An improved flow cytometric assay for detection and discrimination between malignant cells and atypical mesothelial cells, in serous cavity effusions


  • How to cite this article: Kentrou NA, Tsagarakis NJ, Tzanetou K, Damala M, Papadimitriou KA, Skoumi D, Stratigaki A, Anagnostopoulos NI, Malamou-Lada E, Athanassiadou P, Paterakis G. An improved flow cytometric assay for detection and discrimination between malignant cells and atypical mesothelial cells, in serous cavity effusions. Cytometry Part B 2011; 80B: 324–334.



The aim of this study was to evaluate a flow cytometric assay for the detection of malignant effusions.


During the last 4-year period, 125 effusions suspicious for malignancy were prospectively analyzed by flow cytometry and conventional cytology. A three-step flow cytometric assay was performed, beginning with an initial informative panel of two protocols, containing SYTO-16, 7-AAD, CD71-PE, CD45-ECD, and CD66abce-FITC, CD64-PE, CD45-ECD, CD16-PECy5, CD14-PECy7, respectively. This was followed by a basic immunophenotypic panel of seven three-color combinations, containing in the first position, EMA, Ber-EP4, CD66abce, CD56, and intracellular desmin-33, combined with CD71-PE and CD45-PeCy5 in each tube. Finally, a cytokeratin-FITC/propidium iodide DNA panel was conducted, for the detection of aneuploidy in cytokeratin positive cells.


The sensitivity and specificity of flow cytometry were 85.1 and 97.8%, and of cytology 93.2 and 95.6%, respectively. A significant association was observed between the results of the two techniques (P < 0.001). Among eight atypical cases detected by cytology, five had been precisely characterized as malignant by flow cytometry. EMA and Ber-EP4 proved the most sensitive markers for malignancy diagnosis, while the detection of desmin-33 negative/cytokeratin positive cells had the simultaneous highest positive and negative predictive values. CD66abce was very specific, although nonsensitive, while DNA ploidy analysis was nonspecific, as hyperploidy was observed in reactive mesothelial cells.


A flow cytometric assay of high sensitivity and specificity is proposed for the routine identification of carcinoma cells in effusions and their distinction from atypical mesothelial cells, as an ancillary to conventional cytology. © 2011 International Clinical Cytometry Society

Although the main current application of flow cytometry (FCM) is in the diagnosis and prognosis of hematolymphoid neoplasia, it has also been extensively used for cell cycle and DNA ploidy investigation of malignant effusions. DNA aneuploidy and the increased percentage of cells in the S phase of the cell cycle (proliferation phase) were the most frequent observations in studies of malignant tumors (1–4). A wide variety of reports published on the determination of the sensitivity and specificity of FCM technology in the detection of malignancy, based on DNA ploidy investigation (5–7), document sensitivity of 55–100% and specificity 86–100%. The combination of detection of aneuploidy with immunophenotyping appears to be needed for definitive diagnosis(8, 9). The positive and negative predictive values of conventional cytology (CC) for the detection of malignancy have been calculated to be 89.3 and 64.9%, respectively (10). However, a “gray zone” sometimes occurs, where CC cannot distinguish between reactive, atypical cells, and malignant cells (11).

The purpose of this study was to evaluate the efficacy of FCM for the rapid identification of malignant effusions in routine clinical practice. The FCM assay comprised a three-step procedure: an initial informative antibody panel, a second panel for extended immunophenotyping of the suspicious cell population, and a DNA ploidy analysis to define the aneuploidy of malignant cells.



The study was conducted on effusions derived from 125 patients hospitalized in the Departments of Internal Medicine, Cardiology and Surgery, of whom 60 were males and 65 females, with a median age of 65 years (range, 15–91 years). Of the 125 patients, 74 had been previously, or were later, diagnosed with malignancy, based on histological diagnosis using hematoxylin and eosin (H and E) staining or on clinical follow-up, and specifically, 18 tumors of the gastrointestinal tract, 8 cases of breast cancer, 9 cases of lung cancer, 8 pancreatic tumors, 8 ovarian tumors, 6 cases of hepatocellular carcinoma, 8 cases of metastatic adenocarcinoma of unknown origin, 2 tumors of the uterus, 2 cases of melanoma, 2 of sarcoma, and 3 cases of disseminated tumor disease.


Pleural, peritoneal, and pericardial effusions were included in the study. Every effusion, which was morphologically considered suspicious for malignancy, or that was derived from a patient with known malignancy, was further processed by FCM and CC, and in total, 59 pleural, 58 peritoneal, and 8 pericardial samples of fluid were analyzed. Effusions derived from hematopoietic neoplasias were retrospectively excluded. The “gold standard” used to define a sample, as being infiltrated by malignant cells was a positive finding on at least two of the following examinations, in any combination: 1) CC, 2) Biopsy, and 3) imaging (computed tomography, ultrasound).


Giemsa morphology

Giemsa stained centrifuged slide preparations were studied. The morphological features taken into account for the characterization of an effusion as “suspicious for malignancy” were: cells in aggregation or clusters, large discrete isolated malignant cells, large cytoplasmic vacuoles, signet-ring cells, basophilic or eosinophilic cytoplasm, presence of mitoses, central or eccentric large nucleus, prominent nucleolus, and presence of nucleoli (1).


Centrifuged slide preparation was performed on each specimen, followed by Papanicolaou staining. Effusions were classified as: 1) malignant, 2) benign/reactive, and 3) atypical/suspicious for malignancy, according to previously described criteria (1).


An initial informative panel, a basic immunophenotype panel and a DNA-ploidy analysis were performed for every sample. The informative panel consisted of two protocols with 4- and 5-color combinations, respectively. In the first protocol, CD71-PE, CD45-ECD, and two fluorescent nucleic acid stains, SYTO-16 and 7-AAD, were used. SYTO-16 (Invitrogen, Eugene, OR) was provided as stock solution of high density, stored at −20°C. A solution of 1 mM concentration was prepared, after the dilution of 5 μL stock solution in 1000 μL PBS. For further analysis, 5 μL of the latter solution (SYTO-16, 1 mM) was used at the first week of storage, while 10 μL/sample were used for later analysis. The second protocol used CD66abce-FITC, CD64-PE, CD45-ECD, CD16-PE-cy5, and CD14-PE-Cy7.

The basic immunophenotype panel consisted of two parts, one surface direct staining and the other intracellular indirect staining, combined with surface direct staining. Specifically, in the first part, 3-color FCM surface direct staining was performed, by the following FITC antibody conjugates: IgG1(isotype control), EMA, Ber-EP4, CD66abce, and CD56, which were combined in every tube with a common backbone of the two conjugates, CD71-PE and CD45-PE-Cy5. Indirect fluorescence was performed intracellularly (see below) using FITC-polyclonal rabbit anti-mouse immunoglobulin with two other unconjugated antibodies, IgG1 (pure isotype control), and desmin-33, combined in each tube with the common backbone conjugates CD71-PE and CD45-PE-Cy5 (Table 1).

Table 1. Antibodies and Stains Used for Flow Cytometric Analysis
EMAMonoclonal MouseDako, DenmarkE29FITC
Anti-Human Epithelial
Membrane Antigen
Ber-EP4Monoclonal MouseDako, DenmarkBer-EP4FITC
Anti-Human Epithelial
Membrane Antigen
CD66Monoclonal MouseDako, Denmarkkat4cFITC
Anti-Human Myeloid Cell
CD56 (N-CAM)Monoclonal MouseBD, NJ, USANCAM16.2FITC
Anti-Human neural cell adhesion molecule
Monoclonal Mouse
Anti-Human Desmin
DesminDako, DenmarkD33Indirect staining with secondary antibody
IgG1(pure)Negative Control MouseDako, Denmark 
Polyclonal Rabbit Anti-Mouse ImmunoglobulinsRabbit F(ab′)2Dako, Denmark Secondary antibody
IgGNegative ControlDako, Denmark FITC
Mouse IgG1
Cytokeratin IO TestMonoclonal MouseImmunotech, Beckman Coulter, Marseille, FranceJ1B3FITC
Anti-Human IgG1
Human CD71Monoclonal MouseInvitrogen, Camarillo, USAT56/14R-PE
Anti-Human Transferrin
Receptor Antigen
CD45-ECD IO TestMonoclonal MouseImmunotech, Beckman Coulter, Marseille, FranceJ.33ECD
Anti-Human Leucocyte
Common Antigen
CD45-PC5 IO TestMonoclonal MouseImmunotech, Beckman Coulter, Marseille, FranceJ.33PE-Cy5
Anti-Human Leucocyte
Common Antigen
SYTO-16Green fluorescentInvitrogen, Eugene, Oregon, USA  
Nucleic acid stain
7-aminoactinomycin D (7-AAD)Viability DyeImmunotech, Beckman Coulter, Marseille, France  

For the detection of surface antigens, after 15-minute incubation at room temperature of fresh cell suspensions (100 μL) with 10 μL labeled antibodies, the cells were lyzed (BD FACS Lysing Solution, BD Biosciences, CA) for 10 minutes and centrifuged (1,800 rpm, for 5 minutes). The supernatant fluid was discarded and the cells were resuspended in 0.5 mL of PBS for analysis. For the detection of desmin-33 intracellular expression, the following procedure was conducted. After the first incubation period with conjugated CD71-PE and CD45-PeCy5, 100 μL Fix & Perm reagent “A” (Fix & Perm reagent kit, Caltag Laboratories, San Francisco, CA) was added for a second 15-minute incubation. After being washed once, the cells were permeabilized by the addition of 100 μL Fix & Perm reagent “B” and the unconjugated antibodies were added; 5 μL of IgG1 (pure isotype control) and 5 μL desmin-33, respectively, followed by a 15-minute incubation period, one wash-step and a 10-minute incubation period with 10 μL polyclonal rabbit IgG1-FITC (diluted, 1/10 with PBS) for indirect immunofluorescence. After one further wash the cells were resuspended in PBS for analysis. DNA index (DI) was evaluated with Coulter DNA Prep Reagent kit, with propidium iodide and RNAse. Two tubes of the sample were prepared according to instructions, while 10 μL cytokeratin-FITC and IgG1-FITC (control tube) were added, respectively (30 minutes incubation). FCM was performed on a 5-color flow cytometer, FC-500 (Beckman Coulter), where at least 60,000 events were counted for each sample. Analysis was made with CXP software (Beckman Coulter).


Initial Panel

The aim of the initial informative panel was the investigation of CD45 negative/CD71 positive cells, suspected to be malignant, or mesothelial cells. In addition, an extended differential leukocyte count of the effusion cells was conducted. Nuclear dyes SYTO-16 (Molecular Probes) and 7-AAD were combined with CD45-ECD and CD71-PE to identify CD45−, SYTO-16+, and 7-AAD− cells, excluding nonspecific interference from unlyzed erythrocytes, platelet aggregates, debris and dead, or apoptotic cells (SYTO-16−, 7-AAD+). After the exclusion of the above elements, suspicious cells were gated, related to the scatter characteristics (usually high forward and side-scatter signals). This strategy ensured avoidance of false positive CD45 events that could have been considered malignant, or mesothelial cells (Fig. 1). CD71 was used as a marker to localize a suspicious CD45 negative cell population, which could be of malignant or other origin. Although it is not a common tumor marker, it is upregulated in proliferating cells (such as tumor cells) (12–14). It is also expressed by normal mesothelial cells (15), so it could operate as a linkage marker for their differential diagnosis. The detection of CD71+ cluster (cluster positivity and not intensity positivity) of high side- and forward scatter-characteristics was used as evidence of need for further analysis with the basic panel.

Figure 1.

Flow cytometry on first sample case: SYTO-16+/7-AAD negative cells were gated (Gate A and C) in order to exclude apoptotic and necrotic elements (upper panels). The remaining viable cells were examined in SS/CD45 diagram for the detection of CD45-cluster. This cluster was easily differentiated from white blood cells, while debris (cell fractions and membrane fragments) was excluded (gating the high side-scatter cluster). The remaining viable cells were examined in CD71/CD45 diagram for the detection of CD71+/CD45-cluster (even with dim CD71 positivity). This atypical cluster was finally backgated in FS/CD45 and SS/CD45 diagram (to confirm large dimensions, due to cell clustering or isolated large malignant cells), not shown. The detection of CD71+ cluster of high side- and forward scatter-characteristics was used as evidence for further analysis with the basic panel. (Upper and middle panels are from the same sample/case, while lower panels are from a different sample/case, FS: forward scatter, SS: side scatter). [Color figure can be viewed in the online issue, which is available at]

The use of CD66abce, CD64, CD16, and CD14 in the second tube was mainly used to distinguish neutrophils (CD66abce+ and CD16+), eosinophils (CD66abce+ and CD16−), lymphocytes (CD45bright+ and side scatter low), and macrophages/monocytes (CD14+ and CD64+); macrophages are CD14+, CD64+, and CD16+(16, 17). The side-scatter characteristics and CD45 expression of the cells were also helpful. It is significant that a high CD64 expression on neutrophils was supportive of an infectious etiology (18), while an increased number of eosinophils was indicative of a possible underlying malignancy (19).

