Genomic and Global Gene Expression Profiling in Pediatric and Young Adult Acute Leukemia with PICALM::MLLT10 Fusion

Abstract

PICALM::MLLT10 fusion is a rare but recurrent genetic driver in acute leukemias. To better understand the genomic landscape of PICALM::MLLT10 (PM) positive acute leukemia, we performed genomic profiling and gene expression profiling in twenty PM-positive patients, including AML (n = 10), T-ALL/LLy (n = 8), Mixed-phenotype acute leukemia (MPAL), T/B (n=1) and acute undifferentiated leukemia (AUL) (n=1). Besides confirming the known activation of HOXA, differential gene expression analysis compared to hematopoietic stem cells demonstrated the enrichment of genes associated with cell proliferation-related pathways and relatively high expression of XPO1 in PM-AML and PM-T-ALL/LLy. Our study also suggested PHF6 disruption as a key cooperating event in PICALM::MLLT10-positive leukemias. In addition, we demonstrated differences in gene expression profiles as well as remarkably different spectra of co-occurring mutations between PM-AML and PM-T-ALL/LLy. Alterations affecting TP53 and NF1, hallmarks of PM-AML, are strongly associated with disease progression and relapse, whereas EZH2 alterations are highly enriched in PM-T-ALL/LLy. This comprehensive genomic and transcriptomic profiling provides insights into the pathogenesis and development of PICALM::MLLT10 positive acute leukemia.

Introduction

Balanced chromosomal rearrangements that lead to the formation of chimeric fusion genes are the most frequent driver genetic alterations in pediatric acute leukemias (1). The PICALM::MLLT10 fusion, resulting from the t(10;11)(p12;q14) chromosomal translocation, is a rare but recurring gene fusion seen in acute leukemias, mainly in T-lymphoblastic leukemia/lymphoma (T-ALL/LLy), but also reported in patients with mixed-lineage leukemia with co-expression of T-cell and myeloid antigens (T/myeloid), acute myeloid leukemia (AML) and acute undifferentiated leukemia (AUL) (2–7). The PICALM::MLLT10 fusion was observed in approximately 6–7% of T-ALL cases and is associated with a poor prognosis in adult patients with ETP subtype (8).

Given the relatively common recurrence of PICALM::MLLT10 fusion in T-ALL/LLy (PM-T-ALL/LLy), its immunophenotypic correlation and prognostic impact in this subgroup of leukemia has been widely studied (8); in contrast, data regarding PICALM::MLLT10-positive AML (PM-AML) are scarce due to its rarity as it occurs in less than 1% of pediatric AML cases (9). Moreover, previous case series studies have focused on clinical and immunophenotypic features, whereas comprehensive molecular profiling of PICALM::MLLT10-positive acute leukemias has not yet been addressed. In this study, we performed genomic and transcriptomic profiling in a series of pediatric and adolescent PICALM::MLLT10 fusion positive acute leukemia cases along with an evaluation of clinical and immunophenotypic characteristics, with an emphasis on contrasting PM-AML with PM-T-ALL/LLy to identify subtype-specific molecular signatures.

Materials and methods

Patients and samples

Twenty-one samples from twenty patients with acute leukemia with PICALM::MLLT10 fusion diagnosed by whole transcriptome sequencing (RNA-seq) clinical testing and/or conventional cytogenetics on patients’ tumor samples at St Jude Children’s Research Hospital (SJCRH) from May 2017 to December 2022 were included in this study. One patient, SJ061501, had both diagnostic (D) and relapse (R) samples available for the study. Conventional cytogenetics was performed on 8 of 20 patients. Molecular profiling was performed by whole genome sequencing and whole exome sequencing in 16 patients (17 samples), whole exome sequencing only in one, targeted panel NGS testing in two patients, and the remaining one had no DNA available for sequencing. Information about samples and testing platforms is presented in Supplementary Table S1. Appropriate informed consent was obtained from patients. This study was reviewed and approved by the Institutional Review Board (IRB) at SJCRH.

Diagnostic analyses

Cytomorphology evaluation and Immunophenotyping studies

The diagnoses were made with the cytomorphologic evaluation and immunophenotypic data based on the 4^th edition/5^th edition WHO classification and International Consensus Classification (ICC). Cytomorphological assessments were performed on the available peripheral blood smear, bone marrow aspirate smears and/or core biopsy obtained at the time of diagnosis. Flow cytometry immunophenotyping was performed using comprehensive panels when material was available. Cases from collaborative institutions with established diagnoses underwent only limited flow cytometry immunophenotypic studies.

Clinical RNA-seq test

The clinical RNA-seq test at SJCRH was validated to identify recurrent gene fusions/rearrangements that have been recognized to be associated with human neoplasms. Total RNA was isolated from patients’ tumor sample and RNA-Seq libraries were constructed using the TruSeq Stranded Total RNA Kit (Illumina, CA, USA). The RNA-seq analysis was performed as previously described (10).

Conventional cytogenetic analysis

Conventional cytogenetic analysis was performed on 24-hour unstimulated bone marrow cultures with or without synchronization according to standard procedures. When available, 20 metaphases were analyzed, and karyotypes were interpreted according to the International System for Human Cytogenetic Nomenclature.

Whole genome sequencing and whole exome sequencing

Whole genome sequencing (WGS) and/or whole exome sequencing (WES) were performed on paired tumor-normal samples for patients as indicated in Supplementary Table 1. WGS libraries were constructed using the TruSeq DNA PCR-Free sample preparation kit (Illumina, CA, USA) and for WES using the Illumina TruSeq Exome Library Prep Kit (Illumina, CA, USA). Sequencing libraries were sequenced on HiSeq4000 or NovaSeq 6000. The average WGS depth of the samples studied was 64X; the average WES depth was 155X. Data process, analysis, variant calling and annotation were performed as previously described (10). In brief, tumor-acquired single nucleotide variations (SNVs) and small insertion-deletions (indels), structural variants as well as copy number variants and copy neutral loss of heterozygosity (LOH) were evaluated. The limit of detection of variant at low allele frequency in coding genes is estimated to be ~5% for SNVs and indels with a positive predictive value of 97–99% (10).

Gene expression profiling by RNA-seq analysis

Whole transcriptome sequencing FASTQ files were processed and aligned to Human Reference genome hg19 using the Illumina DRAGEN RNA pipeline v.4.0.3 with “--enable-rna-quantification true”. The obtained quantification files were read into R v.4.3.0 and differential expression analysis was performed by DESeq2 (11) using publicly available bulk RNA-seq data of a set of CD34+CD38-CD45RA-EPCR+ hematopoietic stem cells (HSC) derived from human cord blood samples (GSE154263) as control. Principle component analysis (PCA) was generated by PlotPCA function within the DESeq2 package using the variance stabilizing transformed (VST) counts and without the constraint of priori group information. GO enrichment analyses were performed using the clusterProfiler package (12) specifically the functional enrichment analysis was conducted using the enrichGO function with pvalueCutoff and qvalueCutoff of 0.01. The gene set enrichment analysis (GSEA) was performed using the gseGo function with p-value cutoff of 0.05. The p-value adjustment method in this study is Benjamini-Hochberg. As for the expression of HOX genes, TPM (transcripts per million) value of whole blood was retrieved from GTEx_Analysis_2017–06-05_v8_RNASeQCv1.1.9_gene_median_tpm.gct (https://www.gtexportal.org/home/datasets).

Results

Clinical, cytomorphologic and immunophenotypic characteristics

The 20 patients with PICALM::MLLT10-positive leukemia were predominantly male (18 male, 2 female) and teenagers at diagnosis (median age 14.5 years, range 5 to 18 years). The demographics and clinical characteristics of the twenty patients are summarized in Table 1, information of treatment and patients’ follow-up/outcome are shown in Figure 1. Ten patients were diagnosed with AML, 1 with acute undifferentiated leukemia (AUL), 8 with T-ALL/LLy, and 1 with mixed phenotype acute leukemia (MPAL), T/B (T/B-MPAL). Among the 8 T-ALL/LLy cases, one fulfilled the diagnostic criteria of early T-cell precursor lymphoblastic leukemia (ETP-ALL), and 7 had extramedullary disease. In the T-ALL/LLy group, 6 of 8 attained complete remission and remain in remission; one patient relapsed and is alive, while the remaining one patient died of sepsis during induction therapy. The 11 non-ALL patients (10 AML and 1 AUL) were characterized by extramedullary disease in 7 cases, hyperleukocytosis in 4 cases, and poor outcomes with a median survival of 14 months. Furthermore, 6 of the AML cases had primary refractory leukemia, 2 had persistent minimal residual disease (MRD) and subsequently relapsed, and 2 achieved remission before relapse. Only 2 of the AML patients are alive. One patient (SJ62998) has remained in remission for more than seven years since receiving haploidentical transplant; the other patient (SJ033585) was treated at their local hospital after initial diagnosis and received hematopoietic stem cell transplant.

Table 1.

Characteristics of patients in the study cohort.

