Advances in germline predisposition to acute leukemias and myeloid neoplasms

Abstract

While much work has focused on the elucidation of somatic alterations that drive the development of acute leukemias and other hematopoietic diseases, it has become increasingly recognized that germline mutations are common in many of these neoplasms. In this review, we will highlight the different genetic pathways impacted by germline mutations that can ultimately lead to the development of familial and sporadic hematological malignancies, including acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML) and myelodysplastic syndromes (MDSs). Many of the genes disrupted by somatic mutations in these diseases (e.g. TP53, RUNX1, IKZF1, and ETV6) are the same as those which harbor germline mutations in children and adolescents who develop these malignancies. Moreover, the presumption that familial leukemias only present in childhood is no longer true, in large part due to the numerous studies demonstrating germline DDX41 mutations in adults with MDS and AML. Lastly, we will highlight how different cooperating events can influence the ultimate phenotype in these different familial leukemia syndromes.

Table of Contents Summary:

This Review highlights the different genetic pathways impacted by germline mutations that can lead to the development of familial and sporadic hematological malignancies, with a particular focus on recent advances in acute lymphoblastic leukemia, acute myeloid leukemia and myelodysplastic syndromes.

INTRODUCTION

Over the last decade, large scale genomic studies have described the landscape of nearly all major types of cancer in both adults and children¹. In addition to providing important diagnostic and prognostic information from the somatic alterations, evaluation of non-tumor or germline material through comprehensive sequencing approaches has transformed our view of how germline variation influences the development of cancer. The incidence of presumably pathogenic germline mutations in children with cancer was estimated to be 8.5% in a large study involving over 1000 children². These data have been supported by follow-up studies on different pediatric cohorts^3,4. Recent studies on over 10,000 adults within The Cancer Genome Atlas (TCGA) data set reported that approximately 8% of adult patients have a pathogenic germline variant⁵. Many of the patients with these germline mutations lack a family history consistent with a cancer predisposition syndrome². These findings not only highlight the importance of comprehensive sequencing of normal tissue from all cancer patients, but also that familial cancer syndromes are much more common than previously recognized.

Genes in a number of different pathways are impacted by germline mutations in individuals with a predisposition to hematologic abnormalities and neoplasms. The recent inclusion of the “Myeloid Neoplasms with Germline Predisposition” category in the revised 4^th Edition of the WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues further emphasizes the increased recognition and importance of germline evaluation in patients with myeloid tumors⁶. Many of the known germline predisposition mutations affect master transcription factors important in lineage development, such as RUNX1 or GATA2; the genes encoding these transcription factors also harbor recurrent somatic mutations in children and adults with myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML)⁷ (Figure 1). However, it is becoming increasingly clear that these observations now extend to lymphoid neoplasms^8–13. Like RUNX1 in myeloid neoplasms, the genes encoding the transcription factors PAX5 and IKZF1 (also known as Ikaros) are examples of genes that may be altered as germline events predisposing to acute lymphoblastic leukemia (ALL), disease-initiating and subtype defining somatic events, or secondary events that cooperate in leukemogenesis.

Figure 1. Association of germline mutations in genes encoding transcription factors with lineage development.

The representative schematic of the hierarchy of hematopoietic development demonstrates the main hematopoietic lineages that are dysregulated in patients with germline mutations in different transcription factors important in hematopoietic growth and differentiation. Germline alterations in ETV6 and TP53 can lead to hematopoietic neoplasms in both myeloid and lymphoid lineages, while disruptions in the genes encoding other hematopoietic transcription factors result in more lineage-restricted malignancies (e.g. B-ALL with PAX5 or AML with CEBPA). tMN: Therapy-related myeloid neoplasm; LH: Low hypodiploid.

Recent comprehensive sequencing studies have expanded the spectrum of predisposition genes in families with leukemia beyond those encoding transcription factors. Not only have these studies opened up new pathways of mechanistic investigation, but they have also shown that clinical suspicion for a familial predisposition syndrome, along with genetic screening, should not be restricted to children or adolescents. For example, germline mutations in the gene encoding the RNA helicase DDX41 define families with an increased risk of MDS and AML that can present in adulthood¹⁴. Further, germline mutations in SAMD9L and SAMD9, two interferon-inducible genes implicated in endocytosis, have recently been shown to underlie MDS as well as other multi-system disorders associated with MDS with monosomy 7^15–19. DDX41, SAMD9 and SAMD9L are involved in innate immunity and antiviral responses, suggesting new pathways that predispose families to hematological malignancies.

This review will address recently described predisposition genes (Table 1) in familial leukemias and hematologic abnormalities, particularly MDS, myeloid proliferations and acute leukemias, as well as the cooperating events that can lead to the fully transformed acute leukemia phenotype. The review is not designed to serve as a comprehensive list of all described predisposition genes in hematological malignancies (such as those important in myeloma or lymphoma) and detailed reviews on specific genes or diseases (i.e. bone marrow failure syndromes) and clinical management are already available^20–23. For clarity, we have grouped genes by different differentiation lineages of the resulting hematological malignancies, while recognizing that these categories at times may be somewhat artificial and not truly exclusive.

Table 1.

Common Germline Predispositions in MDS and Acute Leukemias

Gene or Pathway*	Hematologic Malignancy	Additional Hematologic Abnormality	Other Malignancies	Additional Symptoms	Syndrome||	Refs.
CEBPA	AML	No common abnormalities noted	No other malignancies are common	None observed	Familial acute myeloid leukemia with mutated CEBPA	40–42
DDX41	MDS and AML; less commonly lymphoma	Cytopenia	No other malignancies are common	None observed	Familial AML with mutated DDX41	14,63,64
ETV6	ALL, MDS, AML	Thrombocytopenia and decreased platelet function	No other malignancies are common	None observed	Thrombocytopenia 5 (OMIM 616216)	144,145,215
GATA2	MDS and AML	Immunodeficiency, bone marrow failure, monocytopenia and B cell lymphopenia	No other malignancies are common	Lymphedema, pulmonary alveolar proteinosis and congenital deafness	Emberger Syndrome (OMIM 614038) and MonoMac Syndrome (OMIM 614172)	46,245,246
IKZF1	B-ALL, less commonly T-ALL	Immunodeficiency	No other malignancies are common	Autoimmunity	Not described	8,206,207
PAX5	B-ALL	No common abnormalities noted	No other malignancies are common	None observed	Not described	12,166,178
RAS–MAPK pathway: CBL, NF1 and PTPN11	JMML, ALL, AML	JMML-like proliferation with spontaneous regression	Neurofibroma and embryonal rhabdomyosarcoma	Cardiac abnormalities, skin manifestations, growth retardation and facial dysmorphologies	Noonan Syndrome (OMIM 163950); NF1 (OMIM 162200) and other RASopathies	100;95,104;215
RUNX1	MDS and AML, less commonly T-ALL	Thrombocytopenia and decreased platelet function	No other malignancies are common	None observed	Familial platelet disorder with propensity for myeloid malignancy (OMIM 601399)	30,33,36
SAMD9	MDS and AML with monosomy 7	Bone marrow failure	No other malignancies are common	Normophosphatemic familiar tumoral calcinosis; congenital adrenal hypoplasia, enteropathy and genital abnormalities	MIRAGE Syndrome (OMIM 617053) and Myelodysplasia and Leukemia Syndrome with Monosomy 7(OMIM 252270)	16,17,71
SAMD9L	MDS and AML with monosomy 7	Systemic autoinflammatory disease and bone marrow failure	No other malignancies are common	Ataxia	Ataxia Pancytopenia Syndrome (OMIM 159550) and Myelodysplasia and Leukemia Syndrome with Monosomy 7 (OMIM 252270)	18,19,70
TP53	Low hypodiploid ALL, Therapy Related Myeloid Neoplasm	No common abnormalities noted	Osteosarcoma, breast cancer, soft tissue sarcoma, brain tumors and adrenocortical carcinoma	None observed	Li-Fraumeni syndrome (OMIM 151623)	143,150,181
Trisomy 21	AML and ALL	Transient abnormal myelopoiesis	No other malignancies are common	Mental retardation, cardiac abnormalities and facial dysmorphologies	Down syndrome (OMIM 190685)	135

