Simplified Flow Cytometric Immunophenotyping Procedure for the Diagnosis of Effusions Caused by Epithelial Malignancies

Abstract

Figure 1

Open in new tabDownload slide

Distribution of effusions. A, Causes of the benign serous effusions. B, Organ of origin for the malignant effusions. CHF, congestive heart failure.

Materials and Methods

Pleural, peritoneal, and pericardial cavities are commonly the site of metastasis of malignant tumors. In fact, in many patients, they are the first presentation of malignancy. Cytomorphology is the diagnostic gold standard for the diagnosis of a malignant effusion. Cytology, however, has a moderate false-negative rate attributable to screening, interpretation, and sampling errors.¹ The specificity of cytomorphology is very good and ranges from 91% to 100%, with an average of 97.0%. Sensitivity ranges from 22% to 81%, with an average of 58.2%. Multiple specimens have been shown to increase sensitivity.¹ Morphologic distinction between reactive mesothelial cells, macrophages, and malignant cells can be difficult, and hypocellular and bloody specimens are particularly challenging. Immunohistochemistry, often performed on paraffin-embedded cell block sections, is a common adjunct technique that serves to improve the sensitivity and specificity of cytology diagnosis. However, the use of immunohistochemistry is limited by the number of cells present in the cell block and the vagaries of immunostaining, which can result in delays in diagnosis. Hence, there is an unmet need for an adjunct technique that can increase the sensitivity and specificity.

Flow cytometric immunophenotyping is an integral part of diagnosing hematologic malignancies.² It is possible to render a diagnosis on the same day of receipt of sample. Along with cytogenetics and molecular studies, it is an important ancillary test to H&E-stained examination as outlined in the World Health Organization classification of hematologic neoplasms.³ Recent advances in flow cytometry have made it possible to simultaneously measure 19 different parameters on a single cell at the rate of 10,000 cells per second.⁴ In spite of its obvious advantages, the use of flow cytometry for the diagnosis of nonhematologic malignancies is rare. It is widely thought that size and cohesiveness of malignant epithelial cells would limit its use in flow cytometry. However, many groups have successfully used flow cytometry in malignant effusions in which the epithelial cells are already suspended as single cells.^5–9 A flow cytometric method has the advantage of simultaneously staining for other surface and intracellular markers.

Ber-EP4 (epithelial cell adhesion molecule) is a cell-cell adhesion molecule expressed on all epithelial cells but not on mesothelial cells or hematopoietic cells.¹⁰ Ber-EP4 is also expressed on all epithelial malignancies.¹¹ Epithelial cells are not a normal component of serous cavity fluids; hence, their mere presence in an effusion can be used to indicate a malignancy. Immunophenotyping with the single epithelial marker Ber-EP4, in fact, has high sensitivity and specificity for the detection of malignant epithelial cells.¹²

We have developed and validated a simple, rapid, 3-color flow cytometric panel using Ber-EP4 to detect epithelial cells in effusions. This assay can be routinely used in conjunction with cytology for the diagnosis of malignant effusions.

Results

Sample Preparation

One hundred ninety-five fresh benign and malignant effusions (pleural, pericardial, and peritoneal) received for routine cytologic examination were concurrently analyzed with flow cytometry. Aliquots of 50 mL of effusion fluid were centrifuged at 1,400 rpm for 10 minutes. The supernatant, large clumps, and any grossly visible mucoid material were discarded. The cell pellet was washed and resuspended in RPMI. The cell suspension was filtered with a 50-μm mesh filter to further remove any remaining dead cell clumps. An automated nucleated cell counter (Advia 2120, Siemens Healthcare Diagnostics, Tarrytown, NY) was used to count nucleated cells in the sample. An aliquot of 2 million cells in 100 μL of resuspended phosphate-buffered saline solution was used for immunophenotypic analysis.

Flow Cytometric Immunophenotyping

The samples were immunostained with CD45 peridinin chlorophyll protein complex, CD14 fluorescein isothiocyanate, and Ber-EP4 phycoerythrin or allophycocyanin (all from BD Biosciences, San Jose, CA) for 30 minutes at 4°C. BD FACS lyse buffer (BD Biosciences) was added for 15 minutes after staining to lyse RBCs. A total of 500,000 events were collected for analysis on a 2-laser, 6-color BD FACS Canto device using BD FACS Diva software (both from BD Biosciences). The data were exported as FCS 3.0 files and analyzed using Flowjo (Tree Star, Ashland, OR) analysis software.

