Advantages of flow cytometry immunophenotyping for the diagnosis of central nervous system non-Hodgkin's lymphoma in AIDS patients


Dr Dolores Subirá, Department of Hematology, Fundación Jiménez Díaz, Avda. Reyes Católicos 2, Madrid 28040, Spain. Tel: 34 1550 43 99; fax: 34 1549 66 99; e-mail:



Neurological disorders are common in HIV-infected patients. Central nervous system (CNS) lymphoma should always be considered because it is an important cause of morbidity and mortality.


To investigate the clinical utility of flow cytometry immunophenotyping (FCI) in diagnosing or discarding leptomeningeal involvement in HIV-infected patients and to compare its sensitivity with that of conventional cytological methods.


Fifty-six cerebrospinal fluid (CSF) samples from 29 HIV-infected patients were independently evaluated by flow cytometry and cytology. The description of an aberrant immunophenotype was the criterion used to define the malignant nature of any CSF cell population.


FCI and cytology gave concordant results for 48 of the 56 CSF samples studied: 37 were negative for malignancy and 11 had evidence of CNS lymphoma. Discordant results were obtained for eight CSF samples, and the accuracy of the FCI findings could be demonstrated for four CSF samples described as positive for malignancy according to the FCI criteria.


A high level of agreement was found between the results obtained using the two methods, but FCI gave at least 25% higher sensitivity than conventional cytomorphological methods for the detection of malignant cells. This advantage suggests that, in case of negative flow cytometry results, disorders other than non-Hodgkin's lymphoma should be strongly considered.


Despite the great advances that have been achieved with highly active antiretroviral therapy (HAART) in improving the management of AIDS patients and their outcome [1], non-Hodgkin's lymphoma (NHL) remains an important cause of morbidity and mortality in these patients [2,3]. Invasion of the central nervous system (CNS) in this type of lymphoma is particularly common in the HIV-infected population; in necropsy studies, it has been estimated that it occurs in more than two-thirds of these cases [4]. Forty percent of AIDS patients with NHL and bone marrow invasion at the time of diagnosis show CNS involvement [5]. Therefore, cytological investigation of the cerebrospinal fluid (CSF) is mandatory for the initial staging evaluation of a recently diagnosed NHL, and also in patients with neurological signs or symptoms suggestive of meningeal involvement. Conventional cytological investigation of the CSF has limited sensitivity, and malignant cells will only be detected in 15–25% of patients [4]. An improved diagnostic test would therefore be of great value.

Flow cytometry immunophenotyping (FCI) is an indispensable tool for the diagnosis and ulterior evaluation of many haematological malignancies in routine clinical practice [6,7]. However, few studies have been performed to evaluate the sensitivity of this technique in samples with low numbers of cells [8–13]. The aim of this work was to determine the clinical utility of FCI in diagnosing or discarding CNS involvement in any of the haematological malignancies often described in HIV-1-positive patients, with special emphasis on B-cell non-Hodgkin's lymphoma (B-NHL). Specifically, we attempted to determine flow cytometry specificity and sensitivity as compared to classical CSF cytological examination.


Patients and CSF samples

A total of 29 HIV-1-positive patients were included in the study at the Infectious Diseases Division and/or the Hematology Department of our Institution between February 1994 and June 2003. The mean age of the patients was 39.6 years (range 30–51 years). Six were female and 23 male. At the time of the CSF study, six patients were in HIV stage A and 23 were in stage B or C, as defined by the Centers for Disease Control and Prevention (CDC) in 1992 (CDC, Atlanta, GA, USA).

At the time of the study, 17 patients were diagnosed with systemic haematological malignancies (15 with systemic NHL, one with Hodgkin's disease and one with multiple myeloma), one patient was suspicious for primary CNS lymphoma and 11 patients had no previous evidence of haematological disease. The patient who was suspicious for primary CNS lymphoma and eight of the 17 patients with systemic haematological malignancies presented with neurological symptoms or signs (a total of 10 CSF samples were studied). For staging purposes, one CSF sample was obtained from seven of the 17 patients (seven CSF samples), and in 13 of the 17 patients CSF samples were also collected during follow-up (26 CSF samples). In 10 of 11 patients with no known haematological disease, the CSF sample was taken when neurological symptoms appeared (12 CSF samples), and in one patient the CSF sample was obtained in the context of fever of unknown origin (one CSF sample). At the time of evaluation of the CSF samples for diagnostic purposes, 14 patients were receiving HAART therapy and 15 were not. In 11 of 15 patients, the diagnoses of HIV infection and haematological malignancy were simultaneous.

