Françoise Dignat-George, UMR-S 608 INSERM-Université de la Méditerranée, Laboratoire d’Hématologie et d’Immunologie, UFR de Pharmacie, 27 Boulevard Jean Moulin, 13385 Marseille Cedex 5, France. Tel.: +33 1 4 91 83 56 00; fax: +33 1 4 91 83 56 02. E-mail: firstname.lastname@example.org
Summary. Background: Circulating endothelial cells (CECs) have emerged as non-invasive biomarkers of vascular dysfunction. The most widely used method for their detection is CD146-based immunomagnetic separation (IMS). Although this approach has provided consensus values in both normal and pathologic situations, it remains tedious and requires a trained operator. Objectives: Our objective was to evaluate a new hybrid assay for CEC measurement using a combination of pre-enrichment of CD146+ circulating cells and multiparametric flow cytometry measurement (FCM). Patients and methods: CECs were determined in peripheral blood from 20 healthy volunteers, 12 patients undergoing coronary angioplasty, and 30 renal transplant recipients, and blood spiked with cultured endothelial cells. CD146+ cells were isolated using CD146-coated magnetic nanoparticles and labeled using CD45–fluorescein isothiocyanate and CD146–PE or isotype control antibody and propidium iodide before FCM. The same samples were also processed using CD146-based immunomagnetic separation as the reference method. Results: The hybrid assay detected CECs, identified as CD45dim/CD146bright/propidium iodide+, with high size-related scatter characteristics, and clearly discriminated these from CD45bright/CD146dim activated T lymphocytes. The method demonstrated both high recovery efficiency and good reproducibility. Both IMS and the hybrid assay similarly identified increased CEC levels in patients undergoing coronary angioplasty and renal transplantation, when compared to healthy controls. In patients, CEC values from these two methods were of the same order of magnitude and highly correlated. Bland–Altman analysis revealed poor statistical agreement between methods, flerrofluid–FCM providing higher values than IMS. Conclusion: This new hybrid FCM assay constitutes an accurate alternative to visual counting of CECs following CD146-based IMS.
Circulating endothelial cells (CECs) are matured differentiated cells that are shed from the vessel wall as a result of pathophysiologic conditions that affect the endothelium. CECs are present at very low levels in healthy subjects, whereas elevated levels have been reported in various pathologic situations, including cardiovascular disorders [1–6], infectious diseases [7–9], immune disorders [10–12], post-transplantation [13,14] and cancer [15–19]. CECs represent a specific and non-invasive surrogate marker with both prognostic and diagnostic significance, and can be useful to monitor response to therapy .
CECs are phenotypically defined by the expression of endothelial markers such as von Willebrand factor, CD31 or CD146 and the lack of leukocyte and progenitor cell markers. Owing to their scarcity, the most widely used method for CEC enumeration is based on a first step of enrichment by immunomagnetic separation (IMS) followed by visual counting using a fluorescence microscope. First described in 1992 , this approach, using paramagnetic particles coated with antibodies directed against CD146 (S-ENDO1), has emerged as the reference method for quantifying CECs. A consensus definition of CECs and a standardized protocol  have led to good agreement among laboratories with regard to CEC levels in normal populations. However, the identification and counting of CECs using fluorescence microscopy remains tedious and requires a trained operator. To date, conventional flow cytometry measurement (FCM), the main alternative approach for CEC analysis, has been rather disappointing. First, the normal values for CECs derived by FCM are highly heterogeneous, ranging from 10 to 7900 cells mL−1 [16,18] and, in most cases, are much higher than values obtained using IMS. Second, there is no consensus regarding the combination of markers used to identify CECs [15,21,22]. Third, the various published methods do not all take into consideration some of the difficulties associated with enumerating ultra-rare events, as suggested by Khan et al. . One major, and still open, question is whether or not conventional FCM is sensitive enough to count cells in blood at a frequency as low as one cell per 106 leukocytes.
The objective of this study was to evaluate a new hybrid assay for measuring CECs that combines pre-enrichment of CD146+ circulating cells using magnetic nanoparticles and multiparametric flow cytometry analysis. The analytic conditions and performance of this assay were validated by reference to IMS using various sources of cultured endothelial cells and clinical blood samples from patients with pathologies known to be associated with elevated CEC levels.
