Genetic drivers of progression towards TAM and ML-DS

T21 in combination with a somatic truncating mutation in GATA1, acquired prenatally, is required and sufficient for the development of TAM (Refs Reference Yoshida30–Reference Taub32). GATA1 is located on the X-chromosome and encodes a haematopoietic transcription factor with essential functions for the differentiation of haematopoietic stem cells (HSCs) towards the erythroid and megakaryocytic lineages (Ref. Reference Hitzler33). Most mutations are found in exon 2 of GATA1 and result in an N-terminally truncated protein isoform GATA1s because of the utilisation of an alternative start codon located in exon 3 (Refs Reference Antonarakis2, Reference Alford15, Reference Taub32, Reference Garnett, Cruz Hernandez and Vyas34). GATA1s lacks the amino-terminal activation domain and has a reduced transactivation potential, which causes decreased expression of GATA1 target genes (Ref. Reference Sas14). Although the DNA binding domain of GATA1s is intact, GATA1s binding is impaired at specific erythroid regulatory regions and the MYC promoter, whereas binding to megakaryocytic and myeloid target genes is normal (Refs Reference Klusmann35–Reference Byrska-Bishop38). Also, studies have shown that GATA1s is not able to bind to RUNX1, an important protein for megakaryocyte differentiation; however, evidence is inconsistent (Ref. Reference Vyas and Crispino39). As a result, GATA1s in T21 causes aberrant differentiation and proliferation of megakaryocytes, which results in the production of megakaryoblasts (Refs Reference Banno6, Reference Li40). In addition, the expression of GATA1s is increased by the extra copy of the 4 megabase DSCR containing RUNX1, ETS2 and ERG, which further promotes the proliferation of megakaryoblasts (Refs Reference Banno6, Reference Li40).

Upon progression to ML-DS later in life, the same GATA1 mutations are present, originating from a dominant or minor TAM clone, which indicates that ML-DS evolves from a persistent TAM clone during clinical remission (Refs Reference Alford15, Reference Hitzler33, Reference Labuhn41). Nonetheless, additional driver mutations are required for progression towards ML-DS (Refs Reference Alford15, Reference Yoshida30). Thus far identified additional drivers include mutations in cohesins (53%) or CTCF (20%), EZH2, KANSL1 and other epigenetic regulators (45%) and common signalling pathways including JAK family kinases, MPL, SH2B3, CSFR2B and multiple RAS pathway genes (47%), which all provide the cell with a growth advantage (Refs Reference Yoshida30, Reference Nikolaev31, Reference Labuhn41). Cohesins are essential for DNA replication and repair and are thought to have a tumour-suppressor function in cancer (Refs Reference Viny and Levine42, Reference Cuartero, Innes and Merkenschlager43). Models of dose-specific cohesin loss have shown HSC expansion and impairments in differentiation as a result of an open-chromatin state, which causes increased transcription of genes involved in self-renewal (Refs Reference Viny and Levine42, Reference Cuartero, Innes and Merkenschlager43). Mutations in epigenetic regulators and CTCF, which directly interacts with cohesins, are known to disrupt the expression of genes involved in HSC renewal and differentiation by chromatin modifications (Refs Reference Fisher44, Reference Ernst45). Both the JAK-STAT pathway and RAS signalling pathway are involved in the regulation of cell proliferation, survival and differentiation (Refs Reference Vainchenker and Constantinescu46, Reference Mercher and Schwaller47). Mutations in these genes have been identified in different haematological malignancies and have been shown to contribute to the proliferation of leukaemic blasts (Refs Reference Vainchenker and Constantinescu46, Reference Mercher and Schwaller47). In conclusion, mutations in these genes could all result in increased cell survival, cell proliferation and impairments in the differentiation of the oncogenic clone (Refs Reference Viny and Levine42, Reference Vainchenker and Constantinescu46, Reference Mercher and Schwaller47).

Mutations in cohesins, CTCF and EZH2 showed a similar variant allele frequency (VAF) to GATA1 mutations, which indicates that they are the oncogenic mutations driving the progression towards ML-DS (Ref. Reference Yoshida30). On the other hand, the RAS pathway, other tyrosine kinases and cytokine receptor mutations had a lower VAF compared with GATA1 mutations, indicating they originate from a later timepoint (Ref. Reference Yoshida30). These varying VAFs point towards the subclonal development of ML-DS (Ref. Reference Yoshida30). In rare cases, some of the driver mutations identified in ML-DS were present at a low VAF in the (pre)leukaemic TAM blasts (Ref. Reference Labuhn41). This observation supports the hypothesis that in some cases the ML-DS clone has been already present early in life, possibly prenatal. This hypothesis is further supported by a study of monozygotic twins who share the same GATA1 mutation and somatic translocation involving CUX1, identified at the time of simultaneous ML-DS development (Ref. Reference Bache48).

The mutations driving ML-DS have, at lower frequencies, also been identified in non-DS AMKL (Refs Reference Yoshida30, Reference de Rooij49), indicating that ML-DS has a different mutational landscape compared with paediatric non-DS AMKL. Non-DS AMKL frequently presents with specific fusion genes as driving aberrations, such as CBFA2T3-GLIS2 and RBM15-MKL1, which both have not been detected in TAM and ML-DS (Refs Reference Yoshida30, Reference de Rooij49). On the other hand, somatic T21 is frequently seen in children with non-DS AMKL, underlining the role of an extra chromosome 21 in leukaemic development (Ref. Reference Hama28). Interestingly, these AMKL blasts frequently carry a somatic GATA1 mutation and the same spectrum of additional aberrations as seen in ML-DS (Ref. Reference Lopez50).

Genetic drivers of progression towards ALL

DS-ALL patients show a different spectrum of cancer driver gene mutations compared with non-DS ALL patients (Refs Reference Mullighan16–Reference Buitenkamp18, Reference Hertzberg51). High hyperdiploidy, ETV6-RUNX1 and BCR-ABL1 are less common in DS-ALL compared with non-DS ALL (Ref. Reference Buitenkamp18). Up to 62% of all DS-ALL cases show upregulation of CRLF2, caused by rearrangements (i.e. IGH-CRLF2, P2RY8-CRLF2) or mutations in CRLF2, compared with 5–12% of non-DS ALL cases (Refs Reference Mullighan16, Reference Nikolaev17, Reference Hertzberg51, Reference Potter52). Other cancer initiation events in DS-ALL are mutations in chromatin remodelers, classic tumour suppressors and B-lymphocyte differentiation factors (Ref. Reference Nikolaev17). Moreover, DS-ALL displays an increased incidence of activating JAK2 mutations, which are only found in CRLF2 overexpressing cases (Refs Reference Mullighan16, Reference Nikolaev17, Reference Hertzberg51, Reference Bercovich53). This connection suggests a cooperating effect of these genes in leukaemogenesis and CRLF2 upregulation as a potential first event, before additional genetic aberrations in JAK2 (Ref. Reference Hertzberg51). CRLF2 is an atypical type I cytokine receptor and a weak activator of JAK2, and thus the JAK-STAT pathway (Ref. Reference Hertzberg51). Hence positive regulation of CRLF2 by JAK2 may partially explain the identified relation (Ref. Reference Hertzberg51). CRLF2 binds the ligand TSLPR, which promotes early B-cell development (Refs Reference Levin54, Reference Yoda55). Upregulation of CRLF2 likely causes increased proliferation of early B-cells. Gain-of-function mutations have also been identified in IL7R, which forms a heterodimeric complex with CRLF2 (Refs Reference Yoda55, Reference Shochat56). IL7R is required for normal lymphoid development and part of the JAK-STAT signalling pathway, which further stresses the importance of the JAK-STAT pathway in DS-ALL (Refs Reference Nikolaev17, Reference Yoda55, Reference Shochat56). DS-ALL patients more often harbour KRAS and NRAS oncogenic mutations compared with non-DS ALL patients, which are mutually exclusive with JAK2 mutations (Ref. Reference Nikolaev17). Interestingly, a complete or partial gain of chromosome 21 is often seen in non-DS paediatric B-ALL cases, predominantly in hyperdiploid ALL and iAMP21 ALL (Refs Reference Laurent, Kotecha and Malinge27, Reference Paulsson57). Similar to in DS-ALL, NRAS and KRAS mutations are often found in these patients, suggesting that T21 selects for these specific mutations (Refs Reference Paulsson58, Reference Malinowska-Ozdowy59). However, the mechanism behind this cooperation is yet unknown.