If a hematological malignancy was suspected in the initial panel (usually CD45 positive), we did not proceed to the basic panel, but to T-cell and B-cell immunophenotype and clonality assessment, with different panels.

Basic Immunophenotypic Panel

The aim of the basic panel was the immunophenotyping of the suspicious population of CD45−/CD71+ cells, for the expression of EMA, Ber-EP4, CD66abce, CD56, and desmin-33. The expression of EMA, Ber-EP4 and CD66abce was considered positive when >20% (ensuring bright expression and avoiding false positive results), compared with the isotype control (adjusted to be <2%). The expression was evaluated in relation to the intensity of expression, as a higher intensity was observed in malignant cells (Fig. 2). The CD45− cells of high forward scatter characteristics (considered suspicious for malignancy) were gated and examined for CD71 expression (CD71/CD45 plot). The CD71+ gated cluster was further examined for EMA, Ber-EP4, CD66abce, and CD56 expression, according to appropriate orientation of intensity, obtained by the use of isotype control (Fig. 2). Desmin-33 expression was examined with indirect immunofluorescence, as described above, using its specific control tube (Fig. 3). The orientation of control was made to optimize the signal-to-noise ratio, and the intensity was always co-evaluated (Fig. 3).

Figure 2.

Flow cytometry on second sample case: CD45− negative cells of relevant high FS were examined for CD71 expression. The isotype control IgG1-FITC was always used to define the cut-off intensity, so to further evaluate the expression of EMA, BerEp4, and CD66 expression on malignant or mesothelial cells (CD45−/CD71+/high FS). The positivity threshold was defined at 20% expression. [Color figure can be viewed in the online issue, which is available at]

Figure 3.

Flow cytometry on third sample case: CD45-negative cells of relevant high FS were examined for CD71 expression. Through indirect immunofluorescence, IgG pure conjugated to polyclonal rabbit IgG1-FITC was used to define the cut-off intensity, so to further evaluate the expression of desmin (conjugated to polyclonal rabbit IgG1-FITC). It was important to obtain a correct orientation gate on the control tube. (Upper and middle panels are from the same sample/case, while lower panels are from a different sample/case). [Color figure can be viewed in the online issue, which is available at]

DNA/Cytokeratin Panel

For the estimation of the DI, the mean intensity of the G0/G1 peak of cytokeratin positive cells was divided by the mean intensity of the G0/G1 peak of cytokeratin negative cells. The DI was evaluated in at least 0.1% cytokeratin positive cells. A DI cut-off value was applied to discriminate between the aneuploidy of mesothelial cells and malignant cells, as has been previously described (20) and confirmed by the present study, which showed hyperploidy in mesothelial cells. DI ≥ 1.4 was considered supportive of malignancy (20), referring to the cytokeratin (FITC) positive cell population (Fig. 4). Conclusions were drawn only on the combined consideration of all three panels.

Figure 4.

For the evaluation of DI, the mean intensity of G0/G1 peak of cytokeratin positive cells (KER) was divided with the mean intensity of G0/G1 peak of cytokeratin negative cells (K). The DI was evaluated in at least 0.1% cytokeratin positive cells. [Color figure can be viewed in the online issue, which is available at]

Inclusion Criteria

The criteria used for samples/cases to be included in the study were: (a) Effusions that were considered morphologically suspicious for malignancy (see above) or/and derived from a patient with known malignancy. (b) Effusions with detectable atypical CD71+/CD45− clusters. Both criteria had to be fulfilled for inclusion.

Exclusion Criteria

The criteria used for samples/cases to be excluded from the study were: (a) Effusions derived from hematopoietic neoplasias (Giemsa morphology, CC, biopsy, and FCM were the methods used for the diagnosis of these retrospectively excluded cases/samples). (b) Effusions with no detectable CD71+/CD45− cluster.

