Absolute lymphocyte counts at end of induction correlate with distinct immune cell compartments in pediatric B cell precursor acute lymphoblastic leukemia

Abstract

Several retrospective studies in children with B cell precursor (BCP) acute lymphoblastic leukemia (ALL) provided clinical evidence that higher absolute lymphocyte counts (ALC) early into treatment significantly correlated with improved relapse-free and overall survival. It still remains unknown, however, whether the predictive role of higher ALCs reflects general bone marrow recovery or a more specific attribute of immune function. To investigate this question, we implemented a prospective observational cohort study in 20 children with BCP ALL on day 29 (D29) of induction chemotherapy and immunophenotyped their lymphoid (T, B and natural killer cells) and myeloid (neutrophils, monocytes, dendritic cells) compartments. In a first evaluation of a cohort treated with Children’s Oncology Group-based induction chemotherapy, the immune cell compartments were differentially depleted at D29. Neither gender, risk status, minimal residual disease, nor bone marrow recovery markers correlated with D29 ALC. In contrast, both CD3⁺ T cell and dendritic cell compartments, which did not correlate with age, significantly correlated with D29 ALC (p < 0.0001). In addition, subset complexity of cellular immune compartments was preserved at D29. This study reveals that D29 ALC significantly correlates with distinct immune cell compartments but not with bone marrow recovery markers, suggesting that higher D29 ALCs may contribute to leukemia control by inducing specific host immune activity.

Electronic supplementary material

The online version of this article (doi:10.1007/s00262-017-2070-3) contains supplementary material, which is available to authorized users.

Introduction

Early risk stratification of newly diagnosed childhood B cell precursor (BCP) acute lymphoblastic leukemia (ALL) has become a cornerstone of pediatric chemotherapy protocols and has significantly contributed to improved outcomes [1, 2]. Current risk-stratifying criteria are derived from leukemia genetics and phenotype, patient age, and disease features at presentation, such as initial white blood cell (WBC) count and involvement of sanctuary sites [1, 2]. In addition to factors identifiable at diagnosis, response to treatment is also used to determine a child’s risk group, as measured by achievement of morphologic remission and minimal residual disease (MRD) burden at the end of induction chemotherapy [3, 4]. Characteristics of the endogenous immune system, however, are not currently considered, even though the absolute lymphocyte count (ALC) early in therapy is emerging as a strong prognostic variable with the potential to inform risk stratification.

Several retrospective studies over the past decade have revealed that higher ALC during or at end of induction was a significant independent predictor of better relapse-free and overall survival in children with BCP ALL [5–11]. Despite differences in induction regimens, and assessment time points and cut-off values, there was high concordance among the studies suggesting that ALC may be a novel host factor with prognostic utility. The prognostic significance of higher ALCs, observed in cohorts of de novo BCP ALL patients without sub-stratification, was not dependent on the initial ALC value at diagnosis and so far has only been reported for the earliest treatment phase. In addition, high ALC during induction chemotherapy was found to be predictive of low MRD status at the end of induction, further supporting that MRD negativity may not be entirely determined by the chemo-responsiveness of the leukemic cells [10, 12]. Finally, ALC stratification was recently found to further discriminate distinct prognostic subgroups within both the MRD-negative and MRD-positive cohorts revealing a significant difference in overall survival between ALC low and ALC high patients [6]. Even in cases of a positive MRD status at the end of induction, usually considered the strongest predictor for subsequent relapse [4], a high ALC was associated with a significantly reduced risk of relapse in a subset of patients [6]. Collectively, these observations suggest the host immune environment early in therapy may contribute to leukemia control.

The mechanism underlying this clinical association with higher ALCs is not well understood. While it has been posited that ALC is simply a marker of general bone marrow (BM) recovery [7, 11, 13, 14], the potential that ALC reflects a specific immunological quality has been raised [6] but not yet investigated. Most immune composition studies in BCP ALL primarily focused on the longitudinal effects of long-term chemotherapy and investigated the depletion and recovery kinetics of the various immune cell compartments after completion of therapy. The few studies that also analyzed the impact of induction chemotherapy on cell numbers revealed that immune cell compartments were differentially affected during induction, some of them severely [15–19]. As identifying the mechanisms underlying ALC-associated outcomes may reveal novel strategies for enhanced immune-mediated control of BCP ALL, we implemented a prospective, observational cohort study involving 20 consecutive children with de novo BCP ALL. Using a comprehensive flow cytometric approach, we investigated the immune cell composition at the end of induction (day 29, D29) in peripheral whole blood with the aim to assess whether distinct immune cell compartments and/or markers of BM recovery were associated with D29 ALC.

Materials and methods

Study set-up

Patients were eligible for enrolment in this prospective study if they were between 1 and 21 years of age at diagnosis of de novo BCP ALL, received COG protocol-defined induction chemotherapy at BC Children’s Hospital, and had adequate medical records available for review. Patients with concurrent infections were excluded. Peripheral blood samples were obtained at end of induction and immediately analyzed by immunophenotyping whole blood. Differential complete blood cell count (CBC) and MRD status (determined by flow cytometry) were simultaneously assessed for each patient by the clinical laboratory as part of the routine clinical care. Leukemia burden < 0.01% of the mononuclear cells was considered MRD negative, while 0.01–5% was MRD positive. At the end of this study, the de-identified patient and disease characteristics, CBC and MRD status at end of induction, and current clinical status were retrieved from the BC Children’s Biobank database with informed consent.

Sample preparation and flow cytometry

Whole blood samples were collected in sodium-heparin tubes. To avoid artificial skewing of the immune cell composition, whole blood instead of isolated peripheral blood mononuclear cells was immunophenotyped. Whole blood was stained within 4 h of procurement to minimize potential artifacts of extended storage [20], isolation technique or multiple washing steps [21], followed by red blood cell lysis with BD© FACS Lysing Solution (San Jose, CA, USA). To allow for comparability with published data, cell-surface marker combinations established by international immunophenotyping consortia [22, 23] were used whenever possible. To assure inter-sample consistency, we adhered to recently published standardization criteria [21, 23]. We designed three panels to immunophenotype the T cell and NK cell compartments (panel A), B cell compartment (panel B), and DC compartment (panel C), the latter also including myeloid markers of bone marrow recovery, neutrophils and monocytes (Supplementary Table S1).

All antibodies and fluorochromes are provided in Supplementary Table S2. Irrespective of ALC, 600 μl of whole blood was used for the entire immune cell subset analysis. All immune cell subsets were named according to their immunophenotypic description (Supplementary Fig. S1) and were not characterized functionally. Compensation beads (CompBeads; BD Biosciences) stained with a representative fluorochrome were used to standardize voltage settings and used as single-stain-positive and -negative controls as per manufacturer’s protocol. Of note, CD14 was added to Panel A in later samples as an additional layer of accuracy to verify that NK cell detection did not include monocytes, which was consistently confirmed (data not shown). The addition of CD14 to Panel A also facilitated the identification of three subsets of monocytes (classical, intermediate and non-classical) in the remaining patients.

Representative dot plots of the gating strategy are provided in Supplementary Fig. S2. Data acquisition was performed with LSR II flow cytometer/Diva software (BD, San Jose, CA, USA) collecting > 1,500,000 events, or the available sample volume in case of lower cell numbers, and analyzed using Flowjo v10.1 (Tree Star, Ashland, OR, USA). Absolute cell counts were determined using counting beads (CountBright, Life Technologies, Carlsbad, CA, USA). To determine the degree of immunosuppression, we compared the patients’ absolute cell counts to published age-dependent normal reference values measured in whole blood [24, 25].

Statistical analysis

Statistical analyses were performed using Prism 6 (GraphPad Software, San Diego, CA, USA). Day 29 cell counts and immune cell subsets were compared by t-tests. Statistics for comparisons of unpaired medians were made using the Mann–Whitney test and for comparisons of paired medians with the Wilcoxon matched-pair signed-rank test, both non-parametric tests that assume non-Gaussian distributions and are the most conservative choices for a small sample size. For analysis of the relationship between two continuous variables, correlation coefficients were determined by Pearson’s method. Comparison of slopes was performed by the linear regression method. In experiments with multiple independent factors, two-way analyses of variance (ANOVAs) were performed to determine row, column, and interaction statistics. Statistical significance level was set at p < 0.05.

