Bone marrow minimal disseminated disease (MDD) and minimal residual disease (MRD) in childhood T-cell lymphoblastic lymphoma stage III, detected by flow cytometry (FC) and real-time quantitative polymerase chain reaction (RQ-PCR)†
Batia Stark, Smadar Avigad, and Drorit Luria contributed equally to this work.
Despite overlapping features of T-cell lymphoblastic lymphoma (T-LLy) and T-cell acute lymphoblastic leukemia (T-ALL), which respond favorably to T-ALL treatment, clinical and biological differences exist. We retrospectively assessed the prevalence of submicroscopic bone marrow (BM) minimal disseminated disease (MDD) at diagnosis and the early response to treatment (minimal residual disease—MRD) and their prognostic significance in 17 children with stage III T-LLy treated according to Berlin-Frankfurt-Munster (BFM) non-Hodgkin lymphoma protocols.
Four-color flow cytometry (FC) was used for lymphoma associated immunophenotype and real-time quantitative polymerase chain reaction (RQ-PCR) for T-cell receptor (TCR β/δ/γ) gene rearrangements with at least 0.01% sensitivity.
Two markers per patient were identified in all cases using FC and in 80% using RQ-PCR. BM MDD at diagnosis of ≥0.01% was detected by FC and RQ-PCR in 88% and 80% of patients, respectively, and by at least one of the methods in all patients. A significant correlation was achieved between the methods by Pearson correlation analysis (P = 0.004). MRD levels significantly decreased to very low levels on day 33 in 9 out of 10 patients studied. The only patient that remained positive relapsed.
T-cell lymphoblastic lymphoma (T-LLy) accounts for around 30% of non-Hodgkin's lymphoma (NHL) in children and is classified as a high-grade tumor, typically presenting with mediastinal mass, indicating stage III lymphoma 1–3. T-LLy and T-cell acute lymphoblastic leukemia (T-ALL) are often considered to be part of a spectrum of a single disease 1. The malignant cells in T-ALL and T-LLy are morphologically indistinguishable, and the immunophenotype and genetic abnormalities are similar 1, 4–7. However, differences in predominant sites of involvement, biological behavior 4, 8–10 and distinct gene expression profile 9 have been observed. The outcome of T-LLy with leukemia-type treatment strategies is generally favorable 11, 12. However, for the 10–20% relapsed T-LLy, the outcome is very poor.
In contrast to ALL, very few prognostic factors have been identified in T-LLy, such as disease stage III versus IV 11, 13–15 and early response to treatment, as measured by resolution of the mediastinal mass 11, 16, 17, and their significance is controversial, depending on the therapy delivered. Early treatment response, expressed by levels of minimal residual disease (MRD) in bone marrow (BM)/peripheral blood (PBL), has proven to be a powerful predictor of treatment outcome 18–24 and was introduced to recent ALL treatment protocols for risk-group stratification 24, 25. The sensitive and specific methodologies used are real-time quantitative polymerase chain reaction (RQ-PCR) for evaluating clonal immunoglobulin (Ig) and T-cell receptor (TCR) gene rearrangements 26–34 and multiparameter Flow Cytometry (FC) for tracing the leukemic-specific immunophenotype 35–39.
We assessed the feasibility of RQ-PCR and FC, as well as the concordance between the two both methods, in evaluating the extent of BM involvement at diagnosis (minimal disseminated disease—MDD) and early BM response kinetics (MRD) during treatment in stage III T-LLy and their clinical relevance.
PATIENTS AND METHODS
Patients and Treatment
This retrospective study included 17 out of a total 31 children newly diagnosed with T-LLy between 1991 and 2006 and treated at the Schneider Children's Medical Center of Israel (SCMCI), for whom sufficient excess of tumor tissue and BM mononuclear cells at diagnosis were available. The local and national Ethics Committees approved the research project. In 10 patients, BM aspirate was available at additional time points during treatment induction (days 15, 33, and 78) for evaluation of MRD.