Forty-two patients were included in the study. Of these, 12 patients suffering from stable angina underwent scheduled angioplasty. Blood samples were collected 6 h after the end of the procedure, a time period previously reported to provide peak concentrations of CECs . Thirty renal transplant recipients were enrolled, and blood collection was performed between 4 and 12 months post-transplantation. Twenty healthy volunteers, with no known illness, were used for comparison with patients. In order to illustrate the detection of CECs, patients undergoing an erythrocytapheresis for treatment of idiopathic hemochromatosis were included, and peripheral blood from the traumatic venepuncture was obtained at the time of connection. Written informed consent was obtained from each patient, and the study protocol was approved by the Local Research Ethical Committees.
Human umbilical vein endothelial cells (HUVECs) were obtained from umbilical cord by collagenase digestion, as previously described . Cells were seeded on 0.2% gelatine-coated culture plates, grown in EGM-2 medium (Clonetics Cambrex, Ermerainville, France), and used at passage 1–3. Human microvascular endothelial cells (HMEC-1 cell line), obtained from E. W. Ades  (Center for Disease Control, Atlanta, GA, USA), were cultured in MCDB 131 medium (Invitrogen, Cergy-Pontoise, France) supplemented with 10% fetal bovine serum, 10 ng mL−1 human recombinant epidermal growth factor (Upstate Cell Signaling Solutions, Lake Placid, NY, USA) and 1 μg mL−1 hydrocortisone (Sigma, St Quentin-Fallavier, France). HMEC-1 cells were used at passage 8–12. Primary human saphenous vein endothelial cells were obtained from Clonetics and cultured until passage 5 in EGM-2MV medium (Clonetics). All cells were maintained under standard cell culture conditions (humidified atmosphere, 5% CO2, 37 °C).
Spiking experiments and determination of intra-assay reproducibility
For recovery experiments, known amounts of HUVECs were spiked into peripheral blood from healthy volunteers. Cells were first harvested using trypsin–EDTA, rinsed, and resuspended in phosphate-buffered saline without Ca2+. After counting in a Malassez hemocytometer, the initial suspension was serially diluted down to about 1000 cells mL−1 in phosphate-buffered saline supplemented with 0.2% bovine serum albumin and 0.1% sodium azide. The actual HUVEC concentration was determined by FCM-based absolute counting of the diluted suspension using Flow CountTM beads (Beckman-Coulter, France) incorporated at a comparable number. Thereafter, an appropriate volume of the diluted suspension was added to blood samples to achieve the theoretical cell concentrations of 20, 50 and 100 cells mL−1 of blood.
Peripheral blood samples from two patients (a traumatic venepuncture and a patient undergoing coronary angioplasty) were used to evaluate the intra-assay reproducibility. Each sample was divided into three aliquots, and each of these was processed using a hybrid assay by the same operator at the same time.
IMS reference assay
CECs were isolated using CD146-based IMS according to the recently proposed consensus protocol . CEC counting was done using a fluorescence microscope (Nikon) equipped with a CCD camera and image analysis software (Lucia, Laboratory Imaging Ltd, Prague, Czech Republic). The criteria for CEC identification included positive staining with acridine orange, rosettes bearing more than five beads, cell size over 15 μm, and binding of fluorescein isothiocyanate (FITC)-conjugated Ulex Europeus lectin.
Hybrid FCM assay
We applied a new hybrid assay combining immunomagnetic pre-enrichment and multiparameter FCM analysis (CELLQUANT FF CD146 kit, Biocytex, Marseille, France). In this method, all blood CD146+ cells, including CECs and a minor subset of activated T lymphocytes, are first enriched from the blood sample. Thereafter, CECs are differentially counted by triple-color FCM.
Reagents included: (i) CD146-coated ferrofluids (FFs) – magnetic nanoparticles  coated with the CD146 antibody Com-7A4 (Biocytex, F), which recognizes a different epitope to S-Endo1; (ii) a proprietary diluent specially formulated to reduce non-specific binding; (iii) a non-fixing red cell lysis reagent; (iv) antibody-based reagents, including a pan-leukocyte fluorescent conjugate CD45–FITC and the CD146 mAb S-Endo1–phycoerythrin (PE) – the negative control reagent is made of the same S-Endo1–PE conjugate inhibited by excess amounts of unlabeled S-Endo1 mAb; (v) a formaldehyde-based fixative reagent; and (vi) a final nuclear cell-counting reagent made of propidium iodide (PI) (1.5 μg mL−1) in a permeabilizing buffer and 3-μm fluorescent beads at known concentration, allowing absolute counting of CECs.
