Pediatric acute myeloid leukemia (AML), with an incidence of approximately seven occurrences per 1 million children annually
[1], still represents a challenge for pediatric oncologists. Even though patient outcome has significantly improved over the past 30 years
[1], as survival rates have reached ~70%, still about 30% of children with AML experience disease recurrence, with a probability of overall survival (pOS) ranging between 29 and 38%
[2][3][4][5]. Thus, given the high frequency of relapse in pediatric AML, a deeper understanding of the mechanisms underlying disease recurrence is necessary.
2. Pediatric AML: Clinical Presentation
AML is a heterogeneous disease, from clinical behavior to morphology, immunophenotyping and genetic abnormalities, due to the abnormal proliferation of myeloid progenitor cells in the bone marrow
[6]. It has some unique presentations, such as granulocytic sarcomas, subcutaneous nodules, infiltration of the gingiva and disseminated intravascular coagulation (DIC)
[7]. Diagnosis is based on leukemic blasts evaluation from the bone marrow, considering the morphology, cytochemistry, immunophenotype, cytogenetics (conventional karyotyping complemented with FISH or reverse transcriptase PCR) and molecular characterization
[6]. AML is then classified according to the 2016 World Health Organization (WHO) classification, which also considers karyotype and molecular aberrations, and has replaced the morphology-based French–American–British (FAB) classification
[6]. Due to the poor outcome of some specific subtypes, risk-group stratification based on genetic abnormalities and response is essential, mainly in order to identify those with high-risk AML, which could benefit from a more intensive treatment
[1] such as hematopoietic stem cell transplantation (HSCT) in first complete remission (CR)
[8]. Despite an increasing knowledge about the landscape of molecular alterations in AML, there are still only few alternatives to standard intensive chemotherapy regimens used for induction and consolidation treatment protocols, considering that ~30% of the pediatric patients relapse after the achievement of CR. The development of new therapies able to overcome resistance/relapse is essential. To this aim, it is necessary to deeply characterize the mechanisms responsible for AML and progression.
Figure 1. Schematic resume of the main recurrent genetic lesions found in pediatric AML.
Other subtypes of pediatric AML, characterized by a favorable prognosis, are those involving core binding factor (CBF), accounting for approximately 20–25% of pediatric AML cases
[11], such as t(8;21) (q22;q22) and inv(16),which lead to the fusion genes
RUNX1-RUNX1T1 and
CBFB-MYH11, respectively.
KMT2A is a histone methyltransferase that is rearranged in 35–60% of infants and 10–15% of children and adolescents
[12], and it has almost 100 different fusion partners
[13]. It represents a subgroup of pediatric AML with intermediate/unfavorable prognosis and for this reason, many specific drugs are in preclinical and clinical trials
[14]. A subtype of pediatric AML with an unfavorable prognosis is that characterized by the fusion gene
CBFA2T3-GLIS2, caused by inv(16) (p13.3q24.3), which is present in 15–20% of pediatric non-Down syndrome acute megakaryoblastic leukemia (non-DS-AMKL) and about 7–8% of pediatric AML with normal karyotype
[15][16], mostly found in infant patients (<3 years)
[17][18]. AMKL occurs particularly in children with Down syndrome, however, in contrast to DS-AMKL which display 80% of overall survival rate, non-DS-AMKL is associated with extremely poor prognosis
[19][20]. Fusions involving the gene nucleoporin 98kD (
NUP98) and over 30 different fusion partners, the most common of which are
KDM5A and
NDS1, can be found in 4–9% of pediatric AML
[21]. The t(6;9) (p22;q34) leads to the expression of the fusion gene
DEK-NUP214, present in just 1–2% of pediatric AML
[22], whose role in leukemogenesis is unknown, but it is associated with a poor prognosis, due to the low rates of remission and high rates of relapse. A very rare cause of pediatric AML (less than 1%) associated with grim prognosis are translocations or inversions involving the
MDS1 and
EVI1 complex locus (
MECOM)
[13]. Aneuploidy, in the form of monosomy 5/5q-, monosomy 7 and abnormal 12p, are present in 4–9% of pediatric AML
[12].
