Neuropilin -1(NRP-1, BDCA4 or CD304) is a non-tyrosine kinase co-receptor for semaphorins and vascular endothelial growth factor (VEGF), implicated in neuronal guidance and angiogenesis (1). CD304 is expressed on endothelial cells (1), blood plasmacytoid dendritic cells (pDC) (2), and diverse human solid tumors (1). Overexpression of CD304 in acute myeloid leukemias (AML) is significantly correlated with survival (3, 4), and might constitute a promising target for anti-angiogenic treatment strategies (5). While evaluating CD304 as a marker for diagnosing pDC leukemia, we found that it was also expressed in 12% of non-pDC acute leukemia, especially B-lineage acute lymphoblastic leukemia (B-ALL) (2). Karjalainen et al. also reported increased levels of NRP-1 in bone marrow specimens from AML and ALL patients, but these results were obtained using a small number of ALL cases (6). Most recently, the St. Jude Children's Hospital group published a very interesting study demonstrating that CD304 was one the most differentially expressed marker between ALL cells and normal B-cell precursors by genome-wide gene expression analysis and by FC (7).
Due to the progress made in the management of ALL, response rates to remission induction therapy have reached high levels, but ∼20% of children and up to 60% of adults relapse (8). Adapting the treatment intensity to the individual relapse risk is critical, and the level of minimal residual disease (MRD) has emerged as a major prognostic factor (9, 10). Quantification of MRD by FC is currently under investigation in European multicenter trials (11, 12), which will hopefully provide new information on its capacity to produce results comparable to a molecular approach (13). The major issue for the FC strategy is to specifically quantify neoplastic B lymphoblasts by differentiating them from their normal counterparts (called hematogones). This is based on the identification of markers defining a leukemia-associated phenotype (LAP) which is expressed on leukemic B lymphoblasts, absent on hematogones, and ideally stable during therapy and at relapse. Because phenotypic switch is common during treatment (14), the identification of new markers is necessary so that multiple LAPs be available and used to avoid pitfalls in MRD detection.
We studied the CD304 expression profile of the normal B lineage and of the leukemic cells from 50 B-ALL patients, which led us to suggest its potential usefulness as a new marker for MRD monitoring. We compared the specificity of CD304 with that of established LAPs and studied its expression during therapy and in particular subsets of B-ALL.
Materials and Methods
Seventy patients diagnosed with B-ALL (2002–2006) at the pediatric [n = 45, EORTC-58951 protocol (11 )] and adult [n = 25, GRAALL-2003 protocol (12)] haematology departments of Besançon University Hospital were evaluated at diagnosis. Fifty-nine (36 children and 23 adults) were monitored by FC at least once during follow-up. This study was approved by the Besançon local ethics committee.
Bone Marrow Collection and Sample Preparation
Flow cytometry analysis was performed on leukemic bone marrow samples collected at diagnosis (n = 70), and after complete remission was achieved (n = 145) at different time-points during follow-up according to the mandatory protocols (11, 12). Normal bone marrow samples were collected from patients without blood disease during cardio-thoracic surgery (n = 15). Regenerative bone marrow enriched with B-cell precursors (hematogones) was obtained from patients after chemotherapy of non-B malignancies [AML (n = 9), solid tumors (n = 3), T-ALL (n = 3)] and from B-ALL patients with negative MRD according to molecular assays (n = 9) (9, 10). B-ALL bone marrow cells collected at diagnosis and stored at −196°C (n = 5) were used for lymphoblast dilution experiments in normal bone marrow (blast concentration 10−1 to 10−4). All FC analyses were performed on ficoll-separated mononuclear cells.
Flow Cytometry Detection of Minimal Residual Disease
The MRD assessment was performed using a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA), Cell Quest Pro software (BD Biosciences), and a previously described strategy (15, 16). The same combinations of four monoclonal antibodies (mAbs) were applied to diagnostic samples, remission bone marrow samples, and normal or regenerative bone marrow samples, to the order of 106 mononuclear cells per combination. A significant abnormal cell population was defined as a homogeneous cluster of at least 20 LAP+ cells(15).
