Almost 20% of Mexican children with acute lymphoblastic leukemia (ALL) die within the first 2 years, and relapses are over 4 times more frequent in Mexico than in high-income countries, mainly early relapses (<36 months from diagnosis or at treatment completion)(Núñez-Enríquez et al., BMC Cancer 2019). It has been observed that gene fusions associated with good prognosis are less common in Mexican ALL cases, but biomarkers associated with older age and poor prognostic molecular subtypes are more frequent. These observations have been reported in Hispanics in the US and in other Latin-American populations. For instance, ETV6-RUNX1, a favorable prognosis marker, shows low frequency in Mexican children and other Native American populations (~10%), but Ph-like subtype is more common in Hispanics (11-15%) and has been associated with ancestry (Lee et al., JAMA Oncol, 2022). This project aimed to identify gene fusions by whole transcriptome sequencing (RNA-seq) in a sub-cohort of Mexican children with ALL. We carried out a retrospective multicenter cohort study as part of the Mexican Interinstitutional Group for the Identification of the Causes of Childhood Leukemia (MIGICCL). From a cohort of 496 children with ALL treated in public hospitals from Mexico City, we selected 49 patients for RNA-seq analysis. All patients were followed-up for at least 36 months. Total RNA was extracted from bone marrow and sequencing was done with Illumina. Fusion detection was performed with FusionCatcher software and differential expression was done with DESeq2. All fusions were manually curated and inspected in IGV software to eliminate false positives. In cases without an identified fusion, raw reads were aligned to GRCh38 genome with STAR software and soft-clipped reads mapping upstream of CRLF2, DUX4, and to IGH locus were searched to identify IGH alterations. Known fusions were validated by RT-PCR and novel ones by RT-PCR and Sanger sequencing. By National Cancer Institute (NCI, USA) risk criteria, thirty-five (71.4%) patients were classified as high-risk, 23 (47%) had early relapse (Table 1). The 4-year overall survival (OS) of the studied population was 50% (95% CI: 37-67). No differences in EFS and OS were observed between patients classified as high- or standard-risk (p=0.87 and p=0.78, respectively). Overall, 65.3% of patients had at least one fusion. Thirty-one different fusions were found and 14 are considered recurrent in B-ALL. Nine fusions (18.4%) were not previously reported or had a novel partner fused to a recurrently mutated gene. By grouping them into well-known B-ALL subtypes, the most common were DUX4 (13%), CRLF2 (10.87%) and BCR-ABL1 (10.87%) (Figure 1). Four DUX4 cases had DUX4-IGH fusion, one case had a novel DUX4-ETV6 fusion and in another one the underlying genetic alteration could not be found but had the highest DUX4 expression. Regarding CRLF2r, P2RY8-CRLF2 was the most common fusion (4/5 cases), and one case had a CRLF2-IGH fusion that was missed by the fusion detection software but identified with targeted search via soft-clipped reads. ETV6-RUNX1 was found in 6.5%, in line with the low frequency reported in Latinos. The frequency of newly discovered subtypes like MEF2D, PAX5 (including a novel 3' partner in PAX5-ST18) and ZNF384 alterations was 6.5%, 4.3% and 2.2%, respectively, similar frequencies to those reported in bigger cohorts. As expected, none of TCF3-PBX1 and ETV6-RUNX1 patients died during follow-up. The remaining genetic subtypes had poor outcome. Four-year OS for kinase-driven subtypes was 50%, P2RY8-CRLF2 having the worst outcome (3/4 died before treatment completion). It was surprising to find the low OS of DUX4, a subtype associated with an excellent prognosis even in adult ALL. The 4-year OS of DUX4 was 33%, and all patients classified as DUX4 who relapsed also died during treatment. The outcome of patients without fusions was also poor, of just 51% after 4 years. We report the first RNA-seq-based analysis in Mexican children with high-risk ALL. We were able to classify around 60% of patients based on transcriptome data alone. Although the implementation of these technologies in a clinical setting is still a challenge, especially for developing countries, some genetic subtypes deserve consideration to explore their prognostic significance in bigger cohorts, perhaps using more feasible techniques to identify them, and to provide a more precise risk stratification. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal