Advances in the Genetics of High-Risk Childhood B-Progenitor Acute Lymphoblastic Leukemia and Juvenile Myelomonocytic Leukemia: Implications for Therapy

BCR-ABL1 leukemia

Expression of the constitutively active tyrosine kinase BCR-ABL1 is a hallmark of chronic myeloid leukemia (CML), an expansion of relatively mature granulocytes, and a subset of ALL, which is typically an aggressive form of leukemia that responds poorly to therapy. Genomic profiling has shown that BCR-ABL1–positive lymphoid leukemia, either de novo ALL or CML at the progression to lymphoid blast crisis, is commonly accompanied by deletion or, less frequently, sequence mutation of the IKZF1 gene (21, 22), which encodes a lymphoid transcription factor required for normal lymphoid development. These alterations are loss of function or dominant negative, usually mono-allelic, and accelerate the onset of leukemia in murine models (23). In contrast, IKZF1 alterations are rare in chronic phase CML and at myeloid blast crisis, indicating that these alterations are important determinants of lymphoid disease lineage and progression. IKZF1 alterations are also associated with poor outcome in BCR-ABL1–negative lymphoid leukemia (24, 25), suggesting an important role in treatment failure, and provide a potential explanation for the inferior outcome of BCR-ABL1 ALL compared with CML. The mechanistic basis for IKZF1 mutation–induced treatment failure is poorly understood, but it may relate to loss of IKZF1 activity, resulting in acquisition of a primitive, stem cell–like phenotype that is more resistant to therapy (24). These findings suggest that modulation of IKZF1 activity, or the downstream pathways regulated by this gene, might further improve the outcome of this subtype of ALL.

“BCR-ABL1–like” acute lymphoblastic leukemia

Two years ago, 2 groups, the Children's Oncology Group (COG)/National Cancer Institute Therapeutically Applicable Research to Generate Effective Treatments (TARGET) consortium (24) and a Dutch group led by Monique Den Boer and Rob Pieters (26), identified a subgroup of BCR-ABL1–negative childhood ALL cases, characterized by a gene expression profile similar to that of BCR-ABL1 ALL, which has poor outcome. These cases usually have alterations of IKZF1, but until recently, the genetic alterations substituting for BCR-ABL1 were unknown. These “BCR-ABL–like” or “Ph-like” ALL cases have at least 2 broad groups of genetic changes, resulting in constitutive activation of cytokine receptor and/or tyrosine kinase signaling.

Up to 50% of Ph-like cases harbor rearrangements of CRLF2 [encoding cytokine receptor-like factor 2 or the thymic stromal lymphopoietin receptor (TSLPR); Fig. 2; ref. 27]. CRLF2 forms a heterodimer with interleukin receptor 7α (IL-7R) for the cytokine TSLP. CRLF2 is located at the pseudoautosomal region of Xp/Yp, and the rearrangements are either translocation to IGH@ at 14q33 (18) or a focal deletion upstream of CRLF2 that results in a novel fusion, P2RY8-CRLF2, in which the first noncoding exon of P2RY8 is fused to the entire open reading frame of CRLF2 (17). Both result in aberrant expression of CRLF2 by leukemic cells that may be detected by routine flow cytometric immunophenotyping at the time of diagnosis (17, 18). Less common is a point mutation, CRLF2 p.Phe232Cys, which also results in receptor overexpression (28, 29). Half of CRLF2-rearranged cases have activating mutations in Janus kinases (17, 18, 29), most commonly at or near R683 in the pseudokinase domain of JAK2, but also in the kinase domain of JAK2, and in both the kinase and pseudokinase domains of JAK1 (30). Coexpression of CRLF2 and JAK mutant alleles transforms model cell lines (17, 29), and human CRLF2-rearranged leukemic cells exhibit JAK-STAT activation (18). This JAK-STAT activation, in either experimental models or primary human leukemic cells, is attenuated by the use of commercially available selective JAK inhibitors (17). Several studies of high-risk ALL cohorts have shown that CRLF2/JAK alterations are associated with IKZF1 deletion and/or mutation and that these cases have poor outcome (27, 31). Thus, the addition of JAK inhibitor therapy to conventional multiagent chemotherapy is attractive and is being explored in a dose-finding phase I trial of the JAK inhibitor ruxolitinib in relapsed and refractory childhood tumors (the COG ADVL1011 trial). These important findings represent one of the first novel targeted therapies to be implemented in ALL since the dramatically effective incorporation of imatinib to the treatment of BCR-ABL1–positive ALL (32).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

