Structural variants involving MLLT10 fusion are associated with adverse outcomes in pediatric acute myeloid leukemia

Current outcomes of pediatric acute myeloid leukemia (AML) remain suboptimal despite treatment intensification, with 5-year overall-survival (OS) rates of 75% to 80% and 5-year event-free survival (EFS) rates of up to 55%.^1-3 Early treatment response and cytogenetic/molecular abnormalities are the most important prognostic factors for survival.^1,4 Because of the significant heterogeneity of pediatric AML, ongoing clinical trials have focused on identifying new prognostic subgroups for whom treatment intensity may be adapted based on novel risk criteria. We, and others, have previously shown that childhood AML is enriched in oncogenic fusions, which can be used to characterize and subclassify patient groups.⁵ In the case of rarer fusions, the incidence can be too low to reliably understand its prognostic impact outside of large clinical trials or international collaboration.

The 11q23/KMT2A-rearranged (KMT2A-r) AML is a heterogeneous genetic subtype, which occurs in 20% to 25% of children with AML. A large International Berlin-Frankfurt-Munster Study Group (I-BFM-SG) collaborative study showed that the prognosis of this subtype is influenced by the fusion partner.⁶ The MLLT10 gene (previously designated AF10), located at 10p12.31, is a known fusion partner of KMT2A, and encodes the AF10 protein, which is a transcription factor that exerts its leukemogenic effect by binding unmodified histone H3 and regulating DOTIL expression.⁷ Van Weelderen et al conducted a more recent I-BFM-SG international study on childhood KMT2A-r AML and assessed the outcome of this subgroup depending on the fusion partner. The KMT2A::MLLT10 fusion was included in the high-risk group that had inferior EFS (30.3% vs 54% in non–high risk) and OS (49.2% vs 70.5%) and a worse cumulative incidence of relapse (59.7% vs 35.2%).⁸ Similarly, the current AML phase 3 Children's Oncology Group (COG) trial, AAML1831 (www.clinicaltrials.gov identifier: NCT04293562), includes patients with rearranged KMT2A::MLLT10 as being at high risk.

However, the role of MLLT10 and its prognostic implications in partnership with other genes have not been well defined. One of these other partners, PICALM::MLLT10 (originally called CALM::AF10) is generated by the t(10;11)(p12-13;q14-21) translocation, and has been rarely reported in patients with AML, acute lymphoblastic leukemia (ALL, especially T-phenotype), and acute undifferentiated leukemia.^9,10 The PICALM::MLLT10 fusion gene has been associated with high relapse rates and poor prognosis in adult AML,⁹ however, data are lacking in children with AML harboring this translocation.

To better understand the impact of MLLT10 structural variants in pediatric AML, we conducted a retrospective study in 2080 children and young adults registered on COG AAML0531 and AAML1031 treatment trials. By comprehensive transcriptome and karyotype evaluation, we defined the full spectrum of MLLT10 fusions, identified new fusion partners, and correlated MLLT10 structural variants with clinical outcomes in pediatric AML. In addition, transcription and methylation profiles were evaluated to identify dysregulated genes in MLLT10 fusions with and without KMT2A.

Patient characteristics and outcome data were collected from 2080 children and young adults aged 0 to 30 years, with AML enrolled on COG AAML0531¹¹ (August 2006-June 2010) and AAML1031¹² (February 2011-June 2016) treatment trials. Collected data included age, sex, race/ethnicity, date of diagnosis, initial white blood cell count, peripheral blood and bone marrow (BM) blast percentages, World Health Organization classification, presence of extramedullary disease (EM; central nervous system [CNS] or chloromas), karyotype (and RNA sequencing to detect fusions and copy number variants), presence of complex cytogenetics, FLT3-ITD with allelic ratio, other FLT3 mutations, and other molecular alterations (NPM1, CEBPA, WT1, KIT, etc); in addition, we collected data on risk category; treatment protocols; receipt of investigational therapy (gemtuzumab ozogamicin [GO] in AAML0531); allogeneic hematopoietic stem cell transplantation (HSCT); response to therapy, including morphology and measurable (minimal) residual disease (MRD) by flow cytometry after induction courses 1 (end of induction 1 [EOI-1]) and 2 (EOI-2); and treatment outcomes including relapse, refractory disease, occurrence of secondary malignancy, death, and status at last follow-up.