Results

Patient cohort and disease characteristics

With a study recruitment rate of 81%, twenty consecutive BCP ALL patients were enrolled to this study through the BC Children’s Hospital Biobank, while one additional patient was excluded due to early signs of an acute infection. Of the 20 patients on study (Supplementary Table S3), fourteen (70%) had standard-risk (SR) and six (30%) high-risk (HR) disease, which required addition of daunorubicin to the SR 3-drug induction (corticosteroids, vincristine, asparaginase) [3]. One patient, initially classified and treated as SR, was later found to have genetic lesions associated with very high-risk BCP ALL that influenced post-induction treatment; as the sample was obtained at the completion of three-drug induction chemotherapy, we kept this patient in the SR group for the purpose of our study. Six patients (30%; 3 SR, 3 HR) were confirmed with positive MRD burden at end of induction, while three of the six HR patients (50%) and 11 of the 14 SR patients (80%) were negative for MRD. The distribution of cytogenetic abnormalities in the study cohort was 25% ETV6-RUNX1 fusion t(12;21), 20% various degrees of hyperdiploidy, 10% Philadelphia chromosome t(9;22), 10% with abnormalities not otherwise specified, 5% various trisomies, 5% KMT2A (MLL) rearrangement t(4;11)(q21;q23), 5% genetic lesions associated with poor outcomes, and 20% had no abnormal findings; this is representative for the total cohort of patients treated on protocol. The study cohort had a median age of 4.6 years (range 1.2–18.7 years) with 12 males and 8 females. Two patients received hematopoietic stem cell transplantations, and while the majority of patients have not yet completed their chemotherapy, all patients are still in remission with a median follow-up period of 554 days (range 230–1021).

Distribution of D29 WBC and its three main subpopulations

The white blood cell count (WBC) at D29 and the distribution of its three main subpopulations of neutrophils (ANC), monocytes (AMC) and lymphocytes (ALC) as enumerated from the patients’ clinical CBCs are shown in Fig. 1a, with patients sorted according to increasing ALCs. Median WBC was 4300 cells/μl (range: 400–9500 cells/μl), median ANC was 2195 cells/μl (range: 50–6110 cells/μl), median AMC was 300 cells/μl (range: 40–3200 cells/μl), and median ALC was 1665 cells/μl (range: 140–4910 cells/μl). Patients showed varying distributions of the three subpopulations with the D29 lymphocyte subset ranging from 13 to 84% (median: 32%), the D29 neutrophil subset from 12.5 to 74.6% (median: 52%) and the D29 monocyte subset from 2.11 to 52.5% (median: 9.1%) of the D29 WBC.

Fig. 1.

a The white blood cell count (WBC) at D29 and the distribution of its three main subpopulations of neutrophils (ANC), monocytes (AMC) and lymphocytes (ALC) as enumerated from the patients’ clinical CBCs are shown together with each patient’s risk status, MRD status and age (years); patients are sorted according to increasing ALCs showing varying distributions of the three subpopulations with the D29 lymphocyte subset ranging from 13 to 84%, the D29 neutrophil subset from 12.5 to 74.6% and the D29 monocyte subset from 2.11 to 52.5% of the D29 WBC. *Patient was initially classified and treated as SR, but later found to have genetic lesions associated with very high-risk BCP ALL that influenced post-induction treatment; as the sample was obtained at the completion of 3-drug induction chemotherapy, we kept this patient in the SR group for the purpose of our study. b The relationship between D29 ALC and age as continuous variables was investigated, revealing a significant inverse correlation (r = − 0.47, p = 0.033) using Pearson’s correlation coefficient. c Assessment of the cellular immune compartments at D29 showing boxplot diagrams of the distribution and whiskers at 95th percentile in 20 patients plotted for each primary immune cell compartment

Relationship of D29 ALCs with patient and disease characteristics

During childhood, lymphocyte numbers vary with age, showing a pronounced decrease in early childhood followed by a moderate decrease with increasing age [24]. In addition, all children over 10 years of age with BCP ALL are classified HR in the COG protocol and receive four-drug induction, potentially leading to increased cytotoxicity that could further decrease cell counts. We, therefore, first investigated the relationship between D29 ALC and age as continuous variables and found a significant inverse correlation (r = − 0.47; p = 0.033) (Fig. 1b). This finding confirms the general observation that lymphocyte counts decrease with increasing age in a disease context that has a strong prevalence in children less than 5 years of age. There was no significant difference in the median D29 ALC when patients were grouped by risk status (p = 0.95) or by MRD status (p = 0.65); also gender (p = 0.84) did not correlate with D29 ALC (Supplementary Fig. S3a).

As in previous studies, the clinically reported prognostic significance of D29 ALC values was not dependent on ALC values at diagnosis (day 0, D0), we investigated the immune composition only at end of induction (D29). In accordance with these findings, we confirmed that D0 ALC values were not correlated with their D29 ALC counterparts (p = 0.26, Supplementary Fig. S3b). Furthermore, when clustering patients’ paired D0 and D29 ALC values according to risk status, MRD status or Rabin’s determined D29 ALC cut-off value of 1500 cells/μl [6], we also did not observe any distinct pattern in any of the groups (Supplementary Fig. S3c). Finally, there also was no significant difference in the median D0 ALC when grouped by risk (p = 0.13) or MRD status (p = 0.13) (Supplementary Fig. S3d); also gender (p = 0.27) did not correlate with D0 ALC (data not shown).

To date, presence of MRD at end of induction remains the strongest predictor for relapse [3, 26–28]. In our patient cohort, 30% of patients were MRD+ with two patients in the SR and four in the HR group (Supplementary Table S3). However, ALC stratification recently allowed discrimination of distinct prognostic subgroups within both the MRD-negative and MRD-positive cohorts [6]. When applying the same D29 ALC cut-off value of 1500 cells/μl to our identically treated patient cohort, we found that only three of the MRD+ patients (50%) displayed D29 ALC values < 1500 cells/μl compared to 5 of the 14 MRD− patients (36%) (Supplementary Fig. S4a).

Whole blood immunophenotyping of immune cell compartments at D29

Using the primary immune compartment gating strategy shown in Supplementary Table S1, we first enumerated the absolute cell counts for the T cell (CD3⁺), B cell (CD19⁺), NK cell (CD56⁺), and DC (total DCs) compartments, as well as neutrophil and monocyte compartments that served as myeloid markers of bone marrow recovery [15] (Fig. 1c). The clinically provided AMC, ANC and ALC values from the differential CBC were highly concordant with the monocyte, neutrophil and lymphocyte numbers obtained by flow cytometry in this study (p < 0.0001 for each, with r = 0.94, r = 0.92, and r = 0.76, respectively; Supplementary Fig. S4b–d). None of the absolute cell counts for individual immune cell compartments (T, B, NK, DC, neutrophils, monocytes) correlated with age, though they showed a slight downward trend with increasing age (data not shown).

We investigated the immune composition only at end of induction (D29) because the previous retrospective studies demonstrated that the prognostic significance of D29 ALC values was not dependent on the initial ALC values at diagnosis (D0 ALC). To determine the degree of immunosuppression, we compared the patients’ absolute cell counts to published age-dependent normal reference values measured in whole blood (Supplementary Fig. S5) [24, 25]. We found a low-normal to moderately depleted CD3⁺ T cell compartment (Supplementary Fig. S5a) in the majority of patients, that did not further decrease with more intense chemotherapy. The CD19⁺ B cell compartment (Supplementary Fig. S5b) was more profoundly affected, with counts significantly more depleted in HR patients (p = 0.0007) (Supplementary Fig. S5c). CD56⁺ NK cell numbers were also severely depleted (Supplementary Fig. S5d) regardless of treatment intensity. The vast majority of our patients (85%) had normal myeloid DC (mDC) counts at D29 (Supplementary Fig. S5e), while they were significantly decreased in three patients; plasmacytoid DC (pDC) counts, on the other hand, were normal in five patients (Supplementary Fig. S5f) and severely decreased in the majority of patients (75%) irrespective of risk status.

