Introduction

Acute Lymphoblastic Leukaemia (ALL) is a genomically complex disease that is aggressive and frequently fatal in some cohorts [1, 2]. ALL can affect people of any age, however, around 80% of cases occur in pediatric patients [2]. While ALL is considered to be a single disease it can be subclassified into T- and B-cell ALL [1]; T-ALL contributes to 15–20% of ALL cases depending on age group [2]. Genomic subclassification of B-ALL is more established as compared with T-ALL, with several recently defined subtypes now incorporated into World Health Organization (WHO) and International Consensus Classification (ICC) systems for acute leukemia [3]. The characterization and clinical significance of these entities have been reviewed elsewhere [4]. In contrast, the subclassification of T-ALL has not yet been widely adopted into the clinic. Reflecting this uncertainty, the WHO only recognizes Early T-cell Precursor-ALL (ETP-ALL) as a distinct subtype [3]. The ICC additionally recognizes BCL11B-activated ETP-ALL, and not otherwise specified (NOS) T-ALL along with 8 provisional entities, several of which are discussed in this review (TAL1/2-rearranged (TAL1/2r), TLX1r, TLX3r, HOXA, LMO1/2r, NKX2r, SPI1r, basic helix loop helix) [5]. Notwithstanding formal T-ALL subclassification systems, there is accumulating data to enhance a deeper understanding of the biology and genomics of T-ALL.

Patients with T-ALL usually have one or two driver genomic alterations capable of causing leukemic transformation, however, unique combinations of co-occurring mutations continue to be identified [6]. Despite the identification of recurrent potentially clinically targetable alterations including NOTCH1 mutations [7], KMT2A rearrangements (KMT2Ar) [8], and the NUP214::ABL1 gene fusion [9], T-ALL patients are routinely treated with intensive chemotherapy regardless of genotype [10]. Where possible, stem cell transplant is routinely utilized in adult patients [10] however, transplants are less frequent in pediatric patients [11]. Considerable research into alternative treatment options has occurred in the B-ALL sphere such as Bispecific T-cell engagers (BiTEs) [12] and chimeric antigen receptor (CAR-T) cells [13] however, innovative therapies for treatment resistant T-ALL patients are currently lacking and those who relapse or are treatment refractory will likely die of their disease.

Traditional, widely utilized methods of subclassification such as immunophenotyping and cytogenetics have yielded several clinically significant subtypes. However, through various next-generation genomic approaches, an increasing number of genomic subtypes of T-ALL have been identified that undoubtedly influence patient response to treatment [14,15,16,17]. These subtypes carry inherent risk and drug sensitivities, however, incorporating the knowledge gained through next-generation sequencing into clinical care presents a major challenge. Also, the systematic identification of druggable targets and effective rational therapies require robust pre-clinical evidence of underlying biology and function, drug efficacy and a comprehensive investigation of other factors that influence response and resistance, including the interplay of novel fusion genes with co-occurring mutations. For example, the efficacy of γ-secretase inhibitors and anti-Notch monoclonal antibodies for NOTCH1 overexpressing cancers have been assessed in vitro and in vivo with some success [18] but severe gastrointestinal side effects and lack of efficacy in NOTCH-mutant breast cancers raised concerns about their clinical utility [19,20,21]. The chemotherapeutic nelarabine is approved for relapsed/refractory T-ALL patients in many countries, and additionally for newly diagnosed T-ALL patients in several others, but beyond this, limited treatment options exist [22]. Despite the variety of genomic drivers of disease offering potentially targetable therapeutic pathways, no such treatments are routinely used for T-ALL and represent a clear unmet need in the clinical setting.

This review covers the diagnostic journey of a T-ALL patient, discussing subtype identification, clinical outcomes and interrogating potential novel treatments. A patient who is feeling unwell may have a blood test ordered by their physician which, if leukaemia is the causative disease, will return an abnormal white blood cell count [23]. Following the initial test, a full blood count including leukocyte differential is ordered to determine blast cell percentage, this may also include morphological analysis of bone marrow [23].

ALL cannot be determined from morphology alone and further diagnostic testing such as flow cytometry is required to define B- or T-cell lineage ALL [23], and karyotyping and fluorescence in situ hybridization (FISH) to assess gross chromosomal abnormalities and specific chromosomal translocations, duplications, or deletions (e.g.: ABL-class, KMT2A-rearrangements). Specialized laboratories also conduct investigations not included in routine testing such as next generation sequencing analyses (transcriptomic, whole exome, and whole genome) and multiplex ligation-dependent probe amplification to detect cytogenetically cryptic genomic alterations and copy number variations in select genes relevant to the leukemic subtype [23]. Additional emerging techniques such as optical genome mapping, capable of detecting large scale structural variations, may prove useful as diagnostic tools in the future [24]. Furthermore, mRNA-sequencing provides insight on driving and co-occurring lesions (discussed later), specifically those amenable to targeted therapies with demonstrated efficacy in other diseases (e.g. FLT3 in Acute Myeloid Leukemia) [25].

We have identified four major T-ALL genomic subtype defining publications [14,15,16,17]. Though there is broad agreement on subtypes such as MLLT10r and KMT2Ar, several genomic subtypes differ between these four papers due to variations in cohorts and methods of classification, summarized in Table 1. One of the first descriptions of the T-ALL genomic landscape came from the North American Children’s Oncology Group (264 children and young adults aged 1-29). Liu et al. utilized whole exome sequencing (WES) and transcriptomic sequencing to define 8 patient subtypes [15]. Two separate subsequent studies focussed on gene fusions and other structural alterations as the main subtype-defining lesions. The first, by Dai et al., was a large international cohort, (707 patients; 510 pediatric aged 1-17, 190 AYA/adults aged 18-75, 7 unknown) using mutational profiles and dysregulated gene expression signatures of leukemogenic factors to define 10 subtypes [14]. These findings were recapitulated by Müller et al. (131 patients; 6 pediatric, 62 adolescent and young adults (AYA), and 63 adults) using diagnostically available techniques including karyotyping and FISH [16]. The latest Children’s Oncology Group study by Pölönen et al. (1309 patients, including those reported in Liu et al. [15]; pediatric and AYA patients up to 30 years) grouped cases into 15 genomic subtypes based on whole genome (WGS), transcriptome (WTS), and WES analyses; the assay for transposase-accessible chromatin with sequencing (ATAC-seq) and sequencing chromosome conformation capture and interaction immunoprecipitation (HiChIP) were utilized for non-coding mutations and structural variations [17]. Details of the patient cohorts are described in Table 2.

Table 1 Genomic subtypes listed in T-ALL genomic landscape publications.
Full size table
Table 2 Number of patients in each of the genomic subtypes reported in T-ALL genomic landscape publications.
Full size table

In addition to the variations in reported subtypes due to analysis techniques employed, different genomic subtypes occur at varying frequencies in different age groups, making the correlation of outcome data and leukemic driver alterations difficult. Here, we draw upon these four key articles and synthesize the genomic landscape of T-ALL to unravel this complexity. We describe the most common T-ALL genomic alterations identified in two or more of the genomic landscape papers as well as several additional high-risk alterations and gene fusions. Finally, we discuss potential targeted therapies, speculate on their clinical use and interrogate survival outcomes and potential targeted treatments.

The old school: subtype classification based on immunophenotype and cytogenetics

Immunophenotyping identifies Early T-cell Precursor ALL (ETP-ALL)

Flow cytometry is an extremely rapid and effective method of diagnosis. T-ALL is usually identified using a combination of markers, though cytoplasmic CD3 (cCD3) is almost always present. The European Group for the Immunologic Classification of Leukemia previously described 4 of the most common patterns of expression: pro-T (cCD3+, sCD3−, CD1a−, CD2+, CD5−, CD7+, CD34−), pre-T/immature (cCD3+, sCD3−, CD1a−, CD2+, CD5+, CD7+, CD34−), cortical T (cCD3+, sCD3+/−, CD1a+, CD2+, CD5+, CD7+, CD34−), and mature-T (cCD3+, sCD3+, CD1a− CD2+, CD5+, CD7+, CD34−) [26]. Most importantly, flow cytometry can identify ETP-ALL. ETP-ALL is defined by an immunophenotype of sCD3-, CD7+, CD1a-, CD8-, CD5-/dim ( < 75%), and positive for one or more myeloid (CD11b, CD13, CD33, CD117) or stem cell (CD34, HLA-DR) antigens; expression of cytoplasmic CD3 is also often observed [3, 27, 28]. Near ETP-ALL cases fulfil the immunophenotypic requirements for ETP-ALL but have CD5 positivity [29]. Reported outcomes of ETP-ALL are varied. Early studies demonstrated poor overall and event free survival in pediatric [27, 30] and adult [28, 31] patients. However, up-front risk stratification to more intensified chemotherapy regimens has improved outcomes, particularly in pediatric patients [29, 32,33,34,35,36,37,38] highlighting the importance of early identification of ETP-ALL status. Interestingly, the type of events differ for ETP-ALL patients compared with non-ETP-ALL patients, with the former more likely to experience refractory disease while non-ETP-ALL patients are more likely to relapse [17]. Two studies have noted that immunophenotypically-confirmed ETP-ALL cases were enriched in the LYL1/LMO2 subtype compared with all other subtypes (37% [15] and 44% [14]). ETP-ALL cases also demonstrated alterations in HOXA13 and MED12 with ETV6 or ZFP36L2 gene fusions (42% of ETP-ALL cases, 12% of all cases) or HOXA9/10/11 deregulation driven by NUP98/NUP214/KMT2A/MLLT10 gene fusions (30% of ETP-ALL cases, 7% of all cases) [17]. BCL11B driven T-ALL has been reported as almost exclusively in ETP-ALL cases [17]. It should be noted that alterations to several of these driver genes (HOXA13, MED12, ETV6, ZFP36L2) have also been observed in near-ETP and non-ETP cases. These observations resulted in delineation of the ETP-like subtype enriched for cases with ETP-ALL and diverse genomic driver alterations to genes involved in HSC development [17]. The ETP-like subtype is discussed in greater detail throughout this review.

Overall, these data indicate ETP-ALL can be effectively managed by contemporary treatments in all age cohorts. However, additional factors, such as remission status [38], may influence ETP-ALL patient outcome and requires further investigation.