Karyotype data for all patients were collected by COG and centrally reviewed using International System for Human Cytogenetic Nomenclature for consistent reporting of lesions. Karyotypes with translocations involving the 10p12.31 locus (MLLT10) were compared with results from transcriptome sequencing (as available) to corroborate findings.

All patients provided informed consent, and the study was conducted in accordance with the principles of the Declaration of Helsinki.

Transcriptome data were available from 1491 pediatric patients with AML enrolled on COG trials AAML0531 and AAML1031. These data were generated by the British Columbia Genome Sciences Center (Vancouver, BC). Total RNA samples were ribodepleted and prepared for sequencing using a strand-specific messenger RNA library construction protocol. Indexed libraries were pooled and sequenced on an Illumina Hiseq, producing 75–base pair paired-end sequence reads. Sequencing data were aligned to human genome assembly GRCh37. Gene-level counts were quantified using Boston Consulting Group Sequencing Gene’s in-house pipeline and annotated using Ensembl version 69 annotations. Gene fusions were identified using 3 calling algorithms: TransAbyss (version 1.4.10), STAR-fusion (version 1.8.1), and Cicero (St. Jude, Memphis, TN). Fusions involving MLLT10, identified by at least 2 algorithms or 1 algorithm with corroborating karyotype information, were considered positive for MLLT10 rearrangement.¹³ Differential expression analyses were performed using the R package Limma and gene enrichment analyses were performed using Gene Ontology enrichment (https://geneontology.org/).

Data from the AAML0531 and AAML1031 trials were current as of 30 September 2020 and 31 March 2022, respectively. Patient characteristics were compared using the χ² test, or Fisher's exact test if data were sparse. The Kruskal-Wallis test was used for comparisons of medians. The Kaplan-Meier method was used to compare OS and EFS. OS was defined as time from study entry until death from any cause. EFS was defined as time from study entry until refractory disease, relapse, or death. Estimates of relapse risk (RR) were obtained by methods that account for competing events. RR was defined as time from EOI-1 for patients in complete remission (CR) until relapse, with deaths without relapse considered competing events. Otherwise, patients were censored at last known contact for OS, EFS, and RR. The statistical significance of predictor variables was tested with the log-rank statistic for OS and EFS, and with the Gray statistic for RR.

Overall, 2080 patients treated on the COG AAML0531 and AAML1031 trials were evaluated by transcriptome profiling and/or karyotyping to identify leukemia-associated fusions and copy number changes linked with prognosis. A total of 127 patients (6.1%) were identified to have primary fusions involving MLLT10. Among 127 cases, 104 (81.9%) had KMT2A::MLLT10, 13 (10.2%) had PICALM::MLLT10, and 10 (7.9%) had MLLT10 fusions (X::MLLT10) with various translocations (Figure 1). Among the latter group, 2 had DDX3X and 2 had TEC fusions, whereas fusions with other partner genes (DDX3Y, CEP164, SCN2B, TREH, NAP1L1, and XPO1) were each identified in single patients (supplemental Table 1).

Approximately 2 of 3 patients had confirmatory evidence of an MLLT10 rearrangement by both RNA sequencing and karyotype data, whereas in another 20% of cases the cytogenetic band reported for the fusion did not map to MLLT10. For some cases for which the karyotype was unavailable, RNA sequencing successfully identified a fusion. Conversely, if a sample had low blast count and RNA sequencing did not identify a fusion, the karyotype was able to provide information. Thus, it is important to note, that both karyotype and RNA sequencing are useful and necessary for characterizing structural alterations in patients.

