Combination of CT and telomerase+ circulating tumor cells improves diagnosis of small pulmonary nodules

BACKGROUND Early diagnosis and treatment are key to the long-term survival of lung cancer patients. Although CT has significantly contributed to the early diagnosis of lung cancer, there are still consequences of excessive or delayed treatment. By improving the sensitivity and specificity of circulating tumor cell (CTC) detection, a solution was proposed for differentiating benign from malignant pulmonary nodules. METHODS In this study, we used telomerase reverse transcriptase–based (TERT-based) CTC detection (TBCD) to distinguish benign from malignant pulmonary nodules < 2 cm and compared this method with the pathological diagnosis as the gold standard. FlowSight and FISH were used to confirm the CTCs detected by TBCD. RESULTS Our results suggest that CTCs based on TBCD can be used as independent biomarkers to distinguish benign from malignant nodules and are significantly superior to serum tumor markers. When the detection threshold was 1, the detection sensitivity and specificity of CTC diagnosis were 0.854 and 0.839, respectively. For pulmonary nodules ≤ 1 cm and 1–2 cm, the sensitivity and specificity of CTCs were both higher than 77%. Additionally, the diagnostic ability of CTC-assisted CT was compared by CT detection. The results show that CT combined with CTCs could significantly improve the differentiation ability of benign and malignant nodules in lung nodules < 2 cm and that the sensitivity and specificity could reach 0.899 and 0.839, respectively. CONCLUSION TBCD can effectively diagnose pulmonary nodules and be used as an effective auxiliary diagnostic scheme for CT diagnosis. FUNDING National Key Research and Development Project grant nos. 2019YFC1315700 and 2017YFC1308702, CAMS Initiative for Innovative Medicine grant no. 2017-I2M-1-005, and National Natural Science Foundation of China grant no. 81472013.


Introduction
Lung cancer is the most common cancer worldwide in both incidence and mortality. It seriously affects public health and increases the socioeconomic burden. Early diagnosis and early treatment are key to the long-term survival of patients with lung cancer (1). The 5-year survival rate of patients with early lung cancer after surgical resection can reach 90% (2). However, early lung cancer mainly manifests as solitary pulmonary nodules (SPNs) in the lungs, which are mainly detected based on screening because they have no symptoms. The National Lung Screening Trial (NLST) shows that low-dose helical computed tomography (LDCT) provides greater sensitivity for the early detection of lung cancer and reduces the risk of lung cancer mortality by 20% compared with chest radiography (3). Although 24% of participants were positive after LDCT screening, the benign proportion was as high as 96.4% after follow-up (3,4).
At present, the clinical diagnosis of benign and malignant pulmonary nodules is mainly based on their size, density, morphology, composition ratio, and signs on computed tomography (CT).
However, due to the diversity of pulmonary nodules, it is often difficult to distinguish benign and malignant diseases on CT. Positron emission tomography (PET)-CT and CT-guided percutaneous biopsy are the methods used to further differentiate small lung nodules that are difficult to diagnose. Alternatively, the growth changes of these nodules can be followed up, and time is considered to be the best diagnostic approach. Although these treatment strategies follow guidelines (5)(6)(7), long waiting times cause anxiety in patients, with the attendant consequences of excessive or delayed treatment. Studies have shown that the number of benign pulmonary nodules after surgical treatment can be as high as 20% (8). Therefore, it is very important to accurately differentiate benign from malignant pulmonary nodules.
Liquid biopsy is a novel and noninvasive method. Compared with traditional invasive biopsy, liquid biopsy has the advantages of better compliance, easy specimen acquisition, repeatability, and comprehensive tumour information. Therefore, it is expected to be an ideal technology for early auxiliary diagnosis, concomitant diagnosis, therapeutic monitoring and prognostic assessment of cancer (9). The liquid part of liquid biopsy includes circulating tumour DNA (ctDNA), circulating tumour cells (CTCs), and exosomes. In the last 10 years, some studies have used CellSearch, folic acid receptor and other detection methods based on CTCs to identify benign and malignant pulmonary nodules (10,11), and some achievements have been made. However, there are still deficiencies in the sensitivity and specificity for the identification of smaller pulmonary nodules. Improving the sensitivity and specificity of CTC detection is also key to the identification of benign and malignant pulmonary nodules. Owing to CTC heterogeneity, the separation methods based on CTC surface biomarkers have some limitations (12). Therefore, other methods are needed to improve the detection ability of CTCs.
In a previous study (13), we have reported a telomerase reverse transcriptase (TERT)-based CTC detection method (TBCD). TERT is the basis for tumours to maintain their limitless replication potential, and upregulation of TERT activity has been detected in 80-90% of malignancies compared with most normal cells, which lack telomerase activity (14,15). TERT has also emerged as a potential diagnostic marker for tumours (16). The pathological type of most malignant lung nodules is adenocarcinoma, in which there is significant activity related to high TERT expression. Accordingly, there is intense interest and intuitive appeal in the discrimination of benign from malignant small pulmonary nodules with TBCD.
In this study, we used TBCD to detect CTCs in the peripheral blood (PB) of patients with pulmonary nodules within 2 cm in size, combined with tumour markers to assist comprehensive CT judgement, which improved the benign and malignant differential diagnosis of pulmonary nodules and helped patients benefit from diagnosis and treatment.