Cytogenetics: Karyotype and FISH-detectable rearrangements

Cytogenetics is used by diagnostic laboratories to detect chromosomal abnormalities in ALL patient cells [23]. A detailed description of the more common cytogenetic abnormalities has already been described [39]. As previously mentioned, cytogenetic karyotyping employs G-banding to detect gross chromosomal abnormalities while FISH detects deletions and chromosomal translocations e.g. CDKN2A/B deletions and KMT2A or ABL1 rearrangements [23]. Subtype defining chromosomal rearrangements, some detectable by FISH/cytogenetic testing, are reported in three of the four genomic landscape papers and involve rearrangement of KMT2A [14, 17], MLLT10 [14, 16, 17], SPI1 [14, 17], NUP214/98 [14, 16], and ABL1 [16, 17], though, in T-ALL, only KMT2A, ABL1, and MYB rearrangements are routinely probed via FISH [23].

Secondary co-lesions identifiable by cytogenetics

ABL-class rearrangements

Aside from BCR::ABL1, ABL1 rearranged (ABL1r) T-ALL most typically arises from the NUP214::ABL1 gene fusion. NUP214 encodes a component of the nuclear pore complex responsible for transporting molecules between the nucleus and cytoplasm [40]. Cases of NUP214::ABL1 are variably reported constituting ~1-6% of T-ALL cases depending on age group, frequently with a cortical immunophenotype and co-occur with dysregulation of TLX1 and TLX3 [9, 14, 16, 17, 41, 42]. Phospho-flow cytometry may be used to screen for ABL1r and other constitutively active kinases and to generate preliminary data on drug sensitivity. The addition of tyrosine kinase inhibitors (TKIs) such as dasatinib to intensive chemotherapy regimens in children with newly diagnosed ABL1r B-ALL has resulted in markedly improved outcomes compared with chemotherapy alone [43]. However, as the use of TKIs in ABL1r T-ALL patients has not been as extensively trialed and the benefit is not as clear, TKIs are not currently included in standard therapeutic regimens in T-ALL even when an ABL-class rearrangement is present. Asciminib, an allosteric inhibitor used to treat BCR::ABL1+ leukemias [44], has demonstrated in vitro [45] and in vivo [46] efficacy for NUP214::ABL1 ALL as long as the ABL1 SH3 domain is present, supporting use of asciminib in NUP214::ABL1 ALL [45, 46].

CDKN2A/B copy number variations

Another common alteration observed by cytogenetics involves chromosome 9, indicative of CDKN2A/B deletions. The CDKN2A/B genes encode cyclin dependent kinase inhibitors that provoke G1 and G2 cell-cycle arrest by blocking degradation of p53, suppressing oncogenesis and inducing apoptosis [47]. CDKN2A/B deletions frequently co-occur and are present in 39-78% of all T-ALL cases [15, 48,49,50,51,52]. Functional inactivation of CDKN2A via deletion of tumor suppressor regions are reported in >70% of CDKN2Aalt patients [53, 54]; whole gene deletion is reported in as many as 42% of CDKN2Aalt patients [55]. Unlike CDKN2A, CDKN2B inactivation occurs via methylation (51%) and deletion (37%) however, no significant difference in outcome has been reported for the two modes of CDKN2B inactivation [55]. Prognosis of patients with CDKN2A/B deletions in T-ALL are variable and likely depend on co-occurring driver alterations. CDKN2A/B alterations conferred slightly inferior prognosis in one study (5-year OS ~ 95% CDKN2A/Bdel (n = 207) vs. 100% non-CDKN2A/Bdel (n = 57; p = 0.0466 NCT00408005), though overall outcomes remained excellent [15]. Similarly, an adult T-ALL patient series reported a 2-year OS of 19% in CDKN2Adel cases (n = 23), vs 47% in non-CDKN2Adel patients (n = 78; p = 0.032 treated with either the CALLG2008 or MDACC Hyper-CVAD program) [56].

CDKN2A/B alterations promote cell cycle progression by phosphorylation of retinoblastoma protein, allowing progression from G1 to S phase [57]. Kinases CDK4/6 are involved in the transition of G1 to S phase and are attractive targets for inhibition [57]. The CDK4/6 specific inhibitors palbociclib (breast cancer phase III trial NCT01942135), ribociclib (melanoma phase II trial NCT01719380), and abemaciclib (breast cancer phase Ib trial NCT02057133) all potently inhibit CDK4/6 [57] and are potentially useful in acute leukemias with CDKN2A/B alterations [58].

MYB

MYB is a transcription factor essential for T-cell lineage commitment. In T-ALL, alterations to MYB constitute 19% of pediatric/adolescent young adult patients, most commonly presenting as amplification (57%), mutation (22%), and rearrangement (19%) [15]. An early report defined the TCRB::MYB fusion as a T-ALL subtype enriched in children aged ~2 years, though specific outcome data were not specified [15, 59]. Conversely, while more recent reports have also identified MYB rearrangements [14,15,16,17, 60], only one study classified TCRB::MYB as a subtype-defining event [16]. Patients with TCRB::MYB fusions frequently harbored co-occurring CDKN2A deletions and NOTCH1 mutations [15, 59]. MYB alteration did not significantly affect 5-year OS (94% MYBalt n = 49 vs. 96% MYBwildtype n = 215; p = 0.12 NCT00408005) [15].