The CSF samples were dispensed for cytological and flow cytometry investigations. The two analyses were performed in parallel and carried out by two independent observers. Criteria for inclusion in the study were red blood cell-free CSF samples and introduction of samples to the flow cytometer within 3 h of CSF collection. CSF sample volume ranged from 0.6 to 2.2 mL and CSF cell count ranged from <1 cell/μL (10 CSF samples) to 251 cells/μL.

Flow cytometry design

All CSF samples were stained with a three-colour combination of fluorescein isothiocyanate (FITC), phycoerythrin (PE) and PE cyanine 5 (PE/Cy5) conjugated monoclonal antibodies (mAb) obtained from the following sources. FITC-conjugated: CD19 (HD37) and HLA DR (CR3/43) from DAKO (Carpinteria, CA, USA); CD3 (SK7) from Becton Dickinson (Immunocytometry Systems, San Jose, CA, USA); PE-conjugated: CD4 (MT310) and CD20 (B-Ly1) from DAKO; CD5 (L17F12) and CD56 (MY31) from Becton Dickinson; CD10 (J5-RD1) from Immunotech (Marseille, France); PE/Cy5-conjugated: CD8 (DK25) from DAKO; CD45 (J.33) and CD19 (J4.119) from Immunotech.

The number of reagents used to search for any malignant population in the CSF was based on the number of cells present in the CSF sample. For those samples with≤10 cells/μL, a three-tube panel of reagents was designed. Additional tubes were only used when >10 cells/μL were present in the CSF.

The CD3 FITC/CD4 PE/CD8 PE/Cy5 tube was used as an internal control in all the CSF samples studied. The two tubes left were selected to search for malignant cells in the CSF. The panel of reagents chosen was based on the immunophenotypic information available at the time of the study concerning the malignant cell population (Table 1). In fact, only four of the 17 patients with systemic haematological malignancies had a complete phenotypic characterization of the malignant cell population obtained from a flow cytometry study performed on peripheral blood or bone marrow specimens. The panel included mAb combinations that could distinguish malignant cells from reactive CSF cells; for example, the CD138/CD56/CD19 tube used in the patient with a multiple myeloma or the CD19/CD10/CD45 tube used in those patients diagnosed with Burkitt's lymphoma. Nevertheless, at the time of the CSF investigation, 13 patients with haematological diseases had no immunophenotypic information available from tissue biopsies of lymph nodes, and the remaining 12 patients had no previous evidence of haematological malignancy. The panel of reagents designed for these CSF samples included cell-antigen combinations that, if present in the CSF B-cell population, would demonstrate its malignant character: for instance, CD10-positive B cells or HLA-DR-negative B cells.

Table 1.   Monoclonal antibody panel selected (fluorescein isothiocyanate/phycoerythrin/phycoerythrin cyanine 5)
 Well-characterized immunophenotypic features
B-NHL with unknown
phenotypic features
  1. B-NHL, B-cell non-Hodgkin's lymphoma.

Tube 1CD8/CD4/CD3CD8/CD4/CD3CD8/CD4/CD3
Tube 2HLA DR/CD20/CD19CD45/CD20/CD19HLA DR/CD20/CD19
Tube 3CD19/CD10/CD45CD138/CD56/CD19CD19/CD10/CD45

The procedure for labelling cells has been previously described [13]. Briefly, CSF samples were equally distributed among the three (or more) tubes and incubated in the dark with the mAb at 4°C for 20 min. After centrifugation, the pellet was resuspended in 0.5 mL of phosphate-buffered saline (PBS) and was acquired on a FACscan or a FACS Calibur flow cytometer (Becton Dickinson, Immunocytometry Systems) using the cellquest software program (Becton Dickinson, Immunocytometry Systems). Acquisition was always stopped when the whole sample volume had been completely acquired. Calibration of the instrument was performed daily prior to data acquisition according to well-established methods [14].