The antibody panel included combinations consisting of three fixed mAbs against CD19 (J4.119), CD45 (J33) (Beckman Coulter, Miami, FL), and CD10 (HI10a, BD Biosciences), coupled with the APC, PC7, and FITC fluorochromes, respectively, and a fourth variable mAb coupled with PE against one of the following antigens: CD22 (S-HCL-1), CD20 (L27), CD34 (8G12), CD38 (HB7) CD13 (L138), CD33 (P67.6), CD123 (9F5), CD117 (YB5.B8) (BD Biosciences), CD15 (80H5), CD65 (88H7) (Beckman Coulter), CD58 (B-L28, Diaclone, Besançon, France), or CD304 (AD.17F6, Miltenyi Biotec, Paris, France). CD24 expression was evaluated with a FITC-conjugated mAb (SN3, Dako, Trappes, France) in association with PE-CD34 (8G12, BD Biosciences), and the same CD45 and CD10 backbone. Finally, an isotype-matched control (CD45-APC/CD19-PC7/CD10-FITC/IgG1-PE) was used to confirm the specificity of the LAP evaluation and to calculate the mean fluorescence intensity ratio (MFIR) by dividing the mean fluorescence intensity of each PE-conjugated mAb by the MFI of the isotype control mAb.
CD304 was considered to be a potential LAP if the CD304 MFIR of leukemia cells was greater than the highest CD304 MFIR value recorded among normal B-cell precursors (7), or if at least 80% of the blasts were localized within the empty gate upstream normal bone marrow B-cells (17).
Compensation was performed with monostained blood cells and the manual method using medians of fluorescence intensity.
Molecular Quantification of Minimal Residual Disease
MRD samples from patients with CD304+BCR-ABL+ B-ALL and dilutions of diagnosis bone marrow cells from these patients in regenerative bone marrow underwent BCR-ABL transcript quantification using a standardized RT-qPCR technique (18) to correlate the results with CD304-based quantification of leukemic cells.
Unless otherwise stated, results are expressed as mean ± standard error of the mean (SEM), [range]. The Student, Mann-Whitney, and Z-tests were used (SigmaStat, Systat Inc, San Jose, California).
Results and Discussion
Potential Usefulness of CD304 in Minimal Residual Disease Detection
We evaluated CD304 expression on mature and immature B-cells in 15 normal or regenerative bone marrow samples. This expression was extremely low or undetectable (mean MFIR ± SEM [range]: 1.3 ± 0.1 [1–2.2], Fig. 1, Table 1). It was important to include in this evaluation not only bone marrow samples from healthy donors (n = 6), but also regenerative bone marrow samples from children and adult patients during chemotherapy (n = 9). As a result, some specimen contained a high proportion of hematogones, which are the most difficult cells to distinguish from ALL lymphoblasts by FCM. What is even more relevant is that B-ALL patients (with negative MRD according to molecular assays) were the most representative of the bone marrow cellularity during MRD studies. On B-ALL cells CD304 was over-expressed in 48% (24/50) of the patients (MFIR>2.2, i.e., highest MFIR value on normal B-cell precursors), and expressed in more than 80% of the CD45lowCD19+ leukemia cells in 36% (18/50) of B-ALL patients (Table 1). In these cases, the difference in fluorescence intensity compared with normal immature B-cells was high (P < 0.001), allowing discriminatory identification of lymphoblasts and suggesting the potential usefulness of CD304 as a new LAP for monitoring MRD in ∼40% of B-ALL patients.
Table 1. CD304 expression on normal bone marrow B cells and B-ALL and applicability of CD304 in B-ALL
A: CD304 expression on hematogones and B-lymphocytes. Mean ± SEM [range] of the mean fluorescence intensity ratio (MFIR) is indicated for each group B and C: Number of cases for which CD304 was considered as a LAP based on MFIR evaluation (B) or by considering that it was expressed on more than 80% of the blasts (C). aHighest value on hematogones >2.2. *Mann–Whitney rank sum test compared to hematogones.
A. Hematogones (MFIR)
1.3 ± 0.1 [1–2.2], n = 15
Mature B lymphocytes (MFIR)
1.7 ± 0.4 [0.8–2.6], n = 15
B. B-ALL cases with CD304 MFIR >2,2a
MFIR, mean ± SEM [range]
20.5 ± 3.5 [2.4–70]
C. B-ALL cases with more than 80% of lymphoblasts expressing CD304
MFIR, mean ± SEM [range]
26.5 ± 3.7 [8.8–70]
Sensitivity and Specificity of CD304-Based Minimal Residual Disease Detection by Four-Color Flow Cytometry
Evaluating of the sensitivity and specificity of this approach is necessary to ensure that it is adapted to the therapeutic stratification thresholds mandated by the treatment protocols (9, 19). The ability to detect rare cells by FC depends on the number of cells that can be analyzed, and the degree of phenotypic difference between target leukemia cells and the remaining cells, particularly hematogones. Although four-color FC has the widely reported ability to detect as few as one leukemic blast among 104 normal cells (15), this extreme limit of sensitivity may not be achieved with every LAP.
