Pathologic, Cytogenetic, and Molecular Features of Acute Myeloid Leukemia with Megakaryocytic Differentiation: a Report from the Children’s Oncology Group

Abstract

Background:

Acute myeloid leukemia (AML) with megakaryocytic differentiation (AMkL) is a rare subtype of AML more common in children. Recent literature has identified multiple fusions associated with this type of leukemia.

Methods:

Morphology, cytogenetics, and genomic sequencing were assessed in patients from Children’s Oncology Group trials AAML0531 and AAML1031 with central-pathology review confirmed non-Down syndrome AMkL. The 5-year EFS, OS, and RR were evaluated in these AMkL subcategories.

Results:

A total of 107 cases of AMkL (5.5%) were included. Distinct fusions were identified in the majority: RBM15::MRTFA (20%), CBFA2T3::GLIS2 (16%), NUP98 (10%), KMT2A (7%), TEC::MLLT10 (2%), MECOM (1%), and FUS::ERG (1%); many of the remaining cases were classified as AMkL with (other) myelodysplasia-related changes (MRC). Very few cases had AML-associated somatic mutations. Cases with CBFA2T3::GLIS2 were enriched in trisomy 3 (p=0.015) and the RAM phenotype with associated high CD56 expression (p<.001). Cases with NUP98 fusions were enriched in trisomy 6 (p<0.001), monosomy 13/del(13q) (p<0.001), trisomy 21 (p=0.026), and/or complex karyotypes (p=0.026). While different 5-year EFS and OS were observed in the AMkL in each trial, in general, those with CBFA2T3::GLIS2 or KMT2A rearrangements had worse outcomes compared to other AMkL while those with RBM15::MRTFA or classified as AMkl-MRC fared better. AMkL with NUP98 fusions also had poor outcomes in the AAML1031 trial.

Conclusion:

Given the differences in outcomes, AMkL classification by fusions, cytogenetics, and morphology may be warranted to help in risk stratification and therapeutic options.

INTRODUCTION

Acute myeloid leukemia (AML) with megakaryocytic differentiation (AMkL) represents less than 5% of all AML, and is defined as a leukemia with at least 20% blasts of which ≥50% are of megakaryocyte lineage.¹ This leukemia has a bimodal age distribution with peaks in children less than 3 years of age and in older adults.^2–4 Many childhood cases of AMkL are associated with Down syndrome (trisomy 21), which is classified as a separate entity in the World Health Organization (WHO) classification of Tumours of Haematopoietic and Lymphoid Tissues as Myeloid leukemia associated with Down syndrome.⁵ Excluding those with Down syndrome, AMkL accounts for 5.9–12% of AML in children, with median ages of onset of ranging from 1.4–1.8 years.^6–14 Bone marrow fibrosis is a common finding in this leukemia subtype.^2,15–21

Prior to more advance molecular methods, AMkL not associated with Down syndrome was most often linked to the translocation t(1;22)(p13;q13) (RBM15::MRTFA previously known as MKL1). Additional rare cases of AMkL have KMT2A (MLL) translocations.^9,22–24 However, with the advent of reverse transcriptase polymerase chain reaction and next generation sequencing, additional fusion proteins in AMkL have been identified. These cryptic abnormalities include inv(16)(p13.3q24.3) (CBFA2T3::GLIS2), and t(5;11)(q35;p15) (NUP98::KMD5A).^7,8,25 These four fusions are now thought to represent the most common rearrangements in AMkL.⁸

Childhood AMkL, excluding cases associated with Down syndrome, has an inferior overall survival compared to other categories of AML,^6,12–14,26 with worse prognoses in AMkL with CBFA2T3::GLIS2, NUP98::KMD5A, and KMT2A rearrangements.^8,25–27 The prognostic significance of RBM15::MRTFA is unclear with studies showing better,^3,7,12 worse,^13,17 or equal^8,9,11 outcomes compared to fusion-negative cases. Herein, our objective was to study the morphologic, immunohistochemical, cytogenetic, and molecular features of non-Down Syndrome AMkL in cases from the Children’s Oncology Group (COG) trials AAML0531 and AAML1031 with the goal of correlating the cytogenetics and molecular classifications with outcomes.

