Fax: (203) 737-5089
Use of magnetic enrichment for detection of carcinoma cells in fluid specimens
Article first published online: 28 DEC 2001
Copyright © 2002 American Cancer Society
Volume 94, Issue 1, pages 205–211, 1 January 2002
How to Cite
Kielhorn, E., Schofield, K. and Rimm, D. L. (2002), Use of magnetic enrichment for detection of carcinoma cells in fluid specimens. Cancer, 94: 205–211. doi: 10.1002/cncr.10193
- Issue published online: 28 DEC 2001
- Article first published online: 28 DEC 2001
- Manuscript Accepted: 27 AUG 2001
- Manuscript Revised: 26 JUL 2001
- Manuscript Received: 26 JUN 2001
Ascites fluid or a pleural effusion are common events in metastatic carcinoma, but they also can be associated with several other medical conditions. The standard for determination of malignancy in these situations is cytologic evaluation of these fluids. Although this method is frequently successful, there are times when it fails, even when the patient has a malignancy, either because of insufficient cells in the fluid or for other reasons. This study addresses this problem taking advantage of the recent advances in technology for detection of rare epithelial cells in liquid specimens.
The authors examined fluid specimens from 59 patients to determine the frequency of recovery of epithelial cells compared with that achieved by conventional cytopathology. The Dynal CELLection Epithelial Enrich (Dynal AS, Oslo, Norway) method was used. This method is based on immunomagnetic selection of cells binding to EpCAM antibodies. Carcinoma cells were confirmed by morphology and, when there was sufficient material, by E-cadherin staining.
Grouping the cases by cytologic diagnosis, the authors found malignant cells using the cell enrichment assay in 11 of 12 malignant cases, 2 of 5 atypical cases, and 3 of 42 negative cases. Further investigations were conducted on the five cases that were cytologically negative or atypical but yielded epithelial cells after immunomagnetic enrichment. Four cases ultimately were proven malignant by other methods and one had incomplete follow-up.
The new methods available for epithelial cell enrichment in liquids may be used successfully on cytologic fluid specimens and may lead to increased sensitivity for detection of malignancy, and consequently more accurate staging. Cancer 2002;94:205–11. © 2002 American Cancer Society.
Ascites fluid may arise from many causes, but metastatic carcinoma is perhaps the most critical. Finding malignant cells often changes the patient's stage. Identification raises two problems: 1) detection of the cells and 2) verification that they are malignant. Determination of malignancy has been addressed by many investigators and several methods have been suggested to distinguish between reactive mesothelial cells from malignant epithelial cells. Beyond morphology, immunohistochemical staining has been used to help distinguish these two cell types. The most commonly used markers include keratin, CD15 (LeuM1),1 carcinoembryonic antigen,2, 3 Ber-EP4,4–6 B72.3,6, 7 and E-cadherin.8, 9 Logistic regression analysis has been used to try to select the best panel of markers.10
Sometimes, the problem is not distinction of cell type, but rather detection of any cells at all. The problem of detection of rare cancer cells has been addressed in the blood11, 12 and bone marrow.13, 14 In blood and bone marrow, studies have investigated the use of several methods for detecting occult metastases including reverse transcription–polymerase chain reaction (PCR),12, 15 flow cytometry,16 and the use of immunomagnetic beads.11, 15, 17 A common method using immunomagnetic beads is the Dynal CELLection system (Dynal AS, Oslo, Norway).18 Synthetic beads (4.5 μm dimension) are linked with anti-EpCAM (Ber-EP4) monoclonal antibody via a DNA oligonucleotide. The DNA linker provides a cleavable site for cell detachment after magnetic enrichment. Samples are treated with beads and placed in a magnet and the supernatant is discarded, leaving only the rosetted cells of interest in the tube. The beads then are cleaved from the cell by treatment with DNase (Fig. 1). The isolated cells are intact, viable, and available for morphologic or immunohistochemical evaluation.
The goal of this study was to assess the use magnetic beads to isolate cells of epithelial origin to find rare metastatic cells in ascites fluids. We show that this technique can detect rare cells that sometimes would be missed in a routine cytologic examination and at the same time help to distinguish benign mesothelial cells from carcinoma cells in fluids. Because clinical staging often is changed when carcinoma is found in fluid specimens, we postulate that routine usage of this technique might lead to more accurate staging.