Patient characteristics
From May 2017 to March 2019, consecutive patients were enrolled in the study. All enrolled patients had been diagnosed by chest CT, were highly suspected of having malignant pulmonary lesions and were prepared for surgery. The details of the enrolled patients are listed in Table 1. Of the 120 highly suspected lung cancer patients enrolled, 89 were pathologically diagnosed with primary lung cancer, and 31 were diagnosed with benign nodules. The majority of patients were diagnosed with p-stage I disease (79/89), 2 patients had p-stage II disease, and 8 patients had p-stage III disease. A flowchart of the diagnoses of the patients enrolled in the study is shown in Fig 1.

TBCD effectively detected CTCs in pulmonary nodule patients
In this study, we used a live CTC detection approach based on telomerase reverse transcriptase (TBCD). First, we used the online platform GEPIA2 (17) to analyse TERT expression in lung carcinoma and normal samples from The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx) project. The results showed that TERT expression was different in lung adenocarcinoma (LUAD) samples, lung squamous cell carcinoma (LUSC) samples, and normal samples, and the expression level in LUAD or LUSC samples was higher than that in normal samples (Fig 2A). In addition, analysis of TERT at different stages of LUAD and LUSC showed that the TERT expression level was independent of stage and was higher at all stages ( Fig 2B and C). These results suggest that TBCD is feasible for detecting CTCs in patients with lung carcinoma.
Subsequently, we performed a statistical analysis of the CTC detection results of 120 patients with pulmonary nodules. The results showed that the number of CTCs in patients with malignant nodules was significantly different from that in patients with benign nodules ( Fig   3A). The average number of CTCs in patients with malignant nodules was 6.05±0.85 cells/4 ml PB, and the average number of CTCs in patients with benign nodules was 1.29±0.43 cells/4 ml PB (p<0.0001) ( Table 2). The details of CTCs detection of enrolled patients are listed in Supplementary Table 1. Next, we analysed the relationship between pulmonary nodule size and the number of CTCs. We divided the patients into the ≦1 cm nodule group and the 1-2 cm nodule group according to the preoperative CT detection results of the size of the pulmonary nodules. The results showed that there was no statistically significant difference in the number of CTCs between patients with ≦1-cm nodules (6.27±1.80 cells/4 ml PB for malignant and 1.11±0.39 cells/4 ml PB for benign) and patients with 1-2 cm nodules (5.97±0.96 cells/4 ml PB for malignant and 1.36±0.58 cells/4 ml PB for benign), regardless of whether the patients had malignant or benign pulmonary nodules (p=0.88 for malignant and p=0.72 for benign) ( Fig   3B). We divided the patients into the ≦1 cm nodule group (31/120) and the 1-2 cm nodule group (89/120) according to the size of the pulmonary nodules. In the above two groups, there were significant differences in the number of CTC tests in patients with benign and malignant pulmonary nodules (p=0.01 and p<0.0001, respectively) ( Fig 3C). As shown in Fig 3D, receiver operating characteristic (ROC) analysis showed that the area under the curve (AUC) was 0.843 (95% CI= 0.759-0.927). When the threshold value was 1, the sensitivity and specificity of detection were 0.854 and 0.839, respectively ( Fig 3E). Moreover, in the analysis of groups by nodule size, TBCD had excellent detection efficiency (Fig 3F and H). The AUCs were 0.798 (95% CI= 0.644-0.952) and 0.858 (95% CI= 0.753-0.963) in the ≦1 cm nodule group and the 1-2 cm nodule group, respectively. When the threshold value was 1, the sensitivity and specificity of the ≦1 cm nodule group were 0.773 and 0.778 and those of the 1-2 cm nodule group were 0.881 and 0.864, respectively (Fig 3G and I). The details are listed in Supplementary Table 2. The results showed that CTC detection could be defined as a positive result when the number of CTCs was greater than or equal to 2. Based on this threshold, this method had a high differentiating efficiency for benign and malignant pulmonary nodules.