For pre-enrichment, two 1-mL aliquots of blood in EDTA were processed in parallel to provide both a test (T2) and a negative control (T1) tube. One milliliter of blood was mixed with 1 mL of diluent and 50 μL of FFs. After thorough mixing, the tubes were placed against the magnet (Dynal MPC-L; Invitrogen) for 25 min at room temperature. The liquid phase was discarded, and the red cell lysis reagent was added for 10 min. The liquid phase was then aspirated off, and the enriched FF-containing cell fraction was resuspended in 150 μL of diluent. For immunostaining, both T1 and T2 tubes received 20 μL of CD45–FITC. The T2 tube received 20 μL of CD146–PE, whereas 20 μL of negative control was added to the T1 tube. Finally, 200 μL of nuclear cell-counting reagent was added, and samples were stored at 2–8 °C until analysis, with a maximum delay of 6 h.
FCM analysis was operated using a Cytomics FC500 flow cytometer (Beckman-Coulter, Villepinte, France) using gating strategies as indicated in Results. CECs were defined as PI-positive (PI+) nucleated cells showing bright CD146 staining and dim CD45 staining due to intrinsic high autofluorescence of endothelial cells.
Statistical analysis was performed using PrismTM software (GraphPad Software, San Diego, CA, USA). As CEC levels in patients were not normally distributed, results were expressed as median and range, and the non-parametric Mann–Whitney U-test was used for comparison with controls. P < 0.05 was considered to be significant. Correlations between CEC values determined using both IMS and FF-FCM assays were evaluated using the Spearman test. To analyze the degree of agreement between the two CEC assays, Bland–Altman plots were generated to relate the interassay difference with the mean CEC count of the two methods for each individual sample.
To evaluate whether the two CEC assays produced results capable of giving similar clinical interpretations, patients’ CEC counts were classified as ‘normal’ or ‘elevated’ according to limit values of clinical interpretation corresponding to the 95th percentile of CEC values in healthy controls (mean + 2 SD) for each assay.
Definition of FCM analysis conditions for CEC determination
FCM conditions were initially adapted using blood samples spiked with HUVECs (∼ 1000 mL−1) and then validated using cultured endothelial cells of different origins (HMECs, saphenous vein endothelial cells) and patient samples.
The enriched cellular suspension was first analyzed on a log scale forward scatter (FS) vs. side scatter (SS) cytogram. As illustrated in Fig. 1A, the suspension contained: (i) CD146dim activated T lymphocytes (pink-colored points); (ii) contaminating polymorphonuclear granulocytes and monocytes (green-colored points); (iii) aggregated FFs; (iv) 3-μm counting beads (red-colored points); and (v) large cells corresponding to spiked HUVECs (blue-colored points). A wide dual scatter gating region (denoted ‘D’) was defined to: (i) gate out submicrometer magnetic particles and cellular debris; and (ii) include even very big cells. It begins in FS below the 3-μm counting bead cloud and in SS before the lymphocyte cloud. FCM analysis was further restricted to nucleated (PI+) cells, as illustrated in Fig. 1B. The optimized positioning of PI+ limit (gate B) was set between polymorphonuclear granulocytes and lymphocytes. The correlated CD45–FITC vs. CD146–PE histograms are illustrated in both Fig. 1C (tube T1) and Fig. 1D (tube T2). Apart from activated T lymphocytes (pink cloud), which are characteristically CD45–FITCbright and CD146–PEneg or dim, the gated cell fraction also contains a few aggregated FFs. These aggregated FFs, which non-specifically bind both FITC-conjugated and PE-conjugated mAbs, create, along the diagonal of the cytogram, a useful spindle-shaped limit separating leukocytes located under the diagonal and spiked HUVECs or potential CECs, which appear CD45neg or dim and CD146bright, located above the diagonal. For detection of CECs, a region, named K, was delineated in the histogram from the T1 tube in order to include all events with a CD146 expression higher than that delineated by the diagonal. The position of the K region was optimized so that it did not contain more than five events for the T1 tube. This position remained unchanged for the analysis of the specific events from the T2 tube.
As illustrated in Fig. 1D, the analysis of blood spiked with HUVECs resulted in the appearance of a clearly detectable cloud of events in the K region. Analysis of peripheral blood spiked with cultured endothelial cells from other origins and peripheral blood from patients with expected high levels of CECs resulted in different patterns of isolated CD146+ cells, as illustrated in Fig. 2. In all cases, specific events were present in gate K with various aspects of the corresponding cloud. FS and SS characteristics of the few CECs defined as CD146bright/CD45low/nucleated cells was possible using back-gating from the K region, as illustrated in Figs 2 and 3. This strategy confirmed that CECs represent a rather heterogeneous population of large cells displaying FS and SS intensities that are often higher than those of granulocytes.
