Analysis of Released Circulating Tumor Cells During Surgery for Non-Small Cell Lung Cancer

Circulating tumor cells (CTCs) can be detected in the peripheral blood with the FDA-cleared CellSearch system. The system identifies cells as CTCs when they have a morphology resembling a cell with a nucleus identified by DAPI, expression of epithelial cell adhesion molecule (EpCAM) and cytokeratins (CK), but not CD45. CTCs have been shown an independent prognostic marker of survival for patients with non–small cell lung cancer (NSCLC) (1–6). In early stage NSCLC, the presence of CTCs detected in the peripheral blood after surgery has been associated with a shorter time to recurrence (7–11). Of note, in advanced disease, CTCs are only observed in around 30% of patients with NSCLC. Patients without detectable CTCs by CellSearch in 7.5 mL of blood may still have CTCs in the bloodstream; however, these are not detected due to the low blood volume screened or the lack of expression of EpCAM or the CK identified by the C11 and A53B/A2 clones (12).

During surgery, we have the unique opportunity to draw blood from the pulmonary vein (PV) draining the lobe containing the primary tumor. As the PV is closer to the source, we hypothesized that the blood in this vein would contain more CTCs. The number of CTCs in the PV may be influenced by several factors. For example, the type of surgery [video-assisted thoracic surgery (VATS) or open thoracotomy] and their associated differences in vessel dissection (PV or pulmonary artery) and manipulation of the tumor while the surgery is ongoing may influence CTC counts.

In this study, we investigated the release of CTCs in the PV and the radial artery at the start of and during surgery. As a secondary goal, we evaluated differences in CTC counts between VATS (PV ligated before pulmonary artery) and open thoracotomy (vice versa, pulmonary artery ligated first). To evaluate the malignant origin of CTCs, we performed low-coverage single-cell whole-genome sequencing (scWGS) to detect copy-number alterations (CNAs).

We prospectively included consecutive patients who were eligible for surgery due to primary NSCLC or a suspected lung malignancy by fast-growing pulmonary nodules after they gave informed consent. Patients undergoing lobectomy, bilobectomy, or pneumonectomy were eligible for inclusion. An important difference between VATS and open surgery is that during open surgery the pulmonary artery is dissected and closed first, whereas during VATS, the PV is the first large vessel to be dissected. For patients undergoing an open thoracotomy, 7.5 mL of blood was drawn from the radial artery at the start of surgery (baseline, T0), followed by blood draws from both the radial artery and the PV that drained the lobe which contained the tumor at two time points: when the draining PV was identified (T1) and after dissection of the pulmonary artery but before the PV was dissected (T2; Fig. 1).

For VATS patients, blood samples were obtained at the start of surgery from the radial artery (T0). Blood samples were obtained from both the radial artery and the PV directly after identification of the draining PV (T1) and after closure of the PV (T2). No sample was obtained after the pulmonary artery was closed.

At all time points, the diseased lobe was still inside the patient with blood circulation in the tumor. Only at T2 would, depending on surgery type, either the PV or artery be dissected and would influence intravenous pressures.

The study was performed according to the Helsinki declaration. It was reviewed and approved by the local medical ethical committee (METc 2015/602) and registered at the Dutch study register (NL55754.042.15).

Blood samples were processed within 96 hours and analyzed blinded to clinical outcomes, as described previously (1, 2). Cells were CTCs when identified by CellSearch, according to the FDA-approved definition, as cells positive for the EpCAM, CK identified by the clones C11 and A.53B/A2 and nuclear stain DAPI, but negative for leukocyte marker CD45 in the blood. CTC numbers are reported as the number per 7.5 mL of blood. Clusters of CTCs (i.e., ≥2 CTCs clustered together) are counted as 1 CTC/7.5 mL. The presence of clusters was noted.

CellSearch cartridges were stored at 4°C up to 24 months before further processing. When a sample had CTC counts ≥4/7.5 mL, the content of the cartridges was transferred to a 1.5-mL tube and washed twice with 300-μL PBS to ensure removal of the majority of cells from the cartridge. Single CTCs were subsequently isolated by FACS. Predefined gates were used for isolation of DAPI⁺/CK⁺/CD45⁻ CTCs. White blood cells were sorted based on a DAPI⁺/CK⁻/CD45⁺ phenotype and served as controls for single-cell sequencing. Cells were collected in a 1x ProFreeze solution as single cells or in groups of 10 cells and stored at −80°C before copy-number aberration (CNA) analysis.