Somatic mutation accumulation during fetal development

As explained before, besides T21, additional oncogenic mutations are needed for leukaemogenesis in DS (Ref. Reference Nikolaev31). Since TAM is already present in DS newborns, mutation accumulation has to occur before birth (Ref. Reference Bhatnagar60). In some TAM cases, multiple GATA1 mutated clones have been detected, characterised by distinct GATA1 mutations (Refs Reference Yoshida30, Reference Labuhn41, Reference Ahmed61). Indeed, it has been suggested that aneuploidy can promote genomic instability, which has been proven in yeast (Refs Reference Matzke62, Reference Sheltzer63). These observations suggest an increased mutation rate of haematopoietic cells of DS fetuses, which could contribute to their higher risk of leukaemia. We have recently tested this hypothesis by whole genome sequencing (WGS) of clonally expanded single haematopoietic stem and progenitor cells (HSPCs) of non-DS and DS human fetuses (Ref. Reference Hasaart20). We have shown that HSPCs of DS fetuses have 34 extra somatic mutations compared with HSPCs of non-DS fetuses (Ref. Reference Hasaart20). This might not seem a lot, however, non-DS fetal HSPCs have already shown a somatic mutation rate of 100 base substitutions per year, which is 5.8 times higher compared with the mutation rate of adult HSPCs (Ref. Reference Hasaart20). This relatively high prenatal mutation rate may contribute to the increased leukaemia risk in children compared with young adults by increasing the chance of acquiring an oncogenic mutation (Ref. Reference Greaves64). The even higher mutation load in HSPCs of DS fetuses could further increase this chance.

Underlying causes of the increased mutation load in DS fetal cells

The driving mechanisms behind the increased somatic mutation load observed in fetal DS stem cells are still unknown (Ref. Reference Hasaart20). Increased reactive oxygen species (ROS) production, caused by oxidative stress, has been found in almost all cancer types and is thought to be an important driving factor for tumour development (Refs Reference Sabharwal and Schumacker65, Reference Schieber and Chandel66). Deficiencies in mitochondrial functioning can result in increased ROS production (Ref. Reference Zorov, Juhaszova and Sollott67). Several cell types of DS individuals show mitochondrial dysfunction and increased levels of ROS (Refs Reference Piccoli68–Reference Busciglio73). Moreover, several genes involved in mitochondrial function are located on chromosome 21 and overexpressed in DS, which further supports deregulation of mitochondrial function (Refs Reference Vilardell74–Reference Izzo76). However, knowledge on an association between increased ROS levels and DS-leukaemogenesis is lacking.

Oxidative stress can damage DNA, which can result in mutations if left unrepaired (Ref. Reference Burrows and Muller77). ROS are mainly mutagenic by inducing 8-oxoguanine lesions, which increase mispairing with adenine instead of cytosine (Ref. Reference Burrows and Muller77). G:C > T:A mutations were identified as the predominant base substitution type causing mutations in GATA1 (Fig. 1b) (Ref. Reference Cabelof78). These predominant G:C > T:A mutations could indicate that ROS production induces GATA1 mutations in DS. Mutational processes, such as oxidative stress, are known to generate a characteristic pattern of mutations, which is called a mutational signature (Ref. Reference Alexandrov79). These signatures are classified in the Catalogue of Somatic Mutations in Cancer (COSMIC) database (Ref. Reference Alexandrov80). COSMIC signature 18, which has previously been associated with oxidative stress mutagenesis (Ref. Reference Viel81), was identified in a subset of HSPCs isolated from DS fetuses (Ref. Reference Hasaart20). However, this signature did not contribute to the somatic mutations observed in the (pre)leukaemic blasts of TAM patients (Ref. Reference Hasaart20) or to the single base substitutions causing GATA1s in TAM (Fig. 1b). Instead, insertions, deletions and duplications are more often the cause of GATA1s in TAM (Fig. 1a). The indel profile of the identified GATA1 mutations in TAM (pre)leukaemic blasts shows that 5 + base pair insertions are the most frequently observed type of indels causing GATA1s (Fig. 1c). The GATA1 indel profile does not show similarity to a known small indel COSMIC signature (Ref. Reference Alexandrov80). Therefore, the mutational process causing this specific indel profile is not yet known and needs further investigation (Fig. 1c). Altogether, there is evidence of mitochondrial dysfunction and increased levels of ROS in DS. However, DNA damage caused by ROS is not necessarily needed to induce the somatic mutations that are necessary for the development of TAM (Ref. Reference Hasaart20).