As a prerequisite for MRD evaluation, the phenotype of normal B-cell precursors was evaluated using normal and regenerative bone marrow, and the same monoclonal antibody combinations and FC protocol as for MRD evaluation (Fig. 2). To assess the specificity of the different LAPs used (including CD304) and so, the maximum sensitivity (or limit of detection) that can be reached for MRD analysis, a quantitative study of normal minority B-cell populations that may express LAPs was performed (Fig. 2, Table 2). Indeed, in normal or reactive bone marrow, some normal B-cells can be quantified in the LAP gates (Fig. 2a–2k). These events limit the sensitivity of MRD detection in a four colors FC setting. We observed a low specificity of CD13 and CD33, or others phenotypes such as CD10+CD22bright or CD10−CD20−, which are expressed in association with CD19 in quantities greater than 10−4 in normal bone marrow (Table 2). These values are consistent with those reported by Ciudad et al. (20). In contrast, the expression of CD304, bright expression of CD58 or CD123, or dim expression of CD38 by B-cells, is indicative of a very specific B-ALL phenotype because it is only found at quantities less than 10−4 in normal bone marrow or close to 10−4 for the bright expression of CD34 or dim expression of CD24 (Table 2).
Table 2. Applicability and specificity of each LAP
A. Frequency in B ALL
B. Quantification of B-cells presenting the LAP in normal BM
C. Maximum sensitivity of LAP-based MRD detection
Number of cases (%)
Mean ± SEM
Mean +2 SEM
A: The applicability of each LAP was the number of B-ALL cases expressing this LAP among the total of B-ALL patients evaluated. B: The FC technique described to assess MRD was applied to quantify the number of normal or regenerative bone marrow B-cells present in the “empty gates” (a to k). Results were expressed as the mean ± standard error of the mean (SEM) number of mononuclear cells expressing each LAP in normal and regenerative bone marrow samples after subtracting the signal obtained in the same gate with the isotype control. C: The maximum sensitivity of MRD detection using each LAP was determined as the mean +2 SEM number of normal cells present in the “empty gates.”
Only cases with CD10 and CD34 expression higher than most immature hematogones were considered.
CD45 was never used alone.
Mature B lymphocytes were excluded by their CD45bright expression.
The sensitivity of CD304-based MRD quantification was confirmed by serial dilutions of CD304+ B-ALL cells (n = 5) in regenerative bone marrow. We obtained a linear quantification and similar results regardless of the LAP used (CD10bright/CD38dim, CD58bright, CD304+), indicating that CD304 allows MRD detection with a sensitivity of at least 10−4, which is in accordance with the low quantification of CD19+CD304+ events in normal bone marrow (3 × 10−5 ± 1 × 10−5; Table 2). In contrast, MRD quantification was not linear until 10−4 with less specific LAPs such as CD10+CD22bright or CD13 (data not shown). For serial dilution of CD304+BCR-ABL+ leukemia cells, CD304-based MRD detection correlated with molecular quantification (Fig. 3A).
Taken together, these data show that some, but not all, of the LAPs commonly used in FC protocols allow for a specific quantification of MRD at the level of 10−4, and suggest that we need to use the most specific LAPs to avoid false-positive results at the level of sensitivity of 10−4, such as CD304.
Stability of Leukemia-Associated Phenotypes During Evolution
Changes in the leukemic immunological fingerprints during treatment or at relapse can be observed in up to 70% of B-ALL cases (14), exposing to false negative assessment of MRD by FC. In our series, the expression intensity of markers was modulated in 8 of 13 patients evaluated by FC at diagnosis and relapse. We noted an up-regulation of CD13, CD22, CD33, and CD20, and down-regulation of CD38, CD13, CD34, CD10, CD22, and CD45.
This problem can be addressed by using multiple markers specific to leukemic blasts and stable during treatment. Therefore, the stability of CD304 expression must be evaluated on follow-up samples. For eight patients with CD304+ leukemic blasts, we assessed the persistence of MRD in samples collected after remission induction chemotherapy (n = 20). MRD was found to be negative (<10−4) in 14 samples based on the lack of CD304 expression and other LAPs among those listed in Table 2. For one patient, MRD was positive with similar levels using CD304 or other LAPs (Fig. 3B). For 2 additional CD304+BCR-ABL+ patients, MRD was positive based on both CD304 and BCR-ABL quantification (Fig. 3C). Note that for the MRD1 point represented in Figure 3C, we observed a deviation of 1 log between MRD levels measured by FC and by RT-qPCR, which is not surprising since the targets measured are distinct (cells by FC and mBCR-ABL transcripts by molecular quantification). Moreover, two patients with CD304+ B-ALL relapsed with persistent expression of CD304 (Fig. 3B). Altogether, these results suggest the stability of CD304 expression in the blastic population during treatment, which must be further evaluated in a larger cohort of patients.