Statistical Analysis

The Chi-Square test was used to compare the positivity of each immunophenotypic marker or specific combinations of immunophenotypes, between true malignant and reactive effusions (Table 2), and also to investigate the association between the diagnostic findings of the two techniques (FCM, CC). P < 0.05 was considered statistically significant.

Table 2. Immunophenotype of Malignant and Reactive Effusions, by Flow Cytometry
Flow cytometryMalignant cells n = 74Mesothelial cells (reactive effusions) n = 45P*Sens (%)Spec (%)PPV (%)NPV (%)
  • A: adenocarcinomas, NS: neuroectodermal sarcomas, M: melanomas, T: total malignancies, Sens: sensitivity, Spec: specificity, PPV: positive predictive value, NVP: negative predictive value.

  • *

    *Significance level in the discrimination between malignant and reactive mesothelial cells.

EMA + (>20%)591060/73 (82.2%)6/45 (13.3%)<0.00182.286.79175
Ber-EP4 + (>20%)580058/72 (80.6%)4/43 (9.3%)<0.00180.690.79474
CD66abce + (>20%)380139/68 (57.3%)1/45 (2.2%)<0.00157.497.89860
CD56 + (>20%)1225/74 (6.7%)0/45 (0%) 6.810010039.5
Desmin-33−651167/74 (9.5%)4/45 (8.9%)<0.00190.591.19485
DNA index ≥ 1.4450045/68 (66.2%)20/39 (51.3%)>0.0566.248.76945
Informative immunophenotypes for adenocarcinomas
 Desmin-33−/cytokeratin+62/73 (84.9%)1/44 (2.3%)<0.00184.997.79880
 EMA+/Ber-EP4+55/71 (77.5%)3/43 (7%)<0.00177.5939571
 Desmin-33−/EMA+/BerEP4+54/71 (76.1%)1/43 (2.3%)<0.00176.197.79871
 EMA+/Ber-EP4+/desmin-33−/cytokeratin+54/70 (77.1%)1/42 (2.4%)<0.00177.197.69872
Informative immunophenotype for NS and M malignancies
Informative immunophenotypes for mesothelial cells in reactive effusions
 Desmin-33+/cytokeratin+6/73 (8.2%)38/44 (86.4%)<.00186.491.88692
 EMA−/Ber-EP4-11/71 (15.5%)36/43 (83.7%)<.00183.784.57790
 EMA−/Ber-EP4−/desmin-33+/cytokeratin+5/70 (7.1%)31/42 (73.8%)<.00173.892.98686


The initial morphological assessment revealed 125 effusions suspicious for nonhematolymphoid malignancies, which were analyzed with FCM and CC.

Initial Panel

The initial panel of FCM performed on all 125 samples detected 119 effusions with an atypical CD71+, CD45− cluster. For the six remaining samples in which no cluster was detected, an inflammatory origin was proposed, and subsequent CC evaluation was also suggestive of an infectious etiology. These samples were not analyzed further with the basic immunophenotypic panel or with the DNA/cytokeratin panel. Amongst the 119 samples with detectable atypical clusters, in 64 effusions this cluster proved to represent malignant cells and in 55 mesothelial cells. The respective CC evaluation revealed 63 malignant, 48 reactive, and 8 atypical specimens. Of the eight atypical cases revealed by CC, six were confirmed as malignancies by biopsy and/or computed tomography (CT), of which five had been diagnosed positive on FCM. Based on 74 cases with verified neoplasm (true malignant effusions) and considering the reports of “atypical cells” by CC as positive reports, the defined sensitivity for FCM and CC was 85.1 and 93.2%, respectively, while the defined specificity was 97.8 and 95.6%, respectively. The positive predictive value (PPV) was evaluated to be 98% for FCM and 97% for CC, while the negative predictive value (NPV) was 80 and 90%, respectively. A significant association was observed between the diagnostic categorization resulting from the two techniques (P < 0.001).

The percentage of atypical clusters in malignant effusions ranged from 0.1 to 79.8%. The second tube of the initial panel detected monocytic/macrophage cells in most effusions (mainly nonmalignant). An increase in neutrophils, especially CD64+, was detected in the six effusions, found negative for an atypical cluster in the first protocol of the initial panel and excluded from further study (data not shown).