After the first CNA analysis results, CTCs ≥4/7.5 mL from four patients were isolated by the puncher system of Vycap (Vycap; ref. 13). This allowed direct comparison of the morphology of the CTCs with the single-cell sequencing result. Cells were removed from the CellSearch cartridge in the same manner as for FACS isolation and transferred to a 1.5 mL tube.

Microwell chips (VyCAP) were degassed in a vacuum chamber at −1.0 bar for 15 minutes in PBS. After degassing, the microwells were placed in a filtration holder. The cells were then transferred and seeded into a microwell chip according to manufacturer's instructions. After seeding, the microwell chip was transferred to the VyCAP Puncher system. The entire chip was scanned using a 20x objective. Images were acquired using the following settings: 100-ms DAPI, 200-ms PE, and 600-ms APC. Cells of interest were selected using the Puncher Software (VyCAP, 64 bit, version 5.3) by using the automatic selection tool. All events with a signal intensity of >2,000 DAPI and >1,500 PE were reviewed by the operator, and CTCs were manually selected. Cells of interest were subsequently punched into a 96-well plate containing 95 μL mineral oil (Sigma). After punching, 5 μL of freeze buffer (1XPBS/42.5% freeze buffer/7.5% DMSO; Sigma) was added. Plates were spun down for 5 minutes at 500 g and subsequently stored at −80°C until further processing.

Single cells and mini bulk cells (n = 10 cells) were stored in freeze buffer after isolation, followed by scWGS as previously described with some minor modifications (14). In short, DNA was fragmented with MNase, after which decrosslinking was performed by incubation at 65°C for 1 hour in the presence of proteinase K (0.025 U) and NaCl (200 mmol/L) followed by Ampure XP bead purification. Hereafter, End-repair, A-tailing, adapter ligation, and PCR amplification were performed as described before (14). During PCR, indexes are introduced to each DNA fragment allowing multiplexing of the libraries for sequencing. All libraries were sequenced on Illumina NextSeq 500. Data analysis was performed with the AneuFinder software package (15).

All patients had tissue of the (suspected) primary tumor frozen (−80°C). Of those patients who had successfully isolated CTCs for CNA analysis, tissue sections were sliced (50 μm) and used to isolate tumor nuclei. These were subjected to scWGS in the same manner as the CTCs and control cells in mini bulk (30 nuclei).

Differences in samples with CTC present between open thoracotomy and VATS were evaluated by Fisher exact tests. Paired analyses were performed by McNemar and Wilcoxon tests. Correlations were tested by Spearman's rho, and if significant further investigated with logistic regressions. Differences were considered significant when P < 0.05.

Systematic differences in CTC counts between different time points, PV versus RA, and type of surgery were evaluated using a longitudinal marginal mixed model, with a variable slope and intersect using type of surgery and location of blood draw as covariables with the repeated measurements in an unstructured matrix. Interaction terms were tested for type of surgery, time point, and location of measurement. All analyses were performed using SPSS version 23.

Thirty-one patients were included (18 open thoracotomy and 13 VATS, Table 1). Patients undergoing VATS were significantly more often female, but showed no other significant differences. Three patients presenting with fast-growing pulmonary nodules had benign disease (two in the open and one in the VATS group), and these patients served as a benign control group. Open surgery patients provided 82 and VATS patients 56 blood samples that were successfully processed by CellSearch. For 28 patients, paired PV and RA T1 samples were collected, and for 25 patients, paired samples were collected at T2.

CTCs were detected in 58 of 138 blood samples (42%). The PV had CTCs in 40 of 57 samples (70%) and the RA 18 of 81 blood samples (22%). Combined, we identified CTCs in any of the two PV samples in 25 of 31 patients (81%) and in any of the three RA samples in 13 of 31 patients (42%). Remarkably, all three patients with benign disease had CTCs (as defined by CellSearch) detected in the PV, but not in the radial artery, indicating a misinterpretation of malign character of the epithelial cells.