METHODS

Patients:

Pediatric and young adults ranging from 1 month to 29.99 years of age with de novo AML and without Down syndrome were eligible for the 2 Phase III randomized COG trials and analyzed in this study. AAML0531 evaluated Gemtuzumab Ozogamicin (GO) with a dose-intensive treatment regimen; results have previously been described.²⁸ The trial included 1022 eligible patients, enrolled between August 2006 and June 2010 from 181 participating institutions.²⁸ AAML1031 compared standard chemotherapy with or without bortezomib and employed sorafenib in patients with high FLT3 internal tandem duplication (ITD) allelic ratios (>0.4).²⁹ This trial risk stratified based upon minimal residual disease (MRD), FLT3 ITD allelic ratio, NPM1 mutations, CEBPα mutations, and other prognostic genetic markers. The trial included 1231 eligible patients, who enrolled between June 2011 and July 2017 from 193 participating institutions. Herein, outcome data were reviewed separately for each trial. Notably, a subset of patients included in this manuscript was previously described in pediatric AMkL/AML cohort studies.^7,8,30 All participating institutions had approval by their institutional review boards (IRB) for these trials. According to institutional regulations, all patients or their parents gave written informed consent before entering this study. The studies were conducted in accordance with the Declaration of Helsinki.

Data availability statement:

The data that supports the findings of this study are available in Supporting Table S1 of this article.

Morphologic assessment:

All cases with an institutional or central pathology review (CPR) diagnosis of AMkL were identified from the AAML0531 and AAML1031 databases (n=1935). Cases were excluded from this study if CPR was not performed or the material submitted was not sufficient for diagnosis. A diagnosis of AMkL required ≥20% blasts of which ≥50% were of megakaryocyte lineage with expression of ≥1 megakaryocytic antigen by flow cytometry or immunohistochemistry. For analyses, morphologic AMkL were classified as AMkL, not otherwise specified (NOS), AMkL with genetic abnormalities, and AMkL with myelodysplasia-related changes (MRC) per the 2017 WHO hematopoietic tissue classification.^1,31,32 CPR evaluated blood smears, bone marrow aspirates and biopsies, flow cytometry reports, and any immunohistochemical stains to confirm the diagnosis. Slides were re-reviewed for multilineage dysplasia and abnormal megakaryocyte maturation. If available, reticulin and collagen stains were reviewed or performed to grade bone marrow fibrosis (MF-0 to MF-3) per standardized guidelines.^33,34

Cytogenetic and molecular assessment:

All cytogenetic results for this study were centrally reviewed and recorded using International System of Human Cytogenetic Nomenclature. The number of cytogenetic abnormalities and/or presence of any recurrent translocations were recorded. Screening for FLT3 ITD, NPM1, CEBPα, and WT1 mutations was performed as previously described.^35–38 RNA-sequencing was performed to identify fusion transcripts using total RNA extracted from patient samples using AllPrep DNA/RNA/miRNA Universal Kit (QIAGEN, Valencia, CA, #80224), by the QIAcube system. The ribodepletion 2.0 protocol (British Columbia Genome Sciences Centre, Vancouver, BC) was employed to prepare the mRNA libraries with 75-bp strand-specific paired-end sequencing. STAR-Fusion v1.1.0 fusion detection algorithm was used, running default parameters with the pre-made GRCh37 resource library with Gencode v19 annotations (https://data.broadinstitute.org/Trinity/CTAT_RESOURCE_LIB/).³⁹ TransABySS v1.4.10 fusion detection algorithm was established to record fusions with breakpoint reads ≥ 1, flanking pairs ≥ 2 counts, and spanning reads ≥ 2 counts.⁴⁰ The dbGaP TARGET: Acute Myeloid Leukemia study (Accession: phs000465.v19.p8) displays the transcriptomic data.³⁰

Outcome assessment and statistical analysis:

AAML0531 and AAML1031 data were current as of September 30, 2018 and June 30, 2021, respectively. EFS and OS were determined employing the Kaplan-Meier method where EFS was defined as time from study entry until failure to achieve complete remission (CR) during induction, relapse or death, and OS was defined as time from study entry to death.⁴¹ Relapse risk (RR) was calculated by cumulative incidence methods defined as time from end of induction I for patients in CR to relapse or death; deaths without a relapse were considered competing events.⁴² Induction I failures were defined as patients who withdrew from therapy due to a) relapse, b) persistent central nervous system disease, and/or c) refractory disease (≥20% bone marrow blasts). Any patient lost to follow-up was censored at their date of last known contact. Log-rank statistic (EFS and OS) and Gray’s statistic (RR) tested the significance of predictor variables. Potential covariates considered were age at diagnosis, morphologic classifications, presence of fibrosis, certain karyotypic abnormalities, and identified fusions. As some of the above subgroups had small numbers, the comparisons were ad hoc analyses. The chi-squared test was employed to test the significance of observed differences in proportions, Fisher’s exact test was used when data were sparse, and Student’s t-test was utilized to compare means and distributions of 2 groups. P-values <0.05 were considered statistically significant.

RESULTS

Patient Characteristics:

A total of 107 AMkL were confirmed by CPR in these trials, accounting for 5.5% of all AML cases which underwent CPR. Table 1 lists the patient demographics of these cases. There were no significant differences in age, gender, race, or ethnicity in AMkL patients between AAML0531 and AAML1031 (Supplemental Table S2). The median age at diagnosis of AMkL was 1.45 years (range 0.08–15.10 years, interquartile range 1.20 years), significantly younger than patients diagnosed with other subtypes of AML in these trials (p<0.001).