Fluid specimens submitted to our cytopathology service between March 1, 2000 and July 7, 2000 was used in this study. Only cases with fluid volumes greater than 100 mL were used to have enough for uncompromised routine processing (50 mL) and an equal volume used in parallel for the enrichment test. The cases obtained included 44 pleural fluids, 14 peritoneal fluids, and 1 pericardial fluid for a total of 59 cases. All fluids were submitted for routine cytologic examination in the usual manner. Specimens were spun down, and the supernatant was discarded. The pellet was resuspended in CytoLyt (Cytyc Corp., Boxborough, MA) and centrifuged. The supernatant was discarded again and two drops of the pellet were added to PreservCyt (Cytyc Corp.) solution. After 15 minutes, the specimen was processed using the Cytyc ThinPrep Processor (Cytyc Corp). Finally, the specimens were stained using the standard Papanicolaou staining technique for routine screening by a cytotechnologist and sign-out by the pathologist. The cases were screened by one of six cytotechs and signed out by one of three pathologists, all of whom are board certified in cytopathology. The standard cytomorphologic criteria were used to define benign and malignant specimens and cases were called atypical when the signout pathologist was uncomfortable providing a definitive diagnosis. Acellular specimens were not included in the study.
To screen and enrich for epithelial cells, we used 50 mL of fluid for each case. On ice, 125 μL of CELLection Epithelial Enrich (Dynal AS, Oslo, Norway) immunomagnetic beads were added to the fluid specimen. The specimen was incubated at 4 °C for 1.5 hours while rotating on a Labquake (PGC Scientifics, Frederick, MD) rotisserie. The fluid then was placed in a Dynal MPC-1 (Dynal AS, Oslo, Norway) magnet at 4 °C for 30 minutes. The fluid was removed by careful pipetting. The tube then was removed from the magnet; the beads were resuspended in 5 mL phosphate-buffered saline (PBS)/0.1% bovine serum albumin (BSA) and transferred to a 6-mL tube. The 6-mL tube was placed in a Dynal MPC-6 magnet on ice for 5 minutes. Again, the fluid was discarded by careful pipetting, and then the tube was removed from the magnet. The beads were resuspended in 1 mL PBS/0.1% BSA and transferred to an Eppendorf tube. After placing the Eppendorf tube in the Dynal MPC-6 magnet for 5 minutes the supernatant was removed, and the beads were resuspended in 1 mL PBS/0.1% BSA. The washing procedure was repeated twice more in the Eppendorf tube before resuspending the beads in 200 μL RPMI-1640/1% FCS preheated to 37 °C. The cells were released from the beads by adding 4 μL releasing buffer and incubating for 15 minutes on a mixing device at room temperature. The mixture was vigorously pipetted and placed in the Dynal MPC-6 magnet for 5 minutes. The media containing the released cells was pipetted away and added to PreservCyt solution. The remaining cells were washed from the bead fraction by adding 200 μL RPMI-1640/1% FCS to the beads. The mixture was again vigorously pipetted and placed in the magnet for 5 minutes. The final 200 μL was removed from the tube and added to the PreservCyt solution. If possible, three slides were made for each case by using the Cytyc ThinPrep Processor. One slide was stained using the standard Papanicolaou staining technique whereas the other two slides were used for E-cadherin staining.
To confirm the cells were of epithelial lineage, we evaluated the remaining two ThinPrep slides for E-cadherin expression as previously described.8 One slide was treated with E-cadherin antibody whereas the other slide was used as a negative control. ThinPrep slides were rinsed for 5 minutes in tap water followed by 5 minutes in Tris-buffered saline (TBS; 150 mM NaCl, 20 mM Tris, pH 8.0). Slides then were blocked for 20 minutes with diluted normal serum from the Vectastain ABC-AP Kit (Vector Laboratories, Burlingame, CA). After blocking, slides were washed once in 1 × TBS. E-cadherin monoclonal antibody (Transduction Laboratories, Lexington, KY) was diluted 1:250 in 1 × TBS, 3–4 drops applied per slide, and then incubated for 30 minutes. The slides then were washed again in 1 × TBS for 5 minutes. Then 3–4 drops of diluted biotinylated antibody from the Vectastain ABC-AP Kit were applied per slide and incubated for 30 minutes, followed by a 5-minute wash in 1 × TBS. Vectastain ABC-AP reagent as added (3–4 drops), and the slides were incubated for an additional 30 minutes followed by a 5-minute wash in 1 × TBS. Vector Red Alkaline Phosphatase Substrate (Vector Laboratories) was applied (3–4 drops), and slides were incubated for 10 minutes. The slides then were washed for 5 minutes in tap water, counterstained with hematoxylin, and coverslipped. Negative control specimens were processed in parallel using the same method, but no E-cadherin antibody was added.