Overall, the median age of our patient cohort was 2.9 years (range, 3 days-21.3 years), 73 were males and 54 females. Patients with PICALM::MLLT10 were older, with a median age of 14.4 years (range, 7.4-18.9 years) vs 6.6 years (range, 1.4-15.2 years) in those with X::MLLT10, and 1.7 years (range, 0-21.3 years) in those with KMT2A::MLLT10. Median initial white blood cell count in the whole cohort was 18.5 × 10⁹/L (range, 0.8 × 10⁹/L - 439 × 10⁹/L), and it was higher at 42.9 × 10⁹/L in the PICALM::MLLT10 group vs 18.4 × 10⁹/L in the KMT2A::MLLT10 group and 8.6 × 10⁹/L in the X::MLLT10 group. Overall, 16 and 20 patients had CNS3 and CNS2 disease, respectively. Overall, CNS3 disease was more common in patients with X::MLLT10, at 30% (n = 3 of 10) vs 15.4% (n = 2 of 13) and 11.1% (n = 11 of 104) in the PICALM::MLLT10 and KMT2A::MLLT10 groups, respectively. Non-CNS EM involvement was overall present in 37 patients (29.1%) and was significantly more common in the KMT2A::MLLT10 group at 34.6% (n = 36) vs 7.7% (n = 1 of 13) and 0% in the PICALM::MLLT10 and X::MLLT10 groups, respectively (Table 1). Sites of EM involvement included skin, gingiva, bone, head/neck, abdomen/pelvis, and spine.

Among patients with MLLT10 fusions, 76 were treated on COG AAML1031, and 51 on COG-AAML0531; in total, 27 patients received GO as part of their treatment. Allogeneic HSCT in first CR (CR1) was performed in 16 patients (12.6% compared with 16.9% of other AML without an MLLT10 rearrangement). Nine patients (75%) with PICALM::MLLT10 fusion underwent HSCT in CR1 vs 5 (50%) and 9 (9.6%) patients in the X::MLLT10 and KMT2A::MLLT10 groups, respectively (Table 1). Analysis of the World Health Organization AML categories showed a predominance of acute monoblastic leukemia subtype in 60% and 43.3% of the X::MLLT10 (n = 6) and KMT2A::MLLT10 (n = 45) groups, respectively; whereas AML with myelodysplastic syndrome-related changes predominated (30.8%; n = 4) in the PICALM::MLLT10 cohort (Table 1).

Molecular data showed mutation enrichment in several genes within the MLLT10 fusion types (Figure 2). PTPN11 mutations were seen at an overall frequency of 12% across the MLLT10 cohort, but the majority of those mutations (14 of 15) occurred in the KMT2A::MLLT10 group.

Additionally, in the KMT2A group, SETD2, FLT3 missense, and U2AF1 mutations were observed at a frequency of 9.6%, 9.6%, and 4.8%, respectively. Mutations in the epigenetic regulator, plant homedomain finger protein 6 (PHF6), and histone methyltransferase gene EZH2 were observed only in the setting of PICALM::MLLT10 (69.2% and 23%, respectively) and 2 of 13 patients also had TP53 mutations. The EZH2 gene mutations included 2 truncating and 1 missense mutation, all which would lead to loss of function mutations, consistent with the report by Grossmann et al.¹⁴ Within the X::MLLT10 cases, SETD2 and WT1 were the most frequently mutated at 50% and 30%, respectively. Of note, only missense mutations in the tyrosine kinase domain of FLT3 were observed; no patients had internal tandem duplication (FLT3-ITD) mutations. Both NRAS and KRAS were frequently observed across all 3 MLLT10 groups, however, these genes are frequently altered across all subtypes of pediatric AML in general, and do not appear to be causally linked to MLLT10 fusions.

