CD38 as a prognostic factor in B cell chronic lymphocytic leukaemia (B-CLL): Comparison of three approaches to analyze its expression




Increased CD38 expression by leukemic cells has been suggested as an adverse prognostic factor in B-CLL. Several approaches have been proposed to quantify its level of expression by flow cytometry.


We compared the use of (i) the percentage of CD38 positive cells, (ii) CD38 antibodies bound per cell (ABC), and (iii) a semi-quantitative method based on the shape of the CD38 histogram, within a cohort of 78 B-CLL patients.


A decreased overall survival was seen with >30% CD38 positivity among B-CLL cells, with CD38 ABC >100, and with bimodal or unimodal, strongly positive CD38 histograms. However, patients with unimodal weakly positive CD38 histograms also showed a significantly reduced survival as did patients with intermediate proportions (i.e. 5–30%) of CD38+ cells. Furthermore, within the group with <5% CD38 positivity among their B-CLL cells, 84% of patients showed prognostically favourable mutated IGVH gene segments and 100% had low ZAP70 gene expression. For 5–30% CD38 positivity, these proportions were 50 and 83%, while for >30% CD38 positivity, these proportion were only 28 and 56%, respectively.


We found a simple method of quantitation of CD38 expression (i.e., >5% CD38 positivity among B-CLL cells) to be sufficient to identify patients with an unfavourable prognosis. The level of CD38 expression as defined with this method correlated well with the IGVH mutation status and ZAP70 gene expression. © 2006 International Society for Analytical Cytology

Chronic B-lymphocytic leukaemia (B-CLL) is a disease with a heterogeneous prognosis. While many patients will never require treatment and have a survival of several decades, others suffer from a much more aggressive, rapidly evolving form. Identification of these subgroups and insight into the prognosis for each individual patient will be important to determine individualized treatment strategies.

In the past five years, several authors have shown the importance of CD38 as an independent prognostic factor in B-CLL (1–7). These studies are unanimous in showing that a high proportion of CD38-expressing cells among the leukemic cells is associated with a significantly poorer prognosis, defined either as overall survival (OS) or as event-free survival or as treatment-free interval. Although most authors use 30% as a cut-off point for CD38 positivity (expressed as fraction of leukemic cells), others have applied a cut-off point of 20%. Furthermore, Ibrahim et al. (4) have shown that the expression patterns of CD38 on B-CLL cells are complex; specifically, the intensity of CD38 expression may be very heterogeneous in spite of the clonal nature of the disease. These observations have raised the question whether or not quantification of CD38 by means of instrument-independent units of fluorescence (FL) intensity would provide a useful, more standardized approach than the arbitrary use of a threshold to distinguish CD38+ from CD38− B-CLL cells. This quantification can be achieved by measuring CD38 ABC (antibodies bound per cell). The first study on CD38 ABC of leukemic cells in CLL patients was published in 2002 by Mainou-Fowler et al. (8). A significant shorter time to first treatment and a shorter disease-specific survival was seen in patients with more than 250 ABC as compared to those with less than 250 ABC. Hsi et al. (9) showed a significant shorter OS for patients with CD38 ABC >100, when compared to a group with <100 CD38 ABC. However, the complex patterns of CD38 expression by B-CLL cells in some patients prompted Ghia et al. (10) to propose an approach in which they divided patients into three groups according to CD38 expression by the leukemic cells: those who were uniformly CD38 negative, those who were CD38 positive, and those with a bimodal profile (i.e., 2 distinct populations, CD38 negative and positive). Indeed, significant differences were seen between the three groups for risk status, disease progression, and treatment requirement. For all these parameters, the patients with a CD38 bimodal profile did worse than the homogeneously negative group, but better than the homogeneously CD38 positive group. The large variation in proportions of CD38+ cells (i.e. between 8 and 78%) in the group with a bimodal profile suggests that even a small percentage of CD38+ cells may be associated with disease progression. In line with this observation, Kröber et al. (11) found that a 7% threshold discriminated best between the two subgroups with different survival probabilities.

To evaluate the most informative and prognostically most relevant way to score CD38 expression, we compared three approaches to study CD38 expression in a cohort of 78 patients with B-CLL: (i) percentage of CD38 positive cells (% CD38+), (ii) CD38 ABC, and (iii) an extended, qualitative method of reporting CD38 expression profile. We confirmed the published CD38 expression patterns, i.e., homogeneously negative, homogeneously positive or bimodal. In addition, we demonstrated that (i) the proportions of CD38+ cells in case of bimodal distributions vary greatly, and that (ii) the intensity of CD38 expression in patients with unimodal CD38+ B-CLL populations shows a wide variation between dull and bright. Importantly, even small percentages of CD38+ CLL cells are associated with decreased survival, and a low cut-off (i.e. 5% instead of 30%) is better able to differentiate between a good and decreased OS rate.