Table 1.

Comparison of pathologic findings between subtypes of AMkL

Subtype	All cases	AMkL, NOS	AMkL, unknown cytogenetics	AMkL with RBM15::MRTFA	AMkL with CBFA2T3::GLIS2	AMkL with NUP98 fusion	AMkL with KMT2A fusion	AMkL with MECOM fusion	AMkL with FUS::ERG	AMkL with TEC::MLLT10	AMkL with (other) MRC
Number of cases	107	11	5	21	17	11	8	1	1	2	30
AAML0531:AAML1031	48:59	4:7	3:2	7:14	7:10	3:8	3:5	0:1	1:0	0:2	20:10 ^
Age Range (years)	0.08–15.10	0.53–6.17	0.84–2.79	0.08–2.49#	0.77–2.07	0.44–13.23	0.87–6.25	0.80	3.54	1.50–12.67	0.11–15.10
Mean	2.12	1.91	1.57	0.91	1.40	3.01	3.00	0.80	3.54	7.08	2.64
Median	1.45	1.95	1.18	0.61	1.35	2.25	3.08	0.80	3.54	7.08	1.48
Male:Female	48:59	5:6	0:5	6:15	7:10	10:1@	4:4	0:1	0:1	0:2	16:14
Fibrosis, any	30/39 (77%)	3/3 (100%)	2/2 (100%)	8/8 (100%)	3/4 (75%)	N/A	4/4 (100%)	N/A	0/1 (0%)	N/A	10/17 (59%)
Fibrosis, MF-2 or MF-3	18/39 (46%)	1/3 (33%)	2/2 (100%)	5/8 (63%)	0/4 (0%)	N/A	4/4 (100%)	N/A	0/1 (0%)	N/A	6/17 (35%)
Abnormal megakaryocytes	23/43 (53%)	4/4 (100%)	1/2 (50%)	2/6 (33%)	3/6 (50%)	2/3 (67%)	1/3 (33%)	N/A	1/1 (100%)	N/A	9/18 (50%)
Multilineage dysplasia	2/45 (4%)	0/4 (0%)	0/2 (0%)	0/5 (0%)	0/6 (0%)	0/4 (0%)	0/3 (0%)	N/A	1/1 (100%)	N/A	1/20 (5%)
Karyotype
Normal	20/100 (20%)	7/11 (64%)*	N/A	2/20 (10%)	6/17 (35%)	0/11 (0%)	3/8 (38%)	N/A	0/1 (0%)	1/2 (50%)	1/30 (3%)
Trisomy 3	7/100 (7%)	0/11 (0%)	N/A	0/20 (0%)	4/17 (24%)*	1/11 (9%)	2/8 (25%)	N/A	0/1 (0%)	0/2 (0%)	0/30 (0%)
Del(5q) / monosomy 5	3/100 (3%)	0/11 (0%)	N/A	0/20 (0%)	0/17 (0%)	1/11 (9%)	0/8 (0%)	N/A	0/1 (0%)	0/2 (0%)	2/30 (7%)
Trisomy 6	16/100 (16%)	0/11 (0%)	N/A	3/20 (15%)	0/17 (0%)	7/11 (64%)*	3/8 (38%)	N/A	0/1 (0%)	0/2 (0%)	3/30 (10%)
Monosomy 7/del(7q)	5/100 (5%)	0/11 (0%)	N/A	0/20 (0%)	0/17 (0%)	0/11 (0%)	1/8 (13%)	N/A	0/1 (0%)	0/2 (0%)	4/30 (13%)*
Trisomy 8	11/100 (11%)	0/11 (0%)	N/A	0/20 (0%)	0/17 (0%)	2/11 (18%)	2/8 (25%)	N/A	0/1 (0%)	0/2 (0%)	7/30 (23%)*
Del(9q)	7/100 (7%)	0/11 (0%)	N/A	0/20 (0%)	0/17 (0%)	2/11 (18%)	0/8 (0%)	N/A	0/1 (0%)	0/2 (0%)	5/30 (17%)*
Monosomy 13/del(13q)	9/100 (9%)	0/11 (0%)	N/A	0/20 (0%)	0/17 (0%)	8/11 (73%)*	0/8 (0%)	N/A	0/1 (0%)	0/2 (0%)	1/30 (3%)
Any chr13 structural ab.	17/100 (17%)	0/11 (0%)	N/A	3/20 (15%)	0/17 (0%)	9/11 (82%)*	0/8 (0%)	N/A	0/1 (0%)	0/2 (0%)	5/30 (17%)
Trisomy 19	16/100 (16%)	1/11 (9%)	N/A	4/20 (20%)	1/17 (6%)	1/11 (9%)	3/8 (38%)	N/A	0/1 (0%)	0/2 (0%)	6/30 (20%)
Trisomy 21	25/100 (25%)	1/11 (9%)	N/A	4/20 (20%)	3/17 (18%)	6/11 (55%)*	4/8 (50%)	N/A	0/1 (0%)	0/2 (0%)	7/30 (23%)
Complex (≥3 abnormalities)	49/100 (49%)	0/11 (0%)&	N/A	6/20 (30%)	1/17 (6%)&	9/11 (82%)*	4/8 (50%)	N/A	0/1 (0%)	1/2 (50%)	28/30 (93%)*
Notable Immunophenotype
RAM phenotype	18/101 (18%)	2/11 (18%)	0/4 (0%)	0/19 (0%)	16/17 (94%)*	0/11 (0%)	0/8 (0%)	0/1 (0%)	0/1 (0%)	0/2 (0%)	0/27 (0%)
CD56	24/47 (51%) ^	3/4 (75%)	1/3 (33%)	0/4 (0%)	17/17 (100%)*	0/2 (0%)	1/3 (33%)	N/A	N/A	N/A	2/14 (14%)

