A lack of standardized assays and consensus of cell definition has lead to a wide variation in the reported range of circulating endothelial cells (CECs).
A lack of standardized assays and consensus of cell definition has lead to a wide variation in the reported range of circulating endothelial cells (CECs).
An automated rare cell analysis system was used to enumerate nucleated, CD146+/CD105+/CD45− CECs in 4 mL of blood.
Recoveries of spiked HUVECs were linear over a range of 0–1,241 cells (R2 ≥ 0.99) with recoveries of ≥70% at each spike level. Correlation coefficient values for interoperator variability and duplicate sample variation were (R2 = 0.99 and 0.90), respectively. Correlation of CEC counts between tubes 1–2 and 2–3 drawn from the same subject in sequence differed (R2 = 0.48 and 0.63, respectively). The normal CEC reference range established in 249 healthy donors was 1–20 CECs/mL blood. CEC counts were significantly higher in the 206 metastatic carcinoma patients (P < 0.0001).
CECs can be accurately and reproducibly enumerated in blood and are elevated in metastatic carcinomas compared with healthy donors. Phlebotomy procedures can affect endothelial cell counts. © 2007 International Society for Analytical Cytology
Proper functioning of the circulatory system is crucial to maintain good health and, assessment of the status of the endothelial cell layer may be an effective means of gauging disease activity. Changes in endothelial cells play an important role, in a variety of disease conditions including cancer, cardiovascular, autoimmune, and infectious disease (1–13).
The presence of mature vascular endothelial cells in the circulation may be due to a variety of factors including traumatic injury, which causes them to slough off and carried away in the circulation (14–19).
As the cause, fate and role of these circulating endothelial cells (CECs) is better understood, their in vitro enumeration and characterization may offer a unique opportunity to noninvasively “biopsy” the vasculature, thereby facilitating our understanding of a variety of homeostatic and disease processes. Indeed, “elevation” of CECs has been observed at some point over the clinical course of all of the above mentioned pathological conditions (4, 17, 19–30). However, the lack of standardized assay methods, the lack of consensus on the definition of a CEC (21, 22, 31) and disease heterogeneity have led to a wide variation in the reported range of CECs in the literature (1–5,700 per mL). All of these factors make interpretation and comparison of existing studies quite difficult if not impossible (21, 25, 30) and more importantly hamper future investigations. To address these issues, an automated rare cell analytical system was developed to accurately and reliably enumerate circulating endothelial cells.
The protocol and informed consent forms were approved by the Independent Investigational Review Board (Plantation, FL) and all donors provided written consent prior to participation in the trial. The healthy donors used for comparison with the patients had no known illness or fever at the time of draw and no known history of malignant disease. No information was collected regarding potential high risk behavior, such as smoking, exercise habits, or obesity. The protocol was designed to survey carcinoma patients, consequently, clinical specimens were drawn from patients from various geographic locations and with a variety of carcinomas at “random” time points during the course of their treatment and regardless of type treatment.
Blood was drawn in CellSave™ Blood draw Tubes (Immunicon, Huntingdon Valley, PA). Samples were maintained at room temperature, shipped via overnight courier to a central laboratory, and processed within 72 h of blood collection.
Human umbilical vein endothelial cells (HUV-EC-C) (ATCC, VA.) were cultured according to ATCC recommendations. The cells were subsequently harvested using trypsin. To determine the actual cell number, a 100 μL aliquot of the HUVEC cells was permeabilized and fluorescently labeled by adding 90 μL of a diluent buffer containing 0.011% saponin and 10 μL CD105 conjugated to phycoerythrin (CD105-PE) at a final reaction concentration of 2.5 μg/mL. After 15 min incubation at room temp, 200 μL of PBS, 25% CytoCheck, 0.1% BSA and 0.1% sodium azide, and 20 μL of fluorescent beads (Beckman-Coulter, FL) containing ∼25,000 total beads were added. Tubes containing beads only were run on a flow cytometer (FACSCalibur, BD Biosciences, San Jose, CA) until 100% of the sample was aspirated. This provided an accurate estimate of the number of beads present in 20 μL. The experimental tubes were then tested in triplicate on the flow cytometer until 20,000 beads were counted for the low spike and 12,000 were counted for the high spike. Using the known number of beads per unit volume, the concentration of cells was determined. The HUVEC cells were fluorescently labeled with DiOC16 (3) (Molecular Probes, CA) before spiking to discriminate these cells from endogenous endothelial cells. For accuracy, linearity, and sensitivity experiments the spiked cell numbers were estimated to be 5, 9, 78, 310, and 1,241, and for assay imprecision the cell numbers were 48 and 1,014 in 4 mL of blood.