Fig. 1.

GATA1 aberrations in transient abnormal myelopoiesis (TAM). Analysis of all GATA1 mutations identified in TAM, which are annotated in the Catalogue of Somatic Mutations in Cancer (COSMIC) Cancer Gene Census database and GATA1 mutations from the recently published study of Labuhn et al. (Refs Reference Yoshida30, Reference Nikolaev31, Reference Labuhn41, Reference Ahmed61, Reference Cabelof78, Reference Xu138–Reference McElwaine149). (a) Pie chart showing the different types of GATA1 mutations observed in TAM (pre)leukaemic blasts. (b) 96-trinucleotide spectrum of all single nucleotide variants (SNVs) observed in GATA1. (c) The spectrum of insertions and deletions (indels) within GATA1.

Besides mitochondrial dysfunction, deficiencies in DNA damage repair may also explain the increased somatic mutation load observed in DS fetal cells (Refs Reference Hasaart20, Reference Morawiec82). Multiple studies have shown a defect in DNA repair in DS cells (Refs Reference Cabelof78, Reference Morawiec82). Also, in the fetal liver, lower base excision repair activity was detected in DS-tissues compared with non-DS tissues (Ref. Reference Cabelof78). These findings support the hypothesis that, already in the prenatal setting, ineffective DNA damage repair in DS could, together with mitochondrial dysfunction, lead to increased mutagenesis and potentially to phenotypic diversity.

Epigenetic and transcriptional alterations in DS haematopoiesis

Besides genetic aberrations, epigenetic changes can contribute to the inheritable phenotypic diversity within cell populations by changing transcription levels. Many studies have shown genome-wide epigenetic alterations in cells of DS fetuses prior to leukaemia development, not solely restricted to chromosome 21 (Refs Reference Letourneau29, Reference Eckmann-Scholz83–Reference Malinge85). Transcriptome analysis of fetal fibroblasts isolated from monozygotic twins, discordant for T21, showed that differential expression is organised in chromosomal domains that vary in size and contain up to 507 genes (Ref. Reference Letourneau29). These well-defined domains that consist of neighbouring genes sharing differential expression profiles are called gene expression dysregulation domains (GEDDs) (Ref. Reference Letourneau29). Syntenic blocks along the mouse chromosomes showed that these GEDDs were conserved in induced pluripotent stem cells (IPSCs) derived from fetal fibroblasts and fibroblasts from the Ts65Dn DS mouse model in which DS is modelled through a partial trisomy comprised of a distal portion of mouse chromosome 16 and a centromeric portion of mouse chromosome 17 (Ref. Reference Letourneau29). This finding indicates a consistent influence of T21 on the transcriptome (Ref. Reference Letourneau29). The identified GEDD pattern in mice and human IPSCs is shown to be associated with chromatin modifications in T21 cells, since the altered H3K4me3 profile observed in these cells followed the GEDD pattern (Ref. Reference Letourneau29). H3K4me3 marks are positively correlated with these gene expression levels, suggesting that the transcriptomic alterations in DS are a result of chromatin modifications (Ref. Reference Letourneau29). Several genes encoding chromatin modifiers, such as HMGN1, DYRK1A, BRWD1 and RUNX1, are located on chromosome 21 and may explain the altered H3K4me3 profile (Ref. Reference Letourneau29). In line with this, B-cells from HMGN1 overexpressing mice show a global increase in hyperacetylation of histone H3K27 (Ref. Reference Mowery86). This hyperacetylation also causes transcriptional changes, which recapitulates the transcriptional changes observed in pro-B-cells of the Ts1Rhr DS mice model in which DS is modelled by triplication of 31 genes on a mouse chromosome orthologous to the DSCR (Ref. Reference Mowery86). Overexpression of several genes that encode chromatin modifiers, located on chromosome 21, may be responsible for global transcriptional changes observed in DS cells.

