Clinico-biological features of T-cell acute lymphoblastic leukemia with fusion proteins

T-cell acute lymphoblastic leukemias (T-ALL) represent 15% of pediatric and 25% of adult ALL. Since they have a particularly poor outcome in relapsed/refractory cases, identifying prognosis factors at diagnosis is crucial to adapting treatment for high-risk patients. Unlike acute myeloid leukemia and BCP ALL, chromosomal rearrangements leading to chimeric fusion-proteins with strong prognosis impact are sparsely reported in T-ALL. To address this issue an RT-MPLA assay was applied to a consecutive series of 522 adult and pediatric T-ALLs and identified a fusion transcript in 20% of cases. PICALM-MLLT10 (4%, n = 23), NUP214-ABL1 (3%, n = 19) and SET-NUP214 (3%, n = 18) were the most frequent. The clinico-biological characteristics linked to fusion transcripts in a subset of 235 patients (138 adults in the GRAALL2003/05 trials and 97 children from the FRALLE2000 trial) were analyzed to identify their prognosis impact. Patients with HOXA trans-deregulated T-ALLs with MLLT10, KMT2A and SET fusion transcripts (17%, 39/235) had a worse prognosis with a 5-year EFS of 35.7% vs 63.7% (HR = 1.63; p = 0.04) and a trend for a higher cumulative incidence of relapse (5-year CIR = 45.7% vs 25.2%, HR = 1.6; p = 0.11). Fusion transcripts status in T-ALL can be robustly identified by RT-MLPA, facilitating risk adapted treatment strategies for high-risk patients.

INTRODUCTION T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive hematological cancer arising from the transformation of T cell precursors arrested at specific stages of differentiation [1,2]. T-ALL represent 15% of pediatric and 25% of adult ALL. Despite modern poly-chemotherapy protocols, followed by allogeneic hematopoietic stem cell transplantation (HSCT) in high-risk cases, outcome of pediatric and adult patients with T-ALL remains poor, particularly in primary refractory and relapsed cases. After a transient initial response, about 20-30% of pediatric [3][4][5] and 40% of adult T-ALL patients relapse [6] with a 5-years overall survival (OS) below 20%. Understanding the mechanisms leading to treatment failure is therefore crucial to identify patients at high risk and adapt treatment in order to improve long term prognosis. Cytogenetic and global transcriptomic analyses led to the classification of T-ALL into molecular subgroups characterized by the abnormal expression of specific transcription factors (TF) (TAL1; LMO1/2; TLX1/3; LYL1; HOXA; MEF2C) and their blocked differentiation at specific stages of maturation [1,7,8]. Aberrant TF expression can be due to their juxtaposition to T Cell Receptor (TCR) loci or to somatic mutations in their enhancer regions. A number of additional, recurrent genetic abnormalities are found, including the loss of major tumor suppressive pathways (e.g., inactivating mutations of PTEN and of the CDKN2A tumor suppressor locus) and activation of oncogenic pathways (e.g., activating mutations in NOTCH1/FBXW7, IL7R/JAK pathway, epigenetic regulators, cell cycle, PI3K, and RAS signaling) [9][10][11][12]. Acquisition and selection of these molecular alterations entails the complex clonal evolution at the cellular level that occurs during T-ALL progression. Unlike acute myeloid leukemia (AML) and B-Cell Precursor acute lymphoblastic leukemia (BCP-ALL), which harbor numerous translocations leading to chimeric protein fusions [13,14] such abnormalities are more rarely reported in large series of T-ALLs [15][16][17][18][19][20]. Recurrent chimeric protein fusions in T-ALL include rearrangements of KMT2A (AFDN (AF6), MLLT1,  ELL), SET-NUP214, ABL1 (NUP214-ABL1, BCR-ABL1), MLLT10 (PICALM,  DDX3X, NAP1L1, XPO1), and the ETS family (SPI and ETV6). Given their individual low frequency, the clinico-biological features of T-ALLs harboring chimeric fusions within a comprehensive series remains elusive.
To address this issue, we designed and developed an RT-MPLA assay allowing identification of the majority of known fusion transcripts leading to chimeric proteins in T-ALLs. Applying this panel to a comprehensive, consecutive series of 522 adult and pediatric T-ALLs, we here report an unexpected overall incidence of 20% of fusion transcript. Their mutational landscape, associated clinico-biological features and prognostic impact on patients enrolled in the French GRAALL protocol for adult patients and the FRALLE 2000T protocol for pediatric patients are described.