^*Only the major genes and pathways discussed in this Review are included in this table.

^||Online Mendelian Inheritance in Man (OMIM) numbers are given for each syndrome, when applicable: https://www.omim.org/

Predisposition to MDS and AML

Alterations in transcription factors

RUNX1

Somatic alterations in RUNX1 are among the most common mutations in both adults and children with ALL, AML or MDS and these include recurrent fusions in B lymphoblastic leukemia (B-ALL) (ETV6-RUNX1) and AML (RUNX1-RUNX1T1), as well as loss of function mutations in myeloid and less commonly lymphoid neoplasms^24–29. In addition, germline mutations in RUNX1 define familial platelet disorder with predisposition to myeloid malignancy (FDP-MM). FDP-MM is an autosomal dominant disorder characterized by variable penetrance of quantitative and/or qualitative platelet defects with a tendency to develop hematological malignancies. When symptomatic, these patients present with thrombocytopenia and although the platelets are normal in size they have decreased function, typically as a result of decreased dense granules³⁰. The germline mutations in RUNX1, first described in 1999 by Song and colleagues³¹, result in loss of RUNX1 protein function, by point mutations that produce proteins that act as dominant negatives, as well as nonsense or frameshift mutations and larger deletions³⁰. The overall lifetime risk of progression to a myeloid disease (or rarely T lymphoblastic leukemia (T-ALL)) is 44%⁷, which most commonly occurs in adulthood³². This risk appears to be higher in patients with missense mutations, presumably due to a higher ability of the resulting proteins to inhibit transactivation of wild-type RUNX1 through a dominant negative activity, as compared to loss of function alleles^30,33. However, eventual progression to a frank neoplasm is associated with acquisition of somatic mutations in the remaining wild-type RUNX1 allele, as well as GATA2 mutations³⁴, and less commonly other genes recurrently mutated in AML and MDS, such as CBL (which encodes an E3 ubiquitin ligase), DNMT3A (which encodes a DNA methyltransferase), KRAS and FLT3 (which encodes a receptor tyrosine kinase)^{30,32,35–37}. Notably, patients with FDP-MM also appear to have a higher incidence of clonal hematopoiesis [G] at a younger age³⁵.

CEBPA

AMLs with biallelic mutations in CEBPA (which encodes CCAAT/enhancer binding protein alpha) are recognized as a unique subtype with good outcome^38,39 and it has been shown that nearly 10% of these cases also harbor a germline CEBPA mutation⁴⁰. These germline CEBPA mutations lead to the development of AML with a near complete penetrance, typically in the second or third decade of life and can occur without a preceding MDS or cytopenic phase [G]^7,41. These germline mutations are commonly frameshift or nonsense mutations near the amino terminus of the encoded protein. This results in a truncated 30 kD C-terminal protein through usage of an internal initiation codon that has dominant negative activity^38,42. However, C-terminal germline mutations have been reported⁴³, including a recent publication describing a multigenerational family with AML⁴⁴. Progression to AML is frequently associated with an acquired mutation in the remaining wild-type CEBPA allele^40,41. Surprisingly, Tawana and colleagues showed that when these patients have disease recurrence after chemotherapy, they do so with new clones that have a different spectrum of acquired mutations, including new somatic CEBPA mutations, demonstrating that these second leukemias are not true relapses⁴¹. This pattern of progression likely underlies the clinical observations that patients with AML and germline CEBPA mutations have a good outcome, yet are prone to second leukemias that continue to be sensitive to chemotherapy, unlike true relapsed disease.

GATA2

Perhaps more so than predisposition syndromes associated with mutations in other transcription factor-encoding genes, patients with germline GATA2 mutations have a marked phenotypic variation with MDS and AML as a main clinical feature. However, these patients can also present with syndromes characterized by mycobacterial infections, monocytopenia [G], B-lymphopenia and pulmonary alveolar proteinosis [G] (MonoMac Syndrome) or primary lymphedema, cutaneous warts and sensorineural deafness [G] (Emberger Syndrome)⁴⁵. Recent studies have identified germline GATA2 mutations in up to 15% of pediatric MDS cases with isolated bone marrow disease, and these germline variants are more frequent in higher grade MDS⁴⁶. Most commonly, these germline GATA2 mutations are missense mutations that affect the second zinc finger of the protein or true null mutations that collectively result in loss of protein function and haploinsufficiency⁴⁷ (Figure 2a). However, other mechanisms of germline GATA2 disruption have been reported, such as disruption of intronic regulatory sequences within intron 4 (ref. ⁴⁸), synonymous coding alterations that disrupt transcript splicing^49,50 and an in-frame insertion that alters the spacing between the GATA2 zinc fingers⁵¹. Progression to AML is commonly associated with acquired mutations in the Polycomb group member ASXL1, as well as monosomy 7 (ref. ⁵²) or trisomy 8, although marked clinical variability exists within families^53–55, some of which has been linked to patterns of allele-specific expression⁵⁶.

Figure 2. Genetics of cancer predisposition genes in pediatric myeloid disorders malignancies.

Schematics and domain architectures of proteins encoded by genes that have germline and somatic mutations in pediatric myeloid disorders, including GATA2 (a), SAMD9 (b), and SAMD9L (c) are shown. Refseq identifiers for transcripts used to generate the protein schematics are shown under the gene symbols. Listed germline mutations in GATA2 were collected from ref ⁴⁶ and from available Pathogenic and Likely Pathogenic variants from ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/). SAMD9 and SAMD9L variants were taken from refs. ^{15,17–19,46,70–73}. ZnF, zinc finger; NTPase, nucleoside triphosphatase; SAM, sterile alpha motif; AlbA, acetylation lowers binding affinity; SIR-2, silent information regulator 2; APAF-1, apoptotic protease-activating factor, TPR tetratricopeptide repeat; OB Fold, oligonucleotide/oligosaccharide-binding fold. SAMD9 and SAMD9L domain structure obtained from ref. 80.

