both of which overlapped after overlying. CD45 staining showed no fluorescence, which suggested that the captured cells were CTCs from GC.

The region with the highest degree of amplification was selected for counting and calculation of the ratio of dichromatic signals of ≥20 consecutive tumor nuclei. When the ratio of total HER2 signals to total CEPl7 signals was ≥2.2, it was considered to be positive for in situ hybridization, i.e., amplification. When multiple signals were connected in clusters or the signal ratio of HER2 and CEPl7 was >20, the ratio could not be calculated and it could be judged as positive for in situ hybridization. When the ratio of total HER2 signals to total CEPl7 signals was <1.8, it was considered to be negative for in situ hybridization, i.e., no amplification. Typical cases are shown in Fig. 4B and C. The representative positive and negative FISH detection of HER2 and EpCAM are displayed.

The detection rate of CTCs using Ep-LMB and H-LMB was 100% (n=50) and 72% (n=36) among the 50 patients with GC who were enrolled, respectively. The results of counting are shown in Fig. 5A. The count of CTCs decreased significantly with respect to prognosis in comparison to the pre- and post-treatment results (Fig. 5B), with the statistical difference (P<0.01) indicating its significance in the dynamic monitoring.

Table II shows that the prepared H-LMB effectively captured CTCs from GC. The positive rate of CTCs was 69.23% (9/13), 100% (6/6) and 100% (14/14) in patients with TNM stage I–II, III and IV, respectively, with obvious differences among the subgroups (P=0.03). The positive rate of CTCs was 66.67% (8/12), 100% (12/12), 100% (8/8) and 100% (1/1) in patients with stage N0, N1, N2 and N3 lymph node metastasis, respectively, showing statistical differences among all subgroups (P=0.047).

FISH was performed to investigate the blood samples of the 50 patients with GC. A total of five cases with HER2 abnormal expressions were detected in the tissue of patients with GC, five in the CTCs separated by H-LMB and six in the blood samples. The HER2-positive CTCs separated by H-LMB were consistent with that of tissue (HER2 expression in tumor samples and paired CTCs were 71.4%, the analysis of HER2 in CTCs showed higher positivity compared with tumor tissues). The mutation rate of HER2 was 10% in both tissue and CTCs of all the patients, while the positive rate of FISH was 14% in the blood samples. The ratio of HER2 and FISH was calculated by using the alternative control probe for one patient with uncertain HER2 status and a false-positive result was determined (Fig. 6A). Considering pathological diagnosis results as the gold standard, the sensitivity and specificity of CTCs detection were 77.1 and 89.6% in cancer diagnosis, respectively, as shown in Fig. 6B, which were higher than those of traditional serum tumor markers of carcinoembryonic antigen (CEA; 69.9 and 77.1%, respectively) and PIVIKA (59.9 and 67.1%, respectively), showing statistically significant differences. The receiver operating characteristic (ROC) curve is shown in Fig. 6B. The area under the ROC curve of CTCs in the diagnosis of gastric cancer was 0.86 [95% confidence interval (CI); 0.777-0.952], which was higher compared with CEA and PIVIKA. ROC curve analysis showed that the diagnostic value of CTC in gastric cancer was higher compared with traditional serum tumor markers.