METHODS T-ALL samples and clinical trials
Diagnostic peripheral blood (PB) or bone marrow (BM) samples from a consecutive series of 522 T-ALL patients, with RNA available, 206 children and 316 adults, were screened for fusion transcripts (fig. S1). Sample collection and analyses were obtained with informed consent in accordance with the Declaration of Helsinki with approval from the institutional review boards of institutions that participated in this study. Diagnosis of T-ALL was based on the World Health Organization 2008 criteria revised in 2016, defined by expression of cytoplasmic and/or surface CD3, and negativity of CD19 and MPO [21].
Definitions: Corticosensitivity was defined as circulating blasts <1 G/L on day 8. Complete remission (CR) was defined as: absence of physical signs of leukemia, BM with active hematopoiesis and <5% leukemic blast cells (identified morphologically), and normal cerebrospinal fluid.

Biological analysis of T-ALL samples
Diagnostic PB or BM T-ALL samples were analyzed for immunophenotype and classified as follows: Immature T-ALL are TCR-and cTCRβ-. αβ-lineage T-ALL are TCRαβ or preαβ (TCR-) but with cTCRβ + . γδ-lineage T-ALL are TCRγδ, SIL-TAL1 and PICALM-MLLT10 detection was performed by RT-PCR as previously described [2,22]). Oncogenic transcripts TLX1 and TLX3 quantification was performed by RQ-PCR (primer and probe sequences as previously described [23,24]). Quantification of HOXA9 expression was performed by qRT-PCR as previously described [25]. All methods are detailed in supplementary methods.
Diagnostic available DNA samples from 273 cases were also analyzed using an 80-gene pan-exon next-generation sequencing capture-panel (details included in supplementary methods and Table S1).

RT-MLPA
Using previously published methods [26], we created a mix of 268 RT-MLPA probes to target 106 different genes (Table S2; Fig. S2). The procedure is detailed in supplementary data. All 141 5′ probes have a GTGCCAGCAAGATCCAATCTAGA tail at their 5′ ends and all 127 3′ probes a TCCAACCCTTAGGGAACCC tail at their 3′ ends to allow for final PCR amplification.

Data analysis
Fisher's exact and Mann-Whitney tests were used for clinical, biological and mutational analyses. Overall survival (OS) was calculated from the date of pre-phase initiation to the last follow-up date. The cumulative incidence of relapse (CIR) was calculated from CR to relapse date, censoring patients alive without relapse at the last follow-up date. The Event-Free-Survival was calculated from diagnosis date to first event among: induction failure, first hematologic relapse and death from any cause in first CR. Relapse and death in CR were considered as competitive events. Univariate analyses assessing the impact of categorical and continuous variables were performed with a Cox model. Statistical analyses were performed with STATA software (STATA 12.0 Corporation, College Station, TX) for survival and R software for others (v4.0.2). All p values were two-sided, with p < 0.05 denoting statistical significance.
The MLLT10 cases were more frequently of TCRγ/δ lineage (p < 0.05) and HOXA9 deregulated (Table 1). They were less likely to have NOTCH1 signaling mutations (p = 0.03) and cell cycle mutation/deletion (p < 0.001) and showed an higher frequency of RAS, IL7R/JAK/STAT and epigenetic regulator mutations compared to negative samples ( SET-NUP214 gene fusions were identified in 18 cases, with uniform breakpoint positions. KMT2A 3′ partners were, by decreasing incidence, AFDN (n = 10), MLLT1 (n = 4) and ELL (n = 1). As for the MLLT10 group, and according to previous reports [20], significantly more of these T-ALL were from the γδ lineage (p < 0.05) and overexpressed HOXA transcripts (Table 1). They were significantly associated with less cell cycle genes mutations/ deletion (p < 0.001) and fewer NOTCH1 signaling genes mutations (53%, p = 0.002) compared to the negative group (Table 2; Fig. 2).

Clinico-biological characteristics of fusion transcripts in GRAALL and FRALLE treated T-ALL
We then investigated the clinico-biological characteristics linked to fusion transcripts in a subset of 235 patients, including 138 adults enrolled in the GRAALL-2003/05 trials and 97 children enrolled in the FRALLE-2000T trial. A fusion transcript was observed in 27% of adults and 22% of children and the overall incidence of fusion transcripts in these cohorts was 25% (58/235).
The incidence of HOXA trans-deregulated T-ALL with identified fusion transcripts was 17% (39/235) distributed as follows: ETS-family (n = 2) and NUP98 (n = 2) fusions were only identified in pediatric patients. Despite high MRD at the end of induction (>10 −4 ) for both patients with NUP98 rearrangements, only 1 patient required allogeneic HSCT. In contrast to previous report [17], all four patients had favorable outcome and remained in CR with a median follow up of 4.5 years (R, 3.4-6.7).  αβ-lineage T-ALL are TCRαβ or preαβ (TCR-) but with cTCRβ+ .