Secondary cytogenetic abnormalities were present in 22 patients with KMT2A::MLLT10; among these, 2 had monosomy 7 and 2 had additional PICALM::MLLT10 fusions (considered as secondary abnormalities). Furthermore, 15% of the PICALM::MLLT10 group had an additional del5q abnormality.

Structurally, MLLT10 retains the highly conserved octapeptide motif and leucine zipper domains in the latter half of the gene when fused as the 3′ partner of the rearrangements (Figure 3). Although MLLT10 reportedly always occupies the 3′ gene position, we observed MLLT10 in the 5′ gene position with 2 partner genes: TREH and NAP1L1. For the genes KMT2A and PICALM, multiple breakpoints occurred when fused to MLLT10; representative diagrams of all fusions are shown in supplemental Figure 1.

Morphologic CR rates after induction cycles 1 and 2 were similar between PICALM::MLLT10, X::MLLT10, and KMT2A::MLLT10 cases (Table 1; P = not significant), and not significantly different from the non-MLLT10 AML cohort. However, MRD as assessed by multidimensional flow cytometry after induction cycle 1 was positive in 60% of patients with PICALM::MLLT10 vs 12.5% in those with X::MLLT10 vs 15.6% in those with KMT2A::MLLT10 vs 29.4% in the non-MLLT10 AML cohort (P = .01). Among patients with MLLT10 fusion and known MRD at EOI-1 (n = 114), there were 22 patients with positive MRD vs 92 patients who were MRD negative. The 5-year disease-free survival (DFS) from EOI-1 for patients who were MRD positive was 9.1% vs 22.4% in those who were MRD negative (P < .001). Five-year OS from EOI-1 for patients who were MRD positive was 13.6% vs 46.4% in those who were MRD negative (P < .001).

Collectively, patients with MLLT10 fusion (n = 127) demonstrated adverse outcomes, with 5-year OS of 38.2% vs 65.7% in patients without MLLT10 fusion (n = 1953; P ≤ .001; Figure 4A), poorer 5-year EFS of 18.6% vs 49% (P < .001; Figure 4B), and a higher RR of 76% (n = 82) vs 38.6% (n = 1434; P < .001; Figure 4C).

Next, we compared the outcomes of patients with other MLLT10 fusion (X:: and PICALM::) with those of patients with KMT2A::MLLT10 fusion. Patients with other MLLT10 fusions had an OS, from study entry, of 27.6% vs 40.4% for those with KMT2A::MLLT10 (P = .361; Figure 4D) and an EFS of 12.7% vs 19.5% for patients with KMT2A::MLLT10 (P = .628; Figure 4E). The 2 cohorts also had similar RR from remission, with 84.7% for KMT2A::MLLT10 vs 74.6% for other MLLT10 (P = .876). Given the fact that PICALM::MLLT10 is the most prevalent fusion in the other-MLLT10 cohort, we further evaluated the outcomes for the other-MLLT10 cohort with and without PICALM fusion partner. Patients with PICALM::MLLT10 and X::MLLT10 had similarly dismal EFS of 9.2% and 20%, respectively (P = .788; Figure 4F).

Overall, relapses occurred in 82 patients (64.6%), of whom 59 (72%) were BM relapses; among these, 37 were isolated BM, and 22 were combined BM and EM relapses (18 with CNS relapse, of whom 2 had also chloromas; and 4 with chloromas); 8 patients had isolated CNS relapse, 11 had isolated chloromas, 1 relapse presented as a combined CNS and skin disease, and 3 relapses were from other sites. Most relapses (63.4%) occurred within 1 year from EOI.