Basic Immunophenotypic Panel


Based on the assumption that malignant cells are expected to be desmin (desmin-33) negative (21), 71 effusions were found positive for malignancy (i.e., desmin-33−) and 48 negative (i.e., desmin-33+). The effusions finally characterized as malignant by FCM were all in the desmin-33− group, while the desmin-33+ group was thought to represent effusions with reactive mesothelial cells. The conclusions based on CC were variable concerning the (desmin-33+) FCM group of patients: 41 patients were considered negative for malignancy, 4 patients were considered positive (i.e., confirmed as malignancy), and 3 patients belonged to the group of cases with detection of atypical cells (one with confirmed malignancy). Based on true positive cases, desmin-33 negativity was evaluated to have 90.5% sensitivity and 91.1% specificity in malignancy detection, with 94% PPV and 85% NPV (Table 2). It should be noted that in this series no case of malignant mesothelioma was observed or confirmed by other clinical or laboratory information.


The expression of EMA was successfully evaluated in 118 effusions and found positive in 66 (55.9%), corresponding to 60 desmin-33− and 6 desmin-33+ effusions. Among 64 cases ultimately diagnosed positive by FCM, EMA was positive in 60 cases, negative in 3 (although confirmed malignancies) and in 1 case, EMA was unsuccessfully determined. These cases were diagnosed positive by FCM, as they fulfilled all the other requirements for malignant characterization (i.e., Ber-EP4+, CD66abce+, desmin-33−, cytokeratin+, CD71+, CD45−, and hyperploid). It should be noted that determinations were unsuccessful due to inadequacy of the samples for the initial measurement or for repetition of a specific measurement. Based on the cases with confirmed malignancy, the sensitivity and specificity of EMA were determined to be 82.2 and 86.7%, respectively, and the PPV and NPV were 91 and 75%, respectively (Table 2).


Ber-EP4 was evaluated in 115 effusions and found positive in 62 (53.9%), corresponding to 59 positive cases obtained by FCM. Among the remaining cases positive on FCM, three were negative for Ber-EP4, and determinations were unsuccessful in two additional cases, but all fulfilled the other requirements for malignancy diagnosis. Based on confirmed malignancies, the sensitivity and the specificity of Ber-EP4 were evaluated at 80.6 and 90.7%, respectively, and the PPV and NPV were 94 and 74%, respectively (Table 2).


CD66 (CD66abce pool) was the least sensitive marker for the detection of malignant cells in effusions. It was evaluated in 113 cases and found positive in 40 (35.4%) samples, all of which corresponded to positive cases on FCM, although one was a true negative case. The sensitivity and specificity of CD66 were determined at 57.4 and 97.8%, respectively, while the PPV and NPV were 98 and 60%, respectively (Table 2).


The importance of the detection of a CD56+/CD45−/cytokeratin− immunophenotypic profile by FCM in neuroendocrine malignancies has been previously underlined (22), and was confirmed in the findings on the cases of melanoma and sarcoma in this study.

DNA/Cytokeratin Panel

Cytokeratin expression was successfully determined in 117 effusions. DNA ploidy analysis was performed in 107 effusions with cytokeratin+, while 10 cases were cytokeratin−. All of these were desmin-33-/cytokeratin−, except for two cases that were desmin-33+/cytokeratin− and which were reported negative by CC. The DI of desmin-33−/cytokeratin+ cells (63 cases found positive by FCM), ranged between 0.97 and 4.5 (median DI 1.6), while the DI of desmin-33+/cytokeratin+ cells (44 cases found negative by FCM), ranged between 1 and 1.8 (median 1.35). DI ≥ 1.4 was detected in 66.2% of true positive for malignancy cases and in 51.3% of true negative cases (66.2% sensitivity and 48.7% specificity for malignancy detection, Table 2). In addition, five of the eight atypical cases obtained by CC and found malignant by FCM had DI > 1.6.

Based on the cases verified for malignancy, the evaluated sensitivity of cytokeratin was 91.8% and its specificity 11.4%, as was to be expected, because mesothelial cells are also cytokeratin positive. The PPV was 63% and the NPV was 45%. The sensitivity of the phenotype desmin-33−/cytokeratin+ was 84.9%, and the specificity was 97.7%, significantly increased because of desmin-33 determination. Accordingly, the PPV of this phenotype was 98% and the NPV 80%.