CRLF2 alterations in B-ALL. A, deletions in the pseudoautosomal 1 region of Xp/Yp. SNP microarray copy number data for 6 ALL cases with matched tumor (T) and normal (N) data are shown. Dark blue shows a focal PAR1 deletion is present, the extent of which is identical in all cases. B, mapping of the PAR1 deletion (probe-level data are shown as vertical red lines) to between intron 1 of P2RY8 and upstream of CRLF2. C, the PAR1 deletion results in a P2RY8-CRLF2 fusion containing the entire CRLF2 open reading frame. D, the P2RY8-CRLF2 fusion results in CRLF2 overexpression that may be detected by immunophenotyping of leukemic cells. Data from Mullighan et al. (17).

Several questions about the role of CRLF2 in leukemogenesis remain unanswered. Although CRLF2 rearrangement is present in only 5% to 7% of general ALL cohorts, it is exceptionally common in B-ALL arising in children with Down syndrome (DS-ALL; refs. 17, 18). The basis for this enrichment of CRLF2 alterations in this unique setting is unknown. Moreover, although data are limited, CRLF2 alterations are not significantly associated with poor outcome in DS-ALL (17), despite a similar frequency of associated JAK2 mutations to that observed in non-DS-ALL (17, 18), perhaps, in part, because of the lack of IKZF1 alterations in this subgroup (17). Moreover, in both DS- and non-DS ALL, half of CRLF2-rearranged cases lack activating JAK mutations and, indeed, seem not to have any additional sequence variants or structural DNA variations activating tyrosine kinases (M.L. Loh and C.G. Mullighan; unpublished observations), despite exhibiting similar activities of downstream signaling pathways to CRLF2-rearranged and/or JAK-mutated cases. The alterations substituting for JAK2 mutations or the reasons why JAK mutations are dispensable for transformation in these cases are unknown, but they should be answered by ongoing genome sequencing efforts and the development of experimental models that carefully model the effect and interaction of each class of lesion (CRLF2, JAK, and IKZF1) in transformation.

A second subgroup of Ph-like cases harbors a remarkably diverse range of cytokine receptor and kinase signaling alterations. RNA-seq and whole-genome sequencing of 12 Ph-like ALL cases, done by the COG/TARGET initiative and St. Jude Children's Research Hospital (Memphis, TN), has shown that cryptic rearrangements, sequence variations, and submicroscopic structural variants are a hallmark of CRLF2 wild-type Ph-like ALL (33, 34). These include chimeric fusions encoding constitutively active tyrosine kinases (EBF1-PDGFRB, NUP214-ABL1, STRN3-JAK2, and BCR-JAK2); dysregulating cytokine receptors (IGH@-EPOR); activating complex sequence mutations in the transmembrane domain of IL7R, encoding the IL-7R chain (also recently reported by other groups in B- and T-lineage ALL; refs. 34–36); and deletions of SH2B3 (LNK), which encodes a negative regulator of JAK2 (37, 38). Recurrence screening of large cohorts of high-risk B-ALL has shown that several of these alterations (NUP214-ABL1, EBF1-PDGFRB, IL7R mutations, and SH2B3 alterations) are recurrent in Ph-like ALL (34). Expression of these alterations in model cell lines and primary hematopoietic progenitors has shown that EBF1-PDGFRB and IL-7R mutant alleles are transforming and result in Jak-Stat activation that is attenuated by Abl1 or Jak inhibitors. Similarly, primary leukemic cells harboring these alterations exhibit activation of downstream signaling pathways that are sensitive to tyrosine kinase inhibitors (34). Thus, although these studies have identified a diverse range of genetic alterations, these converge on common signaling pathways that are potentially amenable to kinase inhibitor therapy.

These findings are important, as Ph-like ALL is common (15%, or at least 3 times as common as BCR-ABL1–positive ALL) and is also present at a similar or higher frequency in adolescents and young adults (M.L. Loh and C.G. Mullighan; unpublished observations), in whom the outcome of ALL therapy is inferior to that of younger children. Moreover, although the genetic landscape of Ph-like is complex, the Ph-like phenotype is readily detectable by expression profiling of a limited set of genes or by flow cytometric analysis of pathway activation (e.g., CRKL phosphorylation for ABL1-rearranged cases and total tyrosine phosphorylation for JAK2-rearranged cases). Implementation of these approaches and the development of trials to treat Ph-like ALL with tyrosine kinase inhibitor therapy are being vigorously pursued.

Hypodiploid acute lymphoblastic leukemia

Hypodiploid ALL is an uncommon subtype of ALL that is associated with a very high risk of treatment failure (39–43). These cases have multiple whole chromosomal losses and have been subclassified according to the degree of aneuploidy: near-haploid cases with 24 to 31 chromosomes and low-hypodiploid cases with 32 to 44 chromosomes. Apart from the documentation of aneuploidy, the genetic basis of this subtype of ALL has been very poorly understood. To investigate this, in conjunction with the COG, we performed detailed genomic profiling of a large cohort of hypodiploid ALL cases, incorporating microarray analysis of structural genetic variation and gene expression, extensive candidate gene sequencing, and next-generation sequencing (44).

These studies have shown that near-haploid ALL cases have a very high frequency of deletions and sequence variations, known or predicted to activate Ras signaling, and that near-haploid and low-hypodiploid cases have distinct and mutually exclusive lesions inactivating lymphoid transcription factors. Several of the lesions, such as alterations of the IKAROS gene family, have only been rarely observed in other leukemias. Conversely, several of the genes involved in regulation of Ras signaling, including NF1, KRAS, NRAS, and PTPN11, are mutated in other disorders, such as juvenile myelomonocytic leukemia (JMML). Moreover, careful analysis of the frequency and nature of these lesions and analysis of leukemic transcriptome have shown that near haploid and low hypodiploid are distinct diseases at the genetic level. As hypodiploidy confers a high risk of treatment failure and satisfactory therapies are lacking, the high frequency of Ras pathway activation suggests that targeted therapies designed to disrupt downstream Ras signaling should be pursued in this disease.

Sequence analysis of relapsed acute lymphoblastic leukemia

Genome-wide studies of sequence variation have lagged behind those of DNA copy number variation, but they are now feasible with advances in the yield and cost-effectiveness of next-generation sequencing. Although whole-genome sequencing studies of ALL have not yet been reported, candidate gene sequencing efforts using Sanger sequencing have identified a number of novel targets of sequence mutation in high-risk B-ALL (Fig. 4; ref. 53). To extend our understanding of the genetic basis of treatment failure in ALL, we sequenced 300 genes in 23 matched diagnosis-relapse ALL samples. A key finding of this study was the presence of deleterious mutations in CREBBP, encoding CREB-binding protein (CREBBP), in almost 20% of relapsed ALL cases. CREBBP is a multifunctional protein with roles in histone and nonhistone acetylation and transcriptional regulation, by acting as a scaffold of a diverse range of transcription factors. Many of the mutations were located in the histone acetyltransferase domain and impaired the normal acetylation activity of CREBBP. We also showed that the CREBBP mutations impaired the normal transcriptional response to glucocorticoids, which are widely used in ALL therapy, suggesting that CREBBP mutations may have an important role in determining resistance to these agents. Moreover, like IKZF1 alterations, CREBBP mutations were preserved from diagnosis to relapse, or they were present in subclones at diagnosis and emerged in the predominant clone at relapse, supporting the notion that these mutations provide a selective advantage during treatment. Histone deacetylase inhibitors are being explored as potential therapeutic agents in a range of tumors, and preliminary studies using leukemic cell lines showed that many, including those with CREBBP mutations, are exquisitely sensitive to these agents. These findings suggest that histone deacetylase inhibitors may represent a novel therapeutic approach in high-risk leukemia.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Recurring sequence mutations in high-risk B-ALL. Sequencing of 125 genes in 187 patients with high-risk B-ALL identified multiple novel targets of mutation. This research was originally published by Zhang et al. (53).