Ab = aberration; AMkL = acute myeloid leukemia with megakaryocytic differentiation; MRC = myelodysplasia-related changes (includes multilineage dysplasia and MDS-related karyotype)

^{^}Significantly different between AAML0531 and AAML1031

^#Significantly younger in age than other subtypes

^@Significantly more males than females compared to other subtypes

^*Significantly more common than other subtypes

^&Significantly less common than other subtypes

Leukemia morphologic, immunophenotypic, and fusion-based classifications:

By morphology, megakaryoblasts are large cells with high nuclear-to-cytoplasmic ratios, round to slightly irregular nuclear contours, fine chromatin, prominent nucleoli (sometimes multiple) and basophilic cytoplasm sometimes having pseudopod or bleb formation (Supplemental Figure S1). Table 1 details the clinical, pathologic, and cytogenetic features associated with the fusion-based classifications (see also Supplemental Tables S3 and S4). A total of 61 cases (57%) had defined fusions including 21 with RBM15::MRTFA, 17 with CBFA2T3::GLIS2, 11 with NUP98 fusions, 8 with KMT2A fusions, 2 with TEC::MLLT10, 1 with a MECOM fusion, and 1 with FUS::ERG (Figure 1). An additional 30 cases (28%) without the aforementioned fusions had either multilineage dysplasia (n=1) or MDS-related cytogenetic abnormalities (n=29), qualifying them for a diagnosis of AMkL-MRC using 2017 WHO criteria. Lastly, 11 cases were classified as AML-NOS, and 5 could not be classified due to unknown cytogenetics.

Figure 1. Fusions present in acute myeloid leukemia with megakaryocytic differentiation.

In this study, 61 of 102 patients with available cytogenetic data had identified fusion proteins.

By flow cytometry and/or immunohistochemistry, the blasts expressed ≥1 megakaryocytic antigens, including CD61 (n=90/92), CD41 (n=60/61), and CD42b (n=37/38). Other variably expressed antigens included CD13 (n=37/60), CD33 (n=72/85), CD34 (n=44/73), CD117 (n=40/60), HLA-DR (n=23/52), CD71 (n=21/24), CD4 (n=34/50), and CD7 (n=37/54). CD56 was also expressed in 24/47 (51%) tested cases, most associated with CBFA2T3::GLIS2 (p<0.001). The RAM phenotype, defined by bright CD56, dim/negative CD45 and CD38, and negative HLA-DR,⁴³ was noted in 18/101 cases, and was significantly more common in those with CBFA2T3::GLIS2 (p<0.001).

Multilineage dysplasia was identified in 2/45 (4%) of evaluated cases, including the case with FUS::ERG. However, 23/43 (53%) marrows assessed for megakaryocytic maturation demonstrated abnormal megakaryocytes, including micromegakaryocytes and forms with separate nuclear lobes (Supplemental Figure S2). These cases were not restricted to any specific subgroup. Bone marrow aspirates in AMkL cases were often hemodilute due to marrow fibrosis. Of the 39 bone marrow biopsies with reticulin staining, 77% had at least mild fibrosis (≥MF-1) (Table 1, Supplemental Figure S2).