Myeloid markers of bone marrow recovery are not correlated with D29 ALC

As most commonly it has been speculated that the improved outcomes observed with higher ALCs are primarily due to better bone marrow recovery [7, 11, 13, 14], we first investigated whether myelopoiesis, a surrogate marker of bone marrow recovery, showed a relationship with ALC. While both neutrophils (r = 0.69, p = 0.0007, Fig. 2a) and monocytes (r = 0.76, p < 0.0001, Fig. 2b) showed a significant correlation with D29 WBC and with each other (r = 0.59, p = 0.006, Fig. 2c), neither correlated with ALC (r = –0.11, p = 0.64, Fig. 2d and r = 0.20, p = 0.40, Fig. 2e, respectively).

Fig. 2.

Myeloid markers of bone marrow recovery are not correlated with D29 ALC. a–e While both neutrophils (a) and monocytes (b) showed a significant correlation with D29 WBC and with each other (c), neither neutrophils (d) nor monocytes (e) correlated with D29 ALC. The Pearson’s correlation coefficient was used for all

Significant correlation of dendritic cells with D29 ALC

While not regarded as a marker of bone marrow function, DCs represent a myeloid lineage capable of significantly influencing lymphocyte populations. In contrast to neutrophils and monocytes, total DCs (median: 15 cells/μl, range: 0.5–78) significantly correlated with D29 ALC (r = 0.86, p < 0.0001, Fig. 3a). DCs, monocytes and neutrophils arise from myeloid precursor populations [29], yet only DCs significantly correlated with D29 ALC; thus we examined the relationship of DCs with their myeloid counterparts. Neither neutrophils (r = −0.29, p = 0.21) nor monocytes (r = −0.14, p = 0.56) significantly correlated with DCs (Fig. 3b, c). As the peripheral DC population contains both myeloid DCs (mDCs, median: 13.4 cells/μl, range 0.44–47) and plasmacytoid DCs (pDCs, median: 0.9 cells/μl, range: 0.06–31.3), we next analyzed their respective relationships with D29 ALC. Both mDCs (r = 0.89, p < 0.0001, Fig. 3d) and pDCs (r = 0.74, p = 0.0002, Fig. 3e) were significantly correlated with D29 ALC.

Fig. 3.

The DC compartment and primary lineage-specific subpopulations, mDCs and pDCs, are significantly correlated with D29 ALC. a There was a significant correlation of total DCs with D29 ALC, but no significant association with b neutrophils and c monocytes. Both d mDCs and e pDCs significantly correlated with D29 ALC. The Pearson’s correlation coefficient was used for all

Correlation of major lymphoid compartments with D29 ALC

Having detected a significant association between D29 ALCs and the adaptive immune-modulating DC compartment rather than general myeloid markers of BM recovery, we next investigated whether CD3⁺ T cells (median 367 cells/μl, range 37–2000), CD19⁺ B cells (median: 85 cells/μl, range 9–412), or CD56⁺ NK cells (median 16 cells/μl, range 1–60) correlated with Day 29 ALC. A significant correlation with D29 ALC was found only for CD3⁺ T cells (r = 0.90, p < 0.0001, Fig. 4a) displaying a high positive linear relationship. Neither CD19⁺ B cells (r = 0.009, p = 0.97, Fig. 4b) nor CD56⁺ NK cells (r = 0.09, p = 0.70, Fig. 4c) showed a significant association with D29 ALC in our patient cohort. As the CD3⁺ T cell compartment is predominantly comprised of CD4⁺ cells (median 171 cells/μl, range 23–1126) and CD8⁺ cells (median 193 cells/μl, range 11–639), we analyzed their respective relationships with D29 ALC and found that both CD4⁺ T cells (r = 0.83, p < 0.0001, Fig. 4d) and CD8⁺ T cells (r = 0.87, p < 0.0001, Fig. 4e) were significantly correlated. When CD4⁺ and CD8⁺ T cells were plotted against their age-dependent normal reference values [24] (Supplementary Fig. S6a, b), it revealed a moderate depletion of both subpopulations, which was more pronounced for CD4⁺ T cells compared to CD8⁺ T cells and irrespective of treatment intensity.