Patients with genetic syndromes provided important clues to unraveling the molecular pathogenesis of juvenile myelomonocytic leukemia

Children affected by specific congenital syndromes, including neurofibromatosis type I (NF1), Noonan syndrome, or Cbl syndrome are at a higher risk of developing JMML or a related transient myeloproliferative disorder in infancy, and the identification of genes associated with these syndromes has provided unique insights into the pathogenesis of this myeloid malignancy.

In 1978, patients with NF1 were noted to be particularly prone to developing a CML or acute myelomonocytic leukemia in the first decade of life (59). Subsequently, NF1 was cloned and found to encode neurofibromin, which hydrolyzes the active, GTP-bound conformation of Ras to the inactive, GDP-bound conformation and, thus, functions as a GTPase-activating protein (60). In 1994, Shannon and colleagues (61) showed loss of the wild-type allele (later shown to identify acquired isodisomy of the mutant allele) in the diseased bone marrows of children with JMML affected by NF1, thus establishing NF1 as a tumor suppressor gene (61, 62). In subsequent mouse models engineered to conditionally delete the Nf1 allele, elevated levels of Ras-GTP occur, and a myeloproliferative neoplasm (MPN) develops (63, 64).

Somatic, oncogenic mutations in NRAS and KRAS2 (RAS) were subsequently identified that year in independent cohorts of patients with JMML, thus providing additional genetic evidence to support that hyperactivation of the Ras/MAPK pathway was essential for the pathogenesis of this disease (Fig. 5; refs. 65, 66). Additional animal models have subsequently revealed that Kras mutant mice also develop fatal myeloid disorders that resemble JMML in vivo and in vitro (67, 68).

Noonan syndrome is another congenital disorder that is inherited in an autosomal dominant fashion 50% of the time (reviewed in ref. 69). Interestingly, some infants with Noonan syndrome also display a hematologic phenotype, including a self-resolving myeloproliferative disorder that resembles JMML (70, 71). In 2001, Tartaglia and colleagues identified germline mutations in PTPN11, a gene encoding the nonreceptor tyrosine phosphatase protein SHP-2, as causative in approximately 50% of children with Noonan syndrome (72). The observations of a JMML-like MPN in patients with Noonan syndrome (primarily those arising as spontaneous mono-allelic germline events) led to the identification of mono-allelic, activating somatic mutations in PTPN11, occurring in as many as 35% of nonsyndromic JMML (73, 74). A comparison between the PTPN11 mutations in de novo and syndromic JMML (Noonan syndrome) reveals that many of the same amino acids in exons 3, 4, and 13 are affected (75), but with different codon substitutions, leading investigators to hypothesize that the transforming ability of mutations in Noonan syndrome are “weaker” and, thus, may be tolerated as germline events. For example, the most common PTPN11 mutation in de novo JMML is the c. 226G > A, resulting in p. Glu76Lys, an alteration that has never been documented as a germline lesion in Noonan syndrome.

To identify additional novel mutations in patients with JMML, we did Affymetrix SNP 6.0 array analysis on samples from JMML patients with and without known Ras pathway abnormalities (76). We identified a region of 11q that acquired uniparental isodisomy in 5 of 27 cases. Importantly, recent reports had uncovered similar lesions in adults with MPN and myelodysplastic syndrome, and subsequent molecular analysis identified homozygous c-CBL (CBL) mutations (77–79). All 5 of our patients also displayed homozygous CBL lesions. The most common CBL mutation in JMML is the c.1111T > C transition, resulting in the substitution of a histidine for a tyrosine residue at codon 371 in the alpha linker region of the protein (76, 80, 81).

A number of children with JMML harboring these homozygous CBL mutations shared phenotypic features, suggesting that these lesions might first occur as germline events (82–84). Additional studies indicate that these mutations are autosomally inherited in a dominant fashion approximately 50% of the time (82). Importantly, genotyping of myeloid colonies grown from 1 patient's cord blood revealed the presence of rare homozygous clones at birth, supporting our previous finding that oncogenic JMML mutations can arise in utero (75, 76).

One striking phenomenon about patients with JMML and homozygous CBL mutations is their high rate of spontaneous resolution (82–84). However, of 5 patients reported with spontaneously resolving JMML and the germline Cbl syndrome, 4 developed signs consistent with serious vasculopathies by the end of their second decade of life, as evidenced by optic atrophy, hypertension, cardiomyopathy, or arteritis (82).