Characterisation of DNA methylation profiles of DS fetuses exhibited inconsistent results, however, these all demonstrated an aberrant methylation pattern in DS newborns (Refs Reference Eckmann-Scholz83–Reference Malinge85, Reference Sailani87–Reference Henneman89). Further research is warranted to clarify the role of these changes in the development of leukaemia in DS.

These findings indicate that besides genetic aberrations, epigenetic alterations can be found in DS cells, which contribute to the increased inheritable phenotypic diversity. However, additional oncogenic mutations and/or epigenetic alterations are needed to drive leukaemia in DS. The chance to acquire an oncogenic mutation causing leukaemia in DS may be increased by the enlarged phenotypic diversity that is already present early during development.

Epigenetic and transcriptional differences in DS-associated leukaemias

RNA expression profiles of TAM and ML-DS have shown to be similar to each other, and distinct from non-DS AMKL (Refs Reference Yoshida30, Reference Malinge85, Reference Bourquin90). However, RNA expression levels of recurrently mutated genes in ML-DS, such as RAD21 and EZH2, appeared to be similar in ML-DS and non-DS AMKL patients (Ref. Reference Yoshida30). Yet, there are some exceptions. Genome-wide DNA methylation analysis demonstrated that mononuclear cells of both TAM and ML-DS are epigenetically deregulated (Ref. Reference Malinge85). Malinge et al. showed that mutations in GATA1 in TAM led to focal DNA hypermethylation of specific target genes, whereas the DSCR remained hypomethylated (Ref. Reference Malinge85). The hypermethylated target genes are related to haematological development and regulation of key cellular processes, such as proliferation, growth, cell cycle regulation and cell death (Ref. Reference Malinge85). In addition, ML-DS and TAM have a similar DNA methylation pattern (Ref. Reference Malinge85). The strong similarities in gene expression and DNA methylation patterns between TAM and ML-DS suggest that the transient nature of TAM and the later development of ML-DS are unlikely to be driven by changes in epigenetic factors. However, mononuclear cells from ML-DS patients were hypomethylated compared with mononuclear cells from non-DS AMKL patients, indicating that ML-DS is epigenetically different from non-DS AMKL (Ref. Reference Malinge85).

RNA expression profiles of DS-ALL cases are very heterogeneous. Some DS-ALL cases cluster together, whereas other cases are more similar to non-DS ALL subtypes, such as the BCR-ABL1 translocated type (Refs Reference Hertzberg51, Reference Kubota91). In line with this, there are no genes on chromosome 21 that show a significantly higher expression in DS-ALL when compared with non-DS ALL (Ref. Reference Kubota91). Cases with increased CRLF2 expression are found to have a Philadelphia-like expression signature (Ref. Reference Kubota91). Nevertheless, the fusion created by the Philadelphia chromosome, BCR-ABL1, is uncommon in DS-ALL (Refs Reference Kubota91, Reference Bernt and Hunger92). The heterogeneity of DS-ALL patients is further supported by the DNA methylation profile, which shows clustering per subtype as opposed to DS-ALL clustering together (Ref. Reference Kubota91). However, in most subtypes, the RUNX1 promoter is hypermethylated in DS-ALL compared with non-DS ALL (with exception of ETV6-RUNX1 non-DS ALL) (Ref. Reference Kubota91). Also, in DS individuals without leukaemia, this promoter is, already congenitally, found to be hypermethylated (Ref. Reference Kubota91). The fact that RUNX1 is essential for B-cell differentiation suggests that RUNX1 hypermethylation is a predisposing factor for BCP-ALL in DS individuals (Ref. Reference Kubota91). Strikingly, another study demonstrated that DS-ALL and non-DS ALL are distinguishable by the overexpression of genes, marked with H3K27me3 (Ref. Reference Lane93). This study showed that H3K27me3 marks are globally suppressed in B-cells of the TsRhr1 DS mice model, which is possibly explained by overexpression of HMGN1 (Ref. Reference Lane93). This suppression, in combination with H3K4me3, leads to overexpression of genes that contribute to B-cell proliferation in Tshr1, pointing out the importance of epigenetic changes in DS-ALL (Ref. Reference Lane93).