DISCUSSION
Last decade transcriptomic and genomic studies identified biological subgroups of T-ALL and uncovered major oncogenic and tumor suppressor pathways [1]. This molecular characterization provided a strong rationale for targeted therapies in T-ALL, such as drugs directed against JAK, NOTCH1, BCL-2 or PI3-AKT signaling pathways. However, contrary to BCP -ALL or AML, prognostic biomarkers identified in large prospective studies are yet lacking or debated to stratify patients at first line and adapt treatment.
In 522 consecutive adult and pediatric T-ALL RT-MLPA identified an unexpected 20% incidence of chimeric fusion protein. Fusion  transcripts were correlated with their respective immunophenotypic, transcriptional and mutational landscapes, resulting in an unprecedented global overview.
Rearrangements involving KMT2A (11q23) [41] (previously MLL) are also recurrent in T-ALL with reported occurrence of 5-8% [7,42]. Consistent with previous reports, the most frequent 3′ partners in our T-ALL cohort were: AFDN (6q27) [43] (previously AF6) and MLLT1 (19p13, previously ENL) [44]. Of note, none of our 34 cases with KMT2A fusion demonstrated KMT2A-AFF1 (previously MLL-AF4), or KMT2A-MLLT3 (previously MLL-AF9) fusions, commonly observed in BCP-ALL and AML respectively. KMT2A rearranged T-ALLs over-express HOXA and are arrested in an early differentiation step after commitment to the γδ-lineage [1,45]. The prognosis of KMT2A-rearrangements in T-ALL was unclear. Contrary to BCP-ALL, KMT2A-MLLT1 in T-ALL has been reported to be favorable in children [46]. In keeping with this, in our series, patients with KMT2A rearrangements do not demonstrate worse outcome neither in terms of initial response to treatment neither in terms of relapse.
SET-NUP214 (previously TAF1-CAN) is a recurrent chimeric protein found in 3-10% of T-ALLs [18,47]. Most SET-NUP214 chimeric proteins result from a cryptic deletion at chromosomic 9q34, leading to loss of one ABL1 allele [48]. SET-NUP214 induces HOXA deregulation by interacting with XPO1, DOT1L and the HOXA promoter [18,49,50]. As confirmed here, this recombination occurs in T-ALL from the γδ-lineage that are associated with a poor outcome [20]. Patients with SET-NUP214 chimeric protein have significantly poor initial treatment response and a trend for a higher risk of relapse without significant prognostic impact on EFS or OS. Thus, the need of intensifying treatment protocol in these patients is still matter of debate.
The ETS family of proteins are TF containing an ETS DNAbinding domain [59] which are detected in pediatric T-ALL, fused  in-frame to different 3′ or 5′ partners. SPI1 (11p11.2 encoding PU.1) is fused in the 3′ position with STMN1 (1p36) or TCF7 (5q31) and ETV6 (12p13) is fused in 5′ with NCOA2 (8q13). SPI1 fusions transcripts have been described in 4% of one pediatric T-ALL series, when they demonstrated a distinct gene expression profile and a poor outcome [17]. In contrast to this, all 4 FRALLE 2000T treated patients in our series had a favorable outcome.
RT-MLPA has the capacity to easily detect many rearrangements and its fast turnaround time and superior exhaustivity and reduced cDNA consumption could make it an interesting alternative to RT-PCR. Such a screening strategy will also identify patients with targetable rearrangement such as ABL1 [57] (with tyrosine kinase inhibitor) or KMT2A (with Dot1l or menin inhibitor) [40,60] or who have poor prognosis and require treatment adaptation. However, RT-MLPA only allows detection of known fusion transcripts. New fusion transcripts detected by T-ALL RNA sequencing can easily be added to the RT-MLPA, thus providing a complement and extension to RNA-Seq approaches. In addition, the excellent specificity of this method was already demonstrated on a previous cohort of 540 patients with acute leukemia, confirming virtually all (98%) fusion transcripts detected by RT-PCR and Sanger sequencing [26]. As such, it is well adapted to routine clinical screening in Acute Leukemia. Table 3. Clinical characteristics and treatment response in GRAALL and FRALLE treated patients according to fusion transcript status.   T. Steimlé et al.