Alterations in non-transcription factors

DDX41

Members of the DExD/H-box helicase family, including DDX41, have been implicated in numerous aspects of RNA biology⁵⁷, as well as antiviral immunity via regulation of signaling pathways or serving as sensors for viral nucleic acids⁵⁸. DDX41 in particular was originally identified as an intracellular DNA sensor in myeloid dendritic cells that binds cytosolic double-stranded DNA and triggers a signaling cascade that ultimately leads to the production of type 1 interferons via the stimulator of interferon genes (STING) pathway^59,60. In addition to the well-documented role of DDX41 in innate immunity, somatic and germline mutations in DDX41 have recently been described in patients with myeloid tumors, and less commonly lymphoid malignancies, including follicular lymphoma and Hodgkin lymphoma^14,61–64. Unlike the majority of other constitutional disorders, patients with germline DDX41 mutations present in adulthood. In fact, the first described family consisted of a father in his 70s and a brother and sister in their 40s¹⁴. Further, there is an estimated 3:1 male predominance in carriers with myeloid tumors⁶⁵. The reported germline alterations are commonly frameshift mutations, most commonly a D140fs mutation. Emerging data is also suggesting that known germline alleles are associated with different ethnicities^14,66, including preliminary data of an A500fs mutation observed in Asian populations⁶⁷. Somatic missense mutations in DDX41 frequently occur in the second allele in patients with a germline DDX41 loss of function mutation, the most common of which is R525H¹⁴. Secondary mutations in TP53 (which encodes p53) have likewise been described in patients with advanced MDS or AML who have germline DDX41 mutations⁶⁵. Clinically silent carriers have also been noted and due to the rarity of these mutations, the overall disease penetrance is currently not known^14,63,64.

While evidence is mounting that DDX41 is important in cell growth, the overall function of this protein in the context of hematopoietic cells and how these mutations lead to hematopoietic neoplasms is still under investigation. Early biochemical studies identified components of the U2 and U5 spliceosome in complex with DDX41 and demonstrated that disruption of DDX41 alters RNA splicing and processing¹⁴, which are pathways known to be dysregulated in MDS⁶⁸. However, non-RNA splicing functions of DDX41, such as its role as a cytosolic censor in the STING pathway to influence interferon signaling, may also be important in the development of MDS in these patients. Functional studies have shown that the common DDX41^R525H missense mutation results in a defect in ribosomal biogenesis and cell cycle progression⁶⁹.

SAMD9 and SAMD9L

Germline mutations in SAMD9 or SAMD9L, two interferon-inducible genes located on chromosome 7, were initially shown to cause a variety of multisystem syndromes characterized by neurological and/or endocrine abnormalities, as well as MDS with monosomy 7^17,18,70,71. Recent studies have shown that germline mutations in these genes can also be found in isolated and familial pediatric MDS and AML^15,16,19,72, as well as inherited bone marrow failure syndromes⁷³ and potentially adult MDS⁷⁴ with a strong association with whole or partial deletions involving chromosome 7. Although the function of the SAMD9 and SAMD9L proteins in hematopoietic cells is not entirely clear, they appear to be involved in endocytosis and cytokine signaling⁷⁵, although like DDX41, a role in antiviral response has also been reported^76–78. Specifically, SAMD9 and SAMD9L are known host restriction factors for Pox virus infections⁷⁹. The germline mutations associated with MDS are heterozygous missense (Figure 2b–c) mutations that lead to growth arrest when exogenously expressed in cells. This phenotype is an enhancement of the effects of wild-type SAMD9 or SAMD9L, suggesting that these may be gain of function mutations¹⁷. Although little is known about the biochemical activity of the SAMD9 and SAMD9L proteins and their domain structure, these missense mutations do appear to cluster in the second half of the proteins within or near a putative P-loop nucleoside triphosphatase domain^19,80 (Figure 2b–c).

Surprisingly, when patients present with monosomy 7, the clone that eventually grows out has consistently lost the copy of chromosome 7 with the mutant allele. This haploinsufficiency of a number of genes on chromosome 7 (e.g. EZH2, SAMD9, SAMD9L, CUX1, and KMT2C (also known as MLL3)) perturbs hematopoiesis and can ultimately lead to MDS and AML^81,82. Many other patients that do not develop monosomy 7 acquire additional somatic mutations in SAMD9 or SAMD9L that abrogate the function of the mutant protein (acquired revertant mutations) or duplicate the wild-type allele through a uniparental disomy event^15,19,71. These observations suggest that there is strong selective pressure to not express the mutant allele and that the clinical phenotype in these individuals is heavily influenced by the method of mutant inactivation. However, clinically silent carriers of pathogenic germline variants that lack known rescue mutations have been observed, implying that other rescue mechanisms exist or that these germline mutations are not fully penetrant^16,19.

To date, there appears to be an increased risk of abnormal blood counts, MDS and AML in children compared to adult mutation carriers, suggesting that developmental stage may influence clinical phenotype. However, adults with hematologic disease have been reported infrequently¹⁵. When patients progress to advanced disease they acquire additional somatic mutations in genes commonly mutated in pediatric and adult MDS and AML, including SETBP1 (which encodes SET-binding protein 1), RUNX1, or ETV6^15,19. In contrast to the presumed gain of function missense mutations in MDS, germline loss of function mutations in SAMD9L have recently been described in children with a systemic autoinflammatory disease, yet no apparent risk of MDS or AML⁸³. Importantly, considering the fairly recent discovery and characterization of these mutations and their associated clinical syndromes, their natural history remains to be determined, especially regarding disease penetrance and phenotypes.

Additional Predisposing Variants

Due to advances in next-generation sequencing approaches, the list of rare variants associated with hematological disorders and an increased risk of MDS and AML is rapidly expanding. This includes recently described mutations in hematopoietic transcription factors like MECOM (also known as EVI1)^84,85, as well as variants in genes not directly involved in gene regulation. For example, germline mutations in the 5’ regulatory region of ANKRD26, which encodes a protein that promotes MAPK signaling and megakaryocyte development⁸⁶, been reported as a cause of thrombocytopenia-2 [G]^87–89 and these families have an approximately 30-fold increased risk of developing AML^88,90. Further, deficiency in MBD4, which regulates 5-methylcytosine deamination, is associated with early onset AML⁹¹. Although beyond the scope of this review, several inherited bone marrow failure syndromes are also associated with variable risks of MDS and AML (Box 1) (ref. ^20–23). These mutations target a myriad of cellular pathways, including ribosomal biogenesis, telomere maintenance and DNA repair.

Box 1. Bone Marrow Failure Syndromes.

Inherited bone marrow failure syndromes (IBMFSs) are a heterogeneous group of genetic disorders typically associated with a decreased production of bone marrow cells, specific clinical phenotypes and variable risk of developing MDS or AML. The different types were traditionally distinguished based on the presence of classical physical manifestations, such as the triad of abnormal nails, reticular skin pigmentation, and oral leukoplakia in dyskeratosis congenita²⁴⁷. However, as sequencing approaches are becoming more mainstream, more and more patients with pathogenic variants in genes implicated in IBMFSs are presenting with unusual or non-classic clinical symptoms and the spectrum of genes implicated in IBMFSs are also expanding^73,248. Within this group of disorders, Fanconi Anemia, an X-linked or autosomal recessive disorder caused by mutations in a group of genes that repair DNA crosslinks has perhaps the highest risk of progression to MDS or AML (cumulative incidence of AML at age 40 years is ~15%–20% and the cumulative incidence of MDS at age 50 is~ 40%)²⁴⁹. Although less common, the risk of developing a myeloid neoplasm is also increased in other IBMFSs like severe congenital neutropenia, dyskeratosis congenita, Shwachman-Diamond syndrome and Diamond Blackfan anemia.

Predisposition to Myeloid Proliferations

RASopathies

Aberrant activation of the RAS–MAPK pathway is a common finding in hematologic malignancies⁹². This is especially true for juvenile myelomonocytic leukemia (JMML), a rare but highly aggressive malignancy in childhood characterized by a marked expansion of granulocytic and monocytic cells in the peripheral blood, spleen, bone marrow and other tissues. Mutations in NF1 (which encodes neurofibromin 1), PTPN11 (which encodes the non-receptor protein tyrosine phosphatase SHP2 that regulates signaling through growth factor receptors), NRAS, KRAS or CBL, which are all part of the RAS pathway, have been identified in nearly 90% of JMML cases^93–96. While many of these are somatic mutations, a subset of patients present with JMML, or JMML-like proliferations, as part of a constitutional syndrome associated with aberrant activation of the RAS pathway (RASopathy). In particular, children with neurofibromatosis type 1 are 200–500 times more likely to develop JMML, and bone marrow cells from these patients typically demonstrate loss of the wild-type NF1 allele and duplication of the mutant allele (acquired uniparental isodisomy)^97,98. Likewise, germline mutations in the gene encoding CBL, which regulates the turnover of components in the RAS pathway, is present in 10–15% of JMML patients⁹⁹. Like NF1-mutant patients, the mutant CBL allele is duplicated in JMML cells through uniparental isodisomy. Curiously, unlike patients with germline NF1 mutations, those with germline CBL mutations show a high rate of spontaneous resolution of JMML and lack additional genetic alterations associated with disease progression^100,101. Despite the decrease in symptoms related to JMML, these patients can ultimately develop vascular complications, including hypertension and optic atrophy^96,.

Noonan Syndrome, the most common of the RASopathies, is frequently associated with germline mutations in PTPN11. Somatic PTPN11 mutations are also the most common alteration in JMML, and children with JMML who have somatic PTPN11 mutations have a poor prognosis with higher rates of relapse⁹⁹. Despite the aggressive nature of neoplasms with somatic PTPN11 mutations, Noonan Syndrome patients with PTPN11 germline mutations commonly develop a JMML-like disease that regresses spontaneously over time¹⁰². These clinical differences between the somatic and germline variants, which occur at different positions within PTPN11, are likely due to differences in the phosphatase activity of the resulting SHP2 protein¹⁰³. Importantly, patients with Noonan Syndrome also have an increased risk of ALL¹⁰⁴. In addition to PTPN11, germline mutations in other genes that encode members of the RAS pathway, including RIT1¹⁰⁵, RRAS¹⁰⁶ and SOS1¹⁰⁷ have also been described in patients with Noonan syndrome, some of which also have features of JMML. Likewise, the spectrum of RASopathies continues to expand with germline mutations in BRAF, SHOC2, HRAS and others being associated with distinct clinical syndromes. Despite activation of the same pathway, many of these lack a propensity to develop hematologic malignancies¹⁰⁸.

While most studies have focused on the impact of secondary mutations in non-syndromic JMML, it is likely that similar pathways are disrupted in those with constitutional syndromes. This includes deletions of chromosome 7 and acquired mutations in JAK3 (which encodes Janus kinase 3), SETBP1 and components of Polycomb repressive complex 2 (PRC2)^109–112 along with distinct patterns of methylation^113,114.

Myeloid Abnormalities in Down Syndrome

Children with Down Syndrome (DS) have an increased risk of developing both ALL and AML, however, the risk is much higher for AML with nearly a 150-fold increase in the first 5 years of life¹¹⁵. These AMLs are commonly megakaryoblastic in morphology and both AML and MDS have been described and collectively are grouped as “Myeloid Leukemia associated with DS (ML-DS)” in the current WHO guidelines. The ALLs that develop in children with DS are frequently characterized by alterations in cytokine receptors or kinase signaling pathways (e.g. Philadelphia chromosome-like ALL [G], also known as BCR-ABL1-like or Ph-like ALL), notably with cytokine receptor-like factor 2 (CRLF2) dysregulation¹¹⁶. Unlike non-DS acute megakaryoblastic leukemias (AMKLs), DS patients with AMKL have a favorable outcome¹¹⁷. At birth, approximately 5–10% of children with DS develop Transient Myeloproliferative Disease (TMD), which is characterized by circulating megakaryoblasts, abnormal blood counts and hepatosplenomegaly¹¹⁵. In addition to trisomy 21, patients with both TMD and ML-DS commonly have a somatic GATA1 mutation^118,119. These somatic GATA1 mutations result in a truncated GATA1 protein (termed GATA1s) that disrupts GATA1-mediated regulation of other transcription factors important in megakaryocyte development and differentiation^119,120. Although TMD commonly resolves spontaneously, nearly 20–30% of children with TMD will develop ML-DS within the first 4 years of life¹¹⁵. The progression from TMD to ML-DS reflects evolution from a TMD clone that has acquired additional somatic mutations, especially in genes that influence the cohesin complex, epigenetic regulation and the RAS pathway¹²¹. Despite the clear association of trisomy 21 and ML-DS, the mechanism of how altered dosage of genes on chromosome 21 results in the leukemic phenotype remains under investigation, with particular focus on RUNX1, ERG (which encodes a transcription factor) and DYRK1A (which encodes a dual-specificity kinase) (reviewed in ref. ¹²²).

Other Predisposing Variants for MPNs

Like MDS, some MPNs may be part of a germline predisposition as relatives of individuals with an MPN are more likely to also develop an MPN as compared to the general population¹²³. In fact, certain common polymorphisms associate with an increased risk of MPN. Notably a germline haplotype in the 3’ region of the JAK2 gene (GGCC haplotype, also known as 46/1) is linked with a three to four-fold increase in acquiring the canonical JAK2^V617F mutation in cis as well as an overall increase in non-JAK2 mutated MPNs^124–127. Despite the association, debate remains regarding the potential mechanistic role for the 46/1 haplotype in the development of MPNs. Potential explanations include the “fertile ground hypothesis” in which cells with the haplotype have a clonal advantage after they also acquire an oncogenic alteration in the JAK–signal transducer and activator of transcription (STAT) pathway and the “hypermutability hypothesis” in which the haplotype leads to genomic instability within the JAK2 locus¹²⁸. Genome-wide association studies (GWAS) have also shown that polymorphisms in TERT (which encodes telomerase reverse transcriptase), MECOM and in the intergenic region between HBS1L and MYB are associated with an increased risk of MPNs^129,130. In addition, germline variants in RBBP6¹³¹ (which encodes an E3 ubiquitin ligase), SH2B3 (encoding SH2B adaptor protein 3 (also known as LNK))^132,133 and duplications of ATG2B (which is involved in autophagy) and GSKIP (which encodes a glycogen synthase kinase 3β-interacting protein)¹³⁴ have also been described.

