Circulating tumor cells with epithelial-mesenchymal transition markers as potential biomarkers for the diagnosis of lung cancer

Sha-Sha Jiang, Chun-Guo Mao, Yong-Geng Feng, Bin Jiang, Shao-Lin Tao, Qun-You Tan, Bo Deng

Sha-Sha Jiang, Chun-Guo Mao, Yong-Geng Feng, Bin Jiang, Shao-Lin Tao, Qun-You Tan, Bo Deng, Department of Thoracic Surgery, Institute of Surgery Research, Daping Hospital, Army Medical University, Chongqing 400042, China

Abstract BACKGROUND Circulating tumor cells (CTCs) can be clustered into three subtypes according to epithelial-mesenchymal transition (EMT) markers: CTCs with epithelial markers(E-CTCs), CTCs with mesenchymal markers (M-CTCs), and CTCs with both markers (E&M-CTCs). CTC detection has clinical implications in the diagnosis of lung cancer (LC).AIM To clarify the diagnostic value of CTCs categorized by EMT markers in LC.METHODS The study included 106 patients with lung adenocarcinoma, including 42 groundglass opacities (GGO) and 64 solid lesions, who underwent surgery between July 2015 and December 2019. Eleven patients with benign tumors and seventeen healthy controls were included. CTCs in peripheral blood and associated EMT markers were detected preoperatively using the CanPatrolTM technique. The diagnostic power of CTCs for discriminating LC cases from controls was analyzed by the receiver operating characteristic (ROC) curve. The CytoploRare technique was used in 20 cases and 18 controls for validation, and Kappa values were calculated to evaluate consistency between techniques.RESULTS Of the 106 LC cases, 94 (89.6%) had at least one CTC. CTCs were detectable in 35(83.3%) of 42 GGO cases. Total CTCs and E&M-CTCs were significantly more frequent in LC cases than in benign or healthy controls. The proportion of MCTCs plus E&M-CTCs increased gradually from healthy controls, to benign controls, to LC cases. The area under the ROC curve of total CTCs and E&M-CTCs was > 0.8 and > 10.75, respectively. The combined sensitivity of total-CTCs and E&M-CTCs was 85.85% for LC patients (80.95% for GGO patients) and the specificity was 78.57%. The Kappa value was 0.415, indicating relative consistency between CanPatrolTM and CytoploRare.CONCLUSION CTC detection is valuable for distinguishing LC from controls, and particularly E&M-CTC detection warrants further study.

Key Words: Circulating tumor cells; Epithelial-to-mesenchymal transition; Diagnosis;Lung cancer; Ground-glass opacities; CanPatrol

INTRODUCTION

Circulating tumor cells (CTCs) destroy the integrity of the base membrane, detach from the primary site, and spread through the peripheral blood, and they have the potential to form distant metastases in the appropriate microenvironment[1].

The epithelial-to-mesenchymal transition (EMT), which is characterized by the upregulation of mesenchymal markers (e.g., cytokeratin [CK] 8, 18, and 19 and epithelial cell adhesion molecule [EpCAM]) and downregulation of epithelial markers (e.g.,vimentin and Twist), can lead to the transformation of epithelial cells into mesenchymal cells[2]. This process can also promote tumor progression, metastasis, and chemotherapy resistance[3]. CTCs include a variety of subtypes and are divided into groups according to the expression of EMT markers[4]. Recent studies suggest that restricted EMT can generate hybrid CTCs with both epithelial and mesenchymal markers, as well as enhance their survival and carcinogenic abilities[5]. The EMT process in these CTCs may lead to cancer metastasis and poor prognosis[3,6]. CTCs are used as biomarkers for the diagnosis of lung cancer (LC), pancreatic cancer, and gastric cancer[7-9]. A meta-analysis found that the sensitivity and specificity (95%confidence interval [CI]) of CTCs for the diagnosis of LC is 75% (95%CI: 54-88) and 92% (95%CI: 82-97), respectively[8]. A pilot study from our group found that CTCs can be detected in early-stage lung adenocarcinoma (LUAD)[10].

MATERIALS AND METHODS

Patients

The study protocol was reviewed and approved by the Ethics Committee of Daping Hospital, Army Medical University 2019, No. 183, and all patients who agreed to participate in the study signed an informed consent.

The study was performed between July 2015 and December 2019 and included 106 patients with LUAD. Computed tomography (CT) performed before thoracoscopic surgery indicated that 42 patients had ground glass opacity (GGO) and 64 had solid lesions. A total of 28 controls were recruited, including 11 benign tumor cases (benign controls) and 17 healthy controls who underwent physical and radiographical examination to exclude the possibility of an occult tumor. Postoperative pathological results were reviewed by professional pathologists. Subjects with bronchiectasis, viral hepatitis, or ischemic cardiac or cerebrovascular disease were excluded from the study. Demographic-clinical data of patients and controls are shown in Table 1.

Detection of CTCs and associated EMT markers by the CanPatrolTM technique

At 1 or 2 d prior to surgery, 10 mL peripheral blood was collected from the median cubital vein of patients and controls and transferred into a sample preservation tube containing ammonium chloride dissolution buffer (Surexam Biotechnology,Guangzhou, China). The samples were incubated for 30 min at room temperature.CTCs and associated EMT markers were detected using CanPatrolTMas follows[1,11]. (1)Red blood cells were lysed and removed using red blood cell lysis buffer, and cluster of differentiation-positive (CD45+) leukocytes were depleted from blood samples using magnetic beads; (2) CTCs were enriched using a calibrated 8 μm pore membrane filter; and (3) RNAin situhybridization (ISH) was used to identify CTCs using the branched DNA signal amplification assay and EMT markers such as CK 8, 18, and 19,EpCAM, vimentin, and Twist were detected according to a recently published protocol[12].

Detection of folate receptor-positive CTCs using CytoploRare validation

Peripheral blood (3 mL) was collected before surgery and stored in a vacuum tube containing ethylenediamine tetraacetic acid. Red blood cell lysis buffer was used to lyse erythrocytes, and magnetic beads were used to deplete CD45+ leukocytes. The folate receptor, which is only expressed on tumor cells, and a synthesized oligonucleotide were used to label the CTCs. The conjugate was annealed and extended using a reverse transcriptase primer prior to amplification. The enriched CTCs were stained by immunofluorescence. Cells expressing the folate ligand and cytokeratin,which were stained with the nuclear stain 4′,6-diamidino- 2-phenylindole, were defined as folate receptor positive (FR+) CTCs. These procedures were completed within 1 h[13,14]. The threshold for positivity was an FR+ CTC value ≥ 8.7 FU/3 mL[13,14].

Data analysis

The Student-Newman-Keuls test was used to analyze the variance among groups.Positive rates were compared using theχ2test, and comparisons between two groups were performed using the Kappa Concordance Test. Cutoff values were determined using receiver operating characteristic (ROC) curves. Statistical analyses were performed using SPSS Version 23.0 software for Windows (SPSS, Inc., Chicago, IL,United States). A value ofP< 0.05 (two-sided) was considered statistically significant.Sensitivity = True positive/(true positive + false negative), Specificity = True negative/(true negative + false positive).

RESULTS

Distribution of CTCs in patients and controls

As shown in Figure 1A, epithelial (E) and mesenchymal (M), and E&M markers were detected in CTCs. Therefore, CTCs were divided into three subtypes according to the associated markers,i.e.E-CTCs, M-CTCs and E&M-CTCs.

Of the 106 LC cases, 94 (88.6%) had at least one CTC, whereas CTCs were only detectable in 4 of 11 benign controls and 6 of 17 healthy controls. The detection rate of CTCs was significantly higher in LC cases than in benign or healthy controls(Figure 1B). E-CTCs were detected in 53 (53/106) LC cases, 3 benign controls (3/11),and 5 healthy controls (5/17) (Figure 1B). There was no significant difference in the incidence of E-CTCs among LC cases, benign cases, and healthy controls. The incidence of M-CTCs was significantly higher in LC cases than in healthy controls(52/106vs0/17;P< 0.05), whereas there was no significant difference between LC cases and benign controls (52/106vs3/11). The incidence of E&M-CTCs was significantly higher in LC cases than in benign controls (74/106vs4/11;P< 0.05) and in healthy controls (74/106vs1/17;P< 0.01). CTCs were detectable in 35 (83.3%) of 42 GGO cases including E-CTCs in 20 cases, M-CTCs in 19 cases, and E&M-CTCs in 28 cases. There was no significant correlation between CTC number and pathological and demographic features such as tumor size, age, and gender (data not shown).

I look at it like this, said the old man. Please a customer with one article, and he will come back when he needs another. Even if it is more costly9. He will save up for it, if necessary.

The number of CTCs, M-CTCs, and E&M-CTCs was significantly higher in LC cases than in healthy controls (P< 0.05), whereas there was no difference between LC cases and benign controls (Figure 1C). The proportions of M-CTCs plus E&M-CTCsincreased gradually from healthy controls, to benign controls, to LC cases (LCvsbenign controls:P= 0.405; benign controlsvshealthy controls:P= 0.086; LCvshealthy controls:P< 0.05) (Figure 1D). The proportion of M-CTCs plus E&M-CTCs in GGO cases was similar to that in other LC cases.

Table 1 Demographic and clinical data of cases and controls

Cut-off and diagnostic value of CTCs determined by ROC curve analysis

As shown in Figure 2 and Table 2, the area under the ROC curve (AUC) of total CTCs and E&M-CTCs was > 0.8 and > 0.75, respectively, indicating their potential diagnostic value for discriminating LC cases from controls. Compared to GGOvscontrols, the AUC of total CTCs or E&M-CTCs remained > 0.75.

The most effective cut-off for positivity according to the AUC was determined to be total CTCs ≥ 2/5 mL and E&M-CTCs ≥ 1/5 mL (Table 2). These cut-off values were used to determine the sensitivity of total CTCs and E&M-CTCs for diagnosing LUAD,which showed that their sensitivity was higher than that of E-CTCs and M-CTCs(Table 2). After defining either total CTCs ≥ 2/5 mL or E&M-CTCs ≥ 1/5 mL for positivity, the sensitivity was 85.85% for LC patients and 80.95% for GGO patients,and the specificity was 78.57%.

Validation of CanPatrolTM using CytoploRare

The CanpatrolTMmethod was validated using CytoploRare in 20 LC cases, 7 benign cases, and 11 healthy controls. As shown in Table 3, the Kappa values reached 0.415,indicating that the detection rate of total CTCs and E&M-CTCs was comparable to that of FR+ CTCs (detected by CytoploRare). Although M-CTCs also had aPvalue of <0.05 compared with FR+ CTCs, this index was excluded because of the low Kappa value (data not shown).

DISCUSSION

Detection and measurement of CTCs is a noninvasive liquid biopsy method that can provide reliable information on the diagnostic, prognostic, and/or predictive biomarkers of different types of cancer compared with invasive diagnostic modalities[15]. CTC detection is performed in two steps: enrichment and identification. CTC enrichment is based on distinguishing tumor cell morphology and sorting cell surface markers. For example, CellSearch enriches CTCs by sorting the surface marker EpCAM using magnetic beads. ISET is a method of enriching cells by filtering cells through filter membranes based on cell size. Ficoll separation enriches cells by density gradient centrifugation based on different cell densities[16]. CTC identification is achieved using polymerase chain reaction (PCR), immunofluorescence, and ISH[17].