	AML	T-ALL	MPAL, T/B	AUL	Total
Sex	1F | 9M	0F | 8M	1F | 0M	0F | 1M	2F | 18M
Age (years)
1–9	0	3	1	0	4
10–20	10	5	0	1	16
WBC count
>100,000	4	1			5
<100,000	5	7			12
N/A	1		1	1	3
Extramedullary disease
Yes	7	6			13
No	3	1	1	1	6
N/A		1			1
Refractory/Relapse
Yes	10	1	1	1	13
No	0	7			7
Survival
Deceased	8	1*	1	1	11
Alive	2	6			8
Lost to Follow Up		1			1
Total # patients	10	8	1	1	20

^*The patient died of sepsis during induction therapy

Abbreviations: AML, Acute myeloid leukemia; AUL, acute undifferentiated leukemia; F, female; M, male; MPAL, mixed-phenotype acute leukemia; N/A, not available; T-ALL/LLy, T-cell acute lymphoblastic leukemia/lymphoma; WBC, white blood cell

Figure 1. Time course of patient treatment.

Time point of diagnosis, clinical course, and therapy are shown for individual patients with PICALM::MLLT10 fusion. Patient information up until December 2023 was included. All patients except SJ033011 were enrolled in clinical trials sponsored by St. Jude Children’s Research Hospital (SJCRH). The clinical trial protocols are listed on the left, and the corresponding trial IDs are as follows, Total 17: NCT03117751, AML16: NCT03164057, VENAML: NCT03194932, M16–106: NCT03181126, AML08: NCT00703820, SELHEM: NCT02212561. Please refer to https://clinicaltrials.gov for detailed information of trial protocols. Timepoint of admission to SJCRH/enrollment in trials are marked as blue lines. N/A, not on clinical trial.

The available immunophenotypic findings of the 20 patients are summarized in Table 2. The PICALM::MLLT10 cohort includes cases of several rare entities within the umbrella category of “Acute leukemia of ambiguous lineage” based on the ICC/WHO classifications. For example, case SJ031965 demonstrated no significant evidence of either myeloid or lymphoid lineage differentiation and is felt to be best classified as AUL; case SJ031798 demonstrated evidence of both B- and T-cell lineage differentiation (CD19+, cCD79a+, surface kappa light chain+, cCD3+) and is felt to be best classified as MPAL, T/B. Among the 8 T-ALL/LLy cases in our cohort of pediatric/young adult patients, 5 were gamma-delta T-ALL/LLy.

Table 2.

Immunophenotypic findings of the study cohort.

Sample	cyMPO	CD13	CD14	CD33	CD42b	CD1a	CD3	cyCD3	CD5	CD8	TCR⍺β	TCRɣδ	CD19	CD34	CD117	HLA-DR	TdT
SJ030459	-	-	-	+	-	-	-	-	+	-	-	-	subset	subset	subset	+	+
SJ030988	-	-	-	+	-	-	-	-	-	-	-	-	-	subset	-	+	-
SJ031965	-	-	-	-	-	-	-	-	subset	-	-	-	subset	subset	subset	-	-
SJ033011	+	-	-	+	-	-	-	-	-	-	-	-	-	-	-	+	-
SJ032188	N/A	+	N/A	+	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	-	+	+	+	N/A
SJ061255	-	-	-	+	-	-	-	-	-	-	-	-	-	+	+	+	-
SJ062998	-	subset	N/A	-	N/A	N/A	-	-	N/A	N/A	N/A	N/A	subset	+	subset	+	N/A
SJ061501 D	N/A	-	N/A	-	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	+	+	N/A	N/A
SJ061501 R	N/A	-	N/A	-	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	-	+	+	+	N/A
SJ074161	+	subset	subset	+	-	N/A	-	-	-	-	N/A	N/A	-	subset	-	subset	-
SJ033167	-	+	N/A	+	N/A	N/A	-	-	N/A	N/A	N/A	N/A	subset	+	+	+	-
SJ033585	+	+	N/A	+	N/A	N/A	-	-	N/A	N/A	N/A	N/A	subset	+	+	+	N/A
SJ030880	-	-	-	-	-	subset	+	+	+	+	-	+	-	-	-	-	subset
SJ030907	-	-	-	-	-	-	subset	subset	+	-	-	-	-	-	-	-	subset
SJ031662	-	-	-	-	-	-	+	+	+	+	-	+	-	-	-	-	+
SJ031798 *	-	-	-	-	-	-	+	+	+	+	-	+	+	-	-	-	-
SJ031904	-	-	-	-	-	+	+	+	+	-	-	+	-	-	-	-	+
SJ031944	-	-	-	+	-	-	-	subset	subset	-	-	-	-	+	+	+	-
SJ032955	N/A	N/A	N/A	N/A	N/A	+	-	N/A	+	+	N/A	N/A	-	-	N/A	N/A	N/A
SJ033522	N/A	-	N/A	-	N/A	+	-	+	+	+	N/A	N/A	-	-	-	-	+
SJ033817	-	-	-	-	-	-	+	+	+	+	-	+	-	-	+	-	-

^*Monotypic surface kappa light chain and cytoplasmic CD79a expression identified.