Seventy-eight patients (51 male, 27 female) were included in the study. CLL was diagnosed based on clinical criteria, peripheral blood and bone marrow morphology, and immunophenotyping (WHO criteria). Median age at the time of diagnosis was 60 years (22–91). Fifty-two percent of patients were in Binet stage A, 38% in stage B, and 20% in stage C. Median time from diagnosis to laboratory analysis was 28 months (range 0–257). None of the patients had received a haematopoietic stem cell transplantation.

Cell Preparation

The QuantiBRITE technique for quantitation of antigen expression (BD Biosciences, San Jose, CA) is designed to stain fresh whole blood. To validate its use on frozen peripheral blood mononuclear cells (PBMNC), we used EDTA-anticoagulated and heparinized peripheral blood samples from 6 CLL patients and 6 healthy volunteers. EDTA anticoagulated samples were used for immediate whole blood analysis, according to the QuantiBRITE manufacturer's recommendations. Mononuclear cells were isolated from heparinized samples using Lymphoprep (Nycomed, Oslo, Norway) with a density of 1,077 g/ml. After washing, the mononuclear cell suspension of each patient or donor was split to two parts: one was stained and analyzed immediately, and the other was cryopreserved at −196°C for at least 24 h in RPMI-1640 (Biowhittaker, Verviers, Belgium) with 10% DMSO (Sigma Chemical, St. Louis, MO) and 50% bovine calf serum (BSC; MultiCell, Wisent Inc., St. Bruno, Quebec, Canada).

We studied cryopreserved PBMC of 78 CLL patients diagnosed according to aforementioned criteria. The ampoules were thawed rapidly by immersion in a 37°C water bath to 0°C and diluted in ice-cold RPMI-1640 containing 20% BCS. After washing in PBS/1% BSA (Sigma), cells were checked for viability by trypan blue staining.

Immunostaining and Flow cytometric Analysis

For the validation study, whole blood (≤100 μl containing a maximum of 20 × 106) was stained with a combination of CD5 FITC, CD38 PE, and CD19 APC monoclonal antibodies (mAb; all from BD Biosciences [San Jose, CA]) for 15 min at room temperature (RT). The CD38 mAb had a ≥95% 1:1 fluorochrome to antibody ratio, which enabled quantification of CD38 in units of ABC. Staining was followed by lysis of red blood cells with ammonium chloride (NH4Cl 8.26 g, KHCO3 1.0 g, Na3EDTA 37 mg (all from Sigma) aqua distillata qs one liter) for 15 min at RT. After washing and centrifugation, the pellet of cells was incubated for 15 min at RT with 500 μl PBS containing 7-AAD (1 mg/l, Sigma). Cells were analyzed within 1 h after staining. Freshly isolated or cryopreserved and thawed PBMC (2 × 106 cells) were also stained for 15 min at RT with the same combination of mAb. After washing and centrifugation, the pellet of cells was dissolved in 500 μl PBS/7-AAD, incubated for 15 min at RT, and analyzed within 1 h after staining.

PBMC of the 78 CLL patients were stained according to the same protocol. Flow cytometric acquisition was performed by collecting 10 × 103 viable B lymphocytes (i.e. CD19+, 7-AAD). CD5+ population was defined either based on the bimodal shape of the CD5 histogram, or the marker was set to include >99% of unstained cells when a single peak of CD5+ B cells was observed. CD38 ABC units were calculated as recommended by the manufacturer (12). The percent CD38+ cells was determined using a fixed threshold. On 10 samples with a uniform negative population (Fig. 2, panel 1), the cursor was set at the right foot of the population so that <0.2% of cells were classified as positive. The fixed threshold was defined as the mean channel number of these 10 analyses. For the qualitative description of CD38 expression, 3 observers independently scored expression patterns of CD38 on B-CLL cells, as described in the legend to Figure 2.

Figure 2.

Assignment of scores to 5 different CD38 histogram types. Type I, unimodal negative; type II, bimodal with less than 50% cells with strong expression; type III, bimodal with more than 50% cells with strong expression; type IV, unimodal dim expression (two examples shown); type V, unimodal strong expression. Expression was considered bimodal when the second population contained >5% but <95% of selected events.

Analysis of IGVH Somatic Mutation Status

Genomic DNA was prepared from cryopreserved cells, and immunoglobulin heavy chain (IGH) gene rearrangements were amplified in duplicate by multiplex PCR, using a mixture of VH-FR1 (VH1-VH6) family primers in combination with a consensus JH primer, as designed by the BIOMED-2 Concerted Action (13). Following amplification, heteroduplex analysis was performed to identify the presence of one or multiple clonal bands (14). In the case of multiple clonal PCR products, bands were cut out from the polyacrylamide gel and eluted for sequencing; single monoclonal PCR products were directly sequenced using the ABI 3100 automated DNA sequencer (Applied Biosystems, Foster City, CA, USA), according to the manufacturer's instructions. Sequencing analysis was started using the consensus JH primer as sequencing primer on PCR products from the two independent reactions. Following identification of the involved VH family member via the IMGT database (, for confirmation, the PCR products were sequenced with the appropriate VH FR1 family primer. From the aligned sequences, a consensus sequence was made and compared with the IMGT data base. Involved VH, DH, and JH gene segments and reading frame were determined. Percentage homology to the closest germ line VH gene segment was calculated as the number of identical nucleotides in the entire VH segment divided by the total number of nucleotides sequenced until the last complete germ line exon that could be recognized, and excluding the primer sequence. Ninety-eight percent homology was taken as the cut-off point to classify cases as non-mutated or mutated.

ZAP70 Gene Expression Analysis

The expression level of the ZAP70 gene was determined by real-time quantitative PCR (RQ-PCR) on an ABI PRISM 7700 sequence detector (Applied Biosystems, Nieuwerkerk aan den Yssel, the Netherlands). A cut-off of 4.6 ddCT was used to distinguish low and high ZAP70 expression (15).

Statistical Analysis

Correlation between data sets was calculated using Spearman's rank correlation method. OS estimates were calculated using the Kaplan–Meier method and differences in OS between patient groups were compared with the logrank test. As it was unknown whether or not the expression of CD38 was stable during the course of the disease, and testing was retrospective, we defined OS as the time lapse between blood sampling and death. Patients who were still alive on the date of last contact were censored from that moment.


Validation of the QuantiBRITE Technique for Quantification of CD38 Expression in Units of ABCon Cryopreserved PBMC

For this purpose, we studied fresh and frozen samples from 6 healthy volunteers and 6 B-CLL patients. We compared leukocytes obtained with the whole blood method, as recommended by the manufacturer, freshly isolated PBMC, and cryopreserved and thawed PBMC (Fig. 1). All approaches showed excellent positive correlations (r = 0.98 for whole blood vs. fresh PBMC, and r = 0.93 for whole blood vs. cryopreserved PBMC). The mean difference between any set of paired observations did not differ significantly from zero.

Figure 1.

Validation of QuantiBRITE CD38 staining on cryopreserved CD19+ PBMC. Comparison of freshly isolated PBMC and fresh whole blood for healthy volunteers (Δ) and CLL patients (□), and comparison of cryopreserved and thawed PBMC and fresh whole blood for healthy volunteers (▴) and CLL patients (▪).

Results of CD38 Analysis on Cryopreserved Cells of B-CLL Patients

Forty-three of the 78 patients in our study group (55%) had less than 30% CD38+ cells among their CD19+ CD5+ B-CLL cells, and the remaining 35 patients presented with more than 30% CD38+ cells. Within the latter group 12/35 patients had more than 90% CD38+ cells. Considering the intensity of CD38 expression (in ABC) on the B-CLL cells, 31 patients presented with a median ABC value smaller than 100, 24 with ABC value between 100 and 499, and 21 with ABC value between 500 and 4,999. We identified only 2 patients with ABC between 5,000 and 10,000. When studying the pattern of CD38 expression within the B-CLL cells, we could identify 5 different histogram types (Fig. 2). We classified these histograms according to the distribution (uni- or bimodal) and intensity of CD38 expression. In this way, we defined histograms that were unimodal negative (type I), bimodal with less than 50% cells with strong expression (type II), bimodal with more than 50% cells with strong expression (type III), unimodal weak expression (type IV), or unimodal strong expression (type V). Table 1 shows the number of patients for each histogram type as well as the median and ranges of % CD38+ cells and median CD38 ABC per histogram type. Obviously, unimodal negative histogram types have low % CD38+ and median CD38 ABC values, while unimodal strong expression histograms have very high % CD38+ and CD38 ABC. When a bimodal histogram was found, the % CD38+ cells could vary markedly, i.e. ranging from 8.1 to 86.4%. There was also a large variation in median CD38 ABC values of the B-CLL cells within this histogram type. Table 1 further shows that patients with a unimodal, weak expression histogram (type IV) could be distinguished from unimodal negative (type I) and unimodal strong expression (type V) on basis of both their median ABC values and percentage positive cells, since there was no overlap between the three groups.

Table 1. % CD38+ and CD38ABC for Each Type of CD38 Histogram
 Number% CD38+ median (range)Median CD38 ABC value median (range)
I (unimodal negative)311.4 (0.3–4.1)80 (66–108)
II (bimodal ≤50% strong expression)1422.1 (8.1–48.3)166 (86–367)
III (bimodal >50% strong expression)770.3 (55.5–86.4)929 (468–1,809)
IV (unimodal weak expression)1336.9 (15.0–82.1)270 (138–942)
V (unimodal strong expression)1397.3 (85.6–99.9)3,049 (1,200–8,348)

OS According to CD38 Percentage, Median CD38 ABC Values, and CD38 Expression Pattern

Figure 3 shows the OS estimates within our patient group. A significantly decreased OS was observed when median CD38 ABC values were >100 (P = 0.003; Fig. 3a) and when %CD38+ was >30% (P = 0.01; Fig. 3b). For the qualitative classification of CD38 histograms, Types II en III (bimodal expression with less and more than 50% cells with strong expression) were pooled because of the too low numbers of patients in the individual groups. Histogram types II, III, IV, and V were all associated with a significantly shortened OS in comparison to type I (P = 0.005; Fig. 3c). Since types II and IV histogram groups included patients with %CD38+ between 5 and 30%, we also analyzed OS for patients with <5%, 5–30%, and >30% CD38+ (Fig. 3d). This analysis shows that both the 5–30% and the >30% group show a significantly reduced OS in comparison to the group with <5% CD38+ cells (P = 0.01).

Figure 3.

OS analysis of 78 B-CLL patients by median ABC values <100/>100 (A), percentage positive cells <30%/>30% (B), type of histogram (C), and percentage positive cells <5%, 5–30%, and >30% (D).

Correlation of % CD38+ Cells with IGVH Mutation status and ZAP70 Gene Expression

The reduced survival for both the 5–30% and the >30% CD38+ groups paralleled the IGVH mutation status and ZAP70 expression within these subgroups (Table 2). Within the <5% group, mutation status was known for 25 patients and 21 (84%) of them showed <98% sequence homology, i.e. were somatically hypermutated, which is prognostically favourable. ZAP70 expression was low in 100% of the 30 patients who were tested for this gene. In the 5–30% group, 50% (5/10) of analyzed patients showed mutated IGVH gene segments, and 83% (10/12) showed low ZAP70 expression. When patients had more than 30% CD38+ cells, 32% (9/28) showed mutated IGVH gene segments and 56% showed low ZAP70 expression.

Table 2. Percentage of Patients Within the Groups of <5, 5–30,and >30% CD38+ Cells, who Have Mutated IGVH Genes or Low ZAP70 Gene Expression
 IGVH status mutated (%)ZAP70 expression<4,6 ddCt (%)
<5% CD38+84 (21/25)100 (30/30)
5–30% CD38+50 (5/10)83 (10/12)
>30% CD38+32 (9/28)56 (18/32)


The relevance of measuring CD38 as a prognostic factor in B-CLL has convincingly been shown over the last few years. While several authors have addressed the issue of measuring either % CD38+ or CD38 ABC and others have shown that the pattern of CD38 expression is also informative, our study compares these three approaches within a single patient group.

Our results confirm earlier reports that both >30% CD38 positive cells and a median CD38 ABC > 100 have predictive value on the duration of OS in B-CLL. Furthermore, we found that several CD38 histogram types can be distinguished, and that these CD38 histogram types are associated with a different outcome. In fact, only the group of patients with a unimodal negative histogram (type I) showed a relatively good OS. All other groups, even those patients with %CD38+ cells between 5 and 30%, showed a decreased survival. Thus, the existence of even low proportions of CD38 positive cells, as can be seen in type II and IV histograms, may have a negative impact on the prognosis of B-CLL patients. Similar results haven been published by Kröber et al (11), who found that the best separation of two prognostic subgroups was achieved for a cut-off value of 7% CD38+ cells. Although many groups now use 30% as a cut-off, we agree with Kröger et al that a low threshold is more informative. We base this conclusion on the significant correlation between low levels of CD38 positivity and mutated IGVH status, which is a very strong, independent prognostic factor in B-CLL (16). Importantly, our findings on ZAP70 expression by the B-CLL cells are in line with these results. Low levels of CD38 expression correlated with low levels of ZAP70, while high levels of ZAP70 expression were observed in patients with >5% CD38+ B-CLL cells. In the past few years, ZAP-70, which is normally expressed only on T and NK cells, has appeared to be an important prognostic factor, and is also functionally important in CLL (17–19).

For routine work-up of B-CLL patients, a cost and time efficient method to obtain optimal prognostic information is desirable. In this context, it is relevant that neither the quantification of CD38 expression in terms of units of ABC, nor the qualitative classification of CD38 fluorescence histograms adds extra information to the simple expression of CD38 results as percent positive cells (i.e. fraction of B-CLL cells). It is of possible pathogenetic interest that the subgroup of patients with 5–30% CD38+ cells among their B-CLL cells already show less often favourable prognostic factors and reduced OS.