FlowSight and FISH confirmed the CTCs detected by TBCD
To verify that the CTCs were derived from the subsection junction, FlowSight imaging and fluorescence in situ hybridization (FISH) experiments were performed. The detection method of CTCs was modified, EpCAM antibodies were added in the detection stage, and finally, the fluorescence tracer of CTCs was displayed through the FlowSight system. As shown in Fig 4A, CTCs were slightly larger than white blood cells (WBCs), and GFP expression under control of the hTERT promoter was observed with or without EpCAM expression (CD45 -/GFP + /EpCAM + or CD45 -/GFP + /EpCAM -). Additionally, WBCs only expressed CD45 (CD45 + /GFP -/EpCAM -). We found that there was a heterogeneity in CTCs among different patients or in the same patient. In patient 1, CD45 -/GFP + /EpCAM + and CD45 -/GFP + /EpCAM -CTCs could be found at the same time. However, in patient 2, we observed the presence of only CD45 -/GFP + /EpCAM + CTCs. We used FISH to detect PTEN deletion or EGFR amplification in patients with CTCs. In patients with positive CTC test results, we found a typical signal configuration with heterozygous deletion of the PTEN gene ( Fig 4B). The detected EGFR amplification was also confirmed by FISH ( Fig 4B). These results suggest that the CTCs detected were derived from malignant nodules.

TBCD CTCs were independent of serum tumour markers in malignant nodule patients
Subsequently, we analysed the correlation between the number of CTCs and serum tumour markers (CEA, NSE, pro-GRP, and CYFRA21-1) in malignant nodule patients. Patients with tumour marker recordings 5 times higher than the detection threshold (all of these patients were positive for CTCs) were excluded because severely deviated sample information would lead to TBCD was better than all other serum models and was significantly different (p<0.0001, p<0.0001, p<0.0001, and p=0.0001, respectively) (Supplementary Table 4). The results also showed that the number of CTCs detected by TBCD was independent of the diameter of pulmonary nodules. These results indicate that CTCs are a useful marker of pulmonary nodule malignancy and that TBCD can effectively discriminate benign and malignant pulmonary nodules.

TBCD could effectively assist CT in pulmonary nodule diagnosis
Although the development of CT technology has made the detection of pulmonary nodules easier, approximately 20% of the pulmonary nodules resected in clinical surgery are still benign lesions, and the qualitative diagnosis of pulmonary nodules is still very difficult. As shown in was 85.4% for malignant nodules and 16.1% for benign nodules ( Supplementary Fig 1).
Therefore, we further compared the correlation between TBCD and CT. According to the CT results, we divided the 120 enrolled patients into the solid nodule group (64/120) and the subsolid nodule group (56/120). The results showed that there was no significant difference in the average number of CTCs detected in patients with solid nodules and in those with subsolid nodules (6.325±1.485 vs 5.816±0.955, p=0.77) in the malignant nodule group ( Fig 6B, Table   2). In the benign nodule group, the average number of CTCs detected was also not significantly different (1.125±0.505 vs 1.875±0.769, p=0.44) ( Fig 6B, Table 2). By comparing malignant and benign nodules, the average number of CTCs detected was significantly different in both the solid nodule and subsolid nodule groups (p=0.0018 and p=0.003, respectively) (Fig 6C, Table 2). ROC analysis showed that the AUC of the solid nodule group was 0.868 (95% CI=0.769-0.967), while the AUC of the subsolid nodule group was 0.773 (95% CI=0.599-0.946), indicating that TBCD was effective in differentiating benign from malignant nodules in different groups (Fig 6D and E). To predict whether a pulmonary nodule is malignant, chest CT features including size, density, growth, and specific morphology features can be used. We obtained parameters for imaging (including spicular sign, lobulation, pleural indentation, vacuole sign, aerial bronchogram, vessel convergence, mean CT value, and node diameter), and logistic regression was used for model fitting. Then, ROC analysis was conducted for the independent CT evaluation and CT combined with CTC evaluation of the enrolled patients.
The results showed that the AUC of CT alone was 0.830 (95% CI=0.741-0.919), and the AUC of CT combined with CTCs was 0.918 (95% CI=0.858-0.977), suggesting that CT combined with CTCs was a better approach for the diagnosis of pulmonary nodules (p=0.040) (Fig 6F,   Table 3). Subsequently, in the solid nodule group and the subsolid nodule group, the AUCs of CT alone were 0.858 (95% CI=0.770-0.947) and 0.653 (95% CI=0.410-0.896), and the AUCs of CT combined with CTCs were 0.923 (95% CI=0.857-0.989) and 0.786 (95% CI=0.580-0.991), respectively, with no significant difference between the two methods (p=0.16 and p=0.37) (Fig 6G and H). Finally, decision curve analysis (DCA) was performed to compare the clinical effects of CT alone and CT combined with CTCs ( Fig 6I). Interestingly, although CT evaluation showed that patients could benefit clinically when the threshold was 0.4, CT combined with CTC evaluation provided clinical benefit to patients at a lower threshold. In addition, within the threshold range of more than 0.4, the clinical benefit of patients evaluated by CT combined with CTCs was significantly higher than that evaluated by CT alone. These results show that TBCD is an effective method to assist CT in the determination of benign and malignant nodules in the lung and can significantly improve the accuracy of CT in the determination of benign and malignant nodules.