Table 2 Area under the receiver operating characteristic curve, P value, sensitivity and specificity of circulating tumor cells

Table 3 Validation of CanpatrolTM by CytoploRare in 20 lung cancer cases and 18 controls

Most of the methods currently used in the clinic provide simultaneous enrichment and identification of CTCs. In this study, CanPatrolTMwas used to enrich CTCs through an 8-micron filter membrane, and RNA ISH was used to identify and characterize the enriched cells based on branch DNA signal amplification technology.In addition, the expanded lanes in the CanPatrolTMsystem allow the detection of EMT markers such as EpCAM, CK 8, 18, and 19, vimentin, and Twist. The expression of these markers allows the classification of CTCs into different subtypes, namely, ECTCs, M-CTCs, and E&M-CTCs[12]. A recent study[18]reported that the sensitivity and specificity of total CTCs detected by CanPatrol? technology for the diagnosis of LC is 81.6% and 86.8%, respectively, which is consistent with the present results.

We used a widely used method, CytoploRare, to validate the CanPatrolTMresults.The FR is a glycoprotein that is highly expressed on the cell surface and can be detected on lung and ovarian cancer cells[19]. Samples consisting of 3 mL peripheral blood were used to enrich FR+ CTCs; white blood cells were removed using immunomagnetic beads, and FR, a tumor-specific ligand, as well as a synthesized oligonucleotide were labeled on the cells. Quantitative PCR was used to analyze the conjugates. FR CTCs were defined as cells expressing folate ligands and cytokeratin with positive nuclear staining[13]. The sensitivity and specificity of this technology for the detection of FR+ CTCs is 81.94% and 73.08%, respectively[20]. Kappa values indicated a relatively similar detection efficacy for CanPatrolTMand CytoploRare,suggesting acceptable consistency. However, further studies with a larger sample size are necessary to confirm these results.

Another advantage of CanPatrolTMis that it can identify CTCs based on mesenchymal markers, whereas current CTCs monitoring methods, including CellSearch and MACS, are based on the detection of EpCAM, and thereby do not detect M-CTCs or E&M-CTCs. CTCs undergoing EMT should receive increased attention because they are characterized by higher aggressiveness and carcinogenicity[21,22]. Tumor cells are more likely to form E&M-CTCs than M-CTCs to obtain both mesenchymal and epithelial cell characteristics[5].

TelomeScan F35, which uses a telomerase-specific replication-selective adenovirus(OBP-1101) to detect CTCs using EMT markers, was used to show that CTCs with EMT-markers are effective for the early diagnosis of non-small cell lung cancer [mean sensitivity: 69.1% (59.6%, 40.0%, 85.7%, and 75.0% for stage I, II, III, and IV,respectively)][23].

Figure 1 Distribution of circulating tumor cells in lung cancer patients and controls. A: Circulating tumor cells (CTCs) with epithelial markers, CTCs with mesenchymal markers (M-CTCs), and CTCs with both marker (E&M-CTCs) can be identified in one CTC; B: Positive rate of CTCs in lung cancer cases was significantly higher than those in benign or healthy controls; C: The numbers of CTCs, M-CTCs, and E&M-CTCs in lung cancer cases were significantly higher than those in healthy controls (P < 0.05); however, there was no disparity between lung cancer and benign controls; D: The proportions of M-CTCs plus E&M-CTCs increased gradually in healthy controls, benign controls, and lung cancer cases. aP < 0.05; bP < 0.01; cP < 0.001. GGO: Ground-glass opacities.

Figure 2 Receiver operating characteristic curve of all lung cancer patients and ground-glass opacities patients. A: Area under the receiver operating characteristic curve (AUC) of total circulating tumor cells (CTCs) and CTCs with both markers (E&M-CTCs) for lung cancer diagnosis were 0.828 and 0.7648, respectively; B: AUCs of total CTCs and E&M-CTCs for ground-glass opacity (GGO) diagnosis were 0.7963 and 0.7504, respectively.