Genetic predisposition to ALL

Historically, outside of the setting of defined cancer predisposition syndromes including DS¹³⁵, Ataxia-Telangiectasia [G] and Nijmegen breakage syndrome [G]^136,137, ALL has been considered to have had a relatively low heritable predisposition, with few familial cases and low twin or sibling concordance. Rarely, constitutional chromosomal rearrangements predispose to ALL, such as the Robertsonian translocation [G] r(15;21) and ring chromosome [G] 21 which are strongly associated with ALL with intrachromosomal amplification of chromosome 21¹³⁸. However, studies utilizing single nucleotide polymorphism (SNP) array-based case-control GWAS and genome sequencing of families and cohorts of sporadic ALL have shown that non-coding variants subtly influence risk of developing ALL in a manner that may be associated with disease subtype, outcome and genetic ancestry^13,139–142. Furthermore, although familial ALL is relatively uncommon, these studies have identified highly deleterious non-silent germline variants, most commonly in genes encoding tumor suppressors and transcription factors, that often predispose to specific subtypes of ALL^{8,12,143–145}, and similarly genes implicated in predisposition to familial ALL commonly also harbor deleterious non-silent germline variants in sporadic ALL that does not have evidence of familial inheritance^8,10. Finally, structural variants represent an important class of predisposing genetic alteration^146–149, and germline variants may influence anti-leukemic drug response and outcome, as well as disease susceptibility^8,150.

Inherited SNPs and ALL susceptibility

GWAS have identified over 10 genetic loci harboring SNPs that are associated with a modest (typically up to 2-fold) increase in ALL risk. The most commonly implicated loci involve genes encoding transcription factors, particularly those known to encode regulators of hematopoietic and early lymphoid development and tumor suppressors: ARID5B, BAK1, CDKN2A, CDKN2B, CEBPE, ELK3, ERG, GATA3, IGF2BP1, IKZF1, IKZF3, LHPP, MYC, PTPRJ, TP63 and the BMI1-PIP4K2A locus^{9,13,139,140,151–159}. Although the risk of ALL associated with a single variant is low, cumulatively the risk genotypes may increase ALL risk up to 9-fold^153,154. Several loci are associated with distinct ALL subtypes and ancestral groups, such as GATA3 with Ph-like ALL in those of Hispanic ancestry^13,160, TP63 and PTPRJ with ETV6-RUNX1 ALL¹⁵⁹ and ERG with TCF3-PBX1 ALL and African-American ancestry^9,161. A variant in USP7 (which encodes a deubiquitinase) is associated with risk of T-lineage ALL¹⁶¹; this gene is also somatically altered in T-ALL, particularly in cases with deregulation of the transcription factor TAL1¹⁶². The enrichment for associations with loci at or near genes with key roles in early hematopoietic and lymphoid development (IKZF1, IKZF3, CEBPE, ERG)¹⁶³, many of which are also targeted by deleterious somatic alterations in ALL¹⁶⁴, suggests that the variants influence timing or amplitude of expression of transcription factor-encoding genes and thus, lymphoid maturation. This is supported by data correlating SNP genotype with gene expression data in cell lines^13,153,156.Alternatively, integration of multimodal data (SNP genotype, gene expression, chromatin accessibility and chromatin conformation) suggests that some SNPs mark promoter or enhancer elements, and may loop to regulate expression of local or more distant loci¹⁴⁰. Preliminary data indicates that GATA3 variants, which are associated with Ph-like ALL, and specifically, dysregulating rearrangements CRLF2, may promote rearrangement by creating a strong GATA3 enhancer and perturbing chromatin topology at the CRLF2 locus¹⁶⁵.

PAX5

Two reports described germline variants in PAX5 in several kindreds with heritable B-progenitor ALL associated with loss of the second, wild-type copy of PAX5 through deletion of chromosome 9p or the formation of a dicentric chromosome 9^12,166 (Figure 3a). PAX5 encodes a DNA-binding transcription factor required for B-lymphoid lineage maturation¹⁶⁷. PAX5 is a common target of somatic mutation in B-ALL, being altered by chromosomal rearrangement, copy number alterations, or sequence mutations in approximately one third of B-ALL cases, resulting in loss of PAX5 function and arrested B-lymphoid development^168–178. PAX5 is involved in somatic chromosomal rearrangements with over 20 partner genes^175,178. Focal intragenic, partial and broad deletion of PAX5 is observed across the spectrum of B-ALL cases, with some subtype specificity – for example, focal internal deletions, but not sequence mutations, are present in one third of ETV6-RUNX1 cases^168,177. Somatic PAX5 missense, frameshift and nonsense mutations are present across the PAX5 protein, with clustering in the paired domain and C-terminal transactivating domains, several of which define B-ALL subtypes, notably PAX5-P80R¹⁷⁸ (Figure 2c).

Figure 3. Genetics of cancer predisposition genes in pediatric acute lymphoblastic leukemia.

Schematics and domain architectures of proteins encoded by genes harboring germline and somatic mutations in pediatric acute lymphoblastic leukemia (ALL), including PAX5 (a), TP53 (b) and IKZF1 (c). The germline mutations are shown below the protein schematic and the somatic mutations on top. (a) PAX5 somatic variants are predominantly missense variants in the DNA binding domain and truncations that remove the transactivation domain. By contrast, a single residue has been observed mutated in familial ALL, G183S in the octapeptide domain. (b) Germline TP53 variants are shown for hypodiploid ALL, which involve the DNA binding domain and nuclear localization sequences. (c) Differences in distribution in IKZF1 variants are observed between germline and somatic variants. Somatic variants are most commonly N terminal truncating mutations, DNA binding domain missense mutations (most commonly in the second and third DNA-binding zinc fingers (ZnFs), which are required for DNA binding), and distal truncating and C-terminal ZnF missense mutation. In contrast, germline variants are predominantly located outside of ZnFs (possibly as these would not be tolerated as germline events) but in functional assays such as protein localization, adhesion and drug resistance, are commonly as deleterious or more deleterious than somatic variants. Germline data taken from refs ^8,12,143; somatic mutation data updated from ref. 180.

A single hotspot at G183 of PAX5 harbors missense alterations, most commonly PAX5-G183S observed in familial ALL^12,166 (Figure 3a). Leukemic cells typically exhibit loss of wild type PAX5, either by deletions of 9p (PAX5 is located at 9p13), or by the formation of iso-/dicentric 9q chromosomes, with few other recurrent somatic mutations. G183 is located in the octapeptide domain of PAX5, which may mediate Groucho family -mediated transcriptional repression¹⁷⁹. The G183S variant is subtly hypomorphic in assays of transcriptional activation and DNA binding, suggesting that it is tolerable in the germline, heterozygous state, but requires inactivation of the wild type PAX5 allele to promote leukemogenesis¹⁷⁸. This model of biallelic PAX5 alteration in leukemogenesis is supported by mouse models, in which chemical or retroviral mutagenesis of Pax5^+/− mice as well as B-ALL spontaneously arising in heterozygous Pax5^P80R/+ knock in mice is accompanied by deletion or mutation of the wild type Pax5 allele^178,180. A minority of cases harbor biallelic alterations of a single gene (PAX5 or IKZF1), and such compound heterozygous mutations result in profound attenuation rather than complete abrogation of activity of a single transcription factor¹⁷⁸. It is possible that a residual degree of activity of B-lineage transcription factors is required for maintenance of B cell lineage in B-ALL.