Among patients who were MLLT10⁺ enrolled on COG AAML0531 trial (n = 51), 27 received GO in Arm B and their 5-year EFS was 22.2% vs 8.3% for Arm A (no GO; P = .094), and 5-year OS for Arm B was 43.2% vs 41.7% for Arm A (P = .566; supplemental Table 2). In summary, GO appears to benefit EFS in these patients, although not statistically significant, similar to the overall results of the AAML0531 study.¹¹ 

When assessing the impact of HSCT in CR1, time-to-event statistical analyses among those who are MLLT10⁺ showed that 5-year OS ± a standard error (SE) of 2 in the HSCT group (n = 16) was 56.3% ± 24.8% vs 43.0% ± 12.9% in the group without HSCT (n = 61; P = .383); whereas the 5-year DFS ± 2 SE in the HSCT group was 37.5% ± 24.2% vs 28.7 ± 11.7% without HSCT (P = .380). Therefore, survival estimates appeared to be higher for HSCT in CR1, but the large SEs result in nonsignificant statistical differences.

We interrogated the transcriptome data from 1491 patients treated on COG AAML0531 and AAML1031 to determine whether those with MLLT10 fusions have a distinct expression profile compared with other AML. Expression profile of 91 patients with MLLT10 fusions were compared with 1400 non–MLLT10 rearranged cases.

In this comparison, 1395 genes were differentially expressed, as presented in Figure 5. The list of top 100 upregulated and downregulated differentially expressed genes is provided in supplemental Figure 2. Differentially expressed genes were enriched in homeobox (HOX) genes. Seven HOXA family genes were among the top 40 upregulated genes, with HOXA11 identified as 4.7-log fold-change higher on average in patients with rearranged MLLT10. Analysis of the top 100 upregulated genes reveals a significant enrichment in the positive regulation of hematopoietic stem cell proliferation for biological processes (22.7-fold enrichment; P < .0001; false discovery rate, 0.00281) and activation of HOX genes during differentiation for pathway prediction (14.42-fold enrichment; P < .0001; false discovery rate, 0.0043).

To further determine whether patients with MLLT10 fusions had distinct epigenetic profiles, we performed differential methylation analyses on samples from normal BM and from patients with MLLT10-rearranged, KMT2A-rearranged, and NUP98::NSD1-fused AML, on the Infinium HumanMethylation EPIC array. After fitting a multivariate model with all interacting molecular features, the 50 most discriminative regions were extracted (Figure 5). Patients with PICALM::MLLT10 and the majority of X::MLLT10 fusions share a distinct hypermethylation signature with patients with high-risk NUP98::NSD1 fusion; this signature is consistent with gross disruption of HOX clusters, at which PICALM::MLLT10, X::MLLT10, and NUP98::NSD1 show severe dysregulation. In contrast, patients with KMT2A::MLLT10 cluster with other KMT2A fusions, and patients with only FLT3 ITD or WT1 frameshift variants (commonly observed alongside NUP98-NSD1 fusions) show only a partial overlap of this hypermethylation signature. Patients with NUP98-KDM5A fusion do not show a hypermethylation signature, as noted by Bertrums et al,¹⁵ reflecting a fundamental difference in leukemogenic mechanism between NUP98 fusions with NSD1 vs KDM5A-like fusion partners.

Gene set enrichment analysis for HOXA family was run between subsets of MLLT10 fusions. No significant differences were observed, but expression trended higher in -KMT2A compared with the other fusion subsets (KMT2A::MLLT10 vs other::MLLT10, P = .11; KMT2A::MLLT10 vs PICALM::MLLT10, P = .22; KMT2A::MLLT10 vs X::MLLT10, P = .13). No significant difference was observed between PICALM::MLLT10 and X::MLLT10 fusions (P = .97) (supplemental Figure 3)

Our study identified new MLLT10 fusion partners not previously described in childhood AML and compared them to other AML subtypes and with known MLLT10 partners like KMT2A and PICALM. All patients with aberrant MLLT10 had a high RR, poor EFS and OS, distinct methylation profiles, and overexpressed HOXA genes. Patients with KMT2A::MLLT10 tend to be younger at diagnosis, have higher rates of EM disease, and exhibit lower EOI MRD levels, whereas patients with PICALM::MLLT10 and MLLT10::X fusions tend to be older with higher MRD levels.

