BM samples were processed using our previously described ‘lyse-stain-wash’ protocol [13]. A minimum of 100,000 events were acquired per tube using a Beckman Coulter Navios EX flow cytometer. Generated list-mode data (LMD) files were analyzed with Kaluza (Version 2.0) software (Beckman Coulter) using our in-house developed analysis templates. The antigen expression profile was reported according to the Associazione Italiana Ematologia Oncologia Pediatrica-BFM (AIEOP-BFM) 2016 recommendations [14]. The expression intensity of each antigen was assessed by the geometric mean (GM) of expression determined by the Kaluza software.

Diagnoses of ETP-ALL and near-ETP-ALL used published criteria [1, 3, 4, 6, 15]. Patients not fulfilling the criteria for ETP-ALL or near-ETP-ALL were designated ‘conventional’ T-ALL (con-T-ALL). The intensity of CD5 expression on blasts was determined as the ratio between CD5-GM of T-lymphocytes within the sample to the CD5-GM of blasts (T-CD5: Bl-CD5 ratio) [15]. Our algorithm for classifying T-ALL patients into immunophenotypic subcategories is described in Supplementary Fig. 1A.

EOI-MRD was assessed in first-pull bone marrow aspiration (BMA) samples. The BMA samples were bulk-lysed with in-house prepared ammonium chloride-based lysis reagent and stained with an 11-antigen, 10-color cocktail (Supplementary Table 1). The processed samples were immediately fixed with 0.5% paraformaldehyde and acquired until the tube ran dry. The generated LMD files were analyzed using an in-house developed “mature antigen-based exclusion” approach adapted from Tembhare et al. [6]. Supplementary Fig. 1B details our sequential gating strategy for leukemia associated immunophenotype (LAIP) identification. A cluster of over 30 events with aberrant immunophenotype was considered for MRD quantification. The sensitivity of our MRD assay was 0.003% with a maximum coefficient of variation of 14.4% (refer to Supplementary Table 1 for our MRD assay validation and formula used for MRD calculation).

Differences in expression intensity between baseline and EOI-residual blasts were analyzed for the following antigens (negative & positive controls): CD7 (B-lymphocytes & T-lymphocytes), CD4 (B-lymphocytes and CD4+ T-lymphocytes), CD8 (B-lymphocytes and CD8+ T-lymphocytes), CD5 (B-lymphocytes and T-lymphocytes), surface-CD3 (B-lymphocytes and T-lymphocytes), and CD38 (granulocytes and monocytes). Normalized mean fluorescence intensity (nMFI) for all these antigens was calculated for baseline and EOI-residual blasts as previously described [16].

For mature antigen-based MRD analysis, we analyzed the stability of mature T-cell associted antigens (CD7, CD4, CD8, CD5 and surface CD3) available in our MRD panel (Supplemental Table 1). CD38 was analyzed to assess the stability of this potentailly targetable antigen by daratumumab. Stability of CD56 could not be analyzed as both CD56 and CD16 were used in the BV510 fluorochrome of our MRD panel.

Statistical Package for Social Sciences (version 23, IBM, Armonk, NY) and MedCalc version 14.8.1 were used for statistical tests. For intergroup comparisons, Chi-squared and Mann-Whitney U tests were used. Occurrence of induction failure (≥5% BM blasts at EOI), relapse, and death were considered events. With the date of disease diagnosis as the starting time point, Kaplan–Meier survival analysis was used to determine 2-year rates of overall survival (OS), relapse-free survival (RFS), and event-free survival (EFS). Wilcoxon’s signed-rank test was used to assess differences in the expression intensity for CD4, CD8, CD5, CD7, CD38, and surface-CD3 (sCD3) antigens between leukemic blasts at diagnosis and residual blast at EOI-MRD. The risks incurred by the presence of mediastinal mass, hyperleukocytosis, immunophenotypic subtype of T-ALL, D8BNC status, and EOI-MRD positive status on OS, RFS, and EFS were determined by Cox proportional hazard model (Wald test). All statistical tests were two-tailed and considered significant at P≤0.05.

FCM determined antigen expression profiles of all 81 T-ALL patients are presented in Supplementary Fig. 2. The median (range) T-CD5: Bl-CD5 expression ratio among con-T-ALL, near-ETP-ALL and ETP-ALL blasts was 1.83 (0.85–8.56), 3.39 (1.43–8.21) and 16.12 (11.06–59.21), respectively. Among con-T-ALL patients, 26.5% (N=13) had isolated CD4 expression, 10% (N=5) had isolated CD8 expression, dual expression for both CD4 and CD8 was observed in 45% (N=22) patients, and 20% (N=9) of the patients did not express either antigen. Expression frequency for myeloid/stem cell antigens (ETP-ALL vs. near-ETP-ALL patients) was CD117 (47% vs. 7%, P=0.011), CD34 (82% vs. 80%, P=0.865), HLA-DR (53% vs. 21%, P=0.073), CD13 (53% vs. 20%, P=0.055), CD33 (47% vs. 73%, P=0.131), and CD11b (29% vs. 21%, P=0.631).

Differences in the percentage of patients expressing immaturity associated antigens (CD10, CD34, and CD117), B-lineage antigens (CD19 and CD79a), myeloid antigens (CD13, CD11b, CD33), and non-lineage-specific antigens (CD123, CD56, and CD38) among our immunophenotypic subcategories of T-ALL are depicted in Fig. 1.

Among 60 patients who completed induction, EOI-MRD was tested in 59 (40 con-T-ALL, 12 ETP-ALL, and 7 near-ETP-ALL). A median of 2.3 million events (range, 0.18 to 7.3 million) was acquired for analysis, and over 1.5 million events were acquired in 68% of the samples.

EOI-MRD was positive in 39% of the samples tested (32% of pediatric and 52% of adult samples). EOI-MRD was frequently positive among ETP-ALL (75%, P=0.001) and near-ETP-ALL (71%, P=0.009) patients, compared to con- T-ALL patients (22.5%).

Median (range) MRD quantified among con-T-ALL, ETP-ALL, and near-ETP-ALL samples was 0.192% (0.015–2.125), 5.360% (0.125–30.306), and 4.250% (0.532–10.436), respectively. There was a significant difference in EOI-MRD quantified between con-T-ALL vs. near-ETP-ALL patients (P=0.019), but not between con-T-ALL vs. ETP-ALL (P=0.074) and ETP-ALL vs. near-ETP-ALL (P=0.898) patients. Comparison of the clinico-laboratory profiles of our EOI-MRD positive and negative T-ALL patients, and their 2-year OS, RFS, and EFS rates are presented in Supplementary Table 3.

Age group-specific analysis revealed significantly different frequencies of EOI-MRD positivity between the subcategories of T-ALL (con-T-ALL vs. ETP-ALL vs. near-ETP- ALL) among pediatric (15% vs. 83.3% vs. 60%, P=0.002), but not adult T-ALL patients (38.5% vs. 67% vs. 100%, P=0.190).

Analysis of the effect of induction therapy on expressions of CD4, CD8, CD7, CD5, CD38, and sCD3 antigens revealed statistically significant upregulations of CD8 (P=0.046) and CD38 (P=0.046) expression in EOI blasts of con-T-ALL and ETP-ALL patients, respectively (Supplementary Fig. 3).

Among the 81 T-ALL patients, 4 died before treatment and 11 left hospital care before initiating treatment (refer to Supplementary Fig. 4 for disease course during follow-up). Among the remaining 66 patients who were treated, 6 died during the induction phase (4.7% of con-T-ALL, 7.6% of ETP-ALL, and 27% of near-ETP-ALL; P=0.068). Induction failure was observed in 8.3% of ETP-ALL, 30% of near- ETP-ALL, and none of our con-T-ALL patients (P=0.005). A mean (±SD) follow-up of 12 (±10) months was available among 60 patients who had completed induction. None of the patients underwent hematopoietic stem cell transplantation.

Irrespective of the age at diagnosis and immunophenotypic subclassification, 2-year OS, RFS, and EFS rates among our T-ALL patients were 65%, 76%, and 64.5%, respectively. The survival profiles of our T-ALL patients pertinent to their immunophenotypic subcategorization are depicted in Table 1 and Fig. 2. Expression of CD56, CD19, and CD79a in the blasts did not have any significant impact (P>0.05) on the OS, RFS, and EFS among any of the immuno-phenotypic subcategories of T-ALL. The 2-year OS, RFS, and EFS specific to our pediatric and adult patients stratified by T-ALL subtypes are summarized in Table 2 and compared in Supplementary Table 2.

Irrespective of immunophenotypic subclassification and age, there were significant differences in 2-year OS (86% vs. 48%, P=0.013), RFS (87% vs. 57%, P=0.022), and EFS (83% vs. 58%, P=0.008) between our EOI-MRD negative vs. positive T-ALL patients (refer to Supplementary Table 3 and Supplementary Fig. 5A for Kaplan–Meier survival curves).

Age group specific analysis revealed significant differences in 2-year OS (95% vs. 53%, P=0.044), RFS (95% vs. 60%, P=0.010), and EFS (95% vs. 60%, P=0.010) among EOI-MRD negative vs. positive pediatric T-ALL patients. However, 2-year OS (58% vs. 41%, P=0.303), RFS (67% vs. 54%, P=0.727) and EFS (50% vs. 44%, P=0.435) was not significantly different between EOI-MRD negative vs. positive adult T-ALL patients (refer to Supplementary Fig. 5B for Kaplan–Meier survival curves).

T-ALL subtype specific analysis revealed significant difference in 2-year OS (94% vs. 37%, P=0.012), RFS (94% vs. 50%, P=0.005), and EFS (94% vs. 50%, P=0.005) between EOI-MRD negative vs. positive pediatric con-T-ALL patients. However, there were no significant differences in 2-year OS (50% vs. 53%, P=0.874), RFS (62% vs. 66%, P=0.584) and EFS (42% vs. 53%, P=0.891) between the EOI-MRD negative vs. positive adult con-T-ALL patients (refer to Supplementary Fig. 5B for Kaplan–Meier survival curves). The lower number of patients available at the EOI timepoint in ETP-ALL (6 pediatric and 6 adult) and near-ETP-ALL (5 pediatric and 2) subtypes precluded analysis of age group specific impact of EOI-MRD status on survival in these categories.

Cox proportional hazard regression analysis was performed to identify risks incurred by the immunophenotypic subtype of T-ALL, presence of mediastinal mass and hyperleukocytosis at diagnosis, and D8BNC and EOI-MRD positive status on 2-year OS, RFS, and EFS on our pediatric and adult patients. The results are presented in Table 3.

The 15% frequency of ETP-ALL documented in our pediatric T-ALL patients is similar to previously observed frequencies of 11% and 14% [4, 6]. As reported in other studies [17, 18], we too observed adult age predilection for ETP-ALL (P=0.039) in our cohort. ETP-ALL compromised 29% of our adult T-ALL patients. The documented frequency is variable, with rates of 17% in the United States, 32% in Germany, and 47% in China [1, 18-22]. This marked heterogeneity might be due either to ethnic predisposition or incongruencies in FCI analysis, where both ETP-ALL and near-ETP-ALL are considered under a common ETP-ALL category.

The exact worldwide frequency of near-ETP-ALL is unknown, as only a few studies have recognized this entity [3, 4, 6, 19]. In the present study, near-ETP-ALL was also frequent among adult T-ALL patients (26%, P=0.040). This frequency is similar to the 33% frequency observed by Van Vlierberghe et al. [19]. The 13% frequency of near-ETP-ALL we observed is similar to the 17% frequency reported by the Children’s Oncology Group [4], but is higher than the 5.4% frequency reported by Tembhare et al. [6] from India.

Our results indicate that these immunophenotypic subcategories of T-ALL cannot be distinguished by the presence of hepatosplenomegaly, lymphadenopathy, or mediastinal mass at diagnosis (P>0.05). Regarding the laboratory parameters at diagnosis, our adult ETP-ALL patients had significantly lower white blood cell (P=0.036) and higher platelet (P=0.016) counts compared to those in our adult con- T-ALL patients. These findings are consistent with those of prior studies [21, 22]. However, Ma et al. [23] had observed significantly low white blood cell counts among their pediatric ETP-ALL patients.

Both ETP-ALL and near-ETP-ALL blasts are proposed to have originated from BM-derived early thymic precursor (ETP) cells that migrated to the thymus. These ETP cells are too immature and have a transcriptome profile enabling differentiation towards T, myeloid, and dendritic cell lineages [1]. The dendritic-lineage orientation of ETP-ALL and near-ETP-ALL blasts was reflected in our results, as we observed a high frequency of CD123 positivity (P<0.001) in these patients as compared to con-T-ALL patients (Fig. 1). In our T-ALL cohort, CD73 expression was restricted only to ETP-ALL and near-ETP-ALL blasts (P=0.007). CD86 was significantly expressed only among ETP-ALL blasts (P< 0.001) (Fig. 1). The diagnostic relevance of this observation has to be verified in a larger cohort.