The CellTracks® system (Immunicon Corp) used for endothelial cell enumeration consisted of CellSave™ tubes, the CellTracks AutoPrep a fully-automated sample preparation system (32, 33), the Endothelial Cell Reagent Kit and the CellSpotter® Analyzer a semiautomated fluorescence microscope. To enrich endothelial cells from 4 mL of blood, ferrofluids coated with monoclonal antibodies directed against the CD146 antigen expressed on endothelial cells, smooth muscle cells and a subset of activated T-lymphocytes (34–36). The enriched cells are then fluorescently labeled with the nuclear stain DAPI, CD105-PE expressed on endothelial cells, activated monocytes, and pre-B-lymphocytes (37, 38) and the pan-leukocyte antibody CD45 conjugated to allophycocyanin (CD45-APC).
Briefly, 4 mL of blood is mixed with 10 mL of buffer, centrifuged at 800g for 10 min, and then placed on the sample preparation system. The instrument aspirates the plasma/buffer layer and adds anti-CD146 ferrofluids. After incubation and subsequent magnetic separation, the unbound cells and remaining plasma are aspirated. Next, staining reagents are added in conjunction with a permeablization buffer to fluorescently label the immunomagnetically bound cells. After incubation, magnetic separation is repeated to remove excess staining reagents. After the final processing step, the cells are resuspended in ∼300 μL of buffer transferred in a chamber that is placed between two magnets that orient the immunomagnetically labeled cells in a monolayer for analysis. This is then placed on a four color semiautomated fluorescent microscope. A gray scale charged coupled device (CCD) camera is used to scan the entire chamber surface once for each of four color fluorescence channels capturing ∼140 frames per channel. Each captured frame is then evaluated for objects that are potential CEC candidates by image analysis software.
To determine the linearity of the assay, 4 mL aliquots of pooled blood from each of five different healthy donors was added to individual AutoPrep tubes containing a known number of DiOC16 (3) prelabeled HUVECs to distinguish them from indigenous CECs. A range of HUVECs was used to establish both a lower and upper limit of detection. Recoveries of the prelabeled HUVECs were compared with the actual number of cells spiked.
Blood from a healthy donor was collected into blood collection tubes and dispensed into eight aliquots of 4 mL. A 50 μL volume containing 48 HUVECs was added to four of the tubes of blood and a 50 μL volume containing 1,014 HUVECs was added to the remaining four tubes. All tubes were mixed by repeated inversion and allowed to sit overnight at room temperature. On the following day, two of the tubes with 48 cells and two of the tubes with 1,014 cells were processed and analyzed in the morning, and the remaining four tubes were processed in the afternoon. All of the samples for the imprecision determination were processed on the same instrument. This entire procedure was repeated each day for 20 days. Calculation of within-run and total imprecision was performed as per the National Committee for Clinical Laboratory Standards (NCCLS) EP-5A guidelines (39).
Blood from 15 healthy individuals was drawn into blood collection tubes and pooled, and 4 mL aliquots of the blood were used to evaluate CECs at 0, 24, 48, and 72 h after blood draw. Interoperator variability was determined by comparing CEC analysis by two different operators who individually reviewed the gallery of images obtained from 100 different specimens. The operators were blinded to the origin of all samples.
CEC variation within aliquots of blood dispensed from a single 10 mL blood collection tube and CEC variation among tubes drawn in sequence at the same time from the same patient were also investigated. To measure the variability of CEC counts obtained from two 4 mL aliquots derived from a single blood draw tube, a single operator assayed blood from 72 different healthy volunteers.
To evaluate the variation among serial draw tubes, CEC counts from a minimum of three tubes of blood drawn from the same vena puncture from 64 healthy individuals, were analyzed. The actual first and second tubes drawn were always represented for the third tube either the actual third or later draw tube was used. All CEC measurements from a single patient were performed by the same operator, and tubes were tested without pooling of the blood. To evaluate variation among serial tubes drawn from a venous port, blood was drawn from the same port from 10 patients with metastatic carcinomas.
The image software used for this analysis identifies objects that stain with both DAPI and PE. It then displays these objects to the operator as browser formatted thumbnail images, Figure 1. Reading from right to left, these thumbnails represent staining of the nucleus (DAPI), cell membrane (CD105-PE), prelabeled HUVECs [DiOC16 (3)], and Leukocyte cell membrane (CD45-APC). The fifth and last image is a false color overlay of the nucleus (DAPI) and cell membrane (CD105-PE) staining—this provides the operator an overall view (nucleus and cytoplasm) of the object and permits confirmation of cell-like morphology. Check boxes beside each image allow the operator to confirm the presence of staining in that channel, whereas the check box adjacent to the composite image is used to confirm whether the object meets all CEC criteria. The software then tabulates the checked boxes for each sample and results are expressed as the number of CECs per 4 mL of blood. In Figure 1, the thumbnails in Row 1 and 2illustrate DiOC16 labeled HUVEC cells identified by DAPI, CD105, and DiOC16 staining and lack of CD45 staining. Rows 3, 4, and 5 show CECs identified by their staining and lack DiOC16 and CD45 staining. Note that the endothelial cell presented in Row 4 is surrounded by two leukocytes stained with DAPI and CD45 but lack DiOC16 and CD105 staining. Rows 6 and 7 show two cells stained with DAPI, CD105, and CD45, but not with DiOC16, these are therefore most likely leukocytes. The images are autoscaled, which results in grey pixilated images when no stained objects are present. In Figure 2 a gallery of typical CEC images is shown and highlights the variety of morphologic appearances. In the figure some sheets of endothelial cells are shown that can be observed occasionally and are counted as one CEC.
HUVEC cells were spiked into aliquots of pooled blood drawn from five healthy donors at levels of 5, 9, 78, 310, and 1,241 cells per 4 mL of blood. In Figure 3A, the approximate number of HUVEC cells spiked into the blood is plotted against the number observed in the samples. Regression analysis using the number of observed CECs versus the number of expected CECs resulted in a best fit line with a slope of 0.72 (95% confidence interval = 0.68–0.76), an intercept of 5.1 (95% confidence interval = −17 to 27), and a R2 of 0.99. As expected, the percent coefficient of variation (%CV) increased as the number of cells spiked decreased, ranging from 12.5% at the 1,241 cell spike to 37.3% at the 5 cell spike. On average, 85.6% of the spiked HUVEC cells were recovered.
The analytical lower limit of detection was measured by spiking a low number of DiOC16 (3) labeled HUVEC cells in the sample processing tube, with verification of the actual number of spiked cells done using an inverted fluorescence microscope before the addition of 4 mL of blood. The samples were processed and the spiked cells were enumerated. The assay efficiency, or percentage of spiked cells recovered, was determined in 60 different samples, and is illustrated in Figure 3B. The actual cell spikes ranged from 2 to 26 (mean 12, SD 5), and the percent recovery ranged from 44 to 100% (mean 86%, SD 14).
The reproducibility of CEC enumeration was measured using a single stock of DIOC16 (3)-labeled HUVEC cells spiked into blood at levels of approximately 48 and 1,014 cells per 4 mL. Duplicate samples at each level were tested twice per day for 20 days, and the results are shown in Figure 3C. The within run %CVs for the 1,014 and 48 cell spikes were 7.2% and 11.7%, respectively. Similar results were found for total imprecision with %CVs of 7.7% and 15.6% for the 1,014 and 48 cell spikes, respectively.
Blood from 15 healthy individuals was drawn into blood collection tubes and pooled. Aliquots of the pooled blood were used to evaluate CECs at 0, 24, 48, and 72 h after blood draw. The differences in the number of CECs at time 0 vs. 24, 48, and 72 h were not significant (P = 0.34, 0.20, and 0.67, respectively, paired t-test), demonstrating that blood samples drawn in the blood collection tubes can be analyzed for CECs for up to at least 72 h after blood draw.
To measure the variability associated with the analysis by different operators, the browser images from 100 different specimens were analyzed blindly by two different operators. Regression analysis demonstrated a best fit line with a slope of 1.03 (95% confidence interval = 1.01–1.05), an intercept of 0.28 (95% confidence interval = 0.72–1.29), and an R2 of 0.99. Figure 4A, shows the correlation between both operators, and in Figure 4B, the same data is shown using a Bland-Altman plot. The error of the CEC determinations by the two operators in each sample is represented as a percentage, and is calculated by dividing the difference of the two Operators CEC counts (Operator 1 − Operator 2) by the average of both CEC counts. These analyses suggest that the criteria for identification of CECs from the images presented in the CellSpotter browser can be taught effectively and reproducibly.
To measure the variability of CEC counts obtained from two 4 mL aliquots derived from a single evacuated blood draw tube, CECs were enumerated in duplicate using the blood in a single draw tube from 72 different healthy volunteers. Regression analysis showed a best fit line with a slope of 0.97 (95% confidence interval = 0.89–1.05), an intercept of 3.55 (95% confidence interval = −0.30 to 7.40), and an R2 of 0.90. Figure 4C shows the correlation between both aliquots, and in Figure 4D, the data is shown using a Bland–Altman plot. The error of each CEC measurement is represented as a percentage calculated by dividing the difference of the CEC counts between the two aliquots (Aliquot 1 − Aliquot 2) by the average of both CEC counts.
To evaluate the effect of the vena puncture and the associated localized turbulence in blood flow on the release of endothelial cells from the local vessel wall into the evacuated blood draw tube, CECs were enumerated in different draw tubes from the same donor. A minimum of three tubes were obtained from 64 healthy individuals Three healthy individuals were excluded from the analysis because the percent difference between at least two tubes were above three standard deviations of the mean (two specimens from Tube 1 vs. Tube 2: 58v230, 144v80. One specimen Tube 2 vs. Tube 3: 26v116). Figure 5A shows the correlation between CECs in the first and second tubes, where regression analysis showed a best fit line with a slope of 0.41 (95% confidence interval = 0.28–0.54), an intercept of 9.79 (95% confidence interval = −2.09 to 17.49), and a R2 of 0.42. Figure 5B shows the corresponding Bland–Altman plot. The median number of 36 CECs in tube 1 was significantly higher compared with the median number of 24 CECs found in tube 2 (P = 0.0070, nonparametric k-sample median test). To determine whether this increase in the number of CECs was due to the actual vena puncture, a similar analysis was done on the CECs enumerated in the second and third blood tubes. Figure 5C shows the correlation between CECs in the second and third tubes drawn, and regression analysis showed a best fit line with a slope of 0.50 (95% confidence interval = 0.40–0.60), an intercept of 8.98 (95% confidence interval = 5.21–12.74), and an R2 of 0.63. Figure 5D shows the corresponding Bland–Altman plot. The median number of 24 CECs in the second tube were not significantly different from the median number of 21 found in the third draw tube (P = 0.15, nonparametric k-sample median test). Although the correlation improved, it did not reach the level obtained when CECs were enumerated in two aliquots of blood from the same tube (R2 = 0.63 and 0.99, respectively). CECs were enumerated in subsequent draw tubes from a venous port from 10 patients to examine whether less variation in CECs numbers would be observed. The mean and median CECs in the first draw tube were 21 and 17 in the second draw tube 20 and 16 and in the third draw tubes were 25, 18. These differences were not significant (tube versus 2, P = 1.0 tube 2 versus 3, P = 0.50, non parametric k-sample median test). The correlation between the first and second draw tube was R2 = 0.72 and R2 = 0.87 between the second and third tube.
CECs were enumerated in blood samples from 255 healthy individuals and 206 patients with various metastatic carcinomas. Healthy individuals (n = 243) ranged in age from 19–63 (mean 38, SD 11, median 39) and patients (n = 206) age ranged from 30–90 (mean 68, SD 12, median 69) The CEC levels in 6 (2.4%) of the healthy volunteers were determined to be outliers (above three standard deviations), and these results (CEC counts = 176, 186, 230, 271, 380, and 916) were excluded from the calculations of the mean and median. In the remaining 249 healthy subjects, the range of CECs detected was 0–97 cells, mean ± SD = 21 ± 18 cells, median = 15 cells. Although blood from all six outliers were from female donors no significant difference in CECs between CECs from male (n = 99, mean 21, median 17) and female donors (n = 150, mean 21, median 17) was noted. The normal reference limits were determined using the CEC results from the remaining 249 healthy individuals (40). The normal range (95% reference limits with 90% confidence intervals) for the CEC assay was determined to be 4 (90% CI = 2–5) to 80 (90% CI = 64–97) CECs/4 mL.
Figure 6 shows a scatter plot comparing CEC counts from healthy individuals and carcinoma patients. In 206 combined cancers patients, the range of CECs detected was 0–1,939 cells, mean ± SD = 111 ± 255 CECs, median = 34 CECs. In 24% of the patients CECs were above the upper reference limit. In 50 breast cancer patients, CECs ranged from 1 to 471 cells, mean ± SD = 78 ± 96 CECs, median = 38 CECs, and in 22% of the patients CECs were above the upper reference limit. In 49 colon cancer patients, CECs ranged from 0 to 1,375 CECs, mean ± SD = 86 ± 204 CECs, median = 29 CECs and in 20% of the patients CECs were above the upper reference limit of 97 CECs. In 35 lung cancer patients CECs ranged from 11 to 1,546 CECs, mean ± SD = 146 ± 270 CECs, median = 75 CECs, and in 40% of the patients CECs were above the upper reference limit. In 48 prostate cancers patients, the range of CECs detected was 3–1,939 CECs, mean ± SD = 82 ± 279 cells, median = 27 CECs, and in 10% of the patients CECs were above the upper reference limit. In 24 patients with other types of cancer, the range of CECs detected was 6–1,499 cells, mean ± SD = 240 ± 427 cells, median = 66, and in 33% of the patients CECs were above the upper reference limit.
