Improved identification of T‐ALL paediatric patients with low MRD but poor prognosis using a transcriptional signature

T-cell acute lymphoblastic leukaemia (T-ALL) is an aggressive haematological malignancy that results from the malignant transformation of immature thymocytes and exhibits a heterogeneous molecular landscape.1 Current standard-of-care treatments for T-ALL mainly consist of high-dose multi-agent chemotherapy and achieve reasonable rates of initial complete responses that can reach up to 90% in children. However, refractory or relapsed patients have a dismal prognosis with survival rates below 10%; hence, most efforts are focused on optimizing first-line treatments and further improving their efficacy. In this context, prognostic markers capable of stratifying patients according to their potential risk are essential, so high-risk patients may benefit from more intensive chemotherapy regimens or haematopoietic stem cell transplantation, while low-risk patients may benefit from less intensive treatment approaches, which are characterized by reduced toxicity levels and fewer side effects.2 Unfortunately, the number of available prognostic factors for T-ALL is significantly reduced, and the most relevant prognostic marker is the assessment of minimal residual disease (MRD), which is usually calculated as the number of leukaemic blasts relative to the number of nucleated cells. MRD is a direct measure of treatment response, so that when treatments are successful, the MRD will be low and the prognosis will tend to be favourable, whereas, for those treatments with only a partial effect, the MRD will be high and the prognosis will tend to be adverse.3 MRD analysis after induction chemotherapy provides valuable information about tumour aggressiveness and efficacy of first-line treatments, becoming a reliable prognostic factor for risk stratification and outcome prediction.4 However, MRD analysis has certain limitations, and although values below 0.01% after induction chemotherapy have been considered a good prognostic marker by several studies,5, 6 there is still a substantial proportion of these patients that subsequently experience an adverse outcome. In this respect, it would be useful to identify novel biomarkers that can be easily implemented in clinical practice and allow for improved discrimination of low-MRD patients according to their prognosis. Despite the fact that MRD is the only prognostic factor routinely used in the clinical management of T-ALL, the molecular bases underlying adverse outcomes in low-MRD patients remain unexplored. Therefore, we hypothesized that analysing the transcriptional profile of low-MRD patients could contribute to the identification of genes that are differentially expressed between cases with good and poor prognosis, thus emerging as potential candidates to become novel biomarkers.3 We initially analysed data from 265 cases enrolled in the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiative with available information regarding the clinical history and the molecular profile of the tumour sample at diagnosis. The results were further validated in 1070 additional cases from the Gabriella Miller Kids First Pediatric Research Program (Figure S1A). All patients received a homogeneous first-line treatment according to the Children's Oncology Group clinical trial AALL0434. MRD was assessed centrally on AALL0434 by flow cytometry at day 29. Statistical analyses were performed using GraphPad Prism version 10, R 4.0.3 and Python 3.13.2. Statistical significance was set at p < 0.05. Additional information is available in Supporting Information Materials and Methods. To explore the molecular basis of patients with low MRD but poor prognosis, we first analysed paired clinical and molecular data relative to 265T-ALL paediatric patients from the TARGET initiative. We specifically focused on T-ALL paediatric patients with MRD below 0.01% after induction chemotherapy (day 29) and therefore considered them as low-MRD patients. These patients accounted for 64.9% of the total (Figure 1A) and showed an 11% probability of experiencing an adverse event (defined as progression, relapse, secondary malignancy or death from any cause) during the clinical course of the disease (Figure S1B). To identify genes potentially deregulated in patients with low MRD but adverse outcomes, we divided low MRD patients into two different groups according to their prognosis and examined their transcriptional profile. In this regard, HTATIP2, IL17RA, KCNAB2 and LINC00426 genes were upregulated in patients with poor prognosis compared to those with good prognosis (Figure 1B) and have been previously implicated in cancer,7-9 becoming promising candidates for subsequent studies. Notably, IL17RA overexpression was identified as an oncogenic event in T-ALL and associated with an increased migration of leukaemic cells.10 We therefore tested whether the differential expression of these genes could have an impact on outcomes and observed that overexpression of all the selected candidates was associated with a higher proportion of adverse events compared to low mRNA expression levels (Figure S1C). To evaluate whether HTATIP2, IL17RA, KCNAB2 and LINC00426 upregulation in T-ALL paediatric patients with low MRD after induction chemotherapy occurs randomly or instead follows a specific pattern, we explored their potential correlations. A substantial number of patients showed high expression levels for HTATIP2, IL17RA and LINC00426, while KCNAB2 overexpression preferentially co-occurred with LINC00426 but not with HTATIP2 or IL17RA (Figure 1C). We next investigated whether the upregulation of the selected genes could result in a transcriptional signature for the efficient discrimination between good and poor prognosis in low MRD patients. Since there is a significant positive correlation between the upregulation of HTATIP2 and IL17RA, as well as between the upregulation of KCNAB2 and LINC00426 (Figure S1D), we hypothesized that these genes might work in pairs and lead to more refined selection criteria. In this regard, low MRD patients with simultaneous overexpression of either HTATIP2 and IL17RA or KCNAB2 and LINC00426 (HTATIP2&IL17RA/KCNAB2&LINC00426) exhibited a significantly shorter event-free survival (EFS) than the remaining low MRD patients (Figure 1D). Since death from any cause may not necessarily be related to the molecular profile of tumour samples, we also confirmed the potential of the transcriptional signature when such end-point is excluded from the list of adverse events (Figure S1E). Notably, those patients who were positive for the transcriptional signature