The paint-a-gate software program (Becton Dickinson, Immunocytometry Systems) was used for further data analysis. Briefly, normal T lymphocytes (including CD4 and CD8 subsets) were first identified on the basis of their antigenic properties and then, once located in the Forward/side angle light sootter (FSC/SSC) dot-plot, they were distinguished from the debris. With this internal control, flow cytometry analysis should be able to identify any other CSF cell population present in a similar percentage or greater than that of CD4 or CD8 T lymphocytes [12,13]. In a second step, B cells were identified on the basis of CD19 expression and an immunophenotypic investigation of their malignant character was performed. This was established on the basis of either the description of an abnormally high percentage of B cells or abnormal scatter (FSC/SSC) properties or the description of an aberrant pattern of antigen expression (i.e. co-expression of CD19 and CD10 or lack of HLA-DR molecules in the CSF B-cell population).


Flow cytometry findings

Flow cytometry immunophenotyping did not detect any cell population suspicious for malignancy in 24 patients (38 CSF samples). These patients were finally diagnosed with criptoccocosis (four patients), tuberculosis (two patients) and lymphocytic meningitis (four patients). A specific aetiology could not be identified in 28 CSF samples, but CSF cells were described as benign lymphoid populations. Immunophenotypic analysis of all these negative CSF samples showed different proportions of T lymphocytes (being the percentage of the T CD8 subset higher than the percentage of the T CD4 subset) and monocytes. A B-cell population was also detected in 16 of 38 CSF samples obtained from 13 patients (patients 15, 16, 18 and 20–29). This B-cell population ranged from 1 to 5% of total cellular events, had intermediate forward and side-angle light scatter properties and showed immunophenotypic features similar to those described for mature polyclonal B cells (CD19+, CD20++, CD45++, HLA DR+ and CD10–).

Flow cytometry analysis detected a phenotypically abnormal B-cell population in eight patients, four of whom (patients 2, 5, 6 and 7) also produced some negative CSF samples during ulterior follow-up, in which the efficacy of intrathecal therapy was monitored. Six patients were diagnosed with Burkitt's lymphoma (patients 1, 2, 4, 5, 6 and 7), one patient was suspicious for NHL (patient 8) and one patient was suspicious for primary NHL (patient 3) (Table 2); at the time of CSF sampling, NHL immunophenotypic information was only available for patient 4. The number of CSF samples examined for these eight patients was 18, with a cell count ranging from <1 cell/μL to 166 cells/μL and the abnormal B-cell population ranging from 1 to 98% of total CSF cellular events. They were considered abnormal populations because of their high percentage (13 of 18 cases), high forward and side-angle light scatter properties (18 of 18 cases), CD10 expression (17 of 18 cases), lack of HLA DR expression (one of 18 cases) and lack of CD19 expression (one of 18 cases). Two cases (8b and 6a) were also evaluated for surface expression of the immunoglobulin light chains, and both of these were kappa positive.

Table 2.   Patients' diagnoses and CSF findings for the 18 CSF samples with flow cytometry data on malignancy
PatientDiagnosisCerebrospinal fluid sample
T cellsLymphoma cellsImmunophenotypeCytological diagnosis
  1. NHL, non-Hodgkin's lymphoma; pos, positive for malignancy; neg, negative for malignancy.

1Burkitt's NHL11663533.8CD19+CD10+CD20+HLA DR+CD45++pos
2Burkitt's NHL2nd37.758CD19−CD10−CD20+dim HLA DR+CD45++pos
3CNS lymphoma338991CD19+CD10+CD20+HLA DR++CD45++CD22+dimpos
4Burkitt's NHL4a97027CD19+CD10+CD45++CD20++neg
 4b85198 pos
 4c6902.5 pos
5Burkitt's NHL5a7649CD19+CD10+CD45++CD20++HLA DR–neg
 5b< 1702 neg
6Burkitt's NHL6a323068CD19+CD10+dim CD20+CD22+dim HLA DR+CD45++Kappa+pos
 6c47518 neg
7Burkitt's NHL7a21590CD19+HLA DR+CD10+CD20++CD22+CD45++pos
 7b1907 neg
 7c1131583 pos
 7d120395 pos
 7end198 pos
 7f20397 neg
8Not located8a85345CD19+CD10+dim CD20++HLA DR+CD45++CD22+Kappa+neg
systemic NHL8b281583 pos

Comparative analysis of the CSF results obtained with flow cytometry and cytology

FCI did not detect aberrant populations in 38 CSF samples. The morphological results for all but one of these CSF samples were also interpreted as negative for malignancy. This one CSF sample was from a patient (patient 9) without CNS symptoms who was being staged for Burkitt's lymphoma. The CSF sample contained 6 cells/μL and, although the cytological evaluation was positive for malignant neoplasm, flow cytometry immunophenotyping detected 93% of T lymphocytes, distributed in both TCD4 and TCD8 cells. A new spinal tap was performed after one dose of intrathecal therapy, and both immunophenotypic and morphological results for this CSF sample were considered negative for malignancy. The patient died a few days later of tumour lysis syndrome.

In this group of patients whose CSF samples were classified as benign according to morphological and immunophenotypic criteria, there was one patient (patient 10) who was being investigated for a cavitated lesion in the CNS that did not improve with steroids. The differential diagnosis of primary CNS lymphoma was being investigated, but the patient died and no brain biopsy could be performed. His CSF sample contained 2 cells/μL and all cells were T lymphocytes; B cells were not detected.

Flow cytometry analysis detected phenotypically abnormal B-cell populations in 18 of the 56 CSF samples studied (Table 2). Morphological examination was positive for lymphoma in only 11 of these 18 CSF samples. The seven CSF samples for which immunophenotypic and morphological results were discordant (cases 4a, 5a, 5b, 6c, 7b, 7f and 8a) were from five patients. Two CSF samples were obtained from the initial lumbar puncture performed on a patient with a recently diagnosed B-cell NHL (cases 4a and 5a), and one CSF sample was collected in an evaluation of a patient with a high clinical suspicion of lymphoma (case 8a). All these patients had different cranial nerve palsies, and a new CSF sample obtained a week later revealed malignant lymphoma by both immunophenotypic and morphological examination in patients 4 and 8 and by morphological evaluation only in case 5 (no sample was sent for FCI).

The four remaining CSF samples belonged to three patients who had received chemotherapy for a B-cell NHL with CNS involvement. Serial lumbar punctures were performed in these patients to assess the efficacy of chemotherapy against the CNS disease. The immunophenotypic analysis detected malignant cells in CSF samples that had been interpreted as benign morphologically. After an additional dose of intrathecal chemotherapy was administered, the immunophenotypic analysis of a new CSF sample was negative for malignancy, except for patient 7 whose positive CSF sample 7f proceded his CNS relapse. According to these data, and always taking the cytological findings as the ‘gold standard’, FCI showed a sensitivity and a negative predictive value of 93.75% and 97.3%, respectively, as compared with 75% and 90.2% for the cytological analysis. Both techniques showed 100% specificity and positive predictive value (Table 3).

Table 3.   Comparison of the flow cytometry immunophenotyping and cytological results
  1. The cerebrospinal fluid (CSF) sample cytological findings were taken as the ‘gold standard’, but note that in patients 4, 5, 7 and 8 a new CSF sample obtained a week later confirmed the flow cytometry findings of the discordant cases 4a, 5a, 7f and 8a.

  2. FCI, flow cytometry immunophenotyping.

Sensitivity (%)93.7575
Specificity (%)100100
Positive predictive value (%)100100
Negative predictive value (%)97.390.2


Flow cytometry is a very useful technique to diagnose CNS involvement as a complication of an AIDS-associated lymphoma, and allows its rapid differentiation from other CNS inflammatory diseases. We have observed a high level of agreement between immunophenotyping and classical cytological examination of CSF samples; however, FCI showed greater sensitivity than conventional morphological methods and a very high negative predictive value. In this sense, ‘double negative’ flow cytometry and cytological findings suggest that other nonmalignant conditions are responsible for the patient's CNS symptoms. We and others have previously obtained similar findings in patients with acute leukaemia [9,13] and NHL [10,11] but, to the best of our knowledge, this is the first report of a study involving a large number of HIV-positive patients.