Although 83% of patients with MLLT10 fusions achieved morphological CR by EOI-2, their RR was still very high at 76%, compared with 39% in other AML subtypes (P < .0001), with most relapses (63.4%) occurring within the first year after diagnosis.

Early response to therapy is currently used for risk stratification of therapy in childhood AML, and flow cytometry–based MRD provides a more accurate measure of therapy response^1,16; despite that, the benefit of MRD measurement is variable among AML subtypes and studies and depends on the sensitivity of the applied method.^17,18 Reverse-transcription polymerase chain reaction analysis of fusion transcripts allows MRD detection with a sensitivity level of up to 0.001%, but it is applicable in only 50% to 60% of pediatric patients with AML with a detectable fusion gene. Next-generation sequencing MRD analysis allows the identification of molecular anomalies, but it is more relevant in AML with normal cytogenetics. Furthermore, the role of next-generation sequencing in MRD detection is still controversial in pediatric AML. Flow cytometry–based MRD has a lower sensitivity than reverse-transcription polymerase chain reaction (up to 0.01%), but it is applicable in >90% of patients with AML, and it is generally the method of choice for MRD detection in clinical AML trials.¹⁹ In our study, flow cytometry–based MRD data were available in most patients (n = 114). Five-year DFS and OS from EOI-1 for patients who were MRD negative (22.4% and 46.4%, respectively) were better than that of patients with EOI-1 MRD positive status (9.1% and 13.6%, respectively).

The study by van Weelderen et al in childhood KMT2A-r AML showed that EOI-2 MRD negativity was associated with better EFS (47.6%, MRD negative vs 16.3%, MRD positive; P < .0001), and OS (66.0% vs 27.9%; P < .0001) and a trend toward lower cumulative incidence of relapse (46.1% vs 65.4%; P = .016).⁸ In our cohort, although 84% of the patients harboring KMT2A::MLLT10 fusion had negative MRD at EOI-2, their EFS was very low at 19.5%. Also, the group with other MLLT10 fusions (PICALM- and X) had an even lower EFS of 12.7% despite 61% of them being MRD negative by EOI-2. These data suggest that MRD does not adequately predict relapse in these cytogenetic groups.

The small group who had PICALM::MLLT10 fusion (13 patients) seem to have the worst EFS (9.2% vs 20% in other MLLT10; P = .788). Our study, to our knowledge, is the first showing outcome data in children with AML harboring this fusion. An adult series of adult patients with AML with PICALM::MLLT10 fusion gene showed more frequent EM disease, younger age, a high percentage of therapy-related cases, and a high relapse rate. In our cohort, the PICALM::MLLT10 group were older, and 2 of 13 (15%) patients had CNS3 disease, 1 (7.7%) had non-CNS EM disease, and 4 (30.8%) had AML with myelodysplastic syndrome-related changes although not therapy related.

A study by Pollard et al in 215 children with KMT2A-r AML, showed that adding GO to conventional chemotherapy was associated with a reduced relapse rate, RR (40% GO vs 66% in patients without GO), improved 5-year DFS (GO 57% vs no GO 33%), and a better 5-year EFS (48% with GO vs no GO 29%). In addition, consolidation with HSCT in that study may further have enhanced outcomes (5-year RR with GO and HSCT, 28% vs 73% without GO and HSCT, P = .006).²⁰ In our study, GO appeared to benefit EFS in patients with MLLT10⁺ AML (22.2% GO vs 8.3% no GO) but was not statistically significant (P = .094; supplemental Table 2).

Allogeneic HSCT in CR1 is considered by some collaborative groups as the recommended approach in high-risk AML subgroups with an EFS of <30%.²¹ Long-term responses were seen in a subset of adult patients with PICALM::MLLT10⁺ AML after HSCT in CR1 but also after high-dose cytarabine.⁹ Among our patients with MLLT10⁺ AML who underwent HSCT in CR1 (n = 16), the 5-year DFS was 37.5% (±24.2%) vs 28.7% (±11.7%) in those without HSCT (P = .380). Although HSCT in CR1 appeared to benefit survival rates, the difference was not statistically significant. Therefore, novel targeted therapies are urgently needed for this group of patients.