An extra chromosome 21 in DS individuals serves as the basis of a broad variety of genetic and epigenetic changes. As mentioned before, we have previously demonstrated that fetal HSPCs of DS fetuses have an increased somatic mutation load compared with HSPCs of non-DS fetuses (Ref. Reference Hasaart20). Using WGS data of fetal HSPCs, we have constructed a model to determine the mutation rate during fetal haematopoeisis (Ref. Reference Hasaart20). For example, HSPCs of a 14 WG (weeks of gestation) DS fetus have 58 somatic mutations (95% confidence interval 40.16–75.39). When we use our model to extrapolate the mutation load for a non-DS fetus of the same gestational age, these HSPCs would have 24 (95% confidence interval 11.88–34.8) somatic mutations. This suggests that HSPCs of a 14 WG DS fetus have a 2.4-fold increase in somatic mutations (Ref. Reference Hasaart20). This increased mutation load combined with the 3.5 times higher number of HSPCs (Ref. Reference Roy94) indicates that DS fetal HSPCs have a higher risk to acquire an oncogenic mutation during fetal haematopoiesis compared with non-DS fetal HSPCs. However, the specific time-based higher incidence of AML and BCP-ALL in these children, with a lower risk of solid tumours (Ref. Reference Hasle, Clemmensen and Mikkelsen10), suggests that other factors may play a role. Additional epigenetic aberrations on top of the increased mutation load further increase the phenotypic diversity, but still do not completely explain the increased incidence of leukaemia in DS children. In conclusion, these findings support our hypothesis that the increased risk of leukaemia in DS children is not solely explained by phenotypic diversity.

The fetal liver niche in DS

During fetal development, the liver is the dominant site of haematopoiesis, which transfers to the bone marrow after birth (Ref. Reference Ivanovs96). As discussed before, the first oncogenic mutations in DS are thought to be acquired during fetal development (Ref. Reference Antonarakis2). These observations suggest that the fetal liver microenvironment, according to the somatic evolutionary model, may have a crucial role in the selection and growth of (pre)leukaemic clones that arise during fetal development in DS. Indeed, it has been shown that megakaryocyte progenitors isolated from the yolk sac or fetal liver of GATA1s mice show a hyperproliferative phenotype (Ref. Reference Li40). Notably, this effect was transient, indicating that GATA1s has a developmental stage-dependent effect (Ref. Reference Li40). Little is known about which cells within the human fetal liver support the expansion and differentiation of HSPCs during fetal haematopoiesis. Mice studies have shown that fetal liver stromal cells and Nestin + NG2 + pericytes support the proliferation of HSPCs (Refs Reference Khan97–Reference Chou and Lodish99). Moreover, fetal liver stromal cells support the expansion of megakaryocyte committed progenitors (Ref. Reference Brouard100).

Both supportive cell types express insulin growth factor-2 (IGF2), which is known to promote the expansion of HSPCs during fetal haematopoiesis (Refs Reference Khan97–Reference Chou and Lodish99). More importantly, the proliferation of mice fetal liver megakaryocyte progenitors is reliant on IGF signalling, whereas proliferation of adult megakaryocyte progenitors isolated from bone marrow is not (Ref. Reference Klusmann35). Several studies have demonstrated that IGF signalling is increased in DS (Refs Reference Klusmann35, Reference Chiang101). Recently, it has been shown that increased IGF signalling is associated with the overproduction of CD43+ haematopoietic progenitors derived from T21 IPSCs (Ref. Reference Chiang101). CD43 is the earliest marker of full haematopoietic commitment after endothelial to haematopoietic transition, which suggests that T21 promotes this transition (Ref. Reference Chiang101). Furthermore, T21 CD43+ haematopoietic progenitors are more sensitive to inhibition of IGF signalling compared with CD43+ haematopoietic progenitors in which the extra chromosome 21 is silenced by site-directed insertion of the X-inactive inactive specific transcript (XIST) gene in one copy of chromosome 21 (Ref. Reference Chiang101). Altogether, these findings suggest that T21 itself increases IGF signalling (Ref. Reference Chiang101).