Abbreviations: N/A, not available; D, diagnosis sample; R, relapse sample;

Conventional cytogenetic analysis

Conventional cytogenetics analysis was performed on 8/11 non-ALL cases as part of the routine diagnostic workup, and the chromosome translocation t(10;11)(p12;q14) was detected in all 8 cases (Supplementary Table S1). Additional abnormalities were observed in all cases except one. The most frequent additional change was abnormalities affecting chromosome 17 with i(17q) found in 3 cases and add(17)(p11.2) in one, all of which resulted in 17p deletion containing TP53 gene locus. The other recurrent secondary chromosome abnormality seen in this cohort is trisomy 19.

Transcriptomic profiling of PICALM::MLLT10 positive acute leukemia

We sought to understand the transcriptomic profile of the PICALM::MLLT10-positive acute leukemias by evaluating the RNA-seq gene expression data of all 20 cases. Principle component analysis (PCA) stratified the cases into two groups, each containing 10 cases (Fig. 2A). Group 1 contained all PM-AML cases; and group 2 contained all PICALM::MLLT-T-ALL/LLy (PM-T-ALL/LLy) cases, one T/B-MPAL, and one AUL. Overall, the transcriptomic profiling indicates that PM-AML and PM-T-ALL/LLy belong to distinct transcriptomic groups. Despite the overall difference in gene expression profile, no HOXA cluster genes were differentially expressed between PM-AML and PM-T-ALL/LLy groups. Instead, we observed elevated expression of HOXA5, HOXA6, HOXA9, and HOXA10 in the vast majority of cases in both transcriptomic groups 1 and 2 compared to whole blood control (Supplementary Figure S1).

Figure 2. Transcriptome profiling of PICALM::MLLT10-positive samples and the comparison between PICALM::MLLT10-positive acute leukemia subgroups.

(A) Distribution of samples based on principal components (PCA) derived from expression data from RNA sequencing. Each dot represents one sample, and colors represent diagnosis. (B) Heatmap with hierarchical clustering across rows and columns showing row-normalized, z-score standardized expression levels (TPM) of significantly differentially expressed genes in each PICALM::MLLT10-positive acute leukemia subgroups in comparison to HSCs. Annotations in the upper panel refer to the expression of selected immunophenotypic markers corresponding to each sample evaluated by flow cytometry analysis. Diagnosis is depicted as colored segments that frame the top and bottom of the heatmap. (C) Pathway enrichment analysis for MSigDB Hallmarks gene sets. The 5 enriched pathways with highest significance are shown. Adjusted p-values are represented within color scale.

To examine in more detail the genes differentially expressed among subgroups, a differential gene expression (DGE) analysis was performed for each subgroup (PM-AML, PM-T-ALL/LLy, the AUL case and the T/B-MPAL case) compared to the control (CD34+CD38-CD45RA-EPCR+ human umbilical cord blood stem cells, “HSCs” for short) (13). Differentially expressed genes identified in each subgroup are listed in Supplementary Table S2). Significantly differentially expressed genes (fold change ≥ 4, adjusted p-value <0.05) for each subgroup were visualized across all samples in a hierarchically clustered heatmap (Figure 2B). Hierarchical clustering produced a similar stratification as observed through PCA, largely separating PM-AML and PM-T-ALL/LLy patients in two groups. Although most of the PM-AML cases cluster separately from PM-T-ALL/LLy, SJ062998 clusters with PM-T-ALL/LLy cases, whereas the ETP-ALL case SJ031944 clusters with PM-AML cases. Gene expression profiling in these two cases that appear to be outliers is likely compromised by the low blast percentage (<10%) in the test samples; besides, only a subset of blasts (~19% of blasts) in the ETP-ALL case expressing cytoplasmic CD3, a marker of T-lineage differentiation. The only case diagnosed with AUL (SJ031965) clustered within the AML group and the T/B-MPAL case (SJ031798) clustered within the T-ALL/LLy group. Although gene signatures could be derived to distinguish PICALM::MLLT10-positive acute leukemias from HSCs, unique signature could be hardly derived for subgroups, indicating genetic similarity across different immunophenotypic subgroups as well as heterogeneity within each subgroup.

Correlation between immunophenotypes and transcriptomic profiles

Immunophenotypic markers commonly used for diagnosis were integrated to assess the correlation with gene expression profiling. CD1A-positive cases appear to group together both in the clustering heatmap and in the unsupervised principal component analysis (PCA) (Fig. 2A and 2B). However, these CD1A-positive T-ALL/LLy cases demonstrated heterogeneity according to other immunophenotypic markers. Two out of the three CD1A-positive cases appeared double positive for CD4 and CD8 but lacked surface CD3, corresponding to the double positive thymocyte. The remaining CD1A-positive case appears to be CD4+, CD8-, with surface CD3 and TCRɣδ expression. No apparent clustering was identified for other immunophenotypic markers, including TCRɣδ. However, the evaluation of association is limited by the number of cases in this study cohort and limited workup in subset of cases.

