New oncogenic subtypes in pediatric B-cell precursor acute lymphoblastic leukemia

When cytogenetic analysis of larger cohorts of malignancies became feasible in the late 1970s and early 1980s, it was demonstrated that cells from acute lymphoblastic leukemia (ALL) contain recurrent chromosomal aberrations that constitute independent prognostic factors.^1-4  This discovery occurred during the introduction of the risk-adapted multiagent treatment regimens that have now immensely improved childhood B-cell precursor ALL (BCP-ALL) survival; soon, cytogenetic information was incorporated in pediatric ALL risk stratification.⁵  Further cytogenetic and molecular genetic studies of BCP-ALL cells have revealed a set of common, nonoverlapping aberrations that are regarded as separate subtypes. The consensus around the subtypes is highlighted by their inclusion in the classifications provided by the World Health Organization (WHO).⁶  The established subtypes are based on the presence of high hyperdiploidy (50-67 chromosomes); hypodiploidy (fewer than 44 chromosomes); the BCR-ABL1, ETV6-RUNX, IL3-IGH, and TCF3-PBX1 fusion genes; and MLL (KMT2A) rearrangements.⁶  Together, these subtypes constitute 70% to 80% of both childhood and adult cases of BCP-ALL, although the frequency of each subtype differs substantially between children and adults.⁷  In addition, 2 provisional entities were added to the 2016 update of the WHO classifications: “B-ALL with intrachromosomal amplification of chromosome 21” and “B-ALL with translocations involving tyrosine kinases or cytokine receptors” (BCR-ABL1–like).⁶  The largest of these 2 subtypes (BCR-ABL1-like, also known as Philadelphia-like or Ph-like ALL) occurs in ∼10% of pediatric cases, and was initially identified by 2 independent groups.^8,9  Mullighan et al and Den Boer et al both identified partially overlapping BCR-ABL1–like gene expression signatures that defined cases with poor outcome.^8,9  When applied to the same patient cohort, both of these signatures identify the cases with gene fusions involving tyrosine kinase genes, but they identify different proportions of cases with JAK2 mutations and CRLF2 fusions.¹⁰ 

The genetