Current treatment protocols for pediatric acute lymphoblastic leukemia (ALL) use information regarding genomic aberrations in the leukemic blasts together with clinical features at presentation, and initial therapy response is used to assign patients to different risk categories and treatment intensities (1). Survival of pediatric acute ALL in risk-adapted chemotherapy trials exceeds 90% (2); however, further improvements in outcome are not likely to be achieved with conventional chemotherapy but will require alternative treatment approaches such as immunotherapy or targeted therapy.

High throughput sequencing technologies have shown enormous potential in the genetic characterization of hematological malignancies, particularly in pediatric ALL (3), and are paving the way to personalized treatment strategies, although implementation in clinical practice is still meager (4). Technologies that interrogate the whole genome with high resolution are increasingly being adopted in the diagnostic workout of pediatric ALL (5). Here, we report a novel JAK2 fusion in a pediatric case of therapy-refractory T-ALL that was detected using paired-end whole genome sequencing (WGS) in the diagnostic setting.

A 7-year-old boy presenting with a leukocyte count 800 × 10⁹/L and peripheral lymphadenopathy was diagnosed with precursor T-ALL with extensive involvement of the bone marrow (BM) and peripheral blood. Multiparameter flow cytometry showed an immature T-lymphoblast population (95%) of cortical phenotype (EGIL III; CD45dim/CD7+/cytCD3+/Tdt+//CD4/CD8dim+/CD10−/CD1ahetero/CD56/16−/CD99++/CD5dim+/ CD2+/CD48+/CD38+//TCRa/b−/TCRg/d−/HLA−DR− {Bene, 1995 #41}). The blasts were negative for myeloid markers (cytMPO and CD33) and for B-cell markers (CD19 and cytCD79a). According to the standard of care (SoC) genomic characterization, the patient had a normal karyotype and none of the targeted fluorescence in situ hybridization (FISH) analysis was positive, i.e., no KMT2A- or BCR::ABL rearrangements and no ABL-class rearrangement were detected. A biallelic CDKN2A/B deletion was the only clonal aberration detected. The patient responded poorly to standard four-drug induction (dexamethasone, pegylated-asparaginase, daunorubicin, and vincristine) and was switched to consolidation at day 19 (6-mercaptopurine, cytarabine, and cyclophosphamide). Two weeks later, he still had 80% blasts in the bone marrow and was considered refractory and treatment with nelarabine initiated. Following two rounds of nelarabine and one additional high-risk chemotherapy block, remission was achieved, and he subsequently underwent hematopoietic stem cell transplantation (HSCT) from his haploidentical mother. The conditioning regimen consisted of ATG (grafalon), fludarabine, and thiotepa together with total body irradiation (TBI) (12 Gy in four fractions) and rituximab after which he received an alpha/beta T-cell-depleted peripheral stem cell graft. Initially, mycophenolate mofetil was switched to cyclosporin together with prednisolone after 45 days due to acute gut and skin graft versus host disease (GVHD). He is currently 8 months post HSCT, with no sign of disease but still on a low dose prednisolone.

At the time of initial failure, the genetic laboratory was engaged to screen for potential targets for experimental therapy. DNA isolated from the bone marrow sample taken at diagnosis was sequenced using a PCR-free, paired-end WGS protocol with a 30× coverage on an Illumina HiSeq X platform at Clinical Genomics (SciLifeLab, Stockholm) and annotated to the Human GRCh37 (hg19) build. Variants were identified using the MIP (6) validated for routine diagnostics and visualized in the SCOUT interface (7). MIP performs structural variant (SV) detection using Delly (8), TIDDIT V2.0 (9), as well as Manta (10) and single nucleotide variant (SNV) detection using GATK HaplotypeCaller (11). Subsequently, filtering was performed using the list of targetable genes described in the INFORM study (12). This analysis resulted in 260 SNVs and 18 SVs. Further filtering of the SNVs based on gnomAD or local observations as well as annotations in ClinVar narrowed the list to 11 SNVs that were manually inspected and dismissed, as no targetable SNV was identified. Similarly, the 11 SVs were manually curated based on recurrency in the local database and inspection in Integrated Genomic Viewer (IGV) (13) leaving only three SVs that interestingly affected regions recurrently involved in ALL.