Functional classification of differentially expressed genes in each subgroup

Functional classification of significantly upregulated genes compared to HSCs (fold change ≥ 4, adjusted p-value <0.05, TPM ≥1) in each subgroup by pathway enrichment analysis highlighted cell proliferation-related MSigDB Hallmark gene sets, G2M checkpoint and E2F targets, for all four subgroups (Figure 2C, Supplementary Table S3). The enrichment of both gene sets in all four subgroups was further supported by gene set enrichment analysis (GSEA) (Supplementary Table S3 and Supplementary Figure S2). GSEA analysis was extended to include the Gene Ontology Biological Function gene sets and demonstrated that genes associated with T cell receptor signaling and immune response process were highly enriched in PM-T-ALL/LLy and the T/B-MPAL case but had no significant correlation with PM-AML or the AUL case (Supplementary figure 2). Refer to Supplementary Table S3 for all gene sets significantly enriched.

Subtype-specific molecular profiling in PM-AML and PM-T-ALL/LLy

Comprehensive genomic profiling was available for 19 out of 20 patients who had samples available for DNA sequencing (Supplementary Table S1). The remaining one AML case had no DNA for genomic profiling, and SNVs and small insertion-deletions (indels) were determined on selected genes (PHF6, TP53, NF1, SUZ12, EZH2 and PTPN11) based on the RNA-seq data as previously reported (14). Overall, PHF6 was the most frequently mutated gene in this study cohort, seen in 12 patients, including 7 transcriptomic group 1 (PM-AML) patients and 5 group 2 patients (3 PM-T-ALL/LLy, 1 T/B-MPAL and 1 AUL). PHF6 is located on the X-chromosome. Among the 12 patients with PHF6 alteration, 11 patients had a complete loss of PHF6, including hemizygous loss of function (LOF) alterations in 10 male patients and bi-allelic LOF in one of the remaining female patients diagnosed with AML, SJ033011 (Fig. 3).

Figure 3. Mutational landscape of PICALM::MLLT10-positive acute leukemia.

Recurrent clinical relevant variations identified in the 21 PICALM::MLLT10-positive samples (two time points for patient SJ061501; D, diagnosis; R, relapse) were presented in the OncoPlot. Upper panel: Large-scale copy number variants (LS-CNVs) revealed by WGS and/or karyotype analysis. SJ074161, SJ033167, and SJ032955 didn’t have whole genome sequencing or conventional cytogenetics performed and therefore data are denoted blank for LS-CNV events. Middle panel: Integrated Alterations. When a given gene was affected by more than one hit, the overall biological effect was presented in the figure. Except KRAS and FBXW7, genes with variants present in at least two cases are included. Abbreviations: LOF, loss of function; GOF, gain of function.

Alterations affecting TP53 and NF1 were hallmarks of PM-AML, seen in 8/10 (80%) and 7/10 (70%) of PM-AML patients, respectively (Fig. 3). Notably, in patient SJ061501 who had both diagnostic and relapsed samples available for testing, TP53 alteration as well as 17p deletion were only detected in the relapsed sample. Aside from TP53 and 17p deletion, all other abnormalities shared by both the diagnostic and relapsed samples were present at similar variant allele frequency in this patient (Supplementary Table S4). Among the eight PM-AML patients’ samples with TP53 alterations, inactivation of both TP53 alleles by “two hits” (one mutation coupled with 17p deletion/copy neutral LOH or two in trans mutations) was observed in five. Similarly, among the seven PM-AML with NF1 alterations, bi-allelic LOF was identified in four cases. In addition to TP53 and NF1, we also observed deletion or LOF alterations of SUZ12 in 60% (6/10) of PM-AML cases, of which one case had biallelic LOF of SUZ12. SUZ12 and NF1 are located ~600 Kb apart at 17q12 and these two genes are often affected by one focal segmental deletion in this cohort. Mutations in PTPN11 (3/10 patients, 30%) are also relatively common in PM-AML.

On the contrary, beyond PHF6 the most recurrent alterations in PM-AMLs were rarely seen in PM-T-ALL/LLy cases. Genetic alterations affecting TP53 were observed in only one case (SJ031904), which was a subclonal 17p deletion event. None of PM-T-ALL/LLy had bi-allelic LOF of NF1, and only one case (SJ031662) had a hemizygous deletion affecting NF1. Indeed, a completely different molecular signature was observed in PM-T-ALL/LLy. This group were specifically enriched for alterations in EZH2 (4/8, 50% versus 10% in PM-AMLs) as well as activating mutations in NOTCH1 signaling (4/8, 50%), followed by alterations activating JAK-STAT pathway (3/8, 37.5%) and biallelic LOF of CDKN2A (3/8, 37.5%) (Fig.3 and Supplementary Table S4).