Regarding cross-lineage antigen expression among T-ALL blasts, expression of the CD56 antigen of natural killer (NK) cells is frequently associated with ETP-ALL blasts and confers a poor prognosis [24-27]. Consistent with the literature, we also observed a higher frequency of CD56 expression in our ETP-ALL and near-ETP-ALL patients (P=0.009). However, in contrast to the available literature, expression of CD56 did not make any difference in the OS, EFS, and RFS of any immunophenotypic subcategories of our T-ALL patients (P>0.05). This contrast might reflect the smaller cohort size, differences in treatment protocols used, and limited follow-up available among our patients. Despite recent observations regarding the expression of B-lineage markers CD19 and CD79a in ETP-derived blasts [6, 28], the prognostic relevance of such aberrant expression is unknown. In the current study, CD19 expression was observed only among ETP-ALL and near-ETP-ALL blasts and not among con- T-ALL blasts (P=0.004). In contrast, CD79a expression was not predilected towards any of the immunophenotypic subcategories of T-ALL (P=0.172). Importantly, aberrant expression of either CD19 or CD79a did not translate into inferior 2-year survival outcomes in our con-T-ALL, ETP-ALL, and near-ETP-ALL patient categories (P>0.05).

Traditional T-ALL MRD assessment by FCM relies on identifying the expression of immaturity associated markers like CD34, TdT, and CD99 on CD7 and cytoplasmic CD3 expressing lymphocytes [4, 8, 9, 29]. This approach is not foolproof as these immaturity-related antigens are frequently down-regulated during treatment [29]. T-ALL MRD analysis by FCM is also hindered by the presence of NK cells and their precursors that can mimic residual disease [6]. Due to these shortcomings, most T-ALL MRD data are by high-throughput sequencing for IgG and TCR rearrangements. The literature on T-MRD by FCM is limited [4, 8, 9, 29, 30]. However, these molecular MRD detection techniques might not be successful in ETP-ALL and near-ETP-ALL samples as these leukemias originate from precursor cells that are too immature to have undergone TCR rearrangement [3].

With the increased availability of ≥8 color flow cytometers, the results and sensitivity of T-MRD assessment by FCM are highly comparable to molecular T-MRD assays [8]. Use of 8–9 color panels by the Children’s Oncology Group (COG) yielded EOI-MRD detection rates of 30.5%, 81.4%, and 64.8% in pediatric (N=1,144) con-T-ALL, ETP-ALL, and near-ETP-ALL patients, respectively [4]. In an Indian study in which 35 T-ALL patients of all age groups were analyzed using an 8-color panel, EOI-MRD was detectable in 37% of the patients [30]. A recent study discussed the experience with T-ALL MRD using an 11-antigen 10-color FCM panel. Use of a mature T-cell antigen-based “exclusion” approach for gating detected EOI-MRD in 46.5% of the pediatric T-ALL cohort (N=269) [6]. In all these studies, cumulative MRD positivity observed among ETP-ALL and near-ETP-ALL patients was higher than the MRD positivity observed among con-T-ALL patients (74% with P<0.001 [6], 73.1% by the COG [4], and 67% with P=0.033 [30]). Our result was similar (73% with P<0.001).

Tembhare et al. [6] observed subtle down-regulation of surface CD3, CD4, CD5, CD8, and CD38 expression in their EOI-residual blasts. In the current study, we observed stable expression of CD4, CD5, CD7, and surface CD3 between the baseline and EOI-residual leukemic blasts (P>0.05) among all immunophenotypic subcategories of T-ALL. However, there was upregulated expression of CD8 (P=0.046) and CD38 (P=0.046) in the EOI blasts of con-T-ALL and ETP-ALL samples, respectively (Supplementary Fig. 3). Tembhare et al. [6] described that these immunophenotype shifts were too subtle to interfere with MRD analysis. Our findings are similar. These collective results indicate that the mature T-cell-related antigen-based approach is reliable for T-MRD analysis by FCM. Importantly, the expression of CD38 in nearly 98% of T-ALL patients at diagnosis in the present and previous [6] studies, along with the stable/upregulated expression of this antigen in the EOI-residual blasts, can be used as a potential target for daratumumab therapy [31]. In our cohort, CD73 expression was specifically associated with ETP-ALL and near-ETP-ALL patients (Fig. 2). Recent optimistic results were obtained with the use of anti-CD73 monoclonal antibody-based treatment in solid tumors [32]. Validation of the stability and specific expression of CD73 in ETP-ALL and near-ETP-ALL patients in a larger cohort would provide a rationale for initiating anti- CD73-based clinical trials in the treatment of these patients.