TP53

TP53 is one of the most frequently mutated genes in cancer, particularly in adult-onset somatic cancers, and germline mutations are the defining feature of Li-Fraumeni syndrome [G] (LFS) and predisposition to a diverse range of tumors, particularly sarcomas, central nervous system tumors, adrenocortical carcinoma and breast cancer, and both myeloid and lymphoid leukemia^181,182. Within the hematopoietic system, germline TP53 alterations are a hallmark of low hypodiploid ALL¹⁴³, and to a lesser extent have been associated with AML and therapy-related myeloid neoplasms^183,184. Hypodiploid ALL comprises up to 5% of pediatric and 10% of adult B-ALL cases, and is characterized by recurring, stereotyped losses of whole chromosomes occurring in two main patterns: near haploidy with a leukemic cell modal chromosome number of 24–31, and low hypodiploid ALL with 32–39 chromosomes¹⁸⁵. Each subtype has distinct additional somatic mutations and deletions, commonly affecting the RAS pathway in near haploid ALL, and IKZF2 and IKZF3 in low hypodiploid and near haploid ALL, respectively¹⁸⁵. Low hypodiploid, but not near haploid ALL, has near-universal mutation of TP53^150,185,186. Most commonly these are sequence mutations in the DNA-binding and nuclear export domains of the protein (Figure 3b), but structural alterations are also observed². In pediatric low hypodiploid ALL, approximately half of cases harbor the TP53 variants in remission hematopoietic cells and so these are presumed to be germline variants with loss of wild type TP53 due to chromosome 17 aneuploidy resulting in hemizygosity for mutant TP53¹⁴³.Germline TP53 structural variants (e.g. deletions) may also be observed². In contrast, while low hypodiploidy is more frequent in adults, the TP53 alterations are rarely germline¹⁸⁷. A broader study of TP53 variants in pediatric ALL has identified germline TP53 variants in approximately 2% of patients, most commonly in low hypodiploid cases, and these variants are associated with poor outcome¹⁵⁰. The germline nature of TP53 mutations in low hypodiploid ALL is supported by the identification of TP53-mutation bearing LFS kindreds with diverse tumor types, including low hypodiploid ALL¹⁸⁸. Thus, the identification of low hypodiploidy in pediatric ALL should prompt the suspicion of a germline TP53 variant.

IKZF1

IKZF1 shares several similarities with PAX5 in the pathogenesis of ALL: both have diverse germline and somatic alterations; and a role as tumor-initiating as well as cooperating events in leukemogenesis. In addition, germline variants of IKZF1 predispose to immunodeficiency syndromes, and influence therapeutic responsiveness in ALL.

IKZF1 encodes the founding member of the Ikaros family of zinc finger transcription factors, at least three of which (IKZF1, IKZF2 (also known as Helios), and IKZF3 (also known as Aiolos)) have roles in hematopoietic development and malignancy¹⁸⁹. IKZF1 is required for the development of all lymphoid cells: T, B, natural killer and dendritic cells¹⁹⁰. Somatic alterations of IKZF1 were first identified in BCR-ABL1 positive (Ph+) B-ALL [G] lymphoid leukemia, either de novo B-ALL or lymphoid blast crisis of chronic myeloid leukemia¹⁹¹. Subsequent studies have shown that IKZF1 alterations are enriched as secondary events in kinase-driven B-ALL (Ph+ or Ph-like ALL) and double homeobox 4 (DUX4)-rearranged ALL^192–194. In kinase-driven^192,195,196, but not DUX4-rearranged B-ALL^194,197, IKZF1 alterations are associated with poor outcome^198,199. A single missense mutation in the second DNA-binding zinc finger of IKZF1, N159Y, defines a subtype of ALL with distinct gene expression profile, and is considered a leukemia-initiating mutation^178,200. Somatic alterations of IKZF1 in ALL are commonly deletions that result in loss of function, or focal internal deletions that act as dominant negative alleles, particularly the IK6 variant that lacks the DNA binding zinc fingers²⁰¹. Somatic sequence mutations are enriched in the central two DNA-binding zinc fingers and C-terminal zinc fingers, and also commonly result in aberrant nuclear and/or cytoplasmic localization and dominant negative effects (Figure 3c). Collectively, IKZF1 deletions and mutations cooperate with fusion oncoproteins such as BCR-ABL1 in lymphoid leukemogenesis, and drive resistance by activation of self-renewal, stemness and cell-cell and cell-stromal adhesion phenotypes, by derepression of stem cell (e.g. CD90, THY1) and integrin proteins^202–204.

Several recent reports have identified germline IKZF1 non-silent mutations in familial and sporadic ALL, and immunodeficiency. We reported a family with multiple individuals with ALL in which affected individuals harbored a germline truncating variant in the second N-terminal zinc finger of IKZF1⁸. Several unaffected carriers were also identified that exhibited features of immunodeficiency. Germline IKZF1 genotyping of almost 5000 children with ALL identified non-silent missense IKZF1 variants not present in control populations in approximately 1% of B-ALL cases. In contrast to somatic variants, the germline missense variants were mostly located outside of the regions of the gene that encode the N- or C-terminal zinc fingers (Figure 3c). Functional assays showed that the majority of germline variants had little direct effect on transcriptional repression, with the exception of truncating mutations, whereas most variants had profound effects on subcellular localization, adhesion, and responsiveness of BCR-ABL1⁺ leukemic cells to conventional chemotherapy and kinase inhibition; indeed, germline variants were the most deleterious in these assays of all germline and somatic IKZF1 alleles tested. In parallel, multiple studies have identified associations between germline IKZF1 alterations and immunodeficiency and autoimmunity, particularly common variable immunodeficiency, a heterogeneous disorder characterized by humoral immunodeficiency and associated autoimmune phenomena^205–208. Germline IKZF1 variants predisposing to immunodeficiency are clustered at N159, a key DNA contact residue of the second zinc finger in the protein, but are commonly serine substitutions, which are rare in ALL; in contrast the ALL subtype-defining somatic IKZF1^N159Y variant has not been reported in the germline, suggesting it is either more deleterious and not tolerated as a germline event, or that there are distinct genotype-phenotype associations. The importance of germline variations in the Ikaros family in hematological phenotypes is reinforced by the recent description of germline mutations in the DNA-binding zinc fingers of IKZF5 (also known as Pegasus) in hereditary thrombocytopenia²⁰⁹.