IGF signalling seems to have a preserved role in the proliferation of TAM and ML-DS (pre)leukaemic blasts (Ref. Reference Klusmann35). Klusmann et al. (Ref. Reference Klusmann35) showed that the IGF receptor 1 (IGFR1) is expressed higher in primary ML-DS leukaemic blasts and ML-DS cell lines, compared with non-DS AMKL leukaemic blasts and cell lines. The same study demonstrated that inhibition of IGF signalling results in impaired proliferation of TAM and ML-DS (pre)leukaemic blasts (Ref. Reference Klusmann35). Moreover, overexpression of IGF2 in GATA1s liver mice megakaryocyte progenitors resulted in increased proliferation, whereas a negligible effect was observed in GATA1 wild-type liver megakaryocyte progenitors and GATA1s bone marrow megakaryocyte progenitors (Ref. Reference Klusmann35). These findings suggest that there might be an interaction between IGF signalling and GATA1 in the fetal liver. It has been shown that the IGF signalling pathway activates E2F target genes through activation of mTOR signalling and upregulation of MYC (Ref. Reference Klusmann35). E2F target genes play a role in the regulation of cell proliferation (Ref. Reference Chen, Tsai and Leone102). Normally, GATA1 represses the expression of E2F target genes by direct interaction with E2F transcription factors and via repression of MYC, this will result in a decrease in cell proliferation (Ref. Reference Klusmann35). However, GATA1s is not able to interact with E2F transcription factors and thereby cannot regulate the expression of E2F target genes, which results in overactive IGF signalling and hyperproliferation (Refs Reference Klusmann35, Reference Maroz36). Together, these findings support that increased IGF signalling in DS has a role in the selection and proliferation of (pre)leukaemic TAM clones within the fetal liver. The role of IGF signalling in the development of DS-ALL has, to our knowledge, not been studied yet. Intriguingly, in non-DS paediatric BCP-ALL, it has been demonstrated that IGF signalling promotes the expansion of leukaemic blasts and therefore IGF signalling might also have a role in the development of DS-ALL (Refs Reference Yamada103, Reference Vorwerk104). Another study found granulocyte-macrophage colony-stimulating factor (GM-CSF), secreted by fetal liver stromal cells, as the main growth factor supporting TAM (pre)leukaemic blasts (Ref. Reference Miyauchi and Kawaguchi105). Here, IGF signalling was not identified to have a supportive role in the growth of TAM (pre)leukaemic blasts (Ref. Reference Miyauchi and Kawaguchi105).

Altogether, these studies indicate that the fetal liver niche cells and IGF signalling pathway in DS contribute to the proliferation and selection of pre(leukaemic) blasts. This supports the hypothesis that alterations in the fetal liver microenvironment in DS can change selection dynamics in such a way that GATA1-mutated clones have a growth advantage, whereas the bone marrow does not provide this advantage. This model may explain why TAM spontaneously disappears when haematopoiesis migrates to the bone marrow after birth (Refs Reference Zipursky13, Reference Sas14). In turn, the additional oncogenic mutations observed in ML-DS leukaemic blasts may subsequently increase the phenotypic diversity among cells (Refs Reference Yoshida30, Reference Labuhn41). This increased phenotypic diversity will then result in the selection of the leukaemic clones and progression to ML-DS.

Chronic inflammation and interferon (IFN) signalling in DS

According to the immune surveillance theory, the immune system acts as a barrier to prevent cancer development (Refs Reference Palucka and Coussens106, Reference Dunn, Old and Schreiber107). However, sequencing studies have demonstrated a neutral drift for most missense and nonsense mutations in normal and cancer tissues, suggesting that selection or removal of these mutations is a stochastic process (Ref. Reference Martincorena108). Of note, not all missense mutations result in a neoantigen. This finding suggests that immune cells may not have a crucial role in the first steps of cancer initiation (Ref. Reference Martincorena108). Several studies have demonstrated that inflammation changes the selection pressure on oncogenic clones (Refs Reference Vermeulen109–Reference de Visser, Eichten and Coussens111). Not surprisingly, chronic inflammation is a hallmark of cancer (Ref. Reference Coussens and Werb112). In addition, chronic inflammation can result in genomic instability (Ref. Reference Kawanishi113), which increases the chance to acquire oncogenic driver mutations.