Characterization of CD304+ B-Acute Lymphoblastic Leukemia
We compared the characteristics of CD304+ B-ALL patients with those of the CD304− B-ALL patients (Table 3). The sex ratio was slightly different (P = 0.044) with a higher proportion of females in the CD304+ group (50%). TEL-AML1-positive cases were more frequent in the CD304+ group (P = 0.045). In our series of 50 B-ALL patients, all identified TEL-AML1+ cases (n = 6) were CD304+. Therefore, we analyzed CD304 expression in seven additional TEL-AML1+ B-ALL cases diagnosed before 2002 (data not shown). This marker was over-expressed in all cases (MFIR range from 3.3 to 12.6). Although the frequency of BCR-ABL was higher in the CD304+ group (32%) compared with the CD304− group (5%), it was not significantly different (P = 0.08).
Table 3. Characteristics of CD304+ and CD304− B-ALL patients
CD304+ B-ALL (n = 24)
CD304− B-ALL (n = 26)
Mean ± SEM [range]
Mean ± SEM [range]
SEM, standard error of the mean; CR, postinduction complete remission; EGIL, European Group for the Immunological Characterization of Leukemia (21).
Mann–Whitney rank-sum test.
CD13, CD15, CD33, CD65, and CD117.
The B-ALL fusion transcripts (BCR-ABL, MLL-AF4, E2A-PBX1), as well as hyperdiploidy and hypodiploidy, were not differentially expressed between CD304+ or CD304− B-ALL.
eMRD was quantified with a competitive polymerase chain reaction assay (9).
The t(12;21)(p13;q22) chromosomal translocation, which generates the TEL-AML1 fusion gene and protein, is the most frequent gene recombination in paediatric B-ALL and is associated with favorable prognosis. The relationship between the TEL-AML1 fusion gene and VEGF receptor CD304 is currently not known (22), but enhanced expression of this receptor in relapsed ALL supports its potential role in lymphoid survival (23). As recently described, CD304 could also be used in combination with CD9 to predict TEL-AML1+ ALL (24). Interestingly, the over-expression of CD304 mRNA has been described in several cases of CBFβ-MYH11 AML (25). Because AML1 (CBFα2) is also a member of the core binding factor (CBF) family, whether the elevated expression of CD304 in TEL-AML1+ ALL correlates with the expression of myeloid markers, such as CD13 and CD33, should be investigated in a larger cohort. CD304 also appears to be a promising entry receptor and therapeutic target for pro-apoptotic agents coupled with a specific peptide ligand (6). Our results show that this ligand-directed therapeutic approach could be applicable in a large subset of B-ALL patients.
The identification of leukemic cell specific markers will help us to improve the sensitivity of FC-based MRD quantification. Confirming recently published data (7), our results demonstrate that, in about 40% of B-ALL patients, CD304 expression intensity allowed to discriminate CD304+ leukemic blasts from CD304− normal B-cell precursors. They also provide additional useful information about the reliability of this marker as CD304 expression appeared stable during treatment and at relapse, which should limit false negative MRD results, especially when combined with other stable markers. Overall, our data highlighted CD304 as a promising MRD marker in 40% of B-ALL patients. The potential of this MRD marker must be evaluated now in prospective multicenter trials and a better understanding of the biological function of CD304, especially in TEL-AML1+ ALL, might also reveal its potential as a prognosis factor or therapeutic target.
Author Contributions: FGO and FS initiated and designed the study. PSR, EP, ED, FLa, and FLe provided the clinical samples. FGO, FSc, and CF provided biological data. FS, FA, AD, and MB performed the flow cytometry analysis. CF executed molecular biology studies. FS, FGO, and PSR analyzed the data and wrote the manuscript. CF, PS, RG, and ES contributed to improve the manuscript.
The authors thank Sabeha Biichle and Faezah Benamar for technical support, Sylvain Perruche, Claire Latruffe, and Sarah Odrion for their help in preparing this manuscript. This work was supported by a grant from the Comité Départemental de la Ligue contre le Cancer du Doubs. The BioMonitoring platform was supported by the Centre d'Investigation Clinique Intégré en Biotherapies (CIC-BT506) of Besançon University Hospital, France.