Cytogenetic and molecular characterization:

Of the 107 AMkL, 100 had karyotype data and 2 had FISH or molecular data that allowed classification into the categories depicted in Table 1. Normal karyotypes were identified in 18 cases (18%), but 10 of these cases (56%) had molecularly-identified cryptic translocations (6 with CBFA2T3::GLIS2, 2 with RMB15::MRTFA, and one each of KMT2A::MLLT3 and TEC::MLLT10). CBFA2T3::GLIS2 were more commonly associated with trisomy 3 compared to other classifications (p=0.015); an additional case with CBFA2T3::GLIS2 demonstrated a translocation involving chromosome 3p. The karyotypes in CBFA2T3::GLIS2 were significantly less complex (p<0.001). NUP98 translocations were present in 11 cases, including 9 with a KDM5A partner, one with a NSD1 partner, and one with a BPTF partner. NUP98 fusions were associated with trisomy 6 (p<.001), monosomy 13/del(13q) (p<0.001), and trisomy 21 (p=0.026); the karyotypes were more complex (p=0.026). Note that NUP98-rearranged AMkL with monosomy 13/del(13q) only had KDM5A fusion partners; the single case of NUP98::KDM5A without monosomy 13/del(13q) had a translocation involving 13q. Of the 8 KMT2A-rearranged AMkL, fusion partners included MLLT10 (n=3), MLLT3 (n=3), MLLT11 (n=1), and unknown (n=1). A total of 30 other cases were classified as AMkL-MRC based upon karyotype or multilineage dysplasia; 28 of these cases had complex karyotypes, and an additional case had del(7q) and trisomy 8, the former of which is a myelodysplasia-related cytogenetic abnormality. Monosomy 7/del(7q), trisomy 8, and del(9q) were more common cytogenetic abnormalities identified in these AMkL-MRC (p=0.027, p=0.016, and p=0.024 respectively).

There was a clear paucity of common AML-associated somatic mutations. Of the 101 tested AMkL, no cases had detectable FLT3 ITD, NPM1, or CEBPα mutations. WT1 mutations were identified in 2 of 97 patients, one with NUP98::NSD1, and another with AMkL-MRC. Figure 2 depicts an oncoprint of mutations by fusion groups. More comprehensive genomic screening by whole genome and targeted exome performed in 87 of the patients demonstrated a lack of prominent somatic single nucleotide variants/indels in these patients (Supplemental Table S1); 60 AMkL (69%) lacked such mutations. Some of the more commonly mutated genes included NRAS (n=8), MYH11 (n=5), and PTPN11 (n=3).

Figure 2. Oncoprint illustrating fusions present in pediatric AMKL, and their co-operating cytogenetic abnormalities and mutations.

The abnormalities present in chr13 (Abn13) include del(13q), monosomy 13, trisomy 13, and chr13 translocations. Additional cytogenetic aberrations reported are copy number variations (trisomy 3, trisomy 6, monosomy 7/del7q, and monosomy5/del5q).

Outcomes:

Table 2 lists the 5-year EFS, OS, and RR for these AMkL. The 5-year EFS and OS were different between the two trials, with those treated on AAML0531 having better outcomes than those treated on AAML1031 (5-year EFS of 62 ± 14% vs 37 ± 13%, p=0.009, respectively; 5-year OS of 64 ± 14% vs 45 ± 14%, p=0.069, respectively). While AMkL compared to other FAB subtypes in AAML0531 did not show differences in 5-year EFS or OS, AMkL in AAML1031 had a worse 5-year OS than the other FAB subtypes (44 ± 14% vs 66 ± 3%, p=0.001) (Figure 3).

Table 2.

Outcome measures in pediatric AMkL per Children’s Oncology Group trial

	AAML0531				AAML1031
Variable	N	5-year EFS	5-year OS	5-year RR from end of course 1b	N	5-year EFS	5-year OS	5-year RR from end of course 1b
Morphology
All M7 FAB	48	62 ± 14% ^	64 ± 14% #	31 ± 17%	59	37 ± 13% ^	45 ± 14% *,#	45 ± 16%
M0-M6 FAB	816	53 ± 3%	66 ± 3%	36 ± 4%	1012	47 ± 3%	66 ± 3% *	43 ± 4%
FAB M7 Karyotype
Not complex	20	53 ± 22%	53 ± 22%	55 ± 34%	31	39 ± 18%	38 ± 18%	38 ± 22%
Complex	24	71 ± 18%	75 ± 17%	24 ± 20%	25	36 ± 20%	53 ± 22%	49 ± 26%
FAB M7 fusion-based classification a
AMkL, NOS	4	-	-	-	7	43 ± 37%	42 ± 37%	25 ± 50%
AMkL with RBM15::MRTFA fusion	7	86 ± 26%	86 ± 26%	0 ± 0%	14	50 ± 27%	54 ± 14%	38 ± 29%
AMkL with CBFA2T3::GLIS2	7	43 ± 37%	43 ± 37%	-	10	10 ± 19%	10 ± 19%	60 ± 56%
AMkL with NUP98 fusion	3	100 ± 0%	100 ± 0%	0 ± 0%	8	30 ± 35%	42 ± 44%	-
AMkL with KMT2A fusion	3	-	-	0 ± 0%	5	40 ± 44%	40 ± 44%	25 ± 50%
AMkL with (other) myelodysplasia-related changes	20	75 ± 20%	80 ± 19%	18 ± 19%	10	58 ± 32%	69 ± 30%	35 ± 34%