Determination of CEC absolute value
Most of the available enriched cell suspensions were analysed by FCM using high-level delivery speed and an acquisition time of 200 s. The numbers of events counted in gate K from both T1 (K1) and T2 (K2) tubes were used to calculate the CEC concentration in the blood sample using the following formula: CEC = [(K2 × 2000)/C2] − [(K1 × 2000)/C1], where 2000 is the total number of counting beads added to the final suspension, and C is the number of beads analyzed in the corresponding tube.
Recovery and reproducibility
Recovery For recovery experiments, known amounts of HUVECs (0, 20, 50 and 100 cells mL−1) were spiked into peripheral blood from healthy volunteers before CEC analysis. The number of CECs detected was corrected with the CEC count of the corresponding non-spiked samples and then compared to the known number of spiked cells. Six independent determinations were performed. The mean percentages of spiked HUVECs recovered were 81%, 80% and 90% for the three levels of spiked CECs, respectively. On average, 83% of the spiked HUVECs were retrieved, suggesting that the recovery of the assay was high and stable within the tested range.
Reproducibility In the blood sample from venepuncture artefact, the mean level of CECs was 331 CECs mL−1 (± 49) with a coefficient of variation (CV) of 15.1%. In the blood sample collected after coronary angioplasty, the mean level of CECs was 16.0 CECs mL−1 (± 2.8), and the CV was 17.7%.
CEC enumeration in patients by reference to IMS
The CEC levels were determined in renal transplant recipients, patients with coronary angioplasty and healthy controls using both assays. Consistent with previous reports [2,13], the median CEC count in the control group was 5 cells mL−1 (range 0–9 CECs mL−1) using IMS, and 8 cells mL−1 (0–21 CECs mL−1) using FF-FCM. The difference between these result was not significant (P = 0.0688). Using FF-FCM analysis, significantly higher CEC numbers [median 39 cells mL−1 (0–189 CECs mL−1), P < 0.0001] were observed in renal transplants in comparison to controls (Fig. 4A). A significant difference was also obtained using IMS [median 19 cells mL−1 (2–152 CECs mL−1), P = 0.0015] (Fig. 4B). CEC counts for the two methods were highly correlated (r = 0.842, P < 0.0001) (Fig. 4C).
CEC counts estimated by FF-FCM were also significantly higher in patients with coronary angioplasty [median 26.5 cells mL−1 (8–80 CECs mL−1)] than in healthy subjects (Fig. 5A), a difference that was also observed using IMS [median 11.5 cells mL−1 (4–60 CECs mL−1)] (Fig. 5B). CEC counts for the two methods were again highly correlated (r = 0.888, P = 0.0001) (Fig. 5C).
To evaluate the agreement between FF-FCM assays and IMS, a Bland–Altman plot was used. In renal transplant recipients (Fig. 4D) and patients with coronary angioplasty (Fig. 5D), more than 95% of events were included in the confidence interval. The bias (estimated by the mean difference) was 26.1 and 16.4 CECs mL−1, respectively, demonstrating that the hybrid method consistently gave higher values than IMS. In both patient groups, the size of the 95% confidence interval was wide, suggesting a poor agreement between methods.
To further investigate whether results from the two methods enabled similar clinical interpretation, despite this poor agreement, each patient from both groups was classified as ‘elevated CEC’ or ‘normal CEC’ according to cut-off values. These cut-off values were defined by reference to the 95th percentile of CEC determinations in healthy subjects: CEC count > 95th percentile of control values defined an ‘elevated CEC’ status, whereas CEC count ≤ 95th percentile of control values defined a ‘normal CEC’ status. In 93% of patients, CEC status was similar between the two assays. It is noteworthy that the three patients (7%) displaying a scoring disagreement had a CEC level very close to the cut-off.
This study describes a new hybrid assay for CEC detection that combines a first step of enrichment that is compatible with CEC scarcity in peripheral blood, followed by FCM of enriched cells that uses a multiparametric characterization of CEC and an automated counting alternative to microscopic observation.
In many respects, FCM is performed in non-conventional ways in this hybrid test. First, PI is used as a simple nuclear stain ensuring the simultaneous counting of both viable and non-viable cells, and not for exclusion of dead cells. Second, this method takes into account the well-recognized high autofluorescence background of endothelial cells that arises from their large size, the occurrence of cell clumps, or PI staining. Third, our FCM strategy opens the dual scatter ‘window of analysis’ by using log scales for both size-related scatter parameters and fluorescence signals. This log scale provides an a priori unrestricted view of all analyzed events, ranging roughly from 1 μm up to 100 μm. Using back-gating analysis, we showed that CECs, as defined in the assay on the basis of a high FL2 to FL1 ratio, appear in a dual scatter region located over the granulocyte region, whether they are from pathologic blood samples or from various types of cultured endothelial cells. Therefore, for the first time, this study provides FCM-based size data for CECs that are consistent with: (i) well-known heterogeneous morphologic features of endothelial cells; (ii) previous microscopic observations of CECs recovered using IMS [1,20]; and (iii) image analysis of CECs isolated by an alternative automated method using the same CD146-coated FFs . In contrast to the flow cytometry assay recently described by Goon et al. , our protocol is compatible with the detection of cultured endothelial cells added to blood. By analysis of normal blood samples spiked with HUVECs, the recovery rate of the hybrid assay ranged from 80% to 90%, indicating comparable efficiency when compared to the standardized IMS method . In addition, the good reproducibility of this rare cell-counting assay for both high and intermediate levels of CECs was attested by CVs lower than 20%.
Clinical specimens were also used to evaluate the accuracy of the hybrid assay for enumerating CECs by reference to IMS. In healthy adult subjects, CEC values determined by FF-FCM ranged from 0 to 21 cells mL−1, with a median value of 8 cells mL−1. These values were highly comparable to those from IMS (< 10 cells mL−1) and to those recently reported from an automated rare cell enumeration system . As compared to normal values from conventional FCM approaches, which gave ten-fold to thousand-fold higher CEC counts, these data emphasize the importance of a prior CD146-based cell enrichment to ensure the specificity of the analysis. Consistent with this, it has been recently demonstrated that the putative CECs identified as CD45−, CD31++ and CD146+ events using flow cytometry were mainly giant platelets .
Previous work, including some from our group [1,2], has demonstrated that percutaneous coronary intervention and renal transplantation are associated with increased CEC levels . As expected in these two clinical settings, CEC counts obtained with FF-FCM and IMS were in the same order of magnitude and highly correlated. Bland–Altman analysis revealed that the two methods are not interchangeable and that FF-FCM provided higher CEC values than IMS in the clinical situations tested. Such differences could be related to: (i) a reduced loss of CECs as compared to the IMS technique, which involves multiple washing steps; (ii) an optimized capture efficiency of CD146-coated nanobeads as compared to 5-μm microbeads, providing constant recovery over a large CD146 antigen density; and (iii) the recovery of nucleated fragments of CD146+ cells, due to a size threshold of 3 μm. The possibility that CD146-expressing endothelial progenitors may interfere with CEC counts can be ruled out, as it has been demonstrated that such endothelial progenitor cell subpopulations express CD45 . Limitations in comparing the two methods can also come from missing some anucleated cell carcasses, using FF-FCM assays, due to the absence of PI staining, or fragile cells, due to the putative deleterious impact of the red cell lysis step. Nevertheless, in most tested cases, the two approaches allow the similar classification of patients between normal or elevated CEC levels according to cut-off values defined with healthy controls, suggesting that both methods have similar clinical applicability.
In conclusion, our study demonstrates that the new hybrid assay combining CD146+ cell enrichment and multiparametric flow cytometry analysis constitutes an accurate alternative method for CEC detection. We also consider that, for the first time, this hybrid assay reconciles all previously defined features of CECs, that is, big cells (over 10 μm in diameter and up to 50–100 μm), nucleated, viable or not, with the practical advantages of flow cytometry-based counting of circulating cells. This approach is particularly attractive for large-scale clinical studies, which still need to firmly establish the clinical value of CECs in diseases with endothelial damage.
The authors thank Biocytex and Beckman-Coulter companies for their technical support, the clinical staffs of the Nephrology Unit (Professor Berland) and of the Hemapheresis Unit (Dr Lefèvre) for their contribution to patient recruitment, and the clinical staff of Gynaecology Obstetrics for collection of umbilical cords. They are also grateful to Patricia Stellmann and Laurent Giraudo for their assistance in endothelial cell culture.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.