Parenchymal cerebral masses with deep locations do not seed malignant cells into the subarachnoid space, and this might explain the negative results of the cytological and flow cytometry analyses in some cases [15,16]. Further studies are needed to investigate whether FCI may increase the sensitivity of conventional cytological CSF analysis for patients diagnosed with a primary NHL lymphoma.

With regard to those cases in which the cytological examination detected a malignant population in the CSF, all samples but one were also described as malignant according to the immunophenotypic criteria. It is difficult to determine which technique gave the right result in this discordant CSF sample because the cytological study described a unique malignant population that was not consistent with 90% of T lymphocytes identified in the flow cytometry examination, where no B cells were detected. Although this could be a rare example of a T-cell non-Hodgkin's lymphoma (T-NHL), the discordance may also have been a result of a reactive T-lymphocyte infiltration, as its morphological appearance often resembles that of malignant cells [17]. Perhaps the use of monoclonal antibodies specific for the different TCRVβ, TCRVα, TCRVγ and TCRVδ (T-cell receptor V repertoire) family members might represent a valid alternative approach [18,19] to classical antigen immunophenotyping.

FCI was also able to detect a malignant population in seven CSF samples that did not fulfil the criterion of malignancy according to the cytological analysis. All these discrepant CSF samples had at least one of two factors in common: (1) a percentage of malignant cells lower than that of reactive CSF T lymphocytes, and (2) a very low CSF cell count. Either of these factors may explain the negative cytological findings, because the main T-lymphocyte population hides the lymphoma cells. From a clinical point of view, the FCI analysis of the CSF simplified the diagnostic procedures in three patients, avoiding the need to perform a brain biopsy and allowing chemotherapy to be started quickly. The four remaining discordant CSF samples were from patients receiving intrathecal therapy against a NHL with leptomeningeal involvement and, again, a low percentage of malignant cells impaired the accuracy of classical cytological studies. These findings suggest that the possibility of using FCI to investigate the clearance of malignant cells in the CSF and to monitor the efficacy of chemotherapy should be fully explored.

The present study appears to solve a problem that often arises in the study of CSF samples by flow cytometry, related to the best choice of a restricted panel of monoclonal antibodies. Most of the samples in our work were from patients with a NHL with unknown immunophenotypic features. However, the antibody panel proposed here, which was designed to detect aberrant immunophenotypic features in CSF lymphoid populations, successfully identified malignant cells in most cases. In this sense, the co-expression of CD10 and mature B-cell antigens (CD19 and CD20) seems particularly useful as it was found in 17 of 18 cases with a malignant B-cell population. CD10 is a molecule expressed in the early stages of lymphoid differentiation, that expression becoming negative in mature B and T lymphocytes [20], but several B-cell malignancies such as diffuse large B-cell lymphoma and Burkitt's lymphoma may frequently show CD10 expression on the cell surface [8]. This high frequency of CD10 expression is important because identification of CD10 can be substituted for identification of surface immunoglobulin light chain restriction, especially in those CSF samples with a low number of cells or a low percentage of B cells where it is very difficult to detect sIg light chains. In addition, use of CD10 expression may overcome the low sensitivity of sIgκ/sIgλ/CD19 triple-staining to detect a monoclonal B-cell population once mixed with a polyclonal B-cell population [21]. This finding needs to be taken into account in HIV-positive patients, because it has been found that CSF B lymphocytes are present in higher numbers in HIV-positive patients [22] than in HIV-negative patients [12,23]. However, in all patients with a high percentage of B cells and normal immunophenotypic features we discovered other nonmalignant conditions.

In conclusion, despite the low numbers of malignant cells in the CSF samples and the unknown immunophenotypical features of the lymphoma, flow cytometry was found to be of clinical value for the investigation of CNS lymphoma in AIDS patients. It is more sensitive than classical cytomorphological methods for the quick and accurate diagnosis of CNS involvement. In addition, it may provide a simple way of monitoring the efficacy of chemotherapy in eradicating tumour cells from the CSF.


The authors thank Professor Alberto Orfao for his critical review, and José María Subirá for editing the manuscript.