Fig. 4.

Relationship of the primary lymphoid compartments with D29 ALC. a–c The immune compartments a CD3⁺ T cells, b CD19⁺ B cells, and c CD56⁺ NK cells were evaluated for correlation with D29 ALC showing a strong positive relationship only for the CD3⁺ T cell compartment. d, e The primary lineage-specific subpopulations of the CD3⁺ T cell compartment, d CD4⁺ T cells and e CD8⁺ T cells, were also significantly correlated with D29 ALC. The Pearson’s correlation coefficient was used for all

Subset complexity of cellular immune compartments is preserved at D29

The detection of significant correlations with ALC at the level of lineage subpopulations raises the possibility that the influence of minor cell subsets may be missed if the dominant cell population does not correlate with ALC. While it might be anticipated that there is a significant depletion or even elimination of some minor cell subsets by induction chemotherapy, there is limited data on the broad immune composition in children undergoing chemotherapy, including immune cell subsets known for their potential contribution to initiation or maintenance of leukemia control. Therefore, we evaluated for the presence of numerically small, but biologically distinct subsets within the D29 peripheral blood samples. Using the gating strategy shown in Supplementary Fig. S1, we were able to reliably gate on an additional 24 immune cell subsets within the lymphoid and myeloid compartments (Supplementary Table S4) from identical blood volumes without correcting for very low WBC or ALC. This reveals that the complexity of the immune system seems to be preserved even in these immunosuppressed children (Fig. 5).

Fig. 5.

Subset complexity of cellular immune compartments was preserved at D29. Twenty-four additional immune cell subsets could be reliably identified within the investigated cellular immune compartments, from identical small blood volumes in all patients despite varying degrees of cytotoxicity and immunosuppression. We stratified the different immune cell subsets by color-coding them according to increasing D29 ALC counts. The applied color gradient (D29 ALC < 500 cells/μl in blue, 500–1000 cells/μl in purple, 1000–1500 cells/μl pink, 1500–2000 cells/μl orange, > 2000 cells/μl red) both addresses that various studies reported different ALC cut-off levels in the literature as well as Rabin’s study findings [6] that D29 ALC showed prognostic significance as a continuous variable demonstrating that overall survival increased by 4% with each 100 cells/μl increase in D29 ALC. As our study cohort is identically treated as Rabin’s cohort [6], we made sure that their concurrently determined D29 ALC cut-off value of 1500 cells/μl can easily be distinguished (blue-purple vs. orange-red). Of note, the outliers in several scatter plots are derived from different patients and the higher cell counts for immature B cells were detected in patients who were MRD negative. In one case (#16), the analysis of the B cell panel was inconsistent for technical reasons, so that we omitted the results of the entire panel B for this patient. In addition, as CD14 was added to the first panel early into the study, there is no data available for the monocyte subsets for the first eight patients, while the total monocyte count is available for each patient on study in panel C. Data is displayed as scatter plots showing mean with standard deviation

Discussion

Emerging clinical evidence supports a role for ALC in the risk stratification of children with ALL [5–11]. The predictive significance of higher ALC for favorable outcome is not limited to ALL or children; it also has been reported for other hematological malignancies, solid tumors, and after autologous and allogeneic hematopoietic stem cell transplantations for cancers in adults and children [13, 30–37]. However, despite its relevance to several malignancies, the nature of the association between ALC and outcome remains unknown.

Here, we report that bone marrow recovery markers, i.e., neutrophils and monocytes, were not correlated with D29 ALC. This confirms previous clinical observations that there was no association between ANC recovery and survival [5], nor did ANC exhibit any consistent prognostic significance at any time point [6]. In contrast, we found that both the CD3 + T cell and DC compartments, which are known to play a prominent role in tumor immune surveillance, leukemia eradication and long-term anti-leukemic immunological memory [25, 36, 38–40], showed a very strong positive linear relationship with D29 ALC. This significant correlation with D29 ALC was further confirmed at the level of primary lineage-specific subpopulations (CD4⁺ and CD8⁺ T cells; mDCs and pDCs) within the significant compartments. Such a finding is consistent with the reported associations of these immune cell subsets with improved responses in children with BCP ALL in studies performed at diagnosis and early into treatment [11, 41, 42] or during early immune recovery after allogeneic hematopoietic stem cell transplantation [36, 39, 43].