EOI-MRD status among our ETP-ALL and near-ETP-ALL patients was independent of their baseline clinical and laboratory characteristics. However, our EOI-MRD negative con-T-ALL patients were frequently diagnosed with mediastinal mass, high white blood cell count, and hyperleucocytosis at diagnosis as compared to those in the EOI-MRD positive counterparts.

Conter et al. [33] observed D8BNC status among 55% of their pediatric ETP-ALL patients. This result is similar to that in our cohort, where 50% of patients in each of the ETP-ALL and near-ETP-ALL categories were steroid unresponsive on day 8 (Table 2). High incidence of induction failure and disease relapse has been documented among pediatric ETP-ALL patients [2, 4, 23, 34]. However, none of our pediatric con-T-ALL, ETP-ALL, or near-ETP-ALL patients experienced induction failure or had a significantly higher relapse (Table 2). These contrasting observation have to be validated in a larger cohort with extended follow-up.

We observed no differences (P>0.05) in the 2-year OS, RFS, and EFS among our pediatric con-T-ALL (79%, 83%, and 81%, respectively), ETP-ALL (67%, 87%, and 80%), and near-ETP-ALL (100%, 75%, and 75%) patients. Consistently, there were no significant differences in 5-year OS and EFS among the con-T (92% and 86.9%, respectively), ETP-ALL (93% and 87%), and near-ETP-ALL (91.6% and 84.4%) patients documented in the largest ever pediatric T-ALL data by the Children’s Oncology Group [4]. However, inferior 2-, 4-, 5-, and 10-year OS and EFS have also been documented among pediatric-ETP-ALL patients from Italy, Japan, China, and USA, respectively [2, 23, 34]. These different observations might be due to heterogeneity in the sample sizes and ethnicities of the study cohorts. Hence, the true prognostic relevance of both ETP-ALL and near-ETP-ALL will be determined only by prospective studies with a large sample size and longer follow-up.

Irrespective of immunophenotypic sub classification, our EOI-MRD negative pediatric T-ALL patients had significantly better 2-year OS (95% vs. 53%, P=0.044), RFS (95% vs. 60%, P=0.010), and EFS (95% vs. 72%, P=0.010) than their EOI-MRD positive counterparts, consistent with previous observations [7].

In the current study, we did not observe any differences in 2-year OS, EFS, and RFS between our adult con-T-ALL and ETP-ALL patients (Table 2). These findings are consistent with two prior studies [21, 22], but discordant with the inferior OS documented among adult & adolescent ETP-ALL patients by Jain et al. [20]. Importantly, our adult near- ETP-ALL patients had the most inferior 2-year OS, EFS, and RFS as compared to that in the con-T-ALL and ETP-ALL patients diagnosed in this age group (Table 2).

An important observation from our study is the EOI-MRD positive status (irrespective of T-ALL subtype) and the presence of near-ETP-ALL immunophenotype (irrespective of EOI-MRD status) influencing the 2-year survival profile among our pediatric and adult T-ALL patients, respectively. These results are encouraging concerning hematopoietic stem cell transplantation at first remission in EOI-MRD positive pediatric T-ALL patients and all adult T-ALL patients diagnosed with near-ETP-ALL. However, the small age-specific sample size in our study precludes any definite conclusions.

Being a relatively rare disease, the number of patients in our cohort might not be powered for exact outcome analysis across each immunophenotypic subcategory of T-ALL. Hence, the survival outcomes documented in our study have to be validated in a larger cohort. The mutational profile of leukemic lymphoblasts was not evaluated in our patients. Being retrospective data, our study results are purely observational and did not explore the underlying leukemogenic differences between the subtypes of T-ALL.