Abnormalities in immune function are a common characteristic of DS (Ref. Reference Ram and Chinen114). DS individuals have an increased incidence of autoimmune diseases and infections (Ref. Reference Ram and Chinen114). This observation suggests that tissues of DS individuals show signs of chronic inflammation, which may contribute to the increased leukaemia risk. Indeed, inflammatory factors related to chronic inflammation are upregulated in blood cells and plasma of children and adults with DS (Refs Reference Sullivan115–Reference Sullivan118). NK-cells and T-cells are hyperactivated, whereas cells of the myeloid compartment show signs of inflammation and increased cytotoxic potential (Refs Reference Waugh117, Reference Bloemers119). Moreover, several studies have indicated that the blood cell composition is perturbed in DS, which further supports the deregulation of the immune system (Refs Reference Roy94, Reference Tunstall-Pedoe120, Reference Chou121). The number of HSCs and megakaryocyte and erythrocyte progenitors are increased in the DS fetal liver, whereas the number of granulocyte–macrophage progenitors, B-cells, pre–pro-B cells and pro-B cell progenitors are decreased (Refs Reference Roy94, Reference Tunstall-Pedoe120, Reference Chou121). The observed depletion of B-cells, pre–pro-B cells and pro-B cells suggests a maturation defect of B-cells during DS fetal haematopoiesis. Studies in adults have shown that this depletion was preserved over time (Refs Reference Waugh117, Reference de Hingh122).

Altogether, these findings suggest that the haematopoietic system of both children and adults with DS is altered and shows signs of chronic inflammation. However, if an altered immune system and chronic inflammation contribute to the increased incidence of leukaemia in children with DS, it remains unclear why this risk decreases tremendously during adolescence, since alterations of the immune system are preserved during adulthood. Moreover, it remains an open question why this increased risk is limited to the haematopoietic system because chronic inflammation is associated with increased cancer risk in general and DS individuals do not experience an increased risk of other malignancies besides paediatric leukaemia (Refs Reference Hasle, Clemmensen and Mikkelsen10, Reference Coussens and Werb112).

Over the last years, it has become evident that hyperactive immune signalling can induce changes in the haematopoietic system, such as the increased proliferation of HSPCs and a biased differentiation towards the myeloid lineage (Ref. Reference Mirantes, Passegué and Pietras123). Overactive IFN signalling is associated with increased proliferation of fetal and adult HSCs and myeloid skewing in adults (Refs Reference Essers124–Reference Li128). Four of the six IFN receptors are located on chromosome 21 and proteomics and RNA sequencing approaches have shown consistent activation of IFN signalling in immune cells of adults with DS (Refs Reference Sullivan115, Reference Waugh117, Reference Sullivan118). These findings suggest that overactive IFN signalling may contribute to the observed changes in DS haematopoiesis. On the contrary, it has been shown that IFN signalling has an antiproliferative effect on megakaryocyte progenitors, which are enriched in DS compared with karyotypically normal newborns (Refs Reference Woo129, Reference Wang130). IFN signalling is increased in adult bone marrow megakaryocyte progenitors isolated from wild-type mice compared with megakaryocyte progenitors isolated from the fetal liver (Ref. Reference Woo129). Indeed, both bone marrow GATA1s and GATA1 wild-type megakaryocyte progenitors, isolated from IFNAR1−/− mice, show a hyperproliferative phenotype (Ref. Reference Woo129). However, hyperproliferation was more pronounced in GATA1s IFNAR1−/− mice (Ref. Reference Woo129). In line with this, IFN-alpha stimulation resulted in a more pronounced decrease in proliferation of GATA1s megakaryocyte progenitors as compared with GATA1 wild-type megakaryocyte progenitors (Ref. Reference Woo129). Taken together, these findings imply that there is an interplay between GATA1s and IFN signalling, which may contribute to the proliferation and selection of (pre)leukaemic blasts in the fetal liver and the spontaneous remission of TAM (Ref. Reference Woo129). Indeed, (pre)leukaemic blasts from TAM patients show a higher expression of IFN signalling genes, which is, considering the anti-proliferative effect, consistent with the spontaneous remission of TAM within three months after birth (Ref. Reference Woo129). Nonetheless, leukaemic blasts from ML-DS patients also highly express IFN signalling genes, indicating that additional acquired oncogenic mutations in ML-DS are needed for leukaemic progression (Ref. Reference Woo129).