Serial cases of ascites or pleural fluids were collected and analyzed using the bead enrichment assay. Table 1 shows the results of each case and allows a comparison of the results of the cytology and the magnetic enrichment assay. Figure 2 shows example of cells from fluid specimens before and after the enrichment technique, as well as an example of the confirmatory E-cadherin stain performed on one case. The cytologic diagnoses were as issued by the attending pathologist, grouped as negative (or reactive), positive, or atypical. The atypical diagnosis occurs when, in the opinion of the pathologist of record, there was insufficient evidence to make a definitive diagnosis. Of the 59 fluid specimens examined, 12 were positive, 42 were negative, and five were atypical.
|Case no.||Site||Cytologic diagnosis||Diagnosis after enrichment||E-cadherin||Surgical/history|
|11||Pleural||Negative||Negative||NA||History of squamous cell carcinoma|
|17||Pleural||Atypical||Negative||NA||History of Ewing sarcoma|
|26||Pleural||Negative||Negative||−||History of colonic adenocarcinoma|
|41||Peritoneal||Negative||Negative||NA||History of breast adenocarcinoma|
|42||Pleural||Atypical||Negative||NA||Mucoepidermoid carcinoma pleural nodule|
|49||Peritoneal||Negative||Negative||−||Adenocarcinoma in liver (?metastatic)|
|53||Pleural||Negative||Negative||NA||Squamous cell carcinoma|
|Cytologic diagnosis||Cytology cases||Malignant diagnosis after enrichment||E-cadherin positive|
|Positive for malignancy||12||11||10|
|Negative for malignancy||42||3||2a|
All cytologically “positive” or malignant cases (12) but one showed strong tumor cell enrichment. This case (Case 23) did not yield any cells after treatment. Confirmation of epithelial status with E-cadherin staining showed 10 of the 11 enriched malignant cases with strong membranous staining. Forty-two cases were signed-out as negative for malignancy. After enrichment, three cases showed cells that were morphologically malignant. Only one of the three cases contained a sufficient amount of cells for E-cadherin evaluation. This case stained positive for E-cadherin. Similarly, two of the five atypical cases revealed morphologically malignant cells. However, neither malignant case yielded enough cells to evaluate E-cadherin expression. These results are summarized in Table 2.
Although the entire slide is difficult to illustrate, the cells seen on the enriched slides appeared as shown in Figure 2. Malignant cases had minimal or absent cellular background, and the slides of negative cases often showed no cells whatsoever. All cells that were recovered, tumor or otherwise, were intact, unaltered, and immediately ready for staining. The cells in Figure 2 were stained with Papanicolaou stain and immunostained for detection of E-cadherin expression.
There were 10 cases in which the diagnosis, after enrichment, differed from the cytology diagnosis (Table 3). Four cases were signed-out as “negative” by cytology. The magnetic cell enrichment technique captured several clusters of malignant cells in three cases and cells that looked like benign endometrial cells in the fourth case. Follow-up on that patient showed a clinical picture consistent with endometriosis. Of the other three negative cases that showed malignant cells by epithelial cell enrichment, two showed adenocarcinoma on subsequent specimens. The third case represents an 83-year-old patient admitted with nonresolving pneumonia and a pleural effusion, tapped to rule out malignancy. No definitive diagnosis was made, but the patient ultimately declined further evaluation and was placed in a nursing home. No final determination of outcome is available on this patient. Five cases were classified as “atypical” by morphologic criteria by the cytopathologist. Malignant cells were recovered after CELLection epithelial cell enrichment in two cases. One of the cases proved positive for adenocarcinoma on subsequent specimens whereas the other was positive for mesothelioma. Finally, there was one discrepant case that was cytologically malignant, in which no tumor cells were detected by magnetic enrichment.
|Case no.||Cytology diagnosis||Diagnosis after enrichment||Follow-up/history/explanation|
|2||Negative||Positive/endometriosis||History of pelvic pain Most likely endometriosis|
|17||Atypical||Negative||History of Ewing sarcoma|
|42||Atypical||Negative||History of mucoepidermoid carcinoma|
The standard in this study for assessment of malignancy in the cells harvested from the enrichment assay was cytomorphology. However, on some occasions there were sufficient cells harvested by magnetic enrichment that more than one slide could be made. In these cases, E-cadherin staining was used to verify the epithelial lineage of the cells. All cases that stained positively for E-cadherin had been judged to be malignant, except for a single case in which cells were judged as endometrial cells, which also express E-cadherin. Mesothelial cells do not express E-cadherin and the rare clusters of cells that were found in magnetically enriched samples were identifiable both morphologically and by the absence of E-cadherin staining.
Antigen-mediated enrichment of epithelial cells in fluid specimens is sufficiently sensitive to raise the possibility of routine use of this test. We found that in 8.5% of the negative or atypical cases, tumor cells were found that were later proven by subsequent studies. Thus, if the enrichment technique were used on these patients, they would have been more accurately staged by the initial procedures and may have been spared subsequent diagnostic tests or procedures. An obvious concern would be the issue of false-positive tests. In this study, there were two cases that could be interpreted as false-positives. On closer examination, one case showed epithelial cells that were morphologically identifiable as endometrial cells. Although this is not likely to become a primary method of diagnosing endometriosis, the finding of endometrial cells in this case does not represent a false-positive diagnosis because epithelial cells were found but not attributed to malignancy. The second case also appears not to represent a false-positive, but rather a lack of information. That case, as discussed above, was clinically suspicious for malignancy, but not ultimately proven because the patient requested termination of medical intervention. Other than those cases, all cases with epithelial cells recovered represent true cases of malignant cells in fluid specimens.
