Between April 2012 and April 2023, 144 paediatric patients (0–18 years) diagnosed at the Children’s Hospital of Soochow University and who underwent NGS were enrolled. All diagnoses were made according to the MICM criteria. Among them, 97 patients received the Chinese Children’s Leukaemia Group-ALL-2008 (CCLG-ALL-2008) protocol (14) and 47 patients received the Chinese Children’s Cancer Group ALL-2015 (CCCG-ALL-2015) protocol (15). A total of 54 and 90 patients were classified into intermediate-risk (IR) and high-risk (HR) groups, respectively. This study was approved by the Ethics Committee of Children’s Hospital of Soochow University and followed the principles of the Declaration of Helsinki. Written informed consent for participation was obtained from the parents/guardians of all the patients.

The CCLG-ALL-2008 protocol consisted of the following components: Induction therapy: VDLD (vincristine, daunorubicin, L-asparaginase, dexamethasone); Early intensification: CAM ×2 (cyclophosphamide, cytarabine, 6-mercaptopurine); Consolidation: High-dose methotrexate (HD-MTX ×4); Delayed intensification I and II: VDLD plus CAM or D-CAM; Maintenance: 6-mercaptopurine (6-MP) and methotrexate (MTX), with or without additional agents depending on risk group. High-risk patients received intensified treatment modules (HR1–3) adapted from the BFM protocol.

The CCCG-ALL-2015 protocol included: Induction therapy: PVDL (prednisone, vincristine, daunorubicin, L-asparaginase); Early intensification: CAT ×2 (cyclophosphamide, cytarabine, 6-MP); Consolidation: High-dose methotrexate (HD-MTX×4); Continuation (Weeks 16–31): DVDL (dexamethasone, vincristine, daunorubicin, L-asparaginase); Reinduction (Weeks 32–34): High-dose cytarabine (HD-Ara-C); Continuation (Weeks 35–54): 6-MP, MTX, and DVCA (dexamethasone, vincristine, cyclophosphamide, cytarabine); Maintenance: 6-MP and MTX, with or without vincristine and dexamethasone depending on randomised group assignment. High-risk patients under the CCCG-ALL-2015 protocol were eligible for hematopoietic stem cell transplantation.

For T-ALL patients, risk stratification in both protocols was based on a combination of clinical and response-related parameters, including: response to prednisone window therapy, bone marrow morphology on Day 15 (CCLG) or Day 19 (CCCG), and again on Day 33 or 46, initial white blood cell (WBC) count, extramedullary involvement, including central nervous system (CNS) or testicular infiltration, presence of high-risk cytogenetic abnormalities, if available.

TRIzo (invitrogen, Cat 15596026) was used to isolate total RNA from bone marrow samples. Ribosomal RNA (rRNA) was depleted, and the stranded RNA-Seq libraries were rapidly constructed from 1 µg of purified total RNA (KAPA RNA HyperPrep Kit with RiboErase (HMR) KIT) according to the manufacturer’s instructions. After fragmentation, cDNA synthesis, adapter ligation, and library amplification, each library was sequenced on an Illumina NovaSeq 6000 platform with a 150-cycle paired-end sequencing. The raw FASTQ data were filtered using the fastp (version 0.23.4) software, which involved the removal of adapter sequences and low-quality sequencing reads. Subsequently, the filtered data were aligned to the reference genome (GRCh37) using STAR (version 2.7.10a) software. The aligned sequences were sorted using the Sambamba (version 0.6.6) software. Duplicate reads in the sorted sequences were removed using the GATK (version 4.0.5.1) software. After filtering, alignment, sorting, and duplicate removal, the resulting BAM files were used for the following analyses: (1) IKZF1 deletions were calculated using rnapeg (version 2.7.7) software; (2) the fusion genes were analysed using Arriba (version 2.4.0), Cicero (version 1.9.6), and Fusioncatcher (version 1.33) software; (3) gene expression levels were quantified using htseq-count (version 2.0.2) software. (4) INDEL analysis: INDELs (insertions and deletions) were detected using the lianti (revision 142) software. The detected mutations were annotated using annovar (updated on 2020-06-08) and snpeff (version 5.1f). All molecular testing was performed on bone marrow samples to ensure detection accuracy, as peripheral blood may contain fewer tumor cells and yield lower variant allele frequency (VAF).

All statistical analyses were performed using SPSS version 25 and R 4.3.2 software. Associations between specific gene mutations and clinical features were analysed using Fisher’s exact test or chi-square test. To account for multiple testing, false discovery rate (FDR) correction was applied. Overall survival (OS) was defined as the time from diagnosis to death. Event-free survival (EFS) was defined as the time from diagnosis (or treatment initiation) to the occurrence of any of the following events: failure to achieve complete remission (refractory disease), relapse, disease progression, or death from any cause. Cumulative incidence of treatment failure (CITF) is defined as the cumulative probability over time of experiencing either relapse or refractory disease, while accounting for death as a competing event. Median follow-up time was calculated using the reverse Kaplan–Meier method, in which censoring and event statuses are reversed. Survival functions were estimated using the Kaplan–Meier method and compared using the log-rank test. Cox regression analysis was used to identify factors associated with OS and EFS. CITF was estimated using a competing risk model and death in complete remission (CR) was treated as a competing event. Gray’s test was used for univariate comparisons between groups. Variables with P < 0.20 in univariate analysis were included in both the Cox proportional hazards model and the multivariable competing risk model. P<0.05 was considered statistically significant.

Among the 144 paediatric patients with T-ALL, 116 (80.6%) were male and 28 (19.4%) were female, with a ratio of 4.1:1. The median age at diagnosis was 8.0 years (range 1.2–16.8 years). The median white blood cell (WBC) at diagnosis was 106.33 ×10⁹/L (range 1.05–750.20). The median haemoglobin and platelet counts at diagnosis were 98.50 g/L (range 37.00–159.00) and 63.00×10⁹/L (range 5.00–412.00), respectively. Regarding the immunophenotype, 6 patients had early thymic precursor (ETP) (4.2%), and 138 had non-ETP (95.8%) in our cohort. At the end of follow-up, 64 patients (44.4%) underwent haematopoietic stem cell transplantation (HSCT), while 80 patients (55.6%) received chemotherapy alone. Among the 64 patients who received HSCT, 54 were transplanted in first complete remission (CR1), 9 in second complete remission (CR2) following relapse, and 1 patient underwent transplantation without achieving remission (NR). The baseline patient characteristics are summarised in Table 1 .

The gene mutation data from 144 children with T-ALL who underwent second-generation sequencing were collected. Among them, 142 patients (98.6%) harboured at least one mutation, and 125 patients (86.8%) had two or more mutations. After applying the VAF threshold, we focused our analysis on mutations of potential clinical relevance. NGS results revealed that the top 20 genes were NOTCH1 (58.3%, 84/144), FBXW7 (19.4%, 28/144), PTEN (17.4%, 25/144), NRAS (9%, 13/144), PHF6 (9%, 13/144), USP7 (9%, 13/144), WT1 (6.9%, 10/144), IL7R (6.9%, 10/144), JAK3 (6.3%, 9/144), EZH2 (4.9%, 7/144), BCL11B (4.9%, 7/144), DNM2 (4.9%, 7/144), SPATA31E1 (4.9%, 7/144), CCND3 (4.2%, 6/144), KRAS (3.5%, 5/144), PIK3R1 (3.5%, 5/144), CDKN2A (3.5%, 5/144), STAT5B (3.5%, 5/144), SUZ12 (3.5%, 5/144), MKI67 (3.5%, 5/144) ( Figure 1a ). Next, we divided those gene mutations into eight groups according to the signalling pathways involved in gene mutations and their functions, including: NOTCH (62.5%, 90/144), epigenetic regulators (31.9%, 46/144), transcriptional factors (27.8%, 40/144), PI3K-AKT (25%, 36/144), RAS (18.1%, 26/144), JAK-STAT (16.7%, 24/144), cell cycle regulation (15.3%, 22/144), translation and RNA stability (3.5%, 5/144) pathways ( Figure 1b ). In addition, we analysed the co-occurrence of the mutated genes. As shown in Figure 1c , we found pairwise associations between NOTCH1 and PTEN (P<0.01), FBXW7 and USP7 (P<0.05).