Abbreviations: AMkL, acute myeloid leukemia with megakaryocytic differentiation; EFS, event-free survival; FAB, French-American-British classification; N, number; NOS, not otherwise specified; OS, overall survival; RR, relapse risk

^aExcluding those with unknown cytogenetics or fusions identified in ≤2 cases (MECOM, FUS::ERG, and TEC::MLLT10)

^bIncluding only those with a complete remission

^*AAML1031 M7 significantly worse compared to AAML1031 M0-M6 FAB group (p=0.001)

^{^}AAML0531 M7 5-year EFS significantly better compared to AAML1031 M7 group (p=0.009)

^#AAML0531 M7 5-year OS better compared to AAML1031 M7 group but not significantly (p=0.069)

- 5-year estimate is undefined

Figure 3. AMkL vs other AML Kaplan–Meier curves per trial.

(A) Event-free survival and (B) overall survival of all cases of AMkL (FAB M7) versus all other centrally-reviewed AML FAB subtypes from trial AAML0531. As shown, M7 cases have relatively equivalent EFS and OS compared to other FAB subgroups. (C) EFS and (D) OS of all cases of AMkL versus all other centrally reviewed AML FAB subtypes from trial AAML1031; as shown AMkL have decreased 5-year OS compared to the other FAB subgroups. In general, AMkL on AAML0531 had better outcomes than AMkL on AAML1031.

To study the outcomes of AMkL by their fusion classification, the following cases were excluded due to low numbers: 5 with unknown cytogenetics, 2 with TEC::MLLT10, 1 with FUS::ERG, and 1 with a MECOM fusion. The outcomes from the remaining subgroups were determined (Table 2 and Figure 4). In general, cases of AMkL-MRC without recurrent fusions and those with RBM15::MRTFA had better outcomes than the other subgroups, while those with CBFA2T3::GLIS2 or KMT2A fusions had worse outcomes. AMkL, NOS (i.e. AMkL without identified fusions, complex karyotypes, or cytogenetic abnormalities that define myelodysplasia-related changes) had variably decreased EFS and OS, especially when compared with the more favorable AMkL-MRC and AMkL with RBM15::MRTFA. Interestingly, complex karyotypes in general (not considering the presence of any fusion proteins) did not lead to significantly different survival rates. Additionally, AMkL with NUP98 fusions had different outcomes in the trials, with increased EFS and OS in the AAML0531 trial compared to those in the AAML1031 trial, though the number of AMkL with NUP98 fusions in the AAML0531 trial was quite small (n=3).

Figure 4. AMkL subgroup Kaplan–Meier curves per trial.

(A) Event-free survival and (B) overall survival of major AMkL subgroups in trial AAML0531 show statistically significant different survivals depending on the identified fusion, other myelodysplasia-related changes (MRC), or not otherwise specified (NOS). (C) Event-free survival and (D) overall survival of major AMkL subgroups in trial AAML1031 did not show statistically significant different survivals depending on the subcategory.

DISCUSSION

In this study, 107 cases of central pathology confirmed AMkL from two COG trials (AAML0531 and AAML1031) were categorized into subgroups based upon morphology/dysplasia, cytogenetics, and molecular features. While there was a paucity of genetic mutations, these AMkL harbored a variety of fusions which helped to subclassify the leukemias; such subclassifications are helpful in determining outcomes.

AMkL is a rarer subtype of AML in children. The diagnosis of AMkL may be challenging, as bone marrow aspirate smears are often hemodilute and/or aparticulate due to marrow fibrosis. Such marrow fibrosis can also lead to suboptimal specimens for flow cytometry, cytogenetics, and molecular testing, with falsely decreased blast percentages. Bone marrow core biopsies may be helpful in characterizing the blast percentage and immunophenotypes of these leukemias. While only performed in a subset of our cohort (n=39), 77% of core biopsies with reticulin staining had ≥MF-1 fibrosis, while 46% had MF-2 or MF-3 fibrosis, concordant with prior studies.^15,21 Prominent fibrosis has been reported in AMkL with RBM15::MRTFA,^16–20 but this is one of the first reports detailing a high rate of fibrosis in those with KMT2A fusions. In one adult study, fibrosis did not correlate with survival.⁴⁴