A 50-mL aliquot of the fresh specimen was centrifuged at 1,800 rpm for 10 minutes. The resulting cell pellet was resuspended in 15 to 30 mL of CytoLyt (Hologic, Marlborough, MA) and centrifuged again at 1,800 rpm for 10 minutes. The pellet was divided for the preparation of 1 Papanicolaou-stained ThinPrep slide and 2 H&E-stained cell block sections. For the ThinPrep slide, 2 to 4 drops of the sediment were pipetted into a vial containing PreservCyt (Hologic) and incubated at room temperature for at least 20 minutes. A ThinPrep slide was prepared using the automated ThinPrep 2000 or ThinPrep 5000 instruments (Hologic). The residual cell pellet was prepared as a cell block by adding 1 drop each of plasma and a thrombin solution and stirring with a wooden applicator until a clot formed. The clot was wrapped in filter paper, placed in a cassette, fixed in formalin, and embedded in paraffin as a paraffin block for microtome sectioning per standard procedures. Immunohistochemical studies were performed on selected cases (37%) at the discretion of the reviewing cytopathologist, and results were reported as “no malignant cells identified,” “atypical,” “suspicious cells present,” or “positive for malignant cells.” The clinical results of the cytomorphologic examination were used for calculating sensitivity and specificity.

Image 3

Open in new tabDownload slide

Correlation among flow cytometry dot plots, cytology cell block preparation, and Ber-EP4 immunohistochemistry (IHC). A, Malignant peritoneal effusion from a papillary serous adenocarcinoma primary. B, Malignant pleural effusion due to lung adenocarcinoma. APC, allophycocyanin; FSC-A, forward scatter; PerCP, peridinin chlorophyll protein complex; SSC-A, side scatter.

Correlation among flow cytometry dot plots, cytology cell block preparation, and Ber-EP4 immunohistochemistry (IHC). A, Malignant peritoneal effusion from a papillary serous adenocarcinoma primary. B, Malignant pleural effusion due to lung adenocarcinoma. APC, allophycocyanin; FSC-A, forward scatter; PerCP, peridinin chlorophyll protein complex; SSC-A, side scatter.

Cytologic Preparation and Interpretation

Discussion

Follow-up Criteria

A total of 195 effusions were analyzed with flow cytometry. Based on follow-up data, 83 fluid specimens were judged to be benign and 76 malignant. Benign effusions Figure 1A were due to cirrhosis, congestive heart failure, acute renal failure, end-stage renal disease, hypoalbuminemia, surgery, chemotherapy, infections, and trauma. In some cases, the cause of the benign effusions could not be conclusively determined (“miscellaneous”). Malignant effusions Figure 1B were due to breast cancer (ductal and lobular), lung cancer (adenocarcinoma, squamous cell carcinoma, and mesothelioma), ovarian cancer (papillary serous), uterine cancer (endometrial carcinoma, carcinosarcoma, and adenocarcinoma), colon cancer (adenocarcinoma), gastric cancer (adenocarcinoma), renal cancer (renal cell carcinoma), bladder primary (transitional cell carcinoma), and hematologic malignancy (lymphoma). For some cases, a definite primary cause could not be determined from pathologic or clinical data (“miscellaneous”). The distribution of benign and malignant effusions is consistent with that observed in a tertiary care center.¹

An unbiased initial gate is needed to capture malignant epithelial cells because they can widely vary in size and cytoplasmic content. Hence, all collected events were broadly gated in the forward vs side scatterplot, excluding only the dead cell debris and air bubbles at the edges Image 1A (gate 1). Singlet events that lie in the diagonal of the forward scatter area vs forward scatter height plot were then gated (Image 1A, gate 2).¹³ This gate excludes small clusters of malignant epithelial cells that can confound analysis of signals from single cells in other channels. Cells of hematopoietic lineage, including lymphocytes, monocytes, and macrophages/histiocytes, all of which express CD45, were gated out in the CD45 vs CD14 plot (Image 1A, gate 3). The only cells remaining in the analysis represent mesothelial cells and Ber-EP4–positive epithelial cells (if present). The proportion of Ber-EP4–positive cells was then calculated as a percentage of the total CD45 and CD14 double-negative population (Image 1A, gate 4). This value was used for further analyses. The presence of more than 0.5% CD45-negative, CD14-negative, Ber-EP4–positive cells was determined to indicate a malignant effusion. This negative cutoff for Ber-EP4 was based on fluorescence minus one controls and background staining in benign effusion samples. Back-gating analysis confirmed that Ber-EP4–positive epithelial cells vary in size and cytoplasmic content and are widely distributed in the forward vs size scatterplot Image 1B.