The patients were treated according to three modified Berlin-Frankfurt- Munster (BFM ALL/NHL protocols 11 (without preventive cranial irradiation): ALL-BFM 90, NHL-BFM 95, and EURO-LB 02, all medium-risk arm. The patient characteristics are shown in Table I. Median follow-up was 6 years (range 1.2–16 years). Three patients (nos. 1, 6, and 11) relapsed within 42 months and died. This relapse rate was similar to that for the total group of T-LLy patients at our center (5 out of 31).
Table I. Clinical Features of T-Cell Lymphoma Patients at Diagnosis
Tumor tissue and BM mononuclear cells were cryopreserved at −80°C. The same BM samples were used for minimal disease detection by FC and RQ-PCR. The immunophenotype of lymphoma tumor cells at diagnosis was classified according to the European group for the Immunological Characterization of Leukemia (EGIL) 40 and the scheme of van Dongen et al. 41, as shown in Supplemental Table I.
BM MDD/MRD by FC
Samples were thawed, and antibody staining was performed using standard protocols 36. Four-color monoclonal combinations were tailored according to the immunophenotype of lymphoma tumor cells obtained at diagnosis (Supplemental Table II) compared with normal BM patterns 35. Data acquisition was performed using a dual-laser FACSCalibur (Becton Dickinson, Sunnyvale, CA) equipped with CELL QUEST Pro software (Becton Dickinson). A total of 1 × 105–3 × 105 non-gated events were acquired for BM MDD/MRD.
Data analysis was performed with PAINT AGATE Pro (BD) software. The expression of all the markers was analyzed within CD7+ or cyCD3/lymphocyte SSC gate. Analysis of a cluster of 10 aberrant events in at least 105 viable cells results in a sensitivity of 1 cell per 104 (0.01%) normal BM cells 36.
BM MDD/MRD by RQ-PCR
Genomic DNA was extracted from tumors and mononuclear cells with the Genomic Purification Kit (Gentra Systems). The presence of incomplete Ig, TCRG, and TCRD rearrangements in DNA of tumor at diagnosis was determined using various primer combinations 27. The TCRB rearrangments were tested according to the BIOMED-2 primers 32 using the IdentiClone TCRB Gene Clonality Assay (InVivoScribe Technologies). Monoclonal rearrangements were sequenced as noted previously 19 and ASO primers were designed 31. MRD levels in diagnosis and follow- up samples were measured by RQ-PCR using hydrolysis (TaqMan) probes 27–30 on the LightCycler platform (Roche Molecular Systems, Mannheim, Germany). BSA (0.04%) was added to each RQ-PCR study. Standard curves were prepared from the DNA of the tumor samples. The quantity and quality of DNA for the RQ-PCR analyses were assessed by albumin gene expression. The RQ-PCR results for the albumin gene transcripts in the samples were in the range of plus or minus one threshold cycle (1 Ct) of the control, indicating that the amount and the efficiency to amplify DNA samples were very similar. RQ-PCR experiments were designed and interpreted according to the guidelines established by the European Study Group of MRD detection in ALL (ESG-MRD-ALL) 33. Standardization and quality control was supported by the ESG-MRD-ALL.
To measure agreement of dichotomous variables, we used Cohen's Kappa analysis, and for association of continuous two variables (FC and RQ-PCR), we used Pearson's correlation co-efficient.
Marker Characteristics and Sensitivity by FC and RQ-RCR
Lymphoma-associated immunophenotype antigen expression was identified in all 17 patients. In 10 patients, the ectopic co-expression of n-TdT with other specific T-lineage antigens such as CD1a, CD2, CD3, CD4, CD5, and CD8, was detected. The immature antigen CD34 was expressed in two cases. The specific CD1a for immature thymus T-cells was identified in 10 patients and used in eight of them. Additional aberrantly expressed markers such as CD10, CD21, and CD45dim were used in the other seven patients. Two antigen marker combinations with a sensitivity of at least 0.01% were used for each patient (Supplemental Table II).