The distribution of the CEC counts were significantly different among healthy individuals and metastatic cancer patients (P-value = 0.0001, Kruskal–Wallis Test). The median CEC counts were also significantly different between these two groups (P-value < 0.001, nonparametric k-sample median test). Using the upper 90% confidence interval of the upper normal 95% reference limit as a cutoff (i.e., >97 CECs/4 mL of blood), a statistically significant difference was found between the proportions of healthy volunteers and metastatic cancer patients whose CEC levels exceeded this threshold (2% vs. 24%, respectively, Fisher's exact test P-value < 0.001).
Over the clinical course of cancer endothelial cells may find their way into the circulation for a variety of reasons and their enumeration and characterization may provide insights into disease processes and response to treatment. However, the reported ranges of CECs in the literature vary greatly. The principle sources of these discrepancies are the use of manual sample preparation methods along with the variety of immunophenotypic definitions of CECs. We used an automated rare cell enumeration system to immunomagnetic enrich CECs targeting the CD146 antigen and overcome the obstacle of variability due to the processing of larger sample volumes. Addition of CEC morphology to the classification by the use of fluorescence microscopy further increased the assay specificity by converting “events” into cells. Accuracy, sensitivity, specificity, and reliability of the CEC assay were established. The system precision was established by performing the assay at low (∼50) and high (∼1,000) cell spikes over a period of 20 consecutive days. The system consistently recovered >70% of the HUVEC cells with a coefficient of variation of 7.5% for the high and 15.2% for the low cell spike, with recovery remaining linear over the tested range (overall average 86%, R2 = 0.99). Robustness of the system was determined by calculating the agreement between two different operators who analyzed CECs in the same 100 individual samples. Agreement between the two readers was R2 = 0.99.
The antigen density of CD146 on HUVEC cells may not be representative of all CECs. To minimize the influence of varying CD146 density the configuration of the assay is such that the magnetic labeling of low antigen density cells is increased to provide consistent recovery over a large antigen density range (41). CD146 is expressed on endothelial derived microparticles and detection of endothelial cells with an antibody recognizing a different epitope of CD146 results in the detection of an abundance of microparticles among few endothelial cells (42). To avoid this CD105, which is broadly expressed on endothelial cells was used to detect endothelial cells among the CD146 immunomagnetic enriched cells.
Variation in CEC counts in aliquots taken from the same draw tube was minimal (R2 = 0.90, Fig. 4). However, there was a significant difference in CEC counts noted between the first and second blood tube drawn (R2 = 0.48). The agreement improved only slightly when comparing CEC counts from the second and third blood tubes drawn (R2 = 0.63, Fig. 5). The significantly higher number of CECs found in the first collection tube may be due to an initial release of local endothelial cells caused by the vena puncture. This, however, does not explain the variation in second, third, and later collection tubes, which did not, show the same correlation as was seen between the two aliquots of blood from the same draw tube. This discrepancy may be due to dislodgment of local venous endothelial cells as a result of the vacuum force created when a subsequent draw tube is introduced, or from shear forces generated by turbulence created by the continued presence of the needle in the vein. That no significant variation in CEC counts were observed from sequential blood draw tubes obtained from venous ports support the latter hypothesis. Implications for regular venous blood draws are that the first draw tube should not be used to obtain a CEC count and that for longitudinal monitoring, a significant change in CEC counts need to be based on the tube to tube variation, and would have to exceed a change of 88% (Fig. 5B).
Clinical specimens from healthy donors and patients with a variety of metastatic carcinomas were evaluated for CECs. The normal reference range was established at 4–80 CECs per 4 mL or 1–20 per mL of blood with an upper reference limit of 24 CECs per mL of blood. CECs were elevated in a significant percentage of patients compared with healthy controls, (Fig. 6). Whether these CECs are derived from the tumor or are merely the result of damage to the vasculature because of the therapy the patients are receiving is not known. Assessment of tumor associated antigen expression on CECs may allow the identification of CECs derived from the tumor whereas assessment of the age, activation state, and viability status may further help in the characterization of CECs (43–48). The question arises as to what the clinical significance of an elevated CEC count in these patients with metastatic carcinomas and prospective clinical trials will be needed to explore the clinical utility of these findings.