The relationship of Cbl to Ras/MAPK signaling is poorly understood, and furthermore, the function of mutant Cbl proteins in malignancy is likely complex. There are 3 members of the Cbl family of E3 ubiquitin ligases (CBL, CBLB, and CBLC), all of which are known to mark activated protein tyrosine kinases and other associated proteins for degradation by ubiquitination, but Cbl proteins also retain many important adaptor functions (reviewed in ref. 85). Most of the mutant Cbl proteins reported to occur thus far in myeloid malignancies result from missense mutations or splice site variants in CBL, in which the majority have been shown to result in an encoded protein that lacks E3 ligase activity only in the absence of the wild-type protein (79, 82). Two recently generated mouse models indicate possible mechanisms by which mutant Cbl proteins exert their phenotypic effects (86, 87). The first is a Cbl, Cblb double-deficient mouse that develops a rapidly fatal MPN in comparison with a Cbl null mouse, which does not develop disease, suggesting that mutant Cbl proteins may confer a dominant negative effect on Cbl b (86). The second is a heterozygous Cbl RING finger mutant knock-in mouse that develops a more latent, but equally fatal, MPN only on a Cbl-null background (87). This suggests that mutant Cbl proteins may act as gain-of-function oncogenes that continue association with certain proteins but that cannot function to degrade protein tyrosine kinases because of the loss of the negative regulatory E3 ligase function, thus constitutively activating growth pathways. Further work to fully interrogate the consequences of mutant Cbl proteins in myeloid malignancies is under way.

Other mutations have been recently reported to occur rarely in JMML, such as ASXL1 or FLT3 (88, 89). In contrast to other MPNs of adulthood, neither JAK2 nor TET2 mutations occur in JMML (M.L. Loh; unpublished data; ref. 81). Taken together, these data would suggest that the current spectrum of commonly occurring mutations in NF1, RAS, PTPN11, and CBL converge on the Ras/MAPK pathway in a highly specific way in JMML, which remains to be fully elucidated.

Approaches to therapy

Although the current standard of care for JMML is allogeneic HSCT, approximately 50% of children will still suffer from relapse. No standard chemotherapy regimen used prior to HSCT has been shown to affect the incidence of relapse after HSCT (90). Based on the spectrum of mutations that were first characterized by loss of NF1 and oncogenic lesions in RAS, it would be reasonable to hypothesize that suppression of Ras/MAPK signaling might have therapeutic efficacy in JMML. However, 3 issues should be mentioned. First, direct inhibition of hyperactive Ras has been problematic for several reasons. For example, Ras is activated when localized to the cellular membrane, which requires covalent binding of a farnesyl group to a conserved CAAX motif by farnesyl transferases (FTase). As prenylation of Ras is catalyzed by FTase, FTase inhibitors (FTI) were developed to prevent Ras localization to the plasma membrane and inhibit Ras signaling (91). However, alternative geranylation can substitute for prenylation, and thus the clinical promise of FTI therapy as Ras inhibitors has been disappointing. The COG sponsored a phase II trial, AAML0122, incorporating a window phase of an FTI followed by chemotherapy only and HSCT (92). The event-free survival on AAML0122 was 40%, with an overall survival of 55%, with a median follow-up of 4 years (T. Cooper; personal communication). Second, the relationships of SHP2 and CBL to Ras signaling are less well understood, and it may be that the biochemical consequences of these lesions will render them more sensitive to alternative therapeutic approaches. Finally, the importance of Ras signaling in diverse physiologic processes in normal cells raises the concern of nonspecific systemic toxicities with any therapies in which normal Ras proteins or their effectors are inhibited. Ideally, one hopes to capitalize on the existence of “oncogene addiction” and identify inhibitors that preferentially treat mutant rather than normal cells. However, amplification of signals upstream or parallel to Ras may occur, raising the possibility that combination therapy to overcome resistance will be necessary to achieve therapeutic effects (93).