Using the matched samples at T1 and T2 (28 + 25 = 53), CTCs were more often detected in the PV (70%) than in the RA (22%, P < 0.01), and in higher counts (P < 0.01). This difference was also significant when testing was performed in a nonmatched manner, or when comparing only NSCLC samples (71% vs 25%, P < 0.01). When stratifying by time point, CTCs were present more often in blood samples derived from the PV compared with the RA (T1: 64% vs 23%, P < 0.01; T2: 76% vs 7%, P < 0.01, Table 1, Fig. 2). The type of surgery, comparing video-assisted surgery versus open surgery, was not significantly associated with either the presence of CTCs (RA: OR = 1.5, P = 0.45; PV: OR = 0.63, P = 0.43) or their number (RA: ρ = −0.01, P = 0.91, PV: ρ = 0.17, P = 0.21).

CTCs derived from the PV showed morphologic differences from those detected in the radial artery (Fig. 3A). In general, cells identified in the PV were larger and expressed CK more strongly than those detected in radial artery (Fig. 3B).

Clusters of CTCs were detected in 19 blood samples. Eighteen of these blood samples were taken from the PV. CTC clusters were more often detected when CTC counts were high (ρ = 0.48, P < 0.01; OR = 1.02, P = 0.01). The presence of clusters was not associated with any other factors (type of surgery, location of tumor, tumor size, performance status, and histology).

The proportion of patients with CTC was lower at the end of surgery when blood circulation through the primary tumor was ended [T0: 9/24 (38%), T2: 2/27 (7%), P = 0.02], with matched samples having significantly lower counts (P = 0.03).

There was no difference in the proportion of patients with CTCs before and after clamping the tumor vessels (T1: 64% vs T2: 76%, P = 0.16), nor was there a difference in the number of CTCs measured (P = 0.54).

To evaluate patterns in the change of CTCs influenced by manipulation during surgery (difference in CTCs between time points) and type of surgery, we used a mixed model. Changes in CTCs in the PV were significantly higher than those in the RA (P = 0.01). CTC counts were not associated with any clinical parameter. There was no difference between CTCs measured in VATS patients and those undergoing open surgery (P = 0.48). No systematic change in CTC counts was observed in the pulmonary or radial artery (Fig. 2).

Matched primary tumor samples from all patients with NSCLC showed aneuploidy. As expected, normal tissue taken from the controls only showed euploid cells.

From 12 blood samples (11 from the PV and 1 from the RA) obtained from 10 different patients, containing at least four CTCs, CTCs were successfully extracted from the CellSearch cartridges and isolated by FACS for scWGS (Table 2). In three patients with NSCLC, CTCs were aneuploid. In 2 of these patients, only one circulating cell was successfully extracted and analyzed, which was aneuploid (Fig. 3D). The third patient had 81 CTCs that were extracted successfully from three different samples. Of these extracted CTCs, only two showed aneuploidy, both derived from the same sample. The other extracted cells were euploid. The remaining 184 extracted CTCs from the other for patients (including two patients with fast-growing benign nodules) were all euploid. The structural changes of aneuploid CTCs closely resembled those identified in the primary tumor cells (Fig. 3D). In addition, of two patients, samples from the PV at T1 were analyzed for CNA by FISH after being passed over a microsieve. The first sample had 85 CTCs detected by CellSearch, yet all were euploid for chromosomes 1, 8, and 17. The second sample contained 19 CTCs as detected by CellSearch. In two CTCs, aneuploidy was detected with loss of chromosome 1 and one had a trisomy of chromosome 8, changes that were similar to those in the primary tumor (Fig. 3C). No other aneuploidy cells were found. In total, four of 12 (33%) patients had CTCs with proven genomic aberrations.

Another 125 CTCs from 4 different patients were isolated for genomic analysis by the punching system. We aimed to include both subsets of so-called CTC with different morphology, but were only able to successfully extract two CTCs from the radial artery at T1.The other three samples were from the PV. Two samples were derived from the PV at T1 (42 CTCs and two CTCs, respectively), one from the PV at T2 (79 CTCs). All punched CTCs from the PV showed no structural genomic changes (Fig. 3E), indicating that they are circulating epithelial cells rather than tumor cells.

Of one of these patients, primary tumor tissue was frozen and used for genomic analysis by minibulk (2 × 30 cells). The primary tumor showed structural abnormalities (Fig. 3F).