In this study, we found that E&M-CTCs had a greater discriminative power for the diagnosis of LC than E-CTCs or M-CTCs. In addition, we showed that CTCs can be present in benign and healthy controls. However, E&M-CTCs were detected in one case in the control group that corresponded to a smoker. Similarly, CTCs can be recognized as a “sentinel” in cases associated with inflammatory diseases such as chronic obstructive pulmonary disease[24]. In patients with benign colon diseases, CTCs can also be detected with both the CellSearch system (11.3%) and the CK19-EPISPOT assay (18.9%), because inflammatory epithelial cells from these benign lesions may enter the peripheral blood[25]. In 232 patients with benign breast diseases, the CTC detection rate reached 15.95%[26]. Herein, we hypothesized that tumor cells or debris,which can be shed from the benign tumor into the peripheral blood, are captured and recognized as CTCs by CanPatrolTM. However, we cannot completely rule out that these CTCs in the controls were not malignant because CTCs may appear at a very early stage of cancer development, even if the lesion is undetectable by systemic physical and radiological examinations. Whether these “benign” tumors or diseases were in a premalignant stage requires long-term follow-up.

In this study, the proportions of M-CTCs and E&M-CTCs were highest in LC patients compared with benign and healthy controls, supporting the critical role of the EMT process in CTCs in promoting tumor invasion, migration, and metastatic growth.

Detection of E&M-CTCs is effective for distinguishing LC, including GGO and solid lesions, from controls, thus providing a new method for discriminating between malignant and benign pulmonary nodes in clinical practice. Because the study only included two benign cases presented as GGO, we were unable to determine the diagnostic value of E&M-CTCs for discriminating between malignant and benign GGO. Intriguingly, fibrosis can be found in either benign or malignant lung GGO[27,28].However, very few published studies have focused on CTCs and GGO, not to mention CTCs and fibrosis. Furthermore, pathological reports from our hospital did not clearly describe the fibrosis condition of these GGO cases. Therefore, we are going to design further clinical trials regarding CTCs in GGO, which will consider the fibrosis condition and the expressions of relevant biomarkers,e.g., CXCL4 and LECT2[29,30].Additional studies with a large sample size and a multicenter design are warranted in the near future.

CONCLUSION

In conclusion, after defining total CTCs ≥ 2/5 mL or E&M-CTCs ≥ 1/5 mL as the cutoff for positivity, the sensitivity and specificity for LC diagnosis were 85.85% and 78.57%, respectively. CTC detection is valuable for distinguishing LC from controls.

ARTICLE HIGHLIGHTS

Research background

Circulating tumor cells (CTCs) can be clustered into three subtypes according to surface biomarkers. CTC detection has clinical implications in the diagnosis of lung cancer (LC).

Research motivation

CTCs categorized by epithelial-mesenchymal transition (EMT) markers have diagnostic value in LC and ground-glass opacities (GGO) patients.

Research objectives

To clarify the diagnostic value of CTCs categorized by the expression of EMT markers in LC and GGO patients

Research methods

Of the 106 patients with lung adenocarcinoma comprising the study cohort, 42 had GGO and 64 had solid lesions; 11 patients with benign tumors and 17 healthy controls were included as controls. Total CTCs and CTCs with both markers (E&M-CTCs) of these patients were detected, the diagnostic value of CTCs categorized by the expression of EMT markers. CytoploRare technique was used in 20 cases and 18 controls for validation.

Research results

Total CTCs and E&M-CTCs were significantly more frequent in LC cases than in benign or healthy controls. The area under the receiver operating characteristic curve of total CTCs and E&M-CTCs was > 0.8 and > 0.75, respectively. The combined sensitivity of total-CTCs and E&M-CTCs was 85.85% for LC patients (80.95% for GGO patients) and the specificity was 78.57%.

Research conclusions

CTC detection is valuable for distinguishing LC as well as GGO patients from controls.

Research perspectives

Epithelial (E)-, mesenchymal (M)-, and E&M markers can be identified in CTC.Therefore, CTCs were clustered into three subtypes, as per the markers on CTC, i.e. ECTC,M-CTC and E&M-CTC. Detection of E&M-CTC has good diagnostic value in distinguishing lung cancer, providing a new method to discriminate between malignant and benign pulmonary node in clinical practice.

ACKNOWLEDGEMENTS

We thank Dr. Xu-Rui Jin for reviewing the statistical methods of this study.