Previous reports have shown inferior outcomes in AMkL compared to other AML subtypes, with 4- and 5-year EFS ranging from 36.6–51% and 41–47%, respectively, and 4- and 5-year OS ranging from 56–58.6% and 49–60%.^8,9,13,14,26 In this study, the 48 AMkL cases on AAML0531 had equivalent to slightly better 5-year EFS and OS, while the 59 AMkL cases on AAML1031 had equivalent to decreased 5-year EFS and OS compared to these prior studies. The different outcomes of AMkL in these two current COG trials is not surprising due to the differences in therapy in these trials. AAML0531 randomized the enrolled patients to receive standard 5-course chemotherapy with or without two doses of GO. While there were more AMkL patients on the no-GO Arm than on the GO Arm (n=30 vs 18, respectively), this difference was not statistically significant; however, the GO arm generally had significantly improved EFS and decreased RR compared to the no-GO Arm.²⁸ In AAML1031, the patients who were classified as low risk (having favorable cytogenetic/molecular features or uninformative cytogenetic/molecular features) but with negative MRD at end of induction (including 30 of the AMkL in this series), received 4-course chemotherapy with or without bortezomib. Notably, these 30 AMkL did not have favorable cytogenetic/molecular features as there were no cases of t(8;21) (RUNX1::RUNX1T1), inv(16)/t(16;16) (CBFB::MYH11), NPM1, and CEBPα in our AMkL cohort. Getz et al. reported that the reduced cytarabine exposure in these low-risk patients without favorable cytogenetics led to reduced disease-free and overall survival.⁴⁵

In this study, we subclassified the AMkL into separate cohorts based upon pathology, cytogenetic, and involved fusion proteins. Our results suggest that genotype and not phenotype defines the outcome of these cases. Despite the small numbers of cases in these subcategories and different outcomes in the two COG trials, we showed a poor prognosis of AMkL with CBFA2T3::GLIS2 and KMT2A fusions. Previous reports have confirmed the decreased outcomes of CBFA2T3::GLIS2 with 4–5-year EFS ranging from 8–33% and 4–5-year OS of ranging from 14–38%,^7,8,25,27,46 though some of those reports did notably include a subset of patients in this study. Similarly, AMkL with KMT2A fusions have reported poor prognoses, with 5-year EFS ranging from 27–28.5% and 5-year OS ranging from 27–32.4%.^7,9,27

CBFA2T3::GLIS2 is cryptic, as it cannot be detected by routine karyotype. In fact, of the 17 CBFA2T3::GLIS2 cases in this series, none had karyotypic evidence of the fusion and 6 had normal karyotypes. This fusion’s surprisingly high association with trisomy 3 and lack of complex karyotypes also suggests that it is especially important to screen for cryptic fusions in these cases. Other clues come from its immunophenotype, as CBFA2T3::GLIS2 is highly associated with bright CD56 expression and the RAM phenotype. Only 18 of 101 cases in this study displayed the RAM phenotype, 16 of which had CBFA2T3::GLIS2, concordant with previous studies.⁴⁷

The most common fusion identified in this series was RBM15::MRTFA. These 21 RBM15::MRTFA cases were significantly younger in age (mean of 0.91 years), consistent with prior studies showing RBM15::MRTFA patients are amongst the youngest with AMkL.^7–9 Our study confirmed that RBM15::MRTFA AMkL generally have a more favorable outcome than other subgroups of AMkL.^7,8,12,27

In our cohort, NUP98 fusions were seen with KMD5A, NSD1, and BPTF partner genes. Concordant with literature, the most common fusion partner was KMD5A, a partner often associated with AMkL.^7,8,25–27 Some karyotypic findings, including trisomy 6, monosomy 13/del 13(q), and trisomy 21, were significantly increased in the cases with NUP98 fusions. NUP98::KDM5A with trisomy 21, monosomy 13/del 13(q), trisomy 6, and complex karyotypes have previously been reported,^7,8,26 but this is the first study to highlight the significant increases in these abnormalities. Specifically, all 9 NUP98::KDM5A had structural chromosome 13 abnormalities, with 8 having monosomy 13 (n=2) or del(13q) (n=6), and the remaining case having a translocation involving 13q; notably, one case each with monosomy 13 and del(13q) also demonstrated translocations involving 13q.

The 5-year EFS and OS of AMkL with NUP98 rearrangements were quite different between the AAML0531 and AAML1031 trials. Only 3 patients had AMkL with NUP98 rearrangements in the AAML0531 trial, all of whom have survived without relapse, although this favorable survival may be influenced by small numbers. In contrast, the 8 NUP98-rearranged AMkL patients on AAML1031 (7 partnered with KDM5A) had reduced EFS and OS, more consistent with prior studies. Two AMkL cohorts studied by de Rooij et al.^7,27 demonstrated that patients with NUP98::KDM5A had 5-year EFS and OS ranging from 22–25% and 22–35%, respectively, though notably both not statistically significant. However, with a different cohort, de Rooij et al.⁸ found this translocation to be an independent predictor of poor outcome with 4-year EFS and OS both of 36%. Hara et al.²⁶ found similarly worse EFS. Two of the de Rooij et al. cohorts^7,8 included a subset of patients described in this study.