Tumors from 15 patients were evaluated for TCR gene rearrangements, and 27 monoclonal genetic changes were identified: 15 (56%) TCRγ, 3 (11%) TCRδ, and 9 (33%) TCRβ. Two markers were identified in 12 patients (80%) and one marker in three. A sensitive range of 10−5 or 10−4 was achieved for 23 markers (85%). In four samples (nos. 3, 10, 12, and 16), MRD was not quantifiable due to a low quantitative range (≥10−3). When MDD/MRD was studied by two targets, the higher level was used (Supplemental Table III).
Comparison Between Methods and Correlation to Clinical Features for BM MDD at Presentation
Since both FC and RQ-PCR reached a sensitivity of 0.01%, this value was chosen as the cut-off for positivity. By FC, cells with a lymphoma-related phenotype in BM were positive (≥0.01%) in 15 out of 17 patients (88%). MDD levels were high (≥1% to <6%) in 6 (35%), intermediate (≥0.1% to <1%) in 3 (17%), and low (≥0.01% to <0.1%) in 6 (35%) (Supplemental Table II).
By RQ-PCR, positive levels of clonal TCR gene rearrangement were found in 12 out of 15 patients (80%). Levels were high (≥1%) in 5 (33%), intermediate (≥0.1% to <1%) in 2 (13%), and low (≥0.01% to <0.1%) in 5 (33%) (Supplemental Table III). It is of note that although levels of MDD in four patients (nos. 2, 7, 9, and 11) were within the range of the morphological definition of leukemic BM involvement (≥5%), we could distinguish only 1–5% lymphoblasts.
A significant correlation (r = 0.7) was achieved between FC and RQ-PCR by Pearson correlation analysis [95% CI (0.29–0.9), P = 0.004; Fig. 1)]. Similar results (up to 1 log difference) were achieved in 12 out of 15 patients: in five patients (nos. 3, 10, 12, 13, and 14), the results were identical; in six patients (nos. 1, 5, 6, 7, 8, and 9), RQ-PCR was higher, and in one patient (no. 15), FC was higher. In only three patients, two (nos. 4, 16) with a higher FC value and one (no. 2) with a higher PCR value was a difference of more than 1 log detected (Supplemental Tables II and III, Fig. 1). When ≥0.01% was used as the positive threshold, the two methods were concordant in 10 out of 15 patients (67%) studied, which did not reach statistical significance by Kappa analysis (P = 0.4). In five patients (nos. 4, 5, 8, 15, and 16), the results were discordant: positive by FC and negative by PCR in three (nos. 4, 15, and 16) and vice versa in two (nos. 5, 8). However, when the quantifiable range of ≥0.05% was used as the positive threshold, the methods reached significant concordance in 12 out of 15 patients (80%) studied [measure of agreement κ = 0.6, 95% CI (0.2–1.0), P = 0.02].
No correlation between MDD levels and immunophenotype was found. MDD levels in the three relapsed patients (nos. 1, 6, and 11) were highly positive (10−2 in two patients) and low-positive (7 × 10−4 in one patient), but the number of patients was too small to assign any prognostic significance.
Minimal Residual Disease (MRD) by FC and RQ-PCR
MRD levels during treatment were investigated in 10 patients, all of whom were MDD-positive (≥0.01%) at diagnosis by at least one method (Table II).
Table II. BM Minimal Disseminated Disease (MDD) and Minimal Residual Disease (MRD) Results Obtained FC and RQ-PCR at Different Time Points of Induction Treatment. Percentage of Lymphoma Cells
ND, not done; NM, no marker; Neg, negative.
MRD levels markedly decreased by 0.5–2 logs on day 15 in most of the samples examined, and continued to decrease to negative or very low (<0.05%) levels in all but one patient (no. 11). This patient remained highly positive (1.5 × 10−2) on day 33 and was the only one of the 10 MRD-examined patients to relapse.
Although a relatively good outcome for T-LLy stage III in children is achieved with ALL-type treatment, over 10% of patients still die after relapse 1, 11, 12. More accurate methods to define BM microscopic involvement at diagnosis in order to refine current staging and for subsequent evaluation of early response to treatment may improve treatment stratification. Ours is the first study to apply two state-of-the-art methods concurrently in the evaluation of T-LLy, four-color FC, and RQ-PCR, both of which were found to reach a sensitivity of 0.01% blasts in normal BM cells in T-ALL 23, 32, 34, 37, 39–47.