Two of these SVs affected the JAK2 locus on 9p24.1 and inspection in IGV revealed a shared breakpoint suggesting that both SVs had occurred in the same genomic rearrangement (Supplementary Figure 1). The first SV consisted in a 122 kb-long intrachromosomal inversion with breakpoints that mapped to intron 15 (NM_001322194.2) and telomeric to the JAK2 locus. The second SV consisted in an interchromosomal event and shared the breakpoint in intron 15 of JAK2, while the second breakpoint was located at 14q32.11 and mapped to intron 25 in the CCDC88C locus (NM_001080414). Both events are visualized in the Circos plot in Figure 1A. The rearrangement of JAK2 could be verified with metaphase FISH that confirmed that the 5′ signal at the JAK2 locus was translocated to chromosome 14 while the 3′ signal was retained on derivative chromosome 9 (Figure 1B). As the rearrangement involves the most distal regions of chromosomal arms 9p and 14q, it is beyond the resolution of chromosome banding analysis and cannot be detected by karyotype analysis. In addition, SV analysis also detected a deletion on 9p21.3 encompassing 39 kb and supported by 95% of reads that resulted in a biallelic loss of CDKN2A/B that had been previously found in SoC testing.

Taken together the data indicated a reciprocal translocation t(9;14)(p24.1;q32.11) with a concomitant inversion of a 122-kb-long fragment from 9p that resulted in the creation of a fusion gene joining intron 25 of CCDC88C to intron 15 of JAK2 on derivative chromosome 9. As a result of the inversion from the 9p fragment, the genes are arranged in opposite orientations and thus no reciprocal fusion is created at derivative 14 (Figure 2).

According to the WGS findings, the rearrangement juxtaposes exons 1 to 25 of CCDC88C to exons 16 through 25 of JAK2, creating a novel fusion gene (Figure 2). The expression of the fusion transcript joining exon 25 of CCDC88C (3′ partner) to exon 16 in JAK2 gene (5′ partner) was confirmed with RT-PCR followed by Sanger sequencing. The fusion gene will thus result in an in-frame hybrid protein with the kinase domain of JAK2 at the carboxy-terminal.

JAK2 encodes a nonreceptor tyrosine kinase involved in cell proliferation and differentiation through the JAK/STAT signaling pathway. Aberrant JAK2 is found in several cancers including both myeloid and lymphoid malignancies (14). Fusions involving JAK2 were found in Philadelphia-like ALL (15), a subset of high-risk B-ALL characterized by a distinctive gene expression profile resembling Philadelphia-positive ALL, but lacking the canonical BCR::ABL fusion (16, 17). A fraction of Philadelphia-like ALL harbors JAK2 fusions where JAK2 is the 5′ partner and all the fusions described so far include exons 19–25 and retain the kinase domain that is constitutively activated in the fusions (18). To this date, over 30 genes have been found as potential 3′ partners to JAK2, but this is the first report of a CCDC88C::JAK2 fusion. JAK2 inhibitors have shown antileukemic effect in ex vivo experiments in cells carrying JAK2 fusions (19, 20), that include the kinase domain of JAK2; however, the experience in vivo is limited (21). JAK2 fusions have only been described in isolated cases of T-ALL (18), although activating point mutations in members of the JAK/STAT pathway has been reported in up to 25% of patients with T-ALL (22). Patients carrying genomic alterations that activate the JAK/STAT pathway either through gain of function mutations or gene fusion that include the JAK2 kinase domain are potential candidates for targeted therapy (23) (24). Pediatric patients with ALL and genomic features consistent with JAK/STAT pathway activation are currently recruited in ongoing trials (clinicaltrials.gov NCT03117751 and NCT02723994) for targeted treatment with ruxolitinib (selective JAK1/2 inhibitor) in addition to chemotherapy. Since our patient responded to nelarabine, he was never considered for therapy with JAK inhibitors nor were cells available to perform in vitro sensitivity tests.

CCDC88C is a ubiquitous protein, involved in the Wnt signaling pathway, with multiple functions that include determination of cell polarity, development of the nervous system, and tumor suppressor activity (25). CCDC88C protein is highly expressed in B- and T-cell pediatric leukemias (https://pecan.stjude.cloud/proteinpaint/CCDC88C). CCDC88C has been reported a handful occasions in hematological malignancies as fusion partner with FLT3 (26) as well as PDGFRB (27) including a case report of pediatric patient with a myeloproliferative disorder (28). It has also been reported as the 5′ fusion partner to PDGFRB in a young adult with Philadelphia-like ALL that responded to imatinib (29).

The genetic landscape of T-ALL is widely heterogeneous, and despite several recurrent aberrations, to date no single lesion is used to risk-stratify T-ALL patients. However, certain aberrations such as BCR-ABL and other ABL-class fusions or KMT2A rearrangements, although rare, contribute important information to individualized patient management.

Since our patient responded to nelarabine, he could be bridged to transplantation and experimental therapy with JAK2 inhibitors was not considered. However, this report illustrates the power of WGS in the diagnostic setting of ALL and describes a novel CCDC88C::JAK2 fusion in T-ALL, adding CCDC88C to the list of potential JAK2 partners. While the validation of WGS in the diagnostic setting of acute leukemias is still ongoing in our laboratory, the present report serves as a real-life example of the power of WGS in the diagnostic setting of acute leukemia, as it enables timely recognition of potential therapeutic targets in high-risk pediatric ALL.