Discussion
In this study, we used the TBCD assay to distinguish benign from malignant nodules in patients with nodules less than 2 cm in size and compared this method with the pathological diagnosis as the gold standard. Our results showed that the number of CTCs detected based on TBCD was not correlated with serum tumour markers. Thus, TBCD could be used as an independent biological indicator for the determination of benign and malignant nodules and showed a significantly better differentiation ability than serum tumour markers. TBCD improved the detection capability in both the ≦1-cm and 1-2-cm pulmonary nodule groups based on the optimal threshold of 1 recommended by the ROC curve (patients with 2 or more CTCs were considered positive). In addition, logistic regression analysis with multiple CT test indexes showed that the CTC test as an auxiliary CT test could significantly improve the ability to distinguish benign from malignant nodules in lung nodules smaller than 2 cm.
Lung cancer is a malignant tumour with the highest morbidity and mortality in the world, and the most effective control methods are early screening, early diagnosis and early treatment. As the initial stage of pulmonary disease progression, pulmonary nodules may be early lesions of lung cancer. The correct diagnosis of pulmonary nodules and the determination of benign and malignant nodules are of great significance for the corresponding clinical treatment (18).
With the emergence of liquid biopsy techniques represented by CTCs and ctDNA, an increasing number of studies have focused on the relationship between CTCs and the prediction and prognostic assessment of lung cancer (10,11,19,20). At present, the methods of CTC capture and enrichment are basically divided into two categories (21). The first is based on physical methods. Such an approach can be used to screen CTCs by differences in cell size or density. The other is the capture of CTCs by immune binding based on cell surface markers.
Most of the studies reported so far have adopted the combination of the above two approaches, such as positive and negative screening schemes through cell size combined with surface markers. In addition, further identification methods, such as RT-PCR and immunofluorescence labelling, are often used after the capture of enriched CTCs (22,23). These different approaches have their own advantages and disadvantages. The enrichment and separation methods based on physical characteristics tend to be biased, resulting in a decrease in detection sensitivity and specificity. However, the detection method based on surface markers is often highly sensitive but produces false positive signals. The root cause is that there is no recognized accurate tumour surface marker at this stage. A CellSearch-based primary lung cancer diagnostic study reported that although CTC counts were significantly different in lung cancer patients and nonmalignant patients, ROC analysis showed that the sensitivity and specificity of the test were lower than those of the serum marker CEA (30.4 vs 45.6, 88.0 vs 92.0) (11). CellSearch-based CTCs are more suitable as biomarkers for predicting primary metastasis than for early identification. A recent study has also shown that the enrichment and separation of CTCs using the isolation by size of epithelial tumour cell technique was not suitable for lung cancer screening (24). Some studies on the diagnosis of lung cancer based on folic acid receptors have shown that the method has a certain determination efficacy in detection, with a sensitivity between 72.0 and 76.4 and a specificity between 73.8 and 84.1 (10,25,26). However, its detection ability is often related to tumour size and stage, and the detection efficiency in stage I patients is lower than that in other stage patients (sensitivity is 67.  (29). Since ctDNA and exosomes exist in plasma, the combined application of CTCs with ctDNA or exosomes to the same sample may be the way to further improve diagnostic accuracy between benign and malignant pulmonary nodules.
In our study, a CTC enrichment and separation approach based on TERT was adopted. The approach was achieved by using TERT activity, which maintains the unlimited replication potential of tumour cells, as a positive screening marker combined with the negative screening of CD45 antibody markers. The results showed that in LUAD, the expression level of TERT in tumour samples was significantly different from that in normal samples, and the expression level of TERT in tumour samples was not correlated with stage. These results suggest that TBCD is feasible for differentiating malignant and benign lung cancer patients and that the ability to differentiate between patients at different stages should be uniform. The detection results of the 120 enrolled patients with pulmonary nodules less than 2 cm in size showed that with a threshold of 1, the overall detection AUC was 84.3, and the detection sensitivity and specificity reached 85.4 and 83.9, respectively. The enrolled patients were further grouped. In the 1-2 cm nodule group, the detection AUC of this method was 85.8, and the sensitivity and specificity were 88.0 and 86.4, respectively. In the less than 1 cm nodule group, the AUC was 79.8, and the sensitivity and specificity were 77.3 and 77.8, respectively. The results showed that TBCD exhibited robust efficacy in the screening of benign and malignant pulmonary nodules in diagnostic experiments using pathology as the gold standard.
For some patients with pulmonary nodules, conventional diagnostic methods such as fineneedle aspiration and transbronchial biopsy are often only able to obtain small tumour specimens, as it is difficult to obtain samples with these methods (30). Therefore, in PB, the assessment of tumour markers is of clinical value for the determination of benign and malignant tumours (31). Because of the low concentration of clinically common tumour markers in the PB of patients with pulmonary small nodules, it is difficult to use them as independent biomarkers for diagnosis (31). Some studies have shown that the combination of multiple tumour markers in serum can distinguish cancer patients from healthy people (32,33).
Although research on tumour markers for lung cancer has made great progress, there is still no breakthrough in tumour markers with high specificity and sensitivity, especially for early-or precancerous-stage lung cancer, and no suitable tumour markers can be used for diagnosis and application. In our study, we performed ROC analysis of serum tumour markers (CEA, NSE, pro-GRP and CYFRA21-1) in the 120 enrolled patients with pulmonary nodules. The results showed that there was little difference in the individual detection efficacy between these markers, and the AUC values were all less than 0.6, suggesting that they were not suitable as biomarkers for the diagnosis of lung cancer. In some studies, the CEA expression level was shown to have certain application value in the determination of pulmonary nodule benignity and malignancy (11,33,34), while in this study, it was not shown to have application value.
This may be because serum CEA expression levels in patients with pulmonary nodules less than 2 cm tend to be within a normal detection threshold range. Compared with serum markers, TBCD performed significantly better than the above four serum tumour markers (all p < 0.001).
With a threshold of 1, the sensitivity and specificity reached 0.854 and 0.839, respectively. In addition, Pearson's correlation coefficient showed that CTCs had no correlation with the above serum markers and could be used as an independent biomarker to determine benign and malignant lung cancer tumours. Compared with serum tumour markers, TBCD alone or in combination with other markers may be a more effective technical approach for differentiating benign from malignant pulmonary nodules.
At present, early-stage lung cancers are mainly diagnosed as isolated lung nodules by chest CT and are divided into solid and subsolid nodules based on their respective differentiation components, the latter containing pure ground-glass nodules (pGGNs) and part-solid nodules (PSNs). Previous studies have shown that chest CT plays an important role in the differential diagnosis of benign and malignant pulmonary nodules with specific morphological features or CT values (radiology, etc.), and even some CT characteristics, such as components, pathological subtypes, gene mutations, and prognostic correlation (JTCVS) (35,36). With the popularization and application of LDCT in early lung cancer screening, a large number of pulmonary nodules have been found, and the main problem is overdiagnosis and overtreatment, with studies showing false positive rates of lung cancer exceeding 18.5% (37). The reason was that the CT radiologic features of benign and malignant pulmonary nodules overlapped, making them difficult to distinguish (Fig. 6A). In theory, CTC detection by a noninvasive liquid biopsy approach can avoid the interference of the complex signs of CT and serve as an adjunctive approach to help CT differentiate and diagnose pulmonary nodules. In our study, we not only evaluated the efficacy of CT alone in the enrolled patients but also conducted a retrospective evaluation of the combination of TBCD with CT. By regression fitting of imaging parameters (including spicular sign, lobulation, pleural indentation, vacuole sign, aerial bronchogram, vessel convergence, mean CT value and tumour diameter), the sensitivity and specificity of CT as a single diagnostic method in this study reached 83.1 and 80.6 in all patients, respectively. The patient population was divided into a solid nodule group and a subsolid nodule group, and CT alone showed high specificity and low sensitivity (67.5 for sensitivity and 95.8 for specificity) in the solid nodule group or high sensitivity and low specificity (87.8 for sensitivity and 42.6 for specificity) in the subsolid nodule group. The diagnostic efficiency of TBCD combined with CT in the diagnosis of pulmonary nodules was better than that of CT alone (p=0.039), and the sensitivity and specificity improved to 89.9 and 83.9, respectively.
The diagnosis in different subgroups showed that TBCD combined with CT could achieve a more balanced detection sensitivity and specificity. Therefore, these results indicate that TBCD can be an excellent auxiliary technique to improve the accuracy of evaluating the nature of pulmonary nodules by CT, as well as to reduce the anxiety, further costly evaluation, cumulative radiation hazard, and pain caused by invasive examination.
To more comprehensively evaluate the diagnostic value of a test, it is necessary to consider all possible diagnostic thresholds. ROC curve analysis is widely used in the performance evaluation of medical diagnostic tests (38). When the AUC is 0.7-0.9, the diagnostic accuracy is moderate, and when the AUC is above 0.9, the diagnostic accuracy is high. DCA is also used in studies to evaluate the diagnostic value of diagnostic tests (39). To ensure the repeatability and accuracy of TBCD, we also conducted test validation using three batches. The simulated samples were tested for accuracy, precision and specificity. The linear regression coefficients of different batches were all higher than 0.99 (data not shown). The coefficient of variation between test batches and batch precision ranged from 5% to 8% (data not shown). These results demonstrate that TCBD is an accurate and reproducible CTC assay. This study has the following shortcomings and improvements. 1. As an exploratory study, the sample size of enrolled patients was small. It is necessary to further expand the sample size and carry out multi-centre verification. 2. The patients enrolled in this study were highly suspected of having lung cancer by CT diagnosis, and TBCD can also be further verified in LDCT to further explore the ability of TBCD combined with LDCT in lung cancer screening.
3. With the development of artificial intelligence (AI) diagnostic technology, the combination of AI diagnostic technology with CT and TBCD may achieve a better diagnostic performance.
In conclusion, TBCD can improve the diagnosis of pulmonary nodules and can be used as a robust auxiliary diagnostic scheme for CT diagnosis.