ETV6

ETV6 encodes a transcriptional regulator of hematopoiesis and platelet development²¹⁰. Somatic rearrangements (most commonly to RUNX1), deletions and sequence mutations are observed in ALL^168,192,211. Moreover, second hit alterations of ETV6, particularly deletion, are common in ETV6-RUNX1 rearranged leukemia^168,212,213. Germline alterations of ETV6 are recognized in both familial and sporadic ALL. Multiple kindreds with germline ETV6 mutations, ALL, and commonly thrombocytopenia, were described in 2015^144,145, with over 20 kindreds now described²¹⁴. Subsequently, large-scale sequencing of pediatric ALL cohorts identified germline, predominantly missense variants in approximately 1% of ALL cases, which were commonly hyperdiploid¹⁰. More complex germline ETV6 alterations have been identified in kindreds with thrombocytopenia or leukemia, including insertion/deletion mutations and structural variants (deletions, rearrangements) that can result in truncation and loss of function of ETV6^{11,146,147,215,216}. Such structural variants may not be identified in conventional analysis of coding (e.g. exome) sequencing data, reinforcing the importance of structural genomic analysis of exomes and/or whole genome sequencing data in kindreds with a presumptive heritable predisposition to leukemia^147,216. Carriers of ETV6 mutations commonly exhibit mild to moderate thrombocytopenia, which may be associated with an abnormal platelet function and a bleeding diathesis^11,144. It is estimated that leukemia arises in approximately 30% of carriers, and is most commonly ALL, but MDS, AML, mixed phenotype acute leukemia, diffuse large B cell lymphoma, myeloproliferative disease and plasma cell myeloma have also been reported. ETV6 variants are clustered in the DNA-binding ETS domain and predicted to be deleterious, and may have dominant negative effects; tested variants usually result in impaired transcriptional repression²¹⁷.Functional interrogation of ETV6 variants has shown perturbed expression of genes involved in platelet biogenesis and platelet cytoskeletal dynamics, including CDC42 and RHOA²¹⁸. While this provides a plausible explanation for altered platelet development and function, like many of the other predispositions it is likely that additional germline or somatic genetic alterations drive leukemogenesis²¹⁹.

Other genes predisposing to ALL

Several case reports have been informative in identifying putative new predisposition genes in ALL. SH2B3 is a negative regulator of JAK–STAT signaling²²⁰, and somatic SH2B3 deletions and mutations are observed in B- and T-ALL, particularly in Ph-like B-ALL^221–223. Rare kindreds with germline SH2B3 mutations and B-ALL have been identified²²³. Germline amplification of the neuroblastoma breakpoint region, encompassing MEF2D and BCL9, was reported in an individual with recurrence of MEF2D-BCL9 positive ALL, in which the two tumors had distinct genomic and RNA breakpoints, suggesting a (rare) heritable predisposition to ALL arising from this anomaly¹⁴⁹. Germline sequencing of a pediatric patient and both parents (also known as trio sequencing) identified an individual with DS-associated ALL with concurrent JAK2 and STAT3 variants that together, but not alone, induced growth advantage in cell lines²²⁴. Analysis of 5 children with recurrent, clonally unrelated leukemia identified two cases with germline TYK2 mutations²²⁵. TYK2 is a member of the JAK family and has been implicated in the pathogenesis of T-ALL²²⁶. The germline TYK2 variants activated signaling that was abrogated with JAK inhibitors, supporting a tumor-promoting role. Analysis of T-ALL with late relapse (to distinguish recurrence of the primary tumor from a second malignancy) identified germline deletion of LMO2 in a case, suggesting this alteration predisposed to the development of T-ALL¹⁴⁸.

Analyses of cohorts of sporadic ALL have identified additional putative leukemia-predisposing variants. Screening of a cohort of high hyperdiploid ALL identified putative pathogenic mutations in the Nijmegen Breakage Syndrome gene NBN, ETV6, FLT3, SH2B3, CREBBP (encoding a histone acetyltransferase somatically mutated in aneuploid ALL, relapsed ALL and lymphoma)^227,228, PMS1 (encoding a protein involved in DNA mismatch repair), ABL1 and MYH9 (encoding myosin heavy chain 9). Multiple variants were identified in GRB2-associated-binding protein 2 (GAB2), a SHP2-interacting protein, which led to increased AKT signaling in vitro²²⁹. Exome sequencing of a cohort of individuals with DS-associated ALL identified different pathogenic or likely pathogenic variants in 3 of 73 patients, including variants in IKZF1, NBN and the telomere maintenance gene RTEL1²³⁰. Heterozygous deleterious mutations in genes in the Fanconi anemia–BRCA pathway have been identified in approximately 10% of T-lineage ALL, several of which were shown to be germline²³¹.

Models of Disease Progression

From the data presented above, it is clear that several themes are emerging for patients with germline predisposition syndromes and the development of hematological malignancies (Figure 3). The development of a hematological malignancy in patients with a germline predisposition (e.g. mutation in DDX41) can occur as late as the 6^th or 7^th decade of life and thus is no longer restricted to childhood. However, for some predispositions, most notably those involving SAMD9L, SAMD9 or NF1, the risk of hematological malignancies does appear to be restricted to younger ages, suggesting a specific developmental stage may be more susceptible to transformation. This is also true for at least a subset of ALL, where notably, TP53 alterations are nearly universal in low hypodiploid ALL in both children and adults, but are only observed as germline events in children. Another theme is that the different types of germline predispositions are associated with widely variable penetrance for both myeloid and lymphoid leukemias, and the disease phenotypes are likely influenced by both cell intrinsic and cell extrinsic factors, including somatic alterations acquired during disease progression. In addition, although there are rare exceptions, most predisposition syndromes are linked with abnormalities in specific hematopoietic cell lineages and lead to distinct tumor profiles. For example, while germline variants in RUNX1, ANKRD26 and ETV6 all predispose to thrombocytopenia and hematological malignancy, there are striking differences in cancer predisposition: ANKRD26 variants predisposes to myeloid malignancy, ETV6 predominantly to B-lineage ALL, and RUNX1 to myeloid and to a lesser extent T-lineage ALL. Finally, common patterns of disease progression are beginning to emerge in germline predispositions (e.g. RUNX1, PAX5^G183S and NF1), which frequently involve inactivation of the remaining wild-type allele and/or somatic acquisition of canonical mutations that define specific hematological malignancy subtypes.

Future Perspectives

The last decade has witnessed remarkable advances in our understanding of the inherited basis of hematological malignancies, particularly in children (Box 2). However, much remains to be learned.

Box 2. Clinical Management and Testing of Children and Families with Predisposition Syndromes.

It is increasingly being recognized that germline alterations are much more common than previously appreciated. Considering that many patients with pathogenic or likely pathogenic mutations lack a suggestive family history of a germline predisposition, unbiased genetic testing on all patients, especially children, is the best approach to accurately catalogue and estimate the frequency of germline alterations. Indeed tumor and germline exome and/or genome sequencing is increasingly used in pediatric cancer diagnosis, as such sequencing commonly identifies “actionable” variants, and patients and families have a desire to understand the potential heritability of cancer.^4,250,251 This approach will present a challenge to practicing genetic counselors and oncologists as they integrate these data into patient care and management, especially with the interpretation of variants of undetermined significance that lack experimental documentation of deleteriousness in a given tumor context, which are commonly identified by such broad sequencing studies.