Driver lesions identifiable by cytogenetics

KMT2A-rearranged (KMT2Ar) disease

KMT2Ar disease is an established subtype of T-ALL and is the only defined subtype identifiable by cytogenetics [14, 17]. It is estimated that 4-8% of T-ALL patients harbor a KMT2A gene fusion and given KMT2A fusions are detectible via cytogenetics, a subtype can be readily defined for 1:12-1:25 patients [8, 42, 61]. KMT2A plays an essential role in regulating gene expression, notably of MEIS1 and the HOXA gene cluster, which have roles in hematopoietic cell proliferation [62,63,64]. KMT2A rearrangements involve a variety of fusion partner genes though the most common observed in T-ALL include epigenetic regulators MLLT1 and MLLT4, and less commonly, MLLT3 and MLLT10 [65,66,67]. KMT2Ar T-ALL transformation reportedly occurs late in T-cell development and traditionally confers a mature immunophenotype [14]. However, Pölönen et al. characterized a subset of KMT2Ar disease within the ETP-like subtype. In ETP-like KMT2Ar ALL, cases demonstrate maturational stages from HSPC up to pre-T differentiation. Additionally, ETP-like KMT2Ar ALL typically harbors KMT2A::MLLT4 fusions whereas non-ETP-like KMT2Ar cases exclusively contain KMT2A::MLLT1 fusions [17]. Furthermore, ETP-like KMT2Ar disease displayed a higher frequency of alterations to ETV6, GATA3, IKZF1, RUNX1 or RAS signaling and a decreased number of alterations within the NOTCH signaling pathway [17].

ALL driven by KMT2A fusions is considered high-risk with dismal outcomes. In general, KMT2Ar patients have a 6.7 fold worse overall survival compared with age-matched non-KMT2Ar T-ALL patients [68]. Deeper interrogation of the ETP-like subtype revealed ETP-like KMT2Ar ALL patients have adverse outcomes whereas non-ETP-like KMT2Ar patients have higher MRD but more favorable outcomes, demonstrating heterogeneity of outcome even within the KMT2Ar subtype [17]. Broadly, KMT2Ar ALL constitutes 15% of adult cases [69] though current relevant outcome data is limited, especially in a T-ALL context. In all age cohorts remission is usually achieved though rapid relapse often occurs which has been postulated to result from a positive selection environment caused by chemotherapy treatment [66]. The introduction of menin inhibitors may improve KMT2Ar patient outcomes; KMT2A interacts with menin after rearrangement, demonstrating menin’s utility as a pharmacological target [70]. Revumenib selectively targets menin-KMT2A interactions and phase I clinical data from relapsed/refractory patients aged 1-79 demonstrated an overall response rate of 53% (overall response encompassed complete remission inclusive of incomplete hematologic recovery, incomplete platelet recovery, and morphologic leukaemia-free state) (NCT04065399) [70].

The new school: subtype classification based on next-generation sequencing (NGS)

NGS, incorporating WTS, WES, and WGS, is not currently part of routine diagnostic testing for many patients and is predominantly a research endeavor, though this landscape is changing rapidly and targeted NGS gene panels are often used in a diagnostic setting in some parts of the world [71, 72]. Recent studies have emphasized the importance of genomic data for risk stratification in pediatric and AYA cohorts [73,74,75]. Furthermore, an NGS-based classifier was recently described for T-ALL patients that is independently prognostic in both adult and pediatric cohorts, demonstrating the incorporation of NGS into diagnostic risk stratification [76]. NGS can be subdivided into WTS, WES, and WGS [77]. WTS is the most frequently utilized sequencing platform and provides clinically relevant data [78]. In short, WTS reads transcribed messenger RNA, WES analyzes the portion of the genome that is read to create proteins (exons), and WGS sequences both exons and introns where features such as gene promoters are located [77]. WGS may prove particularly important in the context of T-ALL. As Pölönen et al. report, a high frequency of non-coding alterations occur in several subtypes, highlighting the potential of WGS in the future of T-ALL diagnosis and treatment stratification [17]. Previously, studies without WGS used transcriptomic data to define the driver alteration potentially leading to misclassification (e.g. increased expression of LMO2 can result from BCL11B enhancer hijacking resulting in the rare LMO2 γδ-like subtype (Table 1)). Of the genomic landscape papers discussed here, only Pölönen et al. and Müller et al. utilize WGS (Table 1) [16, 17]. We unravel subtypes defined by next generation sequencing and comment on their relevance to the diagnosis and treatment of T-ALL. Briefly, alterations to NOTCH1 and FBXW7 are frequently present in T-ALL though outcomes of patients with these alterations are variably reported [7, 79,80,81]. Because NOTCH1/FBXW7 alterations are ubiquitous in T-ALL, and are not subtype-defining, this review does not discuss them further.

ETP-like ALL including NUP98r, SET::NUP214, and GATA3

The largest T-ALL cohort with associated genomic studies has identified a broader category of ALL, termed ETP-like disease, comprising ETP, near-ETP and non-ETP cases. Rather than immunophenotype alone, these subtypes are grouped by WTS expression profiling [17]. Patients with ETP-like disease harbor multiple recurrent driver alterations affecting hematopoietic stem cell (HSC) development, including activating juxtaposition of the HOXA13 gene to enhancer regions of BCL11B and TCR; deregulation by rearrangement of MLLT10, KMT2A, NUP214 and NUP98; MED12 loss of function alterations; ZFP36L2 rearrangements; and alterations to ETV6 [17].