Patients and sample collection
A total of 120 patients newly diagnosed by CT were recruited, and the clinical information of the included patients was provided by the Department of Thoracic Surgery, Beijing Chest Hospital. This study was approved by the ethics committee of the Beijing Chest Hospital, Capital Medical University (KY-2018-004). Written informed consent was obtained from the enrolled patients. Blood (4 ml) was collected from eligible patients with K2E (EDTA) tubes, kept at 4°C, and transported to the laboratory within 2 hours.

Blood sampling treatment and CTC identification
The 4-ml blood samples were centrifuged at 500 ×g for 5 min, and the plasma was discarded.
Then, red blood cell lysis buffer (NH4Cl, 0.15 M; EDTA, 0.1 mM; KHCO3, 10 mM; pH = 7.2) was added to the samples. The samples were centrifuged at 500 ×g for 5 min, the supernatant was discarded, and 5 ml of 1×PBS was added to wash the cells. The samples were centrifuged again at 500 ×g for 5 min, and 2 ml of serum-free medium was used to resuspend and seed all cells from a 4-mL blood sample. CTCs were detected using reagent (oHSV1-hTERTp-GFP) as previously described (13). Cells transduced with oHSV1-hTERTp-GFP were incubated in a humidified atmosphere of 5% CO2 at 37°C for 24 hours. The transduced cells were harvested and stained with an APC-CD45 antibody (HI30, Invitrogen). CD45 − /GFP + cells were recorded as TBCD-CTCs for TBCD (Supplemental Figure 2).

Identification of CTCs using ImageStreamX ® and FISH
The samples were subjected to a standard CTC identification process.

Statistical analysis
Statistical analysis of the data was carried out with standard software (IBM SPSS Statistics 26.0 and Prism 8, USA). The nonparametric Mann-Whitney U-test was used to test two patient groups (categorical and continuous data). Pearson's chi-squared test was used to test the expected frequencies and the observed frequencies in two categories of a contingency table.
ROC curves were constructed based on the diagnostic efficiency of tumour biomarkers and CTCs, and the AUC represented the diagnostic performance. Logistic regression was used to calculate the predictive probability for the combined methods of CT and CTCs. Clinical usefulness was evaluated by DCA. Data are expressed as the mean ± SEM. All P values were two-sided, with P < 0.05 considered statistically significant.

Study approval
This study was approved by the Beijing Chest Hospital, Capital Medical University, depending on where subjects were recruited and research was carried out. This included the receipt of written informed consent from participants.