One case showed a false-negative result. Case 23 obviously was malignant by conventional cytopathology but showed no cells on the enrichment assay. This could have several causes, but perhaps the most likely is that the tumor cells did not express EpCAM. Estimates in the literature suggest that 90–95% of epithelial neoplasms express this marker.5, 19–21 Thus, our results, in which 1 case in 11 showed no cancer cells retrieved by anti-EPCAM antibodies would be consistent with the literature and suggests that this particular case (Case 23) simply did not express sufficient levels of EpCAM.
A second issue is the range of malignancy detected. The antibody on the beads in designed for detection of epithelial cells. In one case, the test failed to detect epithelial cells when they were present and in another cells detected were not epithelial. In that case, a malignant mesothelioma, cells must have expressed sufficient EpCAM for detection. Although EpCAM (Ber-EP4) is predominantly expressed in epithelial cells, previous studies have shown infrequent expression in mesotheliomas.5, 20, 21 Finally, in one case the patient had a history of Ewing sarcoma. These cells would not be expected to be found by this assay, and they were not.
The concept of identification of rare malignant cells in bodily fluid specimens to predict patient outcome or more accurately assess stage is not new. Malignant cells in blood, bone marrow, and lymph nodes have been sought by a range of methods.11, 15, 17, 22–25 The typical study has been to search for rare events, and frequently, cells found are proven to represent neoplasms by PCR-based methods.15, 24, 25 The reason for this is that cells are present in a dilution of between 1:100,000 to 1:1 million. Most likely, PCR-based assays would be more sensitive than this magnetic enrichment assay. However, in pleural and ascites fluids, the target cell numbers are substantially higher. In this study, we have not attempted to estimate the sensitivity of detection of malignant cells in fluid, but rather addressed the primary question “can cells be found at all?” The high frequency at which cells were found, even in “atypical” or “negative” cases suggests that the numbers of cancer cells must be relatively high compared to cancer cells in bone marrow or circulating blood. The relatively high number of cells in fluids also allowed simple morphologic or immunologic identification for confirmation of malignancy. Future studies may determine exactly what the ratio of malignant to nonmalignant cells is, although in fluid specimens, it probably will be highly variable. Theoretic sensitivities should approach or exceed that observed in blood (1 cell in 1 million nucleated cells), but this may never be relevant in fluid specimens.
Like PCR-based methods, this enrichment method adds time and costs to the process of analyzing a fluid specimen. The method used for this report is that recommended for use for purification of cells from peripheral blood. We strictly followed both incubation and wash times recommended in the Dynal CELLection protocol. The method takes approximately 4 hours (although much of that time is hands off, incubation time). We anticipate development of a more streamlined method for routine fluid specimens. Aside from increased technician time, the costs associated with the materials are minimal for this assay. Only a tiny amount of antibody-labeled beads are necessary for each assay. Furthermore, we anticipate usage of this assay in a manner similar to that for immunostaining or other ancillary tests in which the test is only performed in a fraction of difficult cases.
Although the above paragraphs have examined the issues of false-positive and false-negative tests no attempts were made to calculate sensitivity, specificity, and positive or negative predictive value. Although it may be valuable to determine these numbers, the relatively small sample size presented here, as well as the experimental nature of these studies, suggests that these numbers would have little validity. Specifically, in this study we used cases with variable amounts of starting material and cases were omitted if there was insufficient volume in the originally submitted specimen. Thus, this collection strategy makes calculation of sensitivity and specificity invalid. In the future, it is hoped we will prospectively accrue a large cohort of cases that would be treated in a prescribed and uniform manner and be accompanied by longer-term follow-up. At that time, the predictive value of the test may be more accurately assessed.
Finally, we believe this test has the potential to more accurately stage patients with effusions. Frequently, cancer is found easily in these specimens, but difficult cases are not uncommon. In these cases, better methods are needed. There also may be applications for this test in cases in which a tissue diagnosis is critical before adjuvant therapy. Perhaps the most common example is ovarian carcinoma.26 Protocols often require that a tissue diagnosis be made before commencement of chemotherapy; however, even in patients with clinically apparent ascites, cancer cells are not always found. The ability to use some other modality when traditional examination of the fluid specimens has failed is potentially very attractive. We believe this pilot study provides justification for larger future studies to validate this technology.