Transcriptome analysis of patients with MLLT10 fusions showed that HOXA genes are overexpressed, especially the HOXA11 gene. Indeed, HOXA gene upregulation (as in KMT2A rearrangements) is known to mediate leukemogenesis in PICALM::MLLT10–positive acute leukemia.^22,MLLT10 interacts directly with DOT1L in a complex that regulates the methylation of H3K79. MLLT10 fusions are thought to misdirect DOT1L to the promoters of HOXA genes, leading to hypermethylation of H3K79. This hypermethylation causes constitutive activation of HOXA activity, preventing cell maturation and differentiation.²² The DOT1L-MLLT10 complex has also been implicated in HOXA overexpression in patients with NUP98-NSD1–rearranged and KMT2A-rearranged fusion positivity. The similarities of hypermethylation seen in MLLT10 and NUP98 fusions show severe dysregulation of HOX gene clusters and may share mechanistic features leading to poor response to current therapeutic strategies.

Furthermore, variants in the histone methyltransferase EZH2 were previously identified in patients (27%) with PICALM::MLLT10–positive acute leukemias.¹⁴ EZH2 is the catalytic subunit of the polycomb repressor complex, which decommissions HOX clusters, and thus may synergistically deregulate HOXA gene expression with PICALM::MLLT10 in acute leukemia.

Structurally, the octapeptide motif/leucine zipper domain of MLLT10 (residues 730-795) is required for DOT1L association via the DOT1L coiled-coiled domain and appears to be preserved in every MLLT10 fusion, herein. Importantly, MLLT10 fusions are usually formed by complex reciprocal translocations, such that X::MLLT10 and MLLT10::X are transcribed. The NAP1L1::MLLT10 fusion occurs via reciprocal translocation, and TREH::MLLT10 is most likely present in the patient with MLLT10::TREH.

As documented by Deshpande et al,²³ and more recently by Gilan et al,²⁴ DOT1L (but not full-length MLLT10) is critical to MLL fusion–mediated leukemogenesis. It appears that this interaction is also the critical determinant of MLLT10 fusion–mediated leukemogenesis. With few, if any, exceptions, X::MLLT10 is thus likely the primary determinant of leukemogenesis. It is unclear whether reciprocal translocations (MLLT10::X) disrupt targeting and localization of the DOTCOM/SEC complexes in all, some, or none of the fusions reported herein; development of new cell line models may provide further insight.

Our study highlights the poor outcomes of children with AML and MLLT10 fusions. There are currently no targeted therapies available for these patients, and intensive chemotherapy is the standard of care for MLLT10-rearranged leukemias. Anecdotal reports suggest that allogeneic HSCT could provide DFS benefit in T-cell receptor–negative T-cell ALL (T-ALL), early T-cell precursor ALL, or in adults patients with MLLT10-rearranged AML, but larger studies are warranted.^9,25-27 Furthermore, the benefit of HSCT in CR1 for pediatric AML with MLLT10 fusions remains to be determined.

Early-stage clinical trials (www.clinicaltrials.gov identifiers NCT01684150 and NCT02141828) showed that the DOT1L inhibitor, pinometostat, reduces H3K79 methylation, had acceptable toxicity, and modest efficacy as single agent in adults with KMT2A-rearranged acute leukemia.²⁸ Furthermore, recent work has shown monocytic leukemia zing finger (KAT6A) and eleven-nineteen leukemia chromatin reader (MLLT1) to be essential for MLLT10-driven leukemogenesis.^29,30 Importantly, histone acetyltransferase (KAT) inhibitors have shown preclinical efficacy against PICALM::MLLT10 lines,²⁹ as has depletion of eleven-nineteen leukemia chromatin reader,³⁰ but not direct inhibition.³¹ In addition, preclinical models demonstrate efficacy of DOT1L inhibition in primary PICALM::MLLT10 AML cells³² and in vitro AML models.³³ 