Subtype-specific large-scale copy number variations in PM-AML and PM-T-ALL/LLy

In addition to gene-level alterations, we evaluated large scale (>5 Mb) copy number variations (LS-CNVs) of 18 cases, in which WGS and/or conventional cytogenetics were performed. Integrating the WGS and karyotype findings, we demonstrated that the most frequent LS-CNV was 17p deletion and copy neutral loss of heterozygosity (CN-LOH), seen in 7 patients (6 PM-AMLs and one PM-T-ALL/LLy, SJ031904, as a subclonal event). More than half of the cases with 17p deletion were presented as isochromosome 17q, which also resulted in a gain of 17q. The next frequent CNV event was 12p deletion, in particular involving 12p13, seen in 5 cases (4 PM-AMLs and one PM-T-ALL/LLy, SJ031904). In contrast to 17p deletion/LOH and 12p deletion that were highly associated with PM-AML, 9p deletion/LOH were exclusively seen in PM-T-ALL/LLy (Fig. 3).

Mutation and gene expression profiling of the AUL case

Global gene expression analysis clustered the AUL case SJ031956 between AML and T-ALL/LLy, but slightly toward the PM-T-ALL/LLy group (Fig. 2A), while genomic profiling revealed PHF6 inactivation as well as biallelic LOF of NF1 and SUZ12 in this case. As aforementioned, biallelic LOF of NF1 was exclusively seen in PM-AMLs in contrast to PM-T-ALL/LLy in our study. In the hierarchically clustered heatmap (Fig. 2B), this case clustered close to other cases of PM-AML that harbor alterations in NF1 and/or SUZ12. Together, the genomic and transcriptomic findings are compatible with the ambiguous immunophenotypic characteristic of this case. Functional classification of genes significantly upregulated in SJ031956 showed similar results as seen for other subgroups, i.e. enrichment in gene sets associated with cell proliferation (G2M checkpoint and E2F targets) (Supplementary figure 2, Supplementary table 3).

Mutation and gene expression profiling of the T/B-MPAL case

The T/B-MPAL case SJ031798 clustered within the PM-T-ALL/LLy group on PCA, but showed a greater variance according to PC2, suggesting that while similar to PM-T-ALL/LLy in terms of the dominant pattern represented by PC1, there are additional variances in gene expression profiling distinguished SJ031798 from PM-T-ALL/LLy. Nevertheless, functional classification of significantly upregulated genes in SJ031798 and GSEA demonstrated similar features between SJ031798 and PM-T-ALL/LLy cases. In addition, genomic profiling revealed heterozygous mutations in PHF6, EZH2, and JAK1, as well as a biallelic missense variant in STAT5B that is associated with a CN-LOH of chromosome 17q, all of which were recurrent alterations seen in PM-T-ALL/LLy. Interestingly, the T/B-MPAL case is distinguished from all other cases in our study cohort by the extremely low expression of HOXA9 and HOXA10 (Supplementary Figure S1); however, this observation is based solely on one T/B-MPAL case and remains to be evaluated in future studies.

Discussion

The PICALM::MLLT10 fusion is a rare genetic driver in acute leukemias of a variety of lineages. In this study, we presented clinical characteristics and comprehensive genomic and transcriptomic profile of a cohort of 20 paediatric acute leukemia patients (approximately half AML and half T-ALL/LLy). Our clinical observations are largely in consistent with those of a recent international study, in which the 5-year overall survival rates for patients with PICALM::MLLT10-positive ALL and AML were 76% and 26%, respectively (15). In the present study, 7 of 8 PM-T-ALL/LLy patients are alive, compared to only 2 of 10 PM-AML patients. One notable difference was the presence of extramedullary disease in 2 of 35 AML patients in the international study (15) and in 7 of 10 AML patients in the present study. It is also intriguing to note that the 5 gamma delta T-ALLs with PICALM::MLLT10 rearrangement in our cohort did not demonstrate refractory/relapse diseases despite the prior study (16)suggests pediatric gamma delta T-ALLs tend to show induction failure as well as inferior survival. There is likely genetic heterogeneity that can help further stratify pediatric gamma delta T-ALL. In a prior study focusing on T-ALLs with PICALM::MLLT10 fusion (8), the majority of the 10 pediatric gamma delta T-ALL patients also showed complete remission at the time of report except for 1 case with early mortality during induction and 1 patient who died of GVHD after transplant.

More importantly, our complete genomic characterization extends the recent findings of Mark et al. (15) and provides new biologic insights into this rare leukemia. First, global gene expression profiling indicated that PM-AML and PM-T-ALL/LLy belong to two distinct transcriptomic groups, respectively. Furthermore, our study uncovered previously unappreciated mutational signatures in PM-AML and PM-T-ALL/LLy. In general, PHF6 was the most frequently mutated gene in this study cohort regardless of the immunophenotype. Despite sharing a common driver gene fusion, activation of HOXA genes and the recurrence of PHF6 alteration, PM-AML and PM-T-ALL/LLy have remarkably different spectra of cooccurring mutations. By contrast, alterations affecting TP53, NF1, SUZ12 were almost exclusively seen in PM-AML, whereas recurrent genes/pathways altered in PM-T-ALL/LLy (EZH2, NOTCH, JAK-STAT, CDKN2A) were rarely affected in PM-AML.

Located on the X chromosome, PHF6 belongs to the plant homeodomain (PHD)-like finger (PHF) family and plays a role in regulating chromatin accessibility (17). PHF6 loss of function alterations are often implicated in T-cell ALL and less frequently in AML and other myeloid neoplasms (18). In T-ALL, PHF6 is one of the most frequently mutated genes, presenting in ~20% of pediatric T-ALL, 19–40% of adult patients with T-ALL and ~25% of adults with T-lymphoblastic lymphoma (T-LLy) (18,19). Wendorff et al. demonstrated that loss of PHF6 is an early mutational event in leukemia transformation and enhances the full disease development of T-ALL when cooperating genetic alterations are present (20). Of note, PHF6 appears to play a role in lineage plasticity within hematopoietic malignancies (17) and PHF6 mutations commonly present in mixed phenotype acute leukemias with a predilection for T-lineage marker expression (21). In contrast, the overall frequency of PHF6 mutations in AML is only 2–3% (18,22). In this study, we demonstrated high frequency of PHF6 LOF alterations in PICALM::MLLT10-positive acute leukemias. In the PM-T-ALL/LLy group, the frequency is slightly higher than the overall frequency of PHF6 in pediatric T-ALL (37.5% vs. ~20%); in PM-AML, a strikingly high frequency was observed (70%), suggesting that PHF6 alterations may play a key role in cooperating with the PICALM::MLLT10 fusion oncoprotein to disturb hematopoietic differentiation and force leukemia transformation. A connection between PHF6 alterations and PICALM::MLLT10 positivity may also explain the predominance of male patients in this study as well as several previous clinical investigations (5, 6).

TP53 is the most frequently altered tumor suppressor gene in human cancers (23), and TP53 alterations in hematological malignancies have been characterized in multiple studies (24–26). In pediatric patients, the incidence of TP53 alterations is only 1% in de novo AML and 6.6% in relapsed AML (1); whereas TP53 is not considered to be a recurrently mutated gene in pediatric T-ALL (19). In this study, our data clearly showed that TP53 alterations are highly enriched in PM-AMLs and biallelic/multi-hit TP53 alterations are predominantly associated with relapsed/refractory PM-AML (4/6 relapsed/refractory samples versus 1/5 newly diagnosed samples). In addition, in one patient (SJ061501) who had both diagnostic and relapse samples studied, TP53 alterations were only detected in the relapsed sample, presenting as a biallelic event due to a mutation and a 17p deletion acquired in the relapsed sample. Taken together, we speculate that TP53 alterations are acquired in PM-AML tumor and critically cooperate with PICALM::MLLT10 fusion to facilitate AML tumor progression.

Another hallmark of the PM-AML cohort in this study is NF1 alterations. Previous studies have shown that NF1 alterations are present in ~5% of all AML and further enriched in complex karyotype AML (27). In this study, we demonstrated that NF1 is highly enriched in PM-AML and frequently associated with TP53 alterations. Among the seven cases with NF1 alterations, four showed biallelic/multi-hit LOF of NF1 and surprisingly, three out of four also had biallelic/multi-hit alterations in TP53. Biallelic/multi-hit alterations in both TP53 and NF1 were exclusively identified in relapsed/refractory samples in the PM-AML cohort, which may further explain the extremely poor outcome of these patients.

Both EZH2 and SUZ12 are essential factors of the PRC2 complex which is the “writer” of a major repressive chromatin modification, H3K27me3 (28). Loss of function alterations affecting EZH2 and SUZ12 have been reported in 25% of T-ALL and shown to be frequently associated with oncogenic NOTCH1 mutations (29). Interestingly, in a previous study of 15 PICALM::MLLT10 acute leukemia cases that are dramatically biased for T-ALL (12 T-ALL, 2 AML and one MPAL), EZH2 mutations were identified in 27% of PICALM::MLLT10 fusion positive cases and in 0% of the comparison group of 10 PICALM::MLLT10 fusion negative T-ALL in their study (30). In our current study, we demonstrated that EZH2 alteration is highly associated with PM-T-ALL/LLy; in contrast, SUZ12 is associated with PM-AML and one of the PM-AML cases (SJ032188) had biallelic LOF of SUZ12. Of note, biallelic LOF of SUZ12 was identified in the AUL case as well. Overall, our findings suggest cooperation between PICALM::MLLT10 fusion and PRC2 loss of function in leukemic transformation and specifically, a strong cooperation between EZH2 and PM-T-ALL/LLy and an association of SUZ12 LOF and PM-AML. A reason for the preference for different PRC2 components in different entities is not clear and will require further studies to elucidate the underlying mechanism.