CTCs enumerated with the CellSearch system were present in higher counts in samples obtained from the PV compared with samples from the RA. Most of them are epithelial cells, and few are CTCs with genomic aberrations. Release of these so-called CTCs did not appear to be influenced by manipulation during the course of surgery. The different surgical approaches and their associated differences in vessel ligation also did not influence CTC counts. It appears that large numbers of CTCs disappeared during their travel from the PV through the heart to the RA. CTC numbers in the RA were low and did not differ substantially from those reported in the peripheral venous system. That means that these CTC are lost in the central compartment (heart and large blood vessels) and not so much in the peripheral microcirculation as has been previously theorized (16).

The majority of the CellSearch-identified and -extracted CTCs did not have any structural genomic changes when analyzed by scWGS. This was surprising because all primary lung cancer samples were characterized by a complex karyotype with many structural aberrations, a typical finding in lung cancer. This might be due to the FACS, a method that is less reliable in isolating rare cells (17, 18). Genomic analysis on a limited number of cells isolated by the more reliable cell puncher method showed however similar findings. Given these technical constraints, we concluded that in the PV during surgery many normal epithelial cells are observed and only limited numbers of CTCs with abnormal genomic patterns.

When examining the images from the CellSearch analyzer, two populations of CTCs could be discerned based on their morphology. The one more closely resembling CTCs observed in the peripheral blood by CellSearch is probably a real CTC. The other population analyzed in this study did not have any structural abnormalities. This implies that these cells, though defined as CTCs according to the CellSearch protocol, are more likely to be nonmalignant epithelial cells. Other methods often isolate CTCs by their larger size, sometimes in combination with their expression of CK and EpCAM (19, 20). Therefore, it is likely that they would also have isolated these cells. However, using different markers than CK or EpCAM, e.g., markers used pathology like p40 or TTF-1, might be able to distinguish malign CTCs more accurately (21).

The exact mechanism by which these nonmalignant epithelial cells enter the bloodstream is as yet unclear and deserves further exploration. Endothelial cells occasionally can express CK and may be released into the bloodstream during surgery influencing measured CTC counts by CellSearch (22). However, endothelial cells should be excluded by the EpCAM-based separation. In addition, high numbers of CTCs were identified even when the blood vessels were intact. Therefore, it seems unlikely that contamination by endothelial cells can explain the high number of identified (euploid) CTCs by itself. Probably, a spectrum of epithelial cells is released into the bloodstream from normal and premalignant to malignant epithelial cells. Their presence could explain the large difference between CTC counts as defined by CellSearch in the PV and the RA. Benign epithelial cells are less able to survive in the circulation because of lower tolerance of shearing forces and the mesenchymal environment, leading to a fast clearance or destruction. Therefore, benign cells are not (often) detected in the peripheral bloodstream. That so-called CTC may be present in the peripheral blood of healthy individuals, even without surgery, has been shown by Allard and colleagues, albeit in very low numbers (23). The large decrease in CTC count between PV and RA cannot be explained by the dilution from the four other PVs as the difference is too large. However, the high numbers in the PV could represent a short-lasting peak that was missed in the RA. Other possible explanations are the destruction of (suspected) CTCs by physical forces from the moment these cells enter the bloodstream. The benign epithelial cells maybe more susceptible to such forces than malignant cells.

That epithelial cells are misidentified as CTCs by CellSearch when measured in the PV during surgery would explain previous findings summarized in Table 3. The CTC presence during or shortly after surgery in the peripheral circulation is strongly associated with disease recurrence and recurrence-free survival (7, 9, 10, 24–28). Suspected CTCs identified in the PV were also shown to be associated with survival, but the association is less well defined (16, 29–31). Studies comparing both measurements like Crosbie and colleagues and Li and colleagues show that despite higher cutoff values, the association of suspected CTCs in the PV with disease recurrence is not stronger than the association of CTCs identified in the peripheral system (16, 31). Possibly, CTCs in the PV are a mix of true CTCs with other epithelial cells released into the bloodstream, explaining the high counts in the PV, and are therefore not representative of tumor burden. A large part of these cells, primarily the benign cells, are unable to survive in the bloodstream, explaining the low CTC counts peripherally, which are associated with survival (32–36).