EZH2 mutations were previously found in 27% of patients with PICALM::MLLT10–positive acute leukemias (1 AML and 3 T-ALL),¹⁴ compared with 23% (3 of 13) in our PICALM::MLLT10 cohort. EZH1/2 inhibition with valemetostat is currently being evaluated in phase 2 adult trials for relapsed/refractory T-ALL,³⁴ and could be a promising targeted therapy for patients with PICALM::MLLT10–positive AML.

Azacytidine, a DNA demethylating agent, has shown anecdotal efficacy against promoter-hypermethylated infant KMT2Ar leukemia.³⁵ Other targeted therapies such as inhibitors of BET,³⁶ PARP1 (Olaparib),³⁷ and CDK9 ³⁸ have shown preliminary efficacy in preclinical models of KMT2A-rearranged AML.

In our study, expression of HOXA genes and MEIS1 were higher in the KMT2A::MLLT10 group vs other-MLLT10 cohorts. Blocking the menin–KMT2A interaction disrupts the assembly of oncogenic KMT2A wild-type or fusion complexes on chromatin. Preclinical studies demonstrated that menin inhibition downregulates HOXA and MEIS1 transcription and reverses leukemogenesis in KMT2Ar- or NPM1-mutated leukemia.³⁹ Revumenib (SNDX-5613) is a potent oral inhibitor of the menin–KMT2A interaction. A recently published phase 1 dose-escalation study in children and adults with highly refractory acute leukemia with KMT2Ar or NPM1 mutation (www.clinicaltrials.gov identifier NCT04065399), menin inhibition with revumenib monotherapy was associated with promising antileukemic activity leading to deep and sustained remissions.³⁹ 

Lastly, SEDT2 mutations were found in 11% of our patients with KMT2A::MLLT10 and 50% of those with X::MLLT10. Patients with renal cell carcinoma harboring SEDT2 mutations had a higher response rate and prolonged OS rates when treated with immune checkpoint inhibitors.⁴⁰ Therefore, immune checkpoint blockade is a potential alternative therapy for patients with AML with MLLT10 fusions and SEDT2 mutations.

In summary, children with AML and MLLT10 fusions have very dismal prognosis, regardless of the fusion partner, and should be prioritized for novel and alternative therapeutic interventions. It does not appear that these patients benefit from HSCT in CR1, although more studies are warranted. Large prospective trials with targeted therapies are required, although these cases are quite rare; and, it remains to be determined whether patients with rearranged MLLT10 with different fusion partners would respond differently to specific targeted therapies.

This work was supported by National Clinical Trials Network (NCTN) Operations Center grant 180886, NCTN Statistics and Data Center grant 180889, COG Biospecimen Bank grant U24CA196173 and National Cancer Institute grant T32CA251066 (N.J.B.), an American Society of Hematology Graduate Hematology Award (L.M.H.), the St. Baldrick’s Foundation (T.T.Jr), the Michelle Lunn Hope Foundation (T.T.Jr), and the Jane Anne Nohl Hematology Research Fund (T.T.Jr).

The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Contribution: S.M., R.E.R., and T.T.Jr conceived and led the study; R.E.R., T.T.Jr, N.B., L.M.H., H.W., A.J.D., R.B.G., and T.A. generated, processed, and analyzed the data; O.A., R.E.R., T.T.Jr, R.B.G., J.E.F., E.A.K., and S.M. drafted the manuscript; and all authors edited and approved the final manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Oussama Abla, Division of Hematology/Oncology, The Hospital for Sick Children, 555 University Ave, Toronto, ON M5G 1X8, Canada; email: oussama.abla@sickkids.ca.