For hematopoietic tumors, genomic DNA isolated from a skin biopsy or other non-hematopoietic source is advisable for germline analysis. In the absence of true germline testing, the germline status of some variants can be inferred from “tumor only” testing from the variant allele frequency, especially if analyzed using serial genetic testing during remission. Importantly, the finding of specific mutations in tumor only testing in children, such as those in TP53 or GATA2, should prompt germline evaluation.

While the majority of mutations identified to date are present within coding exons and suitable panels are available for targeted sequencing of relevant genes, unbiased approaches are still identifying new and novel coding mutations (for example in SAMD9L and DDX41) and non-coding alterations (for example ANKRD26 and intronic GATA2 alterations). Importantly, cascade testing of the immediate family members is critical as mutations detected in the germline of patients may reflect a pattern of familial inheritance, even in those lacking a strong family history of disease, or may result from a de novo germline mutation. For most variants, data regarding the relative frequency of de novo germline variants versus inherited variants are sparse. This clinically important distinction will influence patient and family communication and management, including options for future hematopoietic stem cell transplantation. In particular, recent reports have described donor derived AMLs as a result of the transfer of germline DDX41^252,253 or GATA2²⁵⁴ mutations within family members during allogeneic transplantation.

Extent of germline predisposition.

Most studies have examined small numbers of genes, or have limited analysis to the coding exome. In pediatric cancer, it has been described that approximately 8–10% of patients have a heritable predisposition, however that number is biased both by analyses of selected cohorts or patients and panels of genes, despite the interrogation of genome-wide sequencing data^{2–4,232,233}. Until recently, many such germline panels did not include key genes relevant to hematological malignancy (e.g. IKZF1, PAX5 and SAMD9L). The increasing recognition of elevated familial risks for many hematological malignancies²³⁴, ascertainment of kindreds lacking a causal mutation from targeted or exome sequencing, and the ability of genome-wide sequence and structural analyses to identify putative predisposition mutations²³⁵ strongly suggests that careful analysis of whole genome sequencing data is required to identify new predisposition genes. This poses several challenges, including amassing the optimal cohorts for comparative analyses, utilizing novel genomic analysis approaches to map complex variants, including long read sequencing and optical mapping [G]²³⁶, and developing optimal algorithms for detection of putative causal non-coding and structural variants²³⁷. Further, larger efforts are needed to accurately define the frequency of these germline mutations across the spectrum of cancer subtypes as many of the new predisposition genes have been reported in small studies focused on rare but informative families. Thus, the absolute risk associated with individual non-silent coding variants is difficult to quantify. This also applies to identification of predisposing silent or non-coding variants, where cohort size directly influences power to detect significant associations in hematological malignancies¹⁴⁰. Such large cohorts will also be required to elucidate cooperativity of germline variants in leukemia predisposition, which is suggested by emerging reports of multiple predisposing variants in individuals with ALL, the combinatorial effect of GWAS-identified SNPs in ALL risk, and the common observation of incomplete penetrance of familial leukemia genes.

Mechanistic interpretation.

Our understanding of how germline variants predispose to hematological malignancies, and the basis for genotype-phenotype differences, lags considerably behind variant identification. Improved understanding will require a broad, sustained and intensive effort to build faithful in vitro and in vivo models to study leukemia initiation and progression.

Variant significance and clinical utility.

There are ongoing extensive efforts, including systematic cooperative efforts between the Precision Medicine Taskforce of the American Society of Hematology and the Clinical Genome Resource (ClinGen), to rigorously define criteria for germline variants, and to annotate such variants accordingly^238–241. However, these annotation efforts rely on published functional data and in silico prediction of deleteriousness, and the majority of identified germline variants are of unknown significance. One intriguing approach is high-throughput functional testing to systematically characterize a large number of mutations within critical proteins or protein domains. For example, this was rigorously done for BRCA1²⁴² and p53²⁴³. In particular the BRCA1 study generated a comprehensive map of functionally important residues in BRCA1 that could be used for the clinical classification of variants of uncertain significance. Perhaps as a model moving forward, ClinGen variant curation expert panels are being established to systemically evaluate the pathogenicity of germline variants. In light of this need the American Society of Hematology convened a ClinGen Myeloid Malignancy Variant Curation Expert Panel that has published their recommendations for germline RUNX1 variants^239,244, and will expand to perform similar curation for additional genes. A challenge moving forward will be to match the rate of variant discovery in the clinical sphere with robust functional data to properly assign pathogenicity to novel variants.

Figure 4. Model of disease progression.

Progression to advanced disease occurs through multiple different and potentially overlapping mechanisms in patients with MDS or acute leukemia predisposition syndromes. The pathways commonly observed for the different germline predispositions (genes highlighted on the right) are shown. Progression to advanced disease in many predisposition syndromes occurs through a step wise process involving loss of the remaining wild-type allele and acquisition of additional cooperating mutations (top), while others appear to maintain the wild-type allele (middle). SAMD9 and SAMD9L are unique in that disease progression is observed when there is outgrowth of monosomy 7 cells that lack the deleterious germline mutation and then cells can acquire additional somatic mutations for the full disease phenotype.

Glossary

clonal hematopoiesis

The expansion of a clonal population of hematopoietic cells marked by somatic mutations

cytopenic phase

A period marked by a decrease in peripheral blood cell counts (anemia: red blood cells, thrombocytopenia: platelets, leukopenia: white blood cells)

monocytopenia

A decrease in monocytes

pulmonary alveolar proteinosis

A lung disease characterized by a build-up of proteins and lipids in the functional units of the lung (alveoli) that is part of MonoMac syndrome, which results from GATA2 germline mutations

sensorineural deafness

A form of hearing loss that results from damage to the inner ear or auditory nerve that is part of Emberger Syndrome, which results from GATA2 germline mutations

thrombocytopenia-2

An autosomal dominant syndrome resulting from germline mutations in ANKRD26 that is characterized by lifelong mild-to-moderate thrombocytopenia and an increase in the risk for myeloid malignancies

Philadelphia chromosome-like ALL

A group of B lymphoblastic leukemias that have a global expression profile similar to BCR-ABL1 positive (Ph+) ALL but have alterations in other kinase signaling and cytokine receptor pathways

Ataxia-Telangiectasia

An autosomal recessive condition caused mutations in the ATM gene. This syndrome is characterized by immunodeficiency, decreased DNA damage repair and neurological abnormalities

Nijmegen breakage syndrome

An autosomal recessive disorder that leads to chromosomal instability and is clinically characterized by microcephaly, growth retardation, immunodeficiency, and predisposition to cancer

Robertsonian translocation

A non-reciprocal chromosomal translocation in which the long arms of two distinct acrocentric chromosomes become fused and share a single centromere

ring chromosome

An abnormal chromosome in which the ends of a single chromosome fuse to form a ring

Li-Fraumeni syndrome

An autosomal-dominant condition caused by an inherited mutation in TP53 that renders humans highly susceptible to a number of cancers, including breast cancer, leukaemia and sarcoma

BCR-ABL1 positive (Ph+) B-ALL

A class of hematopoietic disorders driven by the BCR-ABL1 fusion oncoprotein

optical mapping

An approach to genome assembly in which DNA molecules are fluorescently labeled and imaged to construct maps