A particularly severe immune impairment has been associated with the early treatment phase of European treatment protocols for ALL [15–17, 19]. Based on these observations, it is rather unexpected that the predictive significance of ALC for better outcome in BCP ALL was reported only for the earliest treatment phase, with similar observations made in other malignancies [13, 31, 33, 34]. Most immune composition studies in BCP ALL have focused on investigating the immune reconstitution after completion of chemotherapy and demonstrated the varied regenerative patterns of individual subsets over time. The few studies that also analyzed the early impact of European protocol-based chemotherapy on cell numbers revealed that immune cell compartments were differentially affected during induction, some of them severely. The degree of differential end of induction immune depletion in our study was comparable to these, often similarly sized studies for the T cell [17, 18], B cell [17–19], NK cell [17, 44], and DC compartments [15, 16]. Yet, to the best of our knowledge our study is the first that investigates the end of induction immune composition in children treated with the Children’s Oncology Group (COG)-based induction chemotherapy protocol that was also used in Rabin’s seminal study [6]. In addition, the DC compartment is known to be significantly depressed in peripheral blood [15, 45] and BM [16] at time of diagnosis. However, the detection of mostly normal mDC counts and wide-ranging pDC counts in peripheral blood at D29 in our study seems to be consistent with the observed normalization of BM DCs by the end of the longer induction chemotherapy in the BFM protocol [16].

To enable comparison with published pediatric BCP ALL studies, we chose the D29 time point used in the 171-patient study by Rabin et al. given the identical COG-based risk-stratification and induction regimen [6] as well as a lack of prognostic significance for D0 ALC in previous studies. Our median D29 ALC of 1665 cells/μl (range 140–4910 cells/μl) differed slightly from Rabin’s study (median 2280 cells/μl, range 192–9890 cells/μl), though a similar percentage of patients (65%) were above the ALC cut-point value of 1500 cells/μl applied in Rabin’s study. In our study, the steady increase in absolute cell counts of CD3⁺ T cells and DCs with increasing ALCs is striking and suggests a potential role for an antigen-specific T cell response underlying Rabin’s observation that with each 100 cells/μl increase in D29 ALC, the overall survival increased by 4% [6]. If a distinct ALC level must be reached for optimal long-term survival as suggested by several studies [6–10], it could indicate that a numeric threshold effect may exist for specific immune system functioning to resume. Our findings of strong linear relationships between the T cell and DC compartment with D29 ALC over a wide range of cell counts provide support for such a model. A prospective COG therapeutic trial (AALL1131) for newly diagnosed HR ALL is ongoing to determine whether a certain ALC level is required to best predict favorable long-term outcome, although the exact numeric ALC cut-off and optimal time point may vary depending on the induction treatment regimen [5, 7, 8, 10, 11, 14].

It has been hypothesized that after successful reduction of the initial leukemia burden by induction chemotherapy, immune regulation could be reset in favor of leukemia control [46]. As such, intermittent profound lymphocytopenia may trigger an enhanced lymphocyte recovery drive, including a time of exaggerated T cell expansion [46]. Indeed, several mouse models demonstrated that when tumor antigens were available during lymphocytopenia-induced homeostatic T cell proliferation, it led to restoration of effective anti-tumor immunity [47, 48]. The early immune reconstitution phase seems to be particularly critical, as a state of augmented immune responsiveness resulting from a clonotypic expansion of T cells was only present early after bone marrow transplantation (3 weeks) and had already subsided by week six [49]. Similarly, enhanced lymphocyte proliferation responses were confirmed in pediatric patients early after myelosuppressive chemotherapy [50]. In children with BCP ALL, T cells have shown a fast recovery potential within a few weeks after the end of the longer BFM induction chemotherapy, leading to an intermittently less impacted T cell compartment [51, 52]; in contrast, B cells remained severely diminished [17]. Similarly, DC numbers regenerate in the bone marrow during induction chemotherapy reaching reference values at day 50 in the BFM induction chemotherapy protocol [16]. While most immune cell subsets vary in childhood with age, for reasons that remain unknown, the speed of recovery of B and T cell compartments after chemotherapy cessation were found to be age-independent in children with BCP ALL [53]. The combined recovery of antigen-presenting DCs and effector T cells observed in our study suggests that a capacity for inducing anti-tumour immune responses may be the relevant immunological quality captured by higher ALC.

