Short- and Long-Read RNA Sequencing Improve Molecular Profiling of Pediatric T-Cell Acute Lymphoblastic Leukemia

Acute lymphoblastic leukemia (ALL) is one of the most common pediatric cancers, accounting for approximately 1/3 of childhood cancer diagnoses. Of these patients, ~15% are diagnosed with T-cell ALL (T-ALL). Pediatric T-ALL is less well-characterized and has a worse prognosis than its B-cell counterpart, particularly after relapse. Although many genetic drivers for pediatric T-ALL have been characterized, this information does not currently inform treatment selection for patients. By integrating transcriptional profiling data with genomic findings from molecular and cytogenetic assays, we aim to better characterize the cellular landscape driving distinct subtypes of pediatric T-ALL.

In this study, we performed short-read (n = 5) and long-read (n = 3) RNA sequencing (RNAseq) on frozen bone marrow aspirate samples collected at diagnosis from pediatric T-ALL patients. For short-read RNAseq, total RNA was isolated and libraries prepared using the Illumina Stranded Total RNA Prep kit, then sequenced on the Illumina NovaSeq. Alignment was performed to GRCh38 using STAR 2.6.0 and read counts estimated using kallisto 0.46.2. For long-read RNAseq, libraries were prepared using Pacific Biosciences IsoSeq method for cDNA synthesis and library construction with SMRTbell Prep Kit 3.0 and sequenced on the Pacific Biosciences Sequel IIe or Revio systems. HiFi reads were processed using isoseq 4.0.0, with FLNC reads aligned to GRCh38 using minimap2 2.28-r1209. Fusion detection was performed using FLNC reads as input to FusionSeeker 1.0.1. Downstream analysis and visualization were performed in R 4.3.3.

We found that integrating both long- and short-read RNAseq data with genomic findings was most informative. For one patient, we were able to use long-read RNAseq to confirm a PICALM::MLLT10 fusion that had been suspected based on microarray data. This fusion is known to drive upregulation of HOXA family genes, and we observed elevated expression of both HOXA9 and HOXA10 (compared to other patients in the cohort) in the short-read RNAseq data. In a different case, a STIL::TAL1 fusion had been suspected based on clinical microarray data. Although this fusion was not called by the fusion detection algorithm we used, manual inspection revealed numerous supporting reads encompassing the STIL promoter and the TAL1 gene. The expected corresponding upregulation of the TAL1 gene compared to other patients was also observed in this patient sample using short-read RNAseq. Short-read RNAseq further supported previously identified loss of CDKN2A in 3/5 patients, and enabled comparison of gene expression known to be implicated in T-ALL pathogenesis such as BCL11B, MEF2C, and LMO2 across the cohort. Long-read sequencing presented a unique opportunity to explore the expression of specific isoforms, such as the TAL1-short isoform, which we uniquely observe in one patient and has been identified as a putative tumor suppressor. Each RNAseq approach provided distinct insights into T-ALL biology.

By integrating short- and long-read RNAseq data with genomic microarray profiling, we were able to not only confirm but also extend and contextualize observations about the cellular state of T-ALL leukemic blasts. We find that a combination of all three data types gives a comprehensive picture of the downstream effects of genetic lesions and suggests mechanisms through which distinct subtypes of pediatric T-ALL may drive cancer progression. Future work extending to high-throughput concatenated long-read RNAseq (Pacific Biosciences Kinnex), using this multi-modal profiling approach across a larger cohort will facilitate improved understanding of genomic drivers, and may improve individualized treatment selection and outcomes for pediatric patients with T-ALL.

Farooqi:Bayer: Honoraria; Pacific Biosciences: Honoraria, Other: travel support; 10X Genomics: Honoraria, Other: travel support.