No significant differences in gene mutation were observed between male and female groups (all P>0.050) ( Figure 2a ). No significant differences in gene mutation were observed between 1–9.9 years and ≥10 years groups (all P>0.050) ( Figure 2b ). Moreover, NOTCH1 mutations were associated with elevated WBC counts (P = 0.018) ( Figure 2c ). Patients with PHF6 mutations were more likely to develop lower haemoglobin levels at diagnosis (P = 0.001) ( Figure 2d ). Additionally, NOTCH1 mutations were associated with a higher platelet count at diagnosis (P = 0.030) ( Figure 2e ).

No significant differences in gene mutation were observed between insensitive and sensitive to steroid treatment in this cohort (all P>0.050) ( Figure 3a ). The NOTCH1 mutation was associated with a better D15/19 BM response with a worse response (P = 0.023, Figure 3b ). No gene mutations were related to other early treatment outcomes, such as D15/19 minimal residual disease (MRD), D33/46 BM response, and MRD ( Figure 3c–e ).

The latest follow-up occurred on 2023 December 31, and the median follow-up was 42 months (range, 1.0–129.0 months). The 5-year OS for 139 paediatric T-ALL was 69.90 ± 4.62% ( Figure 4a ). The ten most common gene mutations were NOTCH1, FBXW7, PTEN, RAS (NRAS/KRAS), PHF6, USP7, WT1, and IL7R. We explored their prognostic roles in the 5-year OS of 139 paediatric patients with T-ALL (5 of 144 patients were lost to follow-up). However, no prognostic biomarker was found among the gene mutations that influenced the survival rate of these patients ( Figure 4b–i ).

Of 84 patients with NOTCH1 mutations, 71 had variant allele frequencies (VAF) data. Based on the maximally selected log-rank statistic, a VAF of 12.850 was the cutoff point for differentiating patient prognosis. Patients were categorised into high (52 patients) and low (19 patients) mutation load groups. Regarding clinical characteristics, the high mutation load group had higher bone marrow blasts at diagnosis, and a lower proportion underwent allogeneic HSCT. Other clinical features were not significantly different between the two groups (all P>0.05). The clinical features of the 71 patients are summarised in Table 2 , and the detailed results with false discovery rate (FDR) correction are provided in Supplementary Table 1 . Regarding the association between NOTCH1 mutation load and early treatment response, we found that a high NOTCH1 mutation load have a lower proportion of MRD ≥1% on D15 ( Table 3 ). Additionally, we examined the impact of the NOTCH1 mutation load on the 5-year OS, EFS and CITF in the cohort of 71 paediatric patients with T-ALL. Our results indicated that patients with a high NOTCH1 mutation load have higher 5-year OS ( Figure 5a ) and EFS ( Figure 5b ) than those with a low NOTCH1 mutation load (P = 0.02 and 0.043, respectively). No statistically significant difference was observed for CITF between the two groups ( Figure 5c ), although a trend toward lower incidence was noted in the high mutation load group (Gray’s test, P = 0.067). Table 4 presents the univariate and multivariate analyses of the factors affecting the 5-year OS, EFS, and CITF. In the univariate analyses, we found that platelet count less than and equal to 50×10⁹ was a risk factor for 5-year OS, BM not CR after induction was a risk factor for 5-year EFS, while a high NOTCH1 mutation load was a protective factor for 5-year OS and EFS. In the multivariate analyses, a high NOTCH1 mutation load was an independent protective factor for the 5-year OS (P = 0.042) and 5-year EFS (P = 0.0104). However, high VAF was not significantly associated with CITF in our univariate and multivariate analysis. We further validated the role of VAF in NOTCH1 mutations on paediatric T-ALL OS. The training set consisted of 50 patients treated with the CCLG-ALL-2008 protocol, and 21 patients who received the CCCG-ALL-2015 protocol were used for the validation set. A VAF of 0.1594 was identified as the cutoff point for differentiating prognosis among patients treated with CCLG-ALL-2008 protocol (83.0 ± 8.39 vs 65.0 ± 11.82, P = 0.23, Figure 5c ). When this cutoff was applied to patients who received CCCG-ALL-2015, we still found superior survival in the high mutation load group (100% vs 71.4 ± 17.1%, P = 0.041, Figure 5d ).