Finally, Rabin’s study revealed that ALC refines MRD-based risk stratification resulting in a remarkable 51% difference in overall survival based upon ALC in MRD+ patients, while identifying additional patients at risk for relapse (42%) despite negative MRD [6]. In this context, it is interesting to observe that in our patient cohort 50% of MRD+ patients had ALC values >1500 cells/μl, while 36% of the MRD-negative patients showed ALC values <1500 cells/μl. This may also explain why in our patient cohort D29 ALC values did not correlate with MRD positivity.

A larger, prospective study with long-term outcome data is required to refine the association between individual immune cell subsets and ALC. Even though our study cohort is considerably smaller, the distribution of the different age groups closely resembled those in Rabin et al.’s 171 patients study, as 55% were 1–4 years of age, 35% 5–10 years of age, and 10% 11–18 years of age [6], while the distribution of cytogenetic abnormalities and SR versus HR is representative for the entire patient cohort on protocol. However, our cohort does not fully represent every aspect of the patient cohort on protocol, likely introducing significant bias including the possibility of skewing by random sampling and over- or under-representation of any specific features, which can only be addressed by a large, multi-center study. In addition, time of follow-up is currently too short to assess relapse-free and overall survival.

Conclusion

Despite its limitations, this study reveals that higher ALCs at end of induction (D29) do not correlate with BM recovery markers, gender, risk status or MRD. While a negative correlation with age was detected, D29 ALC showed a very high positive correlation with distinct immune cell compartments (CD3⁺ T cell and DC, p < 0.0001, respectively) known to mediate potent anti-leukemic activity, which was further confirmed at the level of primary lineage-specific subpopulations. Given the demonstrated preserved complexity of the D29 immune environment, understanding the nature of the immune activities associated with high ALC is critical for identifying the endogenous immunological factors that contribute to better survival and may be an important step in the development of novel approaches to enhance endogenous anti-leukemic immune activity.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ALC

Absolute lymphocyte count

ALL

Acute lymphoblastic leukemia

AMC

Absolute monocyte count

ANC

Absolute neutrophil count

BCP

B cell precursor

CBC

Complete blood cell count

COG

Children’s Oncology Group

D0

Day 0 at diagnosis before chemotherapy

D29

End of induction chemotherapy day 29

HR

High-risk

mDCs

Myeloid dendritic cells

NKs

Natural killer cells

pDCs

Plasmacytoid dendritic cells

SR

Standard-risk

WBC

White blood cell count

Author contribution

N. Rolf designed the study and performed the research experiments, analyzed the data and wrote the manuscript; K. Smolen contributed to research design; A. Kariminia, M. Fidanza and A. Seif contributed to research analysis; A. Velenosi was the coordinator of the clinical study details; A. Seif performed statistical analyses; C. Strahlendorf, K. Smolen, A. Kariminia, M. Fidanza and A. Seif critically reviewed the manuscript, G. Reid supervised the study and wrote the manuscript.

Compliance with ethical standards

Conflict of interest

The authors have no conflict of interest to disclose. The University of British Columbia Research Ethics Board approved the study (H12-01987). Informed consent of human subjects was obtained in all cases in accordance with the Declaration of Helsinki.

Funding

This work was funded in part by a Leukemia and Lymphoma Society of Canada operating Grant (Gregor SD Reid) and a Canadian Institutes of Health Research (CIHR) Fellowship Award (Nina Rolf).