Although the exact mechanism by which PICALM::MLLT10 fusion oncoprotein exerts its oncogenic potential is not fully understood, mouse model studies have shown that the expression of PICALM::MLLT10 led to acute leukemia with Hoxa cluster overexpression in both T-cell and myeloid tumors (31). We observed elevated expression of HOXA5, HOXA6, HOXA9 and HOXA10 in the vast majority of cases across the study cohort, which is consistent with previous studies. Preclinical studies have demonstrated that small-molecule inhibition of menin can induce remission and, in some cases, eradicate disease, in mouse models of AML subtypes that are associated with overexpression of HOXA9 and MEIS1, including KMT2A-rearranged, NPM1-mutated, NUP98-rearranged, and UBTF-TD-positive AML. Although menin inhibitors have not been tested in models of PICALM::MLLT10-positive AML, the high expression of HOXA9 in this subtype, as well as a recent report demonstrating the safety and activity of the menin inhibitor revumenib (32) suggest that menin inhibitors should be evaluated in this patient population.

In addition to HOXA activation, our transcriptomic profiling and differential gene expression analysis identified that cell proliferation-related gene sets, G2M checkpoint and E2F targets, are highly associated with acute leukemia positive for PICALM::MLLT10. In particular, XPO1 is one of the few core enriched genes shared by both G2M checkpoint and E2F target pathways, and significantly higher expression of XPO1 was detected in both PM-AML and PM-T-ALL/LLy cases compared to HSCs (Supplementary Figure S3). XPO1 (also known as CRM1) encodes a protein that mediates nuclear export signal (NES)-dependent protein transport. Recent studies have demonstrated that PICALM possesses an XPO1-dependent nuclear export signal (NES) that is preserved in the oncogenic PICALM::MLLT10 fusion and is required for the interaction of the PICALM::MLLT10 fusion protein with the nuclear export receptor XPO1 to activate the expression of the HOXA genes (33,34). Based on an in vitro cell model study, Conway et al. also demonstrated that XPO1 inhibition prevented the enrichment of PICALM::MLLT10 at HOXA loci, resulting in the loss of HOXA gene transcription. Small molecule inhibitors of nuclear export (SINEs) targeting XPO1 have been developed and tested in preclinical studies in various tumor types and showed promising therapeutic efficacy (35–41). It is worth noting that two PM-AML patients in this cohort (SJ062998, SJ061255) were treated by a SINEs, Selinexor. SJ062998 is the only long-term survivor among the group of PM-AML patients, and SJ061255 passed away due to infection related complications after hematopoietic stem cell transplantation. The utility of XPO1 inhibitors as a novel therapeutic approach for patients with PICALM::MLLT10 fusion is worth further investigation.

In addition, CD33 expressing blasts are found in a wide variety of AMLs and therefore, therapeutic strategies aimed at this promising target have been investigated. Gemtuzumab Ozogamicine (GO; mylotarg), an anti-CD33 immunoconjugate, was an FDA-approved treatment for newly diagnosed and relapse/refractory AML with CD33 expression and usually used in combination with chemotherapy (42,43). In our cohort, 8/10 (80%) of the PM-AML cases showed expression of CD33 (dim to bright), making CD33 also a potential therapy target in this subgroup of AML.

The dismal prognosis of AML patients with PICALM::MLLT10 and the lack of benefit of allogeneic hematopoietic stem cell transplantation (15) indicate that novel agents are urgently needed. In this study, exploring the transcriptomic and genomic profile of PICALM::MLLT10 fusion positive acute leukemias, we demonstrated subtype-specific molecular signatures of PICALM::MLLT10 fusion leukemias, which could be potentially used as adjunctive evidence for molecular classification and risk assessment of PICALM::MLLT10-positive acute leukemias, and gained new biologic insights into this rare leukemia. Besides HOXA activation, our study suggested that XPO1, PHF6, TP53, NF1, EZH2, and SUZ12 may play important roles in PICALM::MLLT10-associated leukemogenesis, either fundamentally or in a lineage-specific manner. Future studies are required to elucidate their potential cooperation with PICALM::MLLT10 fusion in molecular pathogenesis of this rare leukemia and assess the therapeutic potential of targeting XPO1 and menin in acute leukemias positive for PICALM::MLLT10.

Data Availability

The next generation sequencing data for the current study are available from the corresponding author on reasonable request through St. Jude Cloud (https://www.stjude.cloud/).