After AMkL with recurrent fusions has been separated, the two main remaining subgroups are AMkL-MRC and AMkL, NOS. Our study is one of the first to include these classifications in AMkL. In general, AML with complex karyotypes, myelodysplasia-defining cytogenetic abnormalities, and/or morphologic dysplasia have poorer outcomes compared to AML, NOS or AML with certain favorable recurrent cytogenetic abnormalities.^32,48,49 In this study, cases with complex karyotypes, but not having known gene rearrangements were classified as AMkL-MRC, including 28 with complex karyotypes, 1 with myelodysplasia-related cytogenetic abnormalities lacking a complex karyotype, and 1 with multilineage dysplasia. Similar to prior reports, monosomy 5/del(5q) and monosomy 7/del(7q) were not commonly identified in these pediatric cases of AMkL (though this group was still significantly enriched in cases with monosomy7/del(7q)) and the most common gains included trisomies 8, 19, and 21.^2,9 Interestingly, these 30 AMkL-MRC cases had relatively increased 5-year EFS and OS compared to the other subcategories. It may be that purifying the AMkL-MRC cohort, excluding specific fusion products, leads to better prognoses. Notably, in the upcoming WHO hematolymphoid tumor 5^th edition⁵⁰ and International Consensus Classification (ICC) of myeloid neoplasms and acute leukemias,⁵¹ the AML-MRC category is removed in favor of AML, myelodysplasia-related (AML-MR, WHO 5^th edition) and AML with myelodysplasia-related cytogenetic abnormalities or AML with myelodysplasia-related gene mutations (ICC), categories depicting cases of AML with specific cytogenetic and molecular abnormalities but removing cases with only morphologic dysplasia. In this study, such categorization would have reclassified only 1 case of AMkL from the AML-MRC category to AMkL, NOS.

Conversely, cases without defined fusion products and relatively normal karyotypes, classified as AMkL, NOS in this study had relatively poor outcomes. Our cohort of 11 cases may be considered more of a pure group than similar cohorts in past studies, as many cases with cryptic fusions can have normal karyotypes.

The relatively small numbers of AMkL in these COG trials could lead to some false associations. Additionally, the 2 trials could not be combined for outcomes analysis due to their statistically significant differences in 5-year EFS and OS. Considering the small size of these subgroups, the comparisons performed herein are more ad hoc analyses and exploratory in nature. Larger scale studies of similar cohorts are recommended to confirm our findings.

In conclusion, full cytogenetic and RNA fusion testing is recommended for all AMkL to enable classification similar to that presented here. The identification of recurrent fusion proteins should prevent classification into the AMkL, NOS and AMkL-MRC categories, even in the cases with complex karyotypes. Such subgroups can aid in future prognostication and development of therapeutic strategies especially in those with poor outcomes.

Supplementary Material

fS1

Supplemental Figure 1. Acute myeloid leukemia with megakaryocytic differentiation morphology. A-B) Aspirate smears of AML, NOS demonstrate megakaryoblasts as large cells with cytoplasmic pseudopods or bleb formation. They have round, slightly irregular nuclei with fine reticular chromatin and 1–3 nucleoli. Rare cytoplasmic granules can be seen. C) AML with RBM15::MRTFA has blasts in which the differential diagnosis may include a small round blue cell tumor. D) AML with KMT2A rearrangements, classically associated with monocytic differentiation, may also have distinctly megakaryoblastic cells.

fS2

Supplemental Figure 2. Acute myeloid leukemia with megakaryocytic differentiation morphology on bone marrow core biopsies. A) Abnormal megakaryocytic maturation is often present in AMkL, including micromegakaryocytes and forms with separate nuclear lobes. B) On a core biopsy, the megakaryoblasts are large with smooth chromatin and prominent nucleoli. C) Low power may demonstrate increased reticulin fibrosis, here myelofibrosis score 3 (MF-3) (reticulin stain).

Abbreviation key:

AMkL

Acute myeloid leukemia with megakaryocytic differentiation

AML

Acute myeloid leukemia

COG

Children’s Oncology Group

CR

Complete remission

EFS

Event-free survival

FAB

French-American-British

GO

Gemtuzumab Ozogamicin

ICC

International Consensus Classification

IRB

Institutional review board

ITD

Internal tandem duplications

MF

Myelofibrosis

MRC

Myelodysplasia-related changes

NOS

Not otherwise specified

OS

Overall survival

RR

Relapse risk

WHO

World Health Organization