With the advent of large-scale genomic approaches, several molecular aberrations have been likewise identified, such as the gene nucleophosmin 1 (
NPM1) which encodes for a chaperon protein normally localized predominantly in the nucleolus, however, when mutated, it is aberrantly localized in the cytoplasm, causing the activation of oncogenes. This is rarely altered in pediatric AML
[23], but much more often in children (10%) and adolescents (20%)
[12], and it often co-exists with
FLT3-ITD mutations
[24]. Another recurrent mutation in AML, especially in older children and adolescents, is that involving the gene CCAAT-enhancer binding protein alpha (
C/EBPA), present in 5–10% of pediatric AML cases
[12]. This is a very important transcription factor for granulocytic and monocytic differentiation
[25], and inactivating mutations causing a block in granulocytic differentiation likely contribute to leukemogenesis. The prognosis of this genetic aberration in pediatric AML is still a matter of debate but generally, patients harboring this mutation are included in low/intermediate risk groups. Among the molecular lesions with an unfavorable prognosis, FMS-like tyrosine kinase 3 (
FLT3) is a common gene involved in pediatric AML, whose frequency is age-related, being rare in infants and rising as age increases.
FLT3 is essential for the proliferation, survival and differentiation of stem/progenitor cells. The most common type of mutation in
FLT3 is represented by internal tandem duplication (ITD) mutations, but point mutations are also frequent, and both these lesions lead to the constitutive activation of this gene
[26]. The
WT1 gene is expressed in CD34+ hematopoietic stem progenitors and is crucial for the regulation of normal growth and development. Among pediatric AML cases, 15% are characterized by inactivating mutations of
WT1 [21], and associated with a poor prognosis when
FLT3-ITD mutations are also present
[21]. Mutations in epigenetic regulators such as
TET2,
IDH1 and
IDH2 are much less prevalent compared to adult AML, characterizing only 1–2% of pediatric patients
[27], while signaling mutations such as
NRAS,
KRAS,
CBL,
GATA2,
SETD2 and
PTPN11 are more common in younger patients. Moreover, MYC alterations were identified as exclusive in children, suggesting different leukemogenesis mechanisms in children compared to adults
[10][21].Nonetheless, therapies targeting alterations in these factors are showing promising results in adults, thus it is important to look for these lesions in pediatric cases too
[27].
4. Clonal Evolution in Pediatric AML: From Diagnosis to Relapse
The concept of clonal evolution in cancer, which was first conceived in 1976
[28], appears to apply also to AML, as demonstrated by the pivotal work of Ding et al.
[29] in adult AML and also confirmed in the pediatric setting
[2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30]. This term refers to the selection and expansion, over the course of the disease, of subclones that originate from a common ancestral clone but acquire different mutations that confer them a survival advantage, leading to genetic diversity within a cell lineage
[2]. AML arises from the accumulation of “early” DNA mutations in hematopoietic stem cells (HSCs), which enhance their self-renewal potential, leading to leukemia stem cells (LSCs). Additionally, “late” mutations that promote proliferation are necessary for the development of AML
[31]. We can therefore identify “primary events” that are highly penetrant and stable during the course of the disease, and “secondary events” that occur later and only in some cells, leading to the development of different subclones and to a variation in the composition of cancer
[2]. This explains the complexity of the determination of the proper treatment option, because of AML genomic heterogeneity not only at diagnosis, but also during therapy and eventually at relapse.
Several studies
[2][30][32] have globally analyzed, by means of whole exome sequencing (WES), trios of pediatric AML diagnostic, remission and relapse specimens, in order to deeply characterize the key features that shapes clonal evolution over time. They showed that most of the dominant variants, which originate during ancestral leukemic development, persist from diagnosis to relapse, but many subclonal modifications, which develop later in specific subclones, do not, and thus, the genomic landscape at relapse also differs based on the use of different therapeutic agents
[30]. They also demonstrated that, even if recurrent single mutations cannot be found, mutations in specific gene families are typical
[30]. Indeed, aside from the canonical classification of AML, in terms of clonal evolution, we can further classify mutations into three major groups ():
Table 1. List of the main mutated genes, classified in three major groups, with the relative prominent occurrence in terms of timing (mainly at diagnosis and/or relapse).
- i.
-
Cohesin complex gene mutations;
- ii.