In line with previous studies, higher CTC counts in the PV were found compared with the peripheral measurements (16, 29, 31, 37–42). Also, significantly lower CTCs counts after surgery are observed when measured in the peripheral circulation (10, 38–41, 43). However, together with Hashimoto and colleagues, we are the only ones that have performed sequential measurements in the PV (42). Hashimoto and colleagues found significantly higher numbers in the second sample taken from the PV, but they took the sample after the lobe had been completely removed from the body, allowing stagnation of the blood. Hashimoto and colleagues also explored whether sequence of vessel ligation influenced CTC release and found no difference. They also found no difference between VATS and open surgery, where the order of vessel ligation differs. It is also in line with the fact that the outcome of both surgical techniques is comparable and release of CTCs is associated with worse outcome (44, 45). Moreover, no significant differences in survival or disease recurrence were observed between patients who had the PV or artery clamped and dissected first (42, 46–48).

CellSearch is originally developed for measuring CTCs in peripheral blood samples expressing both EpCAM-positive and CK but lacking CD45 and with morphology features consistent with a cell containing a nucleus (DAPI). CTCs detected in this way have been associated with survival many times over, making them a well-validated biomarker. It is known that the CTCs identified by CellSearch in the peripheral bloodstream do exhibit a large variation of genomic abnormalities when analyzed by FISH, scWGS, or NGS (49–55). Our findings do not contradict these previous reports, but do indicate that care should be taken when implementing the CellSearch in different blood compartments, like the PV.

The definition of a CTC, currently described as an EpCAM⁺, CK⁺, DAPI⁺, CD45^- cell identified by CellSearch, may need the addition that the measurement should be performed in peripheral blood. Furthermore, manual classification of CTCs is currently being replaced by automated classification using the ACCEPT imaging analysis program, which could further improve the definition of a CTC (56, 57). CTCs from peripheral blood measured with CellSearch remain, in our studies and many other clinical trials, a strong biomarker and especially a sensitive test for low numbers of CTCs (1–6).

In short, a spectrum of epithelial cells is released into the circulation during surgery. The number of CTCs was higher in the PV compared with the radial artery. Release of CTCs was not influenced by the type of surgical approach and the difference in vessel ligation, or by manipulation during surgery. In the peripheral blood, CTCs were less often detected at the end compared with the start of the surgical procedure. Two morphologically distinct circulating epithelial populations were observed: the majority of these cells were euploid epithelial cells released during surgery, and a small minority were CTCs with structural genomic abnormalities. Normal epithelial cells were likely filtered away during their passage in the central blood compartment. We recommend to include genomic tests to verify the malignant character of circulating cells further than by morphology and fluorescence features, especially when using CellSearch in a manner for which it is not validated.

H.J.M. Groen is a paid consultancy for Eli Lilly and is an unpaid consultant/advisory board member for Bristol-Myers Squibb and Takeda. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Tamminga, T.J.N. Hiltermann, P.M. Lansdorp, L.W.M.M. Terstappen, H.J.M. Groen

Development of methodology: M. Tamminga, S. de Wit, H. van den Bos, T.J. Klinkenberg, T.J.N. Hiltermann, D.C.J. Spierings, P.M. Lansdorp, L.W.M.M. Terstappen, H.J.M. Groen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Tamminga, S. de Wit, C. van de Wauwer, H. van den Bos, T.J. Klinkenberg, T.J.N. Hiltermann, K.C. Andree, D.C.J. Spierings, A. van den Berg, H.J.M. Groen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Tamminga, S. de Wit, H. van den Bos, T.J.N. Hiltermann, D.C.J. Spierings, W. Timens, L.W.M.M. Terstappen, H.J.M. Groen

Writing, review, and/or revision of the manuscript: M. Tamminga, S. de Wit, C. van de Wauwer, H. van den Bos, T.J. Klinkenberg, T.J.N. Hiltermann, K.C. Andree, D.C.J. Spierings, P.M. Lansdorp, A. van den Berg, W. Timens, L.W.M.M. Terstappen, H.J.M. Groen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Tamminga, S. de Wit, H. van den Bos, J.F. Swennenhuis, D.C.J. Spierings

Study supervision: T.J.N. Hiltermann, H.J.M. Groen

The authors participate in the Cancer-ID consortium which has received support from the Innovative Medicines Initiative (IMI) Joint Undertaking under grant agreement No. 115749 and part was from Koningin Wilhelmina Fonds (KWF) subsidy No. 10491 from 2016. Its resources are composed of financial contribution from the European Union's Seventh Framework Program (FP7/2007–2013) and EFPIA companies' in-kind contribution. The funding source had no involvement in study design, collection, analysis, or interpretation of the data or in the writing and submission of the report.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.