Immunophenotyping of malignant cells by FCM has been little used in the detection of malignant effusions(9, 10), although FCM has been extensively used for DNA ploidy investigation (10). Based on earlier considerations that aneuploidy of mesothelial cells complicates the diagnostic process (23) and that immunophenotyping by FCM modifies the provisional cytopathological diagnosis (24), it was considered necessary to combine immunophenotyping and aneuploidy determination in one FCM assay for the assessment of malignant effusions. FCM has been suggested to be comparable to immunohistochemistry in terms of sensitivity and specificity, and this was confirmed in this study(9, 25, 26). A three-step FCM assay for the routine assessment of effusions suspicious for malignancy is presented, which includes an initial informative panel, a basic immunophenotyping panel and a DNA ploidy panel.

SYTO-16 and 7-AAD were used in the initial panel for the analysis of nucleated cells and the exclusion of apoptotic cells, necrotic cells, and debris (27–30). In this way, the possibility of false positive clusters in CD45−/CD71+ gating, was excluded, also eliminating the possibility of compensation errors and making more reliable determinations of small clusters (at least 0.1% of nucleated cells). CD71 was the essential marker used for the identification (in the initial panel) and characterization (in the basic panel) of malignant cells. The transferrin receptor (CD71) is a cell membrane-associated glycoprotein involved in iron homeostasis and cell growth. It has been explored as a target for delivering therapeutic agents into cancer cells, due to its increased expression on malignant cells, accessibility on the cell surface, and constitutive endocytosis (12–14). CD71 was also used here as it is always expressed in normal mesothelial cells, contributing to further identification and immunophenotyping with the basic panel (15). Although CD71 is not a common tumor marker, the identification of a CD71+/CD45−/SYTO-16+/7-AAD− cluster of high forward and side scatter in the initial panel (cluster positivity and not intensity positivity), was considered evidence of possible malignancy and was the most important step for decision on the further assessment of an effusion for possible malignancy. The same gating strategy could also involve mesothelial cells, indicating the need for further sample processing. An appropriate adjustment of compensation was usually needed to avoid false negative or false positive results. Additionally, consideration of the forward scatter range should take account of the possible large size of the observed cells. Often the malignant cells are attached to each other, so they appear larger than expected and may not be included in the FW/SS gate. This appeared to be a common reason for decreased sensitivity, and probably account for the negative for malignancy results obtained by FCM in five truly positive cases, found malignant with CC.

Although all three panels were performed in this patient series, the initial panel alone can lead to the rapid exclusion of the diagnosis of malignancy, but only in cases of undetectable CD71+/CD45− cluster and when an infectious etiology is strongly speculated. In this way, unnecessary, time-consuming and expensive processing can be avoided in future applications of the FCM assay.

Concerning the basic panel, the investigation of desmin-33 expression proved to be the most valuable marker for the discrimination between carcinoma and reactive mesothelial cells. Desmin has been detected in benign mesothelial cells, while minimal expression is documented in cases of carcinoma and malignant mesothelioma(10, 20, 31). In this study, a FCM option of desmin-33 assessment on suspicious cells was provided, which proved to have high sensitivity and specificity for malignancy detection. When desmin-33 negativity was combined with cytokeratin positivity, an increase in specificity and PPV was observed. Concerning malignant mesothelioma effusions, data cannot be provided on desmin-33 expression on neoplastic mesothelial cells, as there was no sample in this series. However, desmin has been studied by immunohistochemistry and found to be positive in only 10% of malignant mesothelioma cases, compared with reactive mesothelial hyperplasias (85%);(32). Thus, differential diagnostic evidence between adenocarcinoma and malignant mesothelioma could theoretically be provided by investigating Ber-EP4 and desmin-33 expression (expected positive and negative, respectively, in adenocarcinoma, while both negative in malignant mesothelial cells)(31, 32). It is also important to underline that among the eight cases rated atypical by CC in this series, three cases were considered mesothelial cells by desmin-33 positivity, and thus FCM contributed to the final diagnosis.