-
Transcription factors and epigenetic regulators mutations;
- iii.
-
Signaling molecules mutations.
Cohesin is a protein necessary for cell division, DNA repair and gene expression, and it is composed of four core subunits, three of which have been found to be mutated in pediatric AML, leading to a loss of cohesin function:
RAD21,
SMC3 and
STAG2 [32].
Shiba et al.
[32] demonstrated that variant allele frequencies (VAFs) of mutated
SMC3 were high, suggesting these occur at an early stage of leukemogenesis.
Among the epigenetic regulators, mutations have been found in the Additional Sex combs-Like (
ASXL) family,
BCOR and
BCORL1,
EZH2.
ASXL1 and
ASXL2 mutations have been found in pediatric de novo AML, mainly in patients with t(8;21)
[32], and an
ASXL3 mutation has been described by
Masetti et al.
[2] at relapse (median frequency or MF of 29.7%), backtracking to a small subclone already present at diagnosis (MF of 0.3%). VAFs of mutated
ASXL2 were lower than others, suggesting that this mutation is a secondary event.
BCOR and its homologue
BCORL1 suppress gene transcription through epigenetic mechanisms
[32].
BCORL1 has been shown to be a tumor suppressor gene, and, as
SMC3,
BCORL1 appears to be mutated at an early stage in leukemic cells
[32].
EZH2 is one of the most commonly mutated genes in pediatric Down syndrome AMKL, and it encodes for a subunit of a complex responsible for methylation
[32].
Signaling pathway mutations include mutations of genes involved in Ras pathways such as
NRAS,
KRAS,
PTPN11, and of tyrosine kinases such as
KIT and
FLT3.
NRAS mutations have been found to be relapse specific, like those in
CREBBP, a coactivator of several hematopoietic transcription factors
[32].
Farrar et al.
[30] described
PTPN11 mutations, lost at relapse when they appeared as subclonal variants at diagnosis.
Masetti et al.
[2] found a
PTPN11 mutation gained at relapse (MF of 31.9%), suggesting its role in increasing cell proliferation and/or survival. As previously described,
FLT3 is essential for proliferation, survival and differentiation of stem/progenitor cells.
Masetti et al.
[2] showed that a small
FLT3-TKD-mutated subclone present at diagnosis (MF of 3.4%) underwent expansion at relapse (MF of 13.3%), leading to increased cell proliferation and/or survival, and this mutation, together with
NRAS and
KIT mutations, was also described as a secondary event contributing to disease progression by
Shiba et al.
[32].
In addition, also genes that do not belong to these groups are frequently mutated. The previously described
C/EBPA and
WT1 mutations exhibit different patterns of recurrency in the study presented by
Masetti et al. A highly penetrant biallelic mutation of
C/EBPA was revealed, and in one patient, a homozygous non-frameshift insertion was present both at diagnosis and relapse in the majority of the tumor-cell population (MF > 80%). On the contrary,
WT1 mutations appeared highly unstable, two different frameshift insertions were detected mainly at diagnosis (MF of 27.6% and 13.9), while a single-nucleotide variant was detected only at relapse (MF of 40%).
SETD2 is a methyltransferase involved in the recruitment of mismatch repair (MMR) machinery, and its mutation results in a loss of function of the methyltransferase activity, leading to the build-up of several subclonal mutations and consequently, the increased plasticity and adaptability of leukemia cells, due to the failure of DNA repair
[2]. A frameshift insertion of the
SETD2 gene has been described both at diagnosis (MF of 32.5%) and at relapse (MF of 31.7%)
[2]. The
TYK2 gene is a member of the Janus tyrosine kinases (JAK) family involved in cell growth, differentiation and survival.
Masetti et al.
[2] found a mutation of
TYK2 both at diagnosis (MF of 43%) and at relapse (MF of 14.9%), which causes a hyperactivation of the
TYK2 pathway, resulting in aberrant cell survival through the upregulation of BCL2 (anti-apoptotic protein).
SALL1 is a member of the transcriptional network that regulates stem cell pluripotency
[33]; it has not been detected at diagnosis but only at relapse (MF of 28.6%) with a clone size at relapse of 50–60% (corrected for copy number variations)
[2] ().