The SET::NUP214 fusion is often associated with an ETP-like immunophenotype ( ~ 45%) [82] and unlike NUP214::ABL1 it is not associated with TLX1 or TLX3 dysregulation [14, 82]. NUP98r T-ALL most frequently results from the NUP98::RAP1GDS1 fusion however, other fusion partners such as CCDC28A, LNP1, and PSIP1 have been reported [15]; NUP98 fusions were also frequently identified in patients with ETP-ALL [14, 15, 83]. More recently, both SET::NUP214 and NUP98r T-ALL form distinct clusters within the ETP-like genomic subtype [17]. Generally, GATA3 expression is elevated in T-ALL and GATA3 mutations, most frequently to the DNA binding domain (91%), were a subtype-defining feature in one genomic landscape study [14]. Diffusion map visualization of the top 5% variance genes revealed GATA3 mutations result in an ETP-like genomic expression profile [14]. Further investigation of these cases also revealed some crossover with other distinct genomic subtypes, suggesting gene expression is determined by stage of cell-cycle arrest and oncogenic driver [17]. As the GATA3 subtype was only identified in one study in low numbers it is likely a co-occurring lesion rather than a driver and so grouping with other genomic alterations is of greater clinical relevance.

BCL11B alterations

BCL11B encodes a C2H2-type transcriptional repressor zinc finger protein. Alterations to BCL11B are related to T-ALL development [84] and recent discoveries indicate frequent noncoding translocations and enhancer hijacking events [17]. BCL11B alterations define a distinct genomic subtype in some T-ALL cohorts (Table 1) and up to 94% of cases with BCL11B activation express an ETP immunophenotype [16, 17, 60]. Patients frequently lack mutations to NOTCH1 (typically found in >50% of patients [85]) and PHF6, and lack CDKN2A deletions, but demonstrate high expression of KIT and LMO2 [16]. Additionally, the BCL11B-altered subtype exhibits decreased expression of RAG1 and RAG2 [16] and features activating FLT3 mutations [16, 17, 60], characteristics shared with ETP-ALL. Expression levels of BCL11B can be linked to clinical outcome in standard risk adult patients: BCL11Blow 5-year OS was 35% (n = 40 vs. 53% BCL11Bhigh n = 129; p = 0.02 NCT00199056, NCT0098991) [86]. No difference in pediatric patient clinical outcome has been noted (5-year OS 95% BCL11Bhigh n = 43 vs. 96 BCL11Bnorm n = 221; p = 0.28 NCT00408005) [15].

BCL11B altered ETP-ALL is associated with FLT3 alterations [14, 17, 87] indicating first or second generation FLT3 kinase inhibitors such as sorafenib or quizartinib as potential therapeutics in this subtype when a FLT3 alteration is present [25, 88]. Sorafenib has demonstrated in vitro efficacy in T-ALL cells at a clinically relevant dose [89]. Sorafenib treatment in post-allogeneic stem cell transplant FLT3-mutated acute myeloid leukaemia (AML) patients resulted in significantly improved 2-year OS (sorafenib treated 38% n = 30 vs. allogenic transplant alone 9% n = 30; p = 0.0001 NCT02997202) [90] suggesting a possible application for T-ALL patients with high FLT3 expression who have received SCT. Newly diagnosed AML patients positive for FLT3-internal-tandem-duplication treated with quizartinib experienced a significantly improved OS compared to placebo treatment (quizartinib treated 32% n = 268 vs. placebo 15% n = 271; p = 0.032 NCT02668653) [91]. In general, it remains to be investigated if ETP-like patients respond well to intensified upfront chemotherapy and whether targeted therapeutics could further improve outcomes, especially in the ETP-like subtypes with specific genomic alterations (Table 1).

SPI1-rearranged disease

The SPI1 gene encodes the PU.1 protein, an important transcription factor and master regulator of hematopoietic cell development [92, 93]. SPI1 rearrangements (SPI1r) result in aberrant expression of PU.1 [92, 93]. The SPI1r subtype expressed an immunophenotype that was similar to that reported for LYL1 dysregulation and included low expression of T-cell related markers (CD1a, CD2, CD3e, CD4, and CD8a) and high expression of the myeloid marker CD33, potentially indicating intensive upfront chemotherapy like that required for ETP-ALL is warranted [14]. Reported fusion partners include TCF7, STMN1 and YWHAE [14, 17, 94]. SPI1 alterations have been reported as high-risk with dismal outcomes in two pediatric cohorts [17, 94]. A study of 120 patients, of which 7 harbored SPI1 rearrangements, noted a poor 2-year OS of ~50% [94]. Though relatively rare, secondary malignant neoplasms occur more frequently in patients harboring an SPI1 alteration (-log10(p)=14.39), however, these were myeloid/dendritic cell neoplasms and were not associated with MRD [17].