With the application of a panel of antibodies for four-color FC, we were able to monitor all patients with two sets of markers. For the detection of T-LLy, we used the antibody combination of immature T-cells, which are normally exclusively confined to the thymus 37, and reached high feasibility, similar to that reported in T-ALL 46, 47.
Using TCR γ/δ/β, we identified at least one sensitive marker (≥0.01%) in all 15 patients studied, and two markers in 60% of patients. These findings are in line with the results achieved in T-ALL of Willemse et al. 42 and Kerst et al. 46 and higher than those of Robillard et al. 47 who did not apply TCRβ.
Previous studies compared MRD results obtained by FC and PCR in ALL, mostly in B-lineage 44–48; some of them included very few patients 2–15 with T-cell ALL 44–47. In our study of T-LLy involving BM at presentation, the positive/negative concordance between the two methods using a threshold of ≥0.01% was 67%. This finding was not statistically significant (P = 0.4). However, when ≥0.05% was used as the cutoff, concordance was 80%, which was statistically significant (P = 0.02). This finding is comparable to the results achieved in ALL of 72–97%, depending on the cut-off point (0.1–0.001%) 44–48. We obtained discordant results in five patients: three were MDD-positive by FC, negative by PCR and two vice versa. Earlier studies 44–48 assumed that in the case of lower RQ-PCR values, it was possible that RQ-PCR detected only a minor clone in the presence of oligoclonality, albeit rare in T-cell ALL; in the case of higher RQ-PCR values, it was possible that the test evaluated all cells, including the non-viable ones, which were eliminated with FC. They proposed that such discordance might be minimized by the same processing of identical samples with a broader antibody panel and at least two sensitive markers 44–48. Nevertheless, in the present study, although we analyzed identical samples for both methods and used two sensitive markers, in keeping with these suggestions, some discordance occurred.
BM MDD was assessed retrospectively in our patients with stage III T-LLy and was positive (≥0.01%) by FC in 88% or RQ-PCR in 80%; all patients (100%) had a positive finding by at least one of the methods. Three patients relapsed, of whom two had a high level of MDD and one, a low level. In the only previous report on BM dissemination in childhood T-LLy, performed on a limited number of patients and using nested PCR for TCRδ/γ, dissemination was detected in five out of seven patients, and two patients relapsed, one of whom was BM-MDD positive 49.
Early response to therapy, measured by MRD at various time points during treatment, was shown to be a strongly significant prognostic impact in acute de novo leukemia 18–22, and has since been incorporated in many front-line ALL protocols for treatment stratification 24, 25. The MRD response of patients with T-cell ALL, as evaluated by RQ-PCR, was found to be slower than in patients with B cell precursor, and their prognosis was worse on the BFM protocol 42. Given that treatment for T-LLy is similar to that for T-ALL, we studied MRD levels at identical time points, like for patients with leukemia on BFM protocols 25. On day 33, MRD levels decreased dramatically in most of the patients, and the one with the remaining highly positive MRD level relapsed in the BM, even though BM became negative on day 78. In the only previous report, mentioned above 49, the treatment MRD response of seven patients with T-LLy was mentioned, and the majority of follow-up specimens were still positive by PCR after 3 months.
In the present study, including a larger group of patients than studied previously and using modern methods, we identified a high prevalence of submicroscopic involvement of BM at diagnosis of stage III T-LLy, but could not prove prognostic significance. This may have been due to the very effective ALL-type treatment, which is aimed to cure totally infiltrated BM. In terms of MRD, evaluated retrospectively in a small cohort, we demonstrated a positive early response on day 33. The only patient who responded poorly by MRD relapsed; RQ-PCR and FC methods yielded similar MDD and MRD findings. Thus, we have shown that both methods may be efficient for MDD and MRD analyses in T-LLy. Further studies in a larger cohort are needed to clarify the preferred method, cut-off value, and time point at which MRD might have a prognostic significance.