A wide variety of epithelial carcinomas were detected with Ber-EP4 flow immunophenotyping Image 2. Ber-EP4 expression on flow cytometry and immunohistochemistry on the cytology cell block preparation showed good correlation despite the additional filtering step for flow cytometry preparation that removes large clumps to prevent clogging of the flow cytometer. Data from a malignant peritoneal effusion from a primary ovarian papillary serous adenocarcinoma are shown Image 3A. Flow cytometry showed that 10.5% of the CD45- and CD14-negative population and 2.62% of total cells in the effusion expressed Ber-EP4. A similar proportion of cells were positive on Ber-EP4 immunohistochemistry. The other CD45- and CD14-negative cells are presumably mesothelial cells. Data from a malignant pleural effusion caused by an adenocarcinoma are shown Image 3B. Ber-EP4 expression was seen in 0.99% of the CD45- and CD14-negative population and 0.21% of the total cells in the effusion. A similar proportion of cells were positive on Ber-EP4 immunohistochemistry.

Significant numbers of CD45-negative, CD14-negative, Ber-EP4–positive cells were detected with flow cytometry in 89.3% of malignant effusions Figure 2A. The sensitivity and specificity of flow cytometry was 88.15% and 97.64% compared with 73.68% and 100% with cytology for the presence of epithelial cells in a malignant effusion. Flow cytometry had a positive predictive value of 97.1% and a negative predictive value of 90.2% compared with cytology, which had a positive predictive value of 100% and a negative predictive value of 80.58%. The difference between flow cytometry and cytology for a “positive for malignancy” diagnosis was statistically significant (Fisher exact t test; P = .0378). Benign effusions had very few CD45-negative, CD14-negative, Ber-EP4–positive cells whereas malignant effusions showed significantly larger numbers of CD45-negative, CD14-negative, Ber-EP4–positive cells (Figure 2A). The proportion of CD45-negative, CD14-negative, Ber-EP4–positive cells in benign effusions was less than 0.5% (background staining). An effusion was considered malignant if it contained greater than 0.5% of CD45-negative, CD14-negative, Ber-EP4–positive cells. However, the qualitative nature of the Ber-EP4–positive population (presence of a cluster and intensity of signal) was also taken into consideration for the decision in certain unusual cases. If a specimen contains a large number of mesothelial cells relative to malignant epithelial cells, then the CD45-negative, CD14-negative, Ber-EP4–positive proportion could be lower than 0.5%. Data from a recurrent malignant pleural effusion in a patient with a history of ductal carcinoma of the breast are shown Figure 2C. The proportion of CD45-negative, CD14-negative, Ber-EP4–positive cells is only 0.42%, but the presence of many cells that occur in clusters with intermediate to high Ber-EP4 staining indicates a true malignant effusion. Conversely, in paucicellular specimens, nonspecific staining can artifactually increase the proportion of Ber-EP4–positive cells. An example of a benign effusion with a Ber-EP4–positive proportion greater than 0.5% are shown Figure 2D, in which most of the Ber-EP4–positive staining is of low intensity, which exclusively lies just above the negative control line and hence is not indicative of true Ber-EP4 expression. In addition, as is standard in flow cytometry, events that occur along the diagonal axes in every dot plot are likely dead cells and should be excluded. Hence, the presence of malignant epithelial cells in an effusion cannot be determined on purely quantitative criteria but should always be made in conjunction with qualitative criteria.

In 11 cases, flow cytometry identified significant numbers of malignant epithelial cells, whereas the cytology findings were not considered positive for malignancy. In these cases, a different (prior or subsequent) cytology specimen or subsequent surgical biopsy specimen was positive for malignancy. In such cases, a review of the cytology slides and a Ber-EP4 immunostain were performed. In some cases (n = 6), epithelial cells were present but overlooked because they were either rare occurrences in hypocellular cytologic preparations or because there were too many obscuring RBCs, mesothelial cells, or inflammatory cells. Image 4A illustrates a pleural effusion in a patient with primary lung adenocarcinoma. The cytologic preparations were hypocellular and bloody and interpreted as negative for malignancy. Flow cytometry showed significant numbers of CD45-negative, CD14-negative, Ber-EP4–positive cells. A Ber-EP4 immunostain performed subsequently on the cell block preparation showed rare Ber-EP4–positive malignant epithelial cells. Radiologic evidence showed pleural involvement by malignancy, and a subsequent pleural biopsy specimen was also positive for adenocarcinoma, proving that effusion was most likely a true malignant effusion. Image 4B illustrates another such case—a new peritoneal effusion sampled in a patient with a uterine mass. Cytologic preparations were hypocellular and interpreted as negative for malignancy. Flow cytometry showed significant numbers of Ber-EP4–positive cells. Ber-EP4 and PAX8 (not shown) immunostains highlighted rare epithelial cells of müllerian origin. A subsequent laparotomy revealed widespread peritoneal metastases, and a peritoneal biopsy confirmed involvement by endometrial adenocarcinoma. In these cases, concurrent flow cytometric analysis increases the detection of malignant epithelial cells in the clinical sample because they were missed on cytologic examination alone.