Ber-EP4 has been proposed as a highly sensitive (82.9%) and specific (95.3%) marker for detecting metastatic adenocarcinoma cells, contributing to their differentiation from reactive mesothelial cells(9, 33, 34). The investigation of EMA, CEA (CD66abce), and cytokeratin expression has also been proposed for improving the cytodiagnosis of effusions(2, 10, 35–38). In malignant effusions from metastatic adenocarcinoma, EMA has been shown to be strongly positive, and Ber-EP4 and CD66abce also expected to be positive. Benign effusions with reactive mesothelial cells have been shown to be EMA negative or weakly positive by immunocytochemistry on the cytoplasmic membrane, while Ber-EP4 and CD66abce are negative. In this series, EMA and Ber-EP4 proved to have higher sensitivity than CD66abce for the detection of malignant cells; however, CD66abce was the most specific marker. All the cases found simultaneously positive for EMA, Ber-EP4, and CD66abce were found desmin-33−/cytokeratin+, and were also found positive by FCM (37 cases). All but one of these cases, found positive for malignancy by FCM corresponded to confirmed malignancies, while the respective CC reports included 3 atypical, 1 negative (the same negative case with CC), and 33 positive reports. The inclusion of three cases atypical on CC provides evidence for the possible value of the simultaneous positivity of all three markers (EMA, Ber-EP4, and CD66abce) in desmin-33−/cytokeratin+ cells, for the characterization of an atypical effusion as malignant. However, this specific immunophenotypic profile was not superior to the respective profile without CD66 (Table 2) in terms of sensitivity, specificity, PPV and NPV (62.5, 97.7, 97, and 67%, respectively).

The DNA panel was included in the study assay, as it has been proposed that an abnormal DNA content strongly supports the diagnosis of a malignant pleural effusion(8, 10). It has also been suggested that the additional diagnostic value of DNA image cytometry in cases with a cytological diagnosis of “atypia” or “suspicious for malignancy” is limited (39), but that visual diagnostic cytology and morphometric digital microscopy miss some cases of malignancy, which can be detected by DNA FCM (4). Moreover, hyperdiploidy can be attributed to mesothelial elements, as was demonstrated by a combined analysis of FISH and immunocytochemistry with staining for cytokeratin (23). Despite the conflicting published data, for this study a DI of 1.4 was used as the cut-off value for the orientation of malignant aneuploidy, as previously suggested (20). Although the study findings supported an increased possibility of malignancy when DI ≥ 1.4 was detected, the DNA panel alone could not support a definitive diagnosis. The use of cytokeratin also contributed to the detection of malignant and mesothelial cells and the discrimination between them, while the combination of cytokeratin and desmin-33, and specifically the phenotype cytokeratin+/desmin-33−, proved to be significantly specific and have the highest positive and NPV for malignancy detection.

In a recent study that compared FCM immunophenotyping and DNA ploidy with serum tumor markers and classic cytology for the detection of malignant cells in pleural and ascitic fluids, a high diagnostic accuracy of FCM was detected (9). Ber-EP4, cytokeratin, CD3, and CD45 were the monoclonal antibodies used, but no comment was made on Ber-EP4 positive mesothelial cells, which were detected in the present series, or on the distinction between reactive effusions and malignant mesothelioma, both expected to be Ber-EP4 negative, (32). In addition, the overlap with mesothelial cells in mean DI determination (23) was not taken into account for the evaluation of specificity of DNA ploidy determination, and the suspicious population was gated for further characterization, ignoring the many apoptotic elements that may interfere in CD45 negative clusters, being also cytokeratin+. Finally, the detection of elevated serum tumor markers indicates neither the obligatory presence of malignancy nor the effusion infiltration by malignant cells(40, 41), but FCM showed a significant correlation with cytological examination, which was verified by this study.

In conclusion, FCM assay was found to be effective for the identification of malignant effusions, with high sensitivity and a PPV comparable with that of CC. With a three-step procedure a suspicious cell population was rapidly identified, and characterized immunophenotypically (based on desmin-33, EMA, Ber-EP4, CD66abce, and CD56 expression), and the diagnostic procedure was completed with DNA ploidy analysis of cytokeratin+ cells. Reactive mesothelial effusions can all be excluded through these interrelated steps, resulting in a highly sensitive and specific procedure for the routine identification of malignant effusions.