One example of a Ras effector pathway involved in JMML is the Raf/MAP–extracellular signal—regulated kinase (ERK) kinase (MEK)/ERK pathway. Ras-GTP induces a cascade of activation from the Raf kinases, to MAPK kinase (MEK1 and MEK2), to ERK, which stimulates diverse nuclear and cytosolic effects. Raf mutations are found in melanoma and renal cell carcinoma, and it is reasonable to hypothesize that inhibiting the Raf/MEK/ERK cascade has therapeutic potential in JMML. To date, the largest clinical experience with Raf inhibitors is with sorafenib (Bay 43-9006), which is U.S. Food and Drug Administration (FDA) approved for treatment of advanced renal cell carcinoma and hepatocellular carcinoma (94, 95). Sorafenib is a novel bi-aryl, multikinase inhibitor with activity against VEGF receptor-2 and receptor-3, platelet—derived growth factor receptor b, fibroblast growth factor receptor 1, FLT-3, and c-KIT (96), as well as Raf-1 and B-raf (97), and it is currently being used in conjunction with chemotherapy for acute myeloid leukemia (AML) in the COG. However, the recent data to support the emergence of skin tumors in patients treated with Raf inhibitors highlight the complexity of signaling networks in cancer cells (93).

MEK is also an attractive target, based on the frequency of Ras and Raf mutations in human cancer. Selective inhibition of MEK, resulting in inhibition of its only known substrate, ERK1/2, can be achieved with non-ATP competitive inhibitors. In recent reports, the use of a MEK inhibitor, PD0325901 (PD901; Pfizer), in a model of JMML and chronic myelomonocytic leukemia, characterized by the conditional expression of the Kras^G12D mutant allele, had encouraging results. PD901 treatment resulted in a sustained reduction of peripheral white blood count, improved anemia, and enhanced survival in the treated mutants compared with the wild-type controls (98). Importantly, PD901 abrogated the classic GM-CSF hypersensitivity exhibited by myeloid progenitors in in vitro assays. MEK inhibition may also have a role for patients with refractory JMML who transform to AML. This is suggested by the inhibition of Nf1 mutant acute myeloid leukemic colony growth at low doses of CI-1040 and prolonged survival of mice transplanted with Nf1 mutant AML when treated with CI-1040 (99). Given the recent description of Ras pathway mutations in hypodiploid ALL, it would be reasonable to assess the clinical utility of MEK inhibition for both of these diseases.

Finally, we and others described the phosphorylation of STAT5 to low-dose GM-CSF in subsets of JMML cells (100, 101). The discovery of activating JAK2 mutations in other MPNs, including the majority of adults with polycythemia vera, essential thrombocythemia, and primary myelofibrosis (102–104), led to the development of a number of specific and nonspecific agents that inhibit wild-type and mutant JAK2. Ruxolitinib (formally INCB018424; Incyte/Novartis) has recently been FDA approved for primary myelofibrosis patients, showing excellent tolerability and clinical responses (105, 106). In collaboration with Incyte/Novartis, the COG is determining the maximum tolerated dose of ruxolitinib in children (ADVL1011) on the basis of recent data showing the presence of JAK lesions in approximately 10% of children with high-risk ALL (30), the known JAK lesions in polycythemia vera and essential thrombocythemia, and the documented signaling abnormalities in JMML (100). Correlating the effect on colony formation, cell growth, and signaling networks in the presence of the inhibitor will help evaluate JAK-STAT signaling in disease pathogenesis and may reveal previously unappreciated relationships between the JAK/STAT5 and Ras/MAPK pathways.

Summary and directions for future work

Advances in genomic technologies have identified a number of lesions that have transformed our understanding of the molecular pathogenesis of leukemia, revealing a number of targetable lesions in both lymphoid and myeloid diseases, particularly in subsets of patients for whom cure is currently elusive or for whom cure is associated with substantial morbidity.

Results from next-generation sequencing efforts have indicated that identification of all genetic alterations contributing to leukemogenesis and treatment failure will require comprehensive analysis of sequence and structural variations with these technologies and deep sequencing of samples at multiple disease time points to identify critical lesions in tumor subclones that confer resistance to therapy.

Ongoing investigations of targeted therapies will require continued study of the biochemical and functional consequences of the genetic lesions occurring in these high-risk diseases and will also require rigorous clinical trials to assess both target inhibition and outcome. However, recent efforts will undoubtedly continue to provide fruitful and productive collaborations among basic, translational, and clinical scientists.