Homeobox (HOX) gene dysregulation

HOX genes all perform similar functions and are involved in regulation of gene expression, morphogenesis, and differentiation [95]. HOXA dysregulation occurs through both coding and noncoding alterations and forms distinct subtypes in the reported datasets [14,15,16,17]. Dysregulation of HOXA genes can result from cis mechanisms (HOXA translocations often involving noncoding regions of the genome e.g. enhancer hijacking events) or trans mechanisms (fusion events involving other oncogenes resulting in HOXA cluster activation) [96]. HOXA dysregulation frequently leads to an immature immunophenotype, likely due to the critical role of HOXA gene expression in early development [17, 95]. Specific outcome data for each HOXA family member is limited aside from HOXA11 which is considered intermediate-risk in both pediatric (HOXA11alt 5-year OS of 89% n = 18 vs. 96% HOXA11norm n = 246; p = 0.29 NCT00408005) [15] and adult cohorts (76% HOXA11high n = 61 vs. 45% HOXA11norm n = 57 vs. 57% HOXA11negative n = 168; p = 0.012 GMALL 5/93 & 6/99) [97]. HOXA13 alterations in the context of ETP-like disease have recently been reported as high-risk in a pediatric/AYA cohort, exhibiting significantly worse EFS and MRD (-log10(p)=1.2 and 11.9 respectively) [17].

Dysregulation of TLX1 and TLX3, plus co-occurrence of PTPN2deletion

TLX1/TLX3 are paralogs encoding important nuclear transcription factors in embryogenesis, specifically splenogenesis and the development of some sensory neurons [98, 99]. Despite a lack of involvement in normal hematopoiesis, aberrant activation of these genes is a recurrent feature of T-ALL, frequently resulting from T-cell receptor (TCR) gene loci rearrangements [98, 100], other noncoding translocations, intergenic inversion and hijacking by enhancers [17]. While TLX1 and TLX3 genes are highly similar, TLX3 alterations often result in a near-ETP-ALL immunophenotype with alterations to PRC2 complex genes [101], whereas TLX1 alterations typically result in non-ETP-ALL, indicating variation in disease pathogenesis [14, 15]. Furthermore, non-coding enhancer hijacking events also differ between TLX1 and TLX3. TLX1 was found to hijack the LINC00592 enhancer or be activated due to TCRβ or TCRδ loci rearrangement [17]. Deregulation of TLX3 was due to hijacking of the BCL11B enhancer or rearrangement to the TCRβ locus, CDK6, or the MYC enhancer [17]. PTPN2 deletions, which result in aberrant activation of the PI3K-AKT-mTOR pathway, have specifically been observed in T-ALL cases with aberrant expression of TLX1 and have a similar incidence in pediatric and adult cohorts [15, 102]. Despite differences in ETP- and PTPN2 deletion-status, CDKN2A/B are frequently deleted and PHF6 is frequently mutated in both TLX subtypes [14,15,16,17]. Recent work indicates that when mutations to PHF6 are present, novel hypomethylating agents may have utility in combination with BCL-2 inhibitors such as venetoclax [103].

Leukemia resulting from either TLX1 or TLX3 dysregulation predominantly affects males regardless of age group [14,15,16,17]. Despite these similarities, outcome data for each gene varies. In AYA/adult populations, patients with increased TLX1 expression are considered intermediate-risk, but high-risk when TLX1 expression is decreased (3-year OS 64% in TLX1high and 36% in TLX1low patients vs. 48% in non-TLX1 altered patients) [104]. The TLX1 subtype is considered low-risk in pediatric patients [17] with a reported 5-year OS of 100% (vs. 95% non-TLX1 altered patients) [15] and disease can be effectively managed with contemporary chemotherapy. Patients with TLX3 dysregulation are variably reported as low-risk (5-year OS 92% TLX3alt n = 31 vs. 78% TLX3norm n = 100; p = 0.099 ALL-BFM-A 1986-2000) [105] and high-risk (5-year OS 45% TLX3alt n = 20 vs. 57% TLX3norm n = 72; p = 0.049 FRALLE-93) [106] in pediatric cohorts. These disparities may be due to differences in treatment regimens and genomic profile. Indeed, recent genomic subtyping highlighted the importance of co-occurring alterations and revealed two subsets of TLX3 dysregulation: TLX3-immature patients, enriched for WT1 alterations, NUP214::ABL1 fusions, loss of 16q22.1, FLT3 internal tandem duplications, and JAK pathway alterations; and TLX3 DP-like patients expressing gain of chr14q and LEF1 and MYB alterations [17]. Furthermore, the TLX3-immature subtype was associated with a negative prognosis whereas patients with TLX3 DP-like disease experienced favorable outcomes [17]. Potential translation of these findings into the clinic is yet to occur but it may indicate that those patients with TLX3 DP-like disease may benefit from a less intense upfront chemotherapy regimen.

NKX2-1 and NKX2-5-rearranged disease

Initially defined as a subtype by Liu et al. (Table 1) [15], the NKX2-1 gene encodes a protein that binds and promotes expression of thyroid specific genes and drives intracellular serine/glycine synthesis, important for epigenetic regulation, in T-ALL [107]. NKX2-1 rearrangements (NKX2-1r) typically result in a mature CD4+/CD8+ double-positive immunophenotype [14, 15, 17], indicating leukemic transformation occurs later in T-cell development and is often observed in pediatric cases [14, 15, 17]. Long term outcome data for the NKX2-1r subtype are somewhat limited, however, it has been noted as a low-risk subtype [17] with good early molecular response (100% NKX2-1r n = 13 vs. 95% non-NKX2-1r n = 251; p = 0.39 NCT00408005) [15] and a 5-year EFS of >98% in pediatric patients if day 29 MRD is <0.01%, suggesting these patients may respond well to chemotherapy intensity reduction [17].

Only recently defined as a separate subtype by Pölönen et al. (Table 1) [17], NKX2-5r patients feature rearrangements to TCR-β/δ loci and BCL11B enhancer hijacking [17]. NKX2-5r harbors a distinct genomic expression profile compared to NKX2-1r, with changes noted to WT1, NRAS, NOTCH1, RB1, LEF1, and PTEN [17]. Pediatric patients with NKX2-5 alterations are typically female and younger at diagnosis with enrichment of a myeloid gene signature [17]. These patients are noted as having increased risk of MRD compared with other subtypes, though the NKX2-5r subtype is rare and findings should be confirmed in additional cohorts [17]. Should these observations be confirmed, patients would benefit from diligent MRD monitoring.

MLLT10-rearranged (MLLT10r) disease

MLLT10-rearrangements occur in both AML and ALL and are relatively common in T-ALL, constituting 8-10% of cases [108,109,110]. Like some KMT2Ar ALL, disease immunophenotype indicates leukemic transformation occurs at a later stage of maturation with the exception of ETP-like MLLT10r disease which demonstrates an ETP to pre-T differentiation stage [14]. ETP-like MLLT10r disease commonly displayed concomitant alterations to PSIP1, ETV6, GATA3, IKZF1, RUNX1 or RAS signaling but decreased alterations to genes involved in NOTCH pathway signaling [17]. MLLT10 partner genes typically include PICALM, DDX3X, and KMT2A [15, 108]. MLLT10r T-ALL confers an intermediate- to high-risk disease phenotype in pediatric/AYA cases depending on the study, with a 5-year OS of 46% (n = 20 vs. 72% non-MLLT10r n = 215; p = 0.052 GRAALL03/05) [42] in one cohort and 83% (n = 13 vs. 96% non-MLLT10r n = 13; p = 0.07 NCT00408005) [15] in another. Disease severity and outcome is highly dependent of ETP immunophenotype, with ETP-ALL cases faring worse than non-ETP ALL patients [17, 108]. Despite this variability in outcome between the different patient cohorts, data indicate this subtype can be effectively managed with contemporary therapeutic regimens [15,16,17, 42].

Basic Helix Loop Helix (bHLH) gene alterations

bHLH genes encode a diverse group of transcription factors; alterations to TAL1, TAL2, and LYL1 are considered subtype defining in T-ALL [14,15,16,17]. bHLH alterations are currently thought to confer a favorable ALL risk profile [14, 17]. Interestingly, bHLH alterations broadly result in different gene expression profiles [14,15,16,17].

TAL1 and TAL2

Over 30 years ago Brown et al. described a genomic alteration observed in ~25% of T-ALL cases, involving the chimeric fusion STIL::TAL1 [111]. This observation has been recapitulated in recent patient cohorts with designation into subtypes involving TAL1 or TAL2 (Table 1) [14,15,16,17]. TAL1 is a bHLH transcription factor involved in mediating hematopoietic stem cell fate by regulating gene expression in megakaryocytic/erythroid progenitors [112]. TAL1 also interacts with other known T-ALL associated genes such as LMO2, which, under normal conditions, plays an essential role in lineage fate determination [112]. PI3K-signaling alterations are frequently enriched in TAL1-altered T-ALL and may be clinically targetable [17]. The fusion of TAL1 with STIL is frequently due to an interstitial deletion between the two genes placing TAL1 under the regulatory control of the STIL promoter resulting in aberrant super expression of TAL1, leading to malignancy [113, 114]. Observed in ~30% of childhood and adult T-ALL cases, STIL::TAL1 requires cooperation from other genomic events involving LMO1 or LMO2 or activation of the NOTCH pathway [16, 53, 85]. Additionally, TAL1 overexpression can be due to formation of a super-enhancer caused by introduction of mutations in a noncoding region upstream of the TAL1 transcriptional start site. These mutations create de novo binding motifs for the MYB transcription factor resulting in super expression of TAL1 [115]. Recently, WGS identified additional noncoding alterations involving TAL1 within a pediatric cohort [17] including TAL1 enhancer gain, enhancer alteration via SNV/Indel mutations, enhancer hijacking events, intergenic inversions and noncoding rearrangements.

STIL::TAL1 is rare in adult T-ALL patients [14], but those with STIL::TAL1 experience similar 4-year OS rates as those without STIL::TAL1 (both 45%) [116] suggesting outcomes are similar with current treatment regimens. Combined with the outcome data mentioned above, survival of TAL1 dysregulated patients varies depending on patient age, the type of TAL1 alteration and co-occurring lesions [14, 17].

LYL1

LYL1 has purported roles in blood vessel maturation and hematopoiesis where it interacts with LMO2 [117]. LYL1 driven T-ALL results from gene activation and was identified as a subtype defining alteration in two of the genomic landscape studies [14, 15]. In both instances LYL1 perturbations co-occurred with LMO2 dysregulation, likely due to established interactions between the two transcription factors [14, 15, 117]. Pediatric patients with alterations to LYL1 are reported to have a worse EFS and MRD [17].

Overall, alterations of bHLH genes TAL1, TAL2 and LYL1 demonstrate mutual exclusivity [17]. Leukemia arising from TAL1 dysregulation has been described in predominantly male pediatric patients [14,15,16]. Recent data demonstrates the importance of co-occurring genomic alterations in TAL1 driven T-ALL [17]. Termed TAL1αβ-like and TAL1 double positive (DP)-like, these subtypes differ in immunophenotype and gene alterations. TAL1 αβ-like disease exhibits a higher frequency of STIL::TAL1 fusions, LMO1 enhancer SNVs, and is enriched for MYCN mutations [17]. Conversely, TAL1 DP-like T-ALL has a mature immunophenotype (CD4+/CD8+), increased expression of RAG1/2, and increased frequency of LMO2 TCR rearrangements [17]. Clinical utility of these subtypes is yet to be fully realized, however, recent data suggest TAL1 αβ-like and TAL1 DP-like patients demonstrate a significantly worse OS (-log10(p)=1.5) and disease free survival (-log10(p)=2.4) respectively compared to other subtypes [17]. If patients can be more accurately grouped based on gene expression profiles rather than individual driver alterations, then appropriate treatment regimens can be selected, potentially significantly improving patient outcomes.

LMO1 and LMO2

LMO1 and LMO2 encode members of a family of nuclear transcription co-regulators involved in mediating protein-protein interactions rather than binding directly to DNA [118]. LMO2 is an essential regulator of hematopoiesis and angiogenesis, and deletion in murine models results in lethality during early development [119]. LMO1 is a paralog of LMO2 but has a distinct function in neurogenesis [118] and LMO1 and LMO2 alterations are mutually exclusive [14, 17]. Dysregulation of LMO2 gene expression is observed in two different subtypes: combined with dysregulation of either LMO1 [15] or LYL1 [14, 15]. Similarly, LMO1 dysregulation forms a distinct subset when combined with TAL1 dysregulation [14, 15]. Activating mutations to the JAK/STAT signaling pathway co-occur in the LYL1/LMO2 (JAK1, JAK3, and STAT5A) and TAL1/LMO1 (JAK1, JAK3, STAT5A, STAT5B) subtypes [14]. Historically, AYA/adult patients with LMO2 alterations were considered high-risk (4-year OS 29% LMO2alt n = 15 vs. 44% LMO2norm n = 54, LALA-94) [116] but pediatric patients low-risk (5-year OS 90% LMO2alt n = 20 vs. 96% LMO2norm n = 244; NCT00408005 p = 0.078) [15]. New insight from a pediatric cohort has identified two additional subtypes of LMO2 termed LMO2 γδ-like and STAG2/LMO2. Patients clustering in the LMO2 γδ-like subtype displayed LMO2 activation from BCL11B enhancer hijacking, TCRδ rearrangements, and LMO2 enhancer SNV/Indels [17]. LMO2 γδ-like is associated with poor outcome and has a 5-year EFS < 60% [17]. The STAG2/LMO2 subtype typically involves a LMO2::STAG2 rearrangement resulting in LMO2 activation and simultaneous STAG2 inactivation [17, 120, 121]. These patients are predominantly female, very young at diagnosis, and constitute a rare subtype of T-ALL ( < 1% of cases analyzed) [17]. Patients with STAG2/LMO2 ALL who are <3 years old and have ≥1% MRD following induction have dismal outcomes [120].

Initial analyses by Liu et al. suggested association of LMO1 dysregulated pediatric patients with a poor outcome [15] however, additional analyses by Pölönen et al., which include patients from the original cohort, observed no association [17]. The severity and risk of relapse from LMO driven disease is highly dependent on patient age, the specific driver alteration or the co-occurring lesion/s contributing to disease pathogenesis [15, 17, 116].

As previously mentioned, activating mutations to the JAK/STAT signaling pathway are frequently observed in the TAL1/LMO1 and LYL1/LMO2 subtypes rendering tyrosine kinase inhibitors as attractive targeted therapies [122]. In vitro studies of the JAK inhibitors ruxolitinib, tofacitinib, and upadacitinib all significantly reduced growth of Ba/F3 cells harboring common JAK/STAT activating mutations such as JAK1 p.A634D, IL7R insSRCL, JAK3 p.M511I, JAK3 p.L857Q (except upadacitinib) [123]. Ruxolitinib has also demonstrated ex vivo efficacy and synergistic action when used in combination with venetoclax in patient samples harboring activating mutations to IL7R pathway genes [124]. The JAK2 inhibitor fedratinib is active against wild-type and mutated JAK2 and also FLT3, suggesting potential utility in leukemias harboring activating mutations in both genes [125].

Conclusion

T-ALL is a genomically complex disease which is frequently fatal upon relapse. Here we have summarized four T-ALL genomic landscape studies and found consensus in subtype defining alterations where possible. We have also described patient outcomes and potential targeted treatments for laboratory and clinical investigation. T-ALL occurs less frequently than B-ALL and, until recently, has remained an under-studied cohort, particularly in terms of genomic interrogation of large patient datasets. The synopsis presented here highlights the genomic complexity of T-ALL and the importance of co-occurring lesions on disease progression and outcome. Of significance is the recent insight gained from WGS analyses and the advantage over traditional diagnostic techniques, particularly in the T-ALL setting. These genomic methodologies have identified the frequency at which noncoding alterations occur in T-ALL, revealing a genomic landscape predominantly comprised of noncoding lesions. Given we now know that noncoding alterations have a role as T-ALL driver lesions as well as secondary co-lesions impacting outcome, in the future, detailed genomic analyses may be required to inform the clinical approach for disease treatment and potential eradication.