showed a higher cumulative incidence of relapse (CIR), further confirming their association with adverse outcome (Figure 1E). We next conducted univariate/multivariate analyses incorporating multiple clinical parameters considered to be risk factors for ALL and observed that only the transcriptional signature yielded statistically significant p-values, emerging as a novel independent prognostic factor for low MRD patients (Figure 1F). Finally, we evaluated the potential association between the transcriptional signature and several genetic alterations that have been recently described as poor prognostic indicators for T-ALL patients.11 Our results showed that the transcriptional signature was significantly associated with different genetic alterations targeting the PI3K pathway, specifically PTEN intragenic loss/deletion and PIK3CD SNV/Indel (Figure S1F). However, patients positive for both events represent only a small fraction, suggesting that the transcriptional signature is not exclusively limited to cases with the aforementioned alterations. To validate the potential of the transcriptional signature, we next analysed paired clinical and molecular data relative to novel 1070 T-ALL paediatric patients from the Gabriella Miller Kids First initiative. In this subset, low MRD patients after induction chemotherapy (day 29) accounted for 57.9% of the total (Figure 2A) and showed an 11.6% probability of experiencing an adverse event (defined as progression, relapse, secondary malignancy or death from any cause) during the clinical course of the disease (Figure S2A). We observed that HTATIP2, IL17RA, KCNAB2 and LINC00426 gene upregulation was preferentially found in those patients with poor prognosis (Figure 2B) and was again associated with a higher proportion of adverse events (Figure S2B). We also confirmed the pairwise correlations described above between HTATIP2 and IL17RA as well as between KCNAB2 and LINC00426 (Figure S2C). We observed that patients positive for the transcriptional signature had shorter EFS, whether or not death from any cause was considered as an adverse event (Figure 2C; Figure S2D), and showed a higher CIR (Figure 2D). Moreover, we confirmed the potential of the transcriptional signature as a novel independent prognostic factor for low MRD patients (Figure 2E), postulating that the differential expression of such genes combined with the information provided by the MRD analysis may become useful for identifying those T-ALL cases with low MRD who are likely to develop an adverse clinical course. Finally, we observed that the transcriptional signature was again associated with different alterations targeting the PI3K signalling pathway, further supporting a relationship between both events (Figure S2E). Although MRD is the most relevant prognostic factor in T-ALL,3 it also has certain limitations that demand the identification of additional prognostic markers for patients who show low MRD values after induction chemotherapy but subsequently experience poor prognosis. Therefore, analysing the molecular profile together with the clinical history emerges as a promising approach for the identification of novel biomarkers that contribute to a more accurate risk stratification of T-ALL patients.2 While the immunophenotype and the mutational background have been traditionally considered the two main sources for identifying novel biomarkers, they drastically reduce the number of potential candidates, since the immunophenotype consists of only five major subgroups and the number of genes affected by recurrent genetic alterations is around 160.11 Consequently, most of the biomarkers proposed for T-ALL have not shown reproducible results across different studies.6, 12, 13 Based on the above, we hypothesized that the transcriptional profile may exhibit greater potential to capture the inherent variability among tumour samples, since the aberrant activity of a single transcription factor leads to multiple regulatory levels and complex signalling networks. Indeed, T-ALL is characterized by the aberrant activity of several transcription factors11 that can be affected by either genetic and epigenetic modifications, as in the case of STAT5, whose overactivation can be caused by missense mutations in JAK3 or PTPRC, as well as by downregulation of SOCS314 or upregulation of IL7R.15 We found that HTATIP2, IL17RA, KCNAB2 and LINC00426 genes are preferentially overexpressed in T-ALL paediatric patients with low MRD but poor prognosis. Moreover, we demonstrated that a transcriptional signature based on the expression levels of those genes may be useful for the identification of patients prone to develop an adverse clinical course despite a favourable response to induction chemotherapy. A.L., L.V.-M. and M.V.-M. conceived and designed the initial hypothesis and the working plan. A.L., L.V.-M. and M.V.-M. performed bioinformatics analysis from patients' data. P.F.-N. assisted with bioinformatics analysis and data interpretation. J.L.L.-L., J.C. and P.L. provided clinical advice. M.F. and J.S. made critical intellectual contributions throughout the project. M.F., J.S. and M.V.-M. obtained funding. A.L., L.V.-M. and M.V.-M. wrote the original draft, which was reviewed and edited by all authors prior to submission. All authors approved the final version of the manuscript. This work was supported in part by funds from Ministerio de Economía y Competitividad (SAF2015-70561-R; MINECO/FEDER, EU to M.V.-M.); Ministerio de Ciencia, Innovación y Universidades (RTI2018-093330-B-I00; MCIU/FEDER, EU to J.S.); Ministerio de Ciencia e Innovación (PID2022-137487OB-I00; MINECO/FEDER, EU to M.F.) and Instituto de Investigación Sanitaria Fundación Jiménez Díaz to J.S. Institutional grants from the Fundación Ramón Areces and Banco de Santander to CBM are also acknowledged. The authors have no competing interests. Data from the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiative are managed by the National Cancer Institute (NCI) and accessible in the database of Genotypes and Phenotypes (dbGaP, https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000464.v23.p8), accession number phs000218.v26.p8 and substudy specific accession number phs000464.v23.p8 (TARGET Acute Lymphoblastic Leukemia (ALL) Expansion Phase 2). Data from Gabriella Miller Kids First Pediatric Research Program are accessible in the dbGaP (https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs002276.v3.p1), specific accession number phs002276.v3.p1 (Gabriella Miller Kids First Pediatric Research Program in Pediatric T-Cell Acute Lymphoblastic Leukemia). Data S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.