Five of the cytologic false-negative cases were reported as atypical (n = 1) or suspicious (n = 4) because the specimens showed rare, poorly preserved atypical cells. In these cases, flow cytometric analysis detected significant numbers of epithelial cells (>0.5%). Here again, concurrent flow cytometric analysis increases the detection of malignant epithelial cells in the clinical sample and could enable a cytology result of atypical or suspicious to be reclassified as positive.

Two cases were falsely positive on flow cytometry. Both cases showed a borderline proportion (around 0.5%) and low-level expression of CD45-negative, CD14-negative, Ber-EP4–positive cells. Cytology was suspicious in one of these cases and negative in the other. Although cytology had no false-positive cases, 3 cases that were called atypical/suspicious were negative using gold standard criteria. Flow cytometry was clearly negative in 2 of those cases. It is possible that the interpretation of flow cytometry results in conjunction with cytology can eliminate potential false positives. Finally, 8 effusions were negative on both flow cytometry and cytology for malignant cells but were considered malignant effusions using the gold standard follow-up criteria. Some of these cases could represent minimal/focal involvement of the serosal membrane by the tumor. It is also possible in at least some of these cases that the effusion was due to indirect effects of a tumor, such as lymphatic blockage or irritation of serosal surfaces.

Exclusion Criteria

We have demonstrated that simple, rapid, 3-color flow cytometric immunophenotyping using Ber-EP4 can be used to detect malignant effusions. Because Ber-EP4–positive cells do not naturally occur in effusions, their detection with flow cytometry can be a useful adjunct to cytology for the diagnosis of a malignant effusion when sufficient numbers of Ber-EP4–positive cells with high levels of expression are detected. The sensitivity and specificity were 88.15% and 97.64%, respectively, with flow cytometry compared with 73.68% and 100%, respectively, with cytology for the presence of epithelial cells in a malignant effusion. We showed that flow cytometry is particularly useful if the cytologic diagnosis is atypical/suspicious or if the cytologic preparations are hypocellular and hemorrhagic. We found 11 cases in which flow cytometry identified significant numbers of malignant epithelial cells but the cytologic results were falsely negative. Of these 11 cases, 5 were called atypical (n = 1) or suspicious (n = 4) and 6 were called negative. Although the clinical usefulness of flow cytometry in those 6 cases is clear, we believe that flow cytometry is clinically useful even in the 5 atypical/suspicious cases because it detected significant numbers of epithelial cells in all those cases. One explanation is that cytologic preparations are limited by the number of cells on the slide, whereas flow cytometric immunophenotypic analysis takes into account all cells in the sample. Flow cytometry is inherently quantitative, and the number of Ber-EP4–positive cells detected can be used as additional information to improve diagnostic accuracy. This information can potentially decrease the number of atypical/suspicious diagnoses rendered because of paucity of cells.

Statistical and Sensitivity/Specificity Calculations

We considered atypical/suspicious cases seen on cytology as negative for statistical calculations because they are considered to be negative by clinicians for practical decision making involving therapeutic interventions. If the atypical/suspicious cases on cytology are considered to be positive, the sensitivity of cytology would increase but its specificity would decrease. In many of the atypical/suspicious cases seen on cytology, flow cytometry was clearly negative or clearly positive and hence would add clinical value if concurrently performed.

Discordant results between cytology and flow cytometry. A, Cytologic preparations and flow cytometric data from a pleural effusion in a patient with a lung adenocarcinoma. B, Cytologic preparations and flow cytometric data from a peritoneal effusion in a patient with a uterine mass. APC, allophycocyanin; IHC, immunohistochemistry; PerCP, peridinin chlorophyll protein complex.

We used a novel gating strategy to accurately identify epithelial cells in contrast to previous studies.^6,8,12 Back-gating of Ber-EP4–positive cells revealed that many malignant cells were in the lower ends of the axes near the region usually occupied by cell debris. This region is gated out during routine flow cytometric analysis. Malignant epithelial cells are fragile and fragment during the sample preparation. Hence, it is important to use a broad initial gate to include even fragmented cells to improve sensitivity. These cells do not nonspecifically pick up other antibodies and do not appear in the diagonal region of flow cytometry plots, thus suggesting that they are not nonspecific debris or dead cells. However, preparing cells for immunophenotyping on ice with gentle manipulation might keep the malignant cells intact and reduce the presence of such events. A forward scatter area vs height plot is then used to gate on singlet events.¹³ The gate is important to exclude multiple signals that can potentially arise from cell aggregates in other channels. Cells of hematopoietic lineage were excluded by gating CD45- and CD14-positive events. This gate isolated mesothelial cells and epithelial cells for further analysis. It is interesting to note that CD45 and Ber-EP4 double-positive events often occur. This phenomenon has been seen and described in the circulating tumor cells.¹⁴ These are probably Ber-EP4–positive cell fragments that adhere to CD45-expressing cells. Our gating strategy excludes such events by gating out all CD45-positive cells.

In contrast to previous studies that relied on purely quantitative criteria, we used a combination of quantitative and qualitative criteria to infer the presence of a malignant effusion.^6,7,12 As a first step, we used a 0.5% level of Ber-EP4 expression to separate benign from malignant effusions. However, the level of Ber-EP4 expression that distinguishes a benign from a malignant effusion depends on a number of factors such as the antibody conjugate used, flow cytometric panel used, number of cells collected, gating strategies, and instrument settings. Hence, such cutoffs should ideally be standardized by every flow cytometry laboratory. In spite of standardization, it can sometimes be truly difficult to differentiate background Ber-EP4 staining from true low-level Ber-EP4 expression in a tumor or presence of benign epithelial cells. In those cases, correlation with cytology findings and clinical data will help arrive at the correct diagnosis.

CME/SAM

Benign and Malignant Effusions

One caveat of our study is the reliance on Ber-EP4 expression as a marker for all epithelial malignancies. It is possible that some poorly differentiated epithelial tumors downregulate Ber-EP4 expression. Additional markers for epithelial cells might be helpful in such situations.¹⁵ Rarely, Ber-EP4 expression is also possible in atypical/reactive mesothelial cells. Markers of mesothelial cells, such as WT-1 or calretinin, can be used to identify and exclude such cells from the analyses. However, antibody conjugates of epithelial and mesothelial markers are not commercially available and would need to be made by the flow cytometry laboratory.

Flow Cytometric Gating Strategy

It is important to note that flow cytometric immunophenotyping in its current form detects the presence of epithelial cells in the effusion and cannot conclusively determine whether the cells are benign or malignant. Cytomorphology examination is the current standard for that determination. A proliferation marker such as Ki-67 or DNA content analysis for aneuploidy can be used in flow cytometry to differentiate a benign epithelial cell from a malignant cell^7,16 and could be combined with the current approach. Reactive mesothelial cells can show a high proliferation index or aneuploidy, but they should be gated out using mesothelium-specific markers. Using DNA content in isolation without using lineage-specific markers did not improve specificity.^7,8,12

Another interesting observation is the frequent presence of Ber-EP4–positive cells in peritoneal washings. The likely origin for such cells is exfoliation of ovarian, tubal, or uterine epithelium. It is conceivable that epithelial proliferations like endometriosis that involve the peritoneal cavity can be a source of benign epithelial cells in peritoneal effusions. Recent surgery in the peritoneal or pleural cavity can also theoretically result in the presence of benign epithelial cells in effusions. However, such cells have low-level expression of Ber-EP4 and occur in much lower numbers compared with malignant epithelial cells in effusions. Previous studies have included peritoneal washings in their analysis, which was probably responsible for their higher false-positive rate.^6,9

A major advantage of flow cytometric immunophenotyping is that panels can be scaled up to gather additional data as needed. Additional markers in a flow panel can also determine the tissue of origin of the neoplasm. In this era of personalized medicine, additional patient-specific data can be determined with flow immunophenotyping of effusions. Hormonal status (estrogen and progesterone receptors) and Her2 status in metastatic breast carcinomas^17,18 or BRAF mutation status (through use of mutation-specific antibodies) in metastatic lung carcinomas can be determined in effusions.^19–22 Highly purified tumor cells can also be isolated by fluorescence-activated cell sorting for subsequent use in molecular diagnostics. Similar strategies can also be used to detect and phenotype malignant epithelial cells in lymph nodes, peripheral blood, and bone marrow specimens.^23,24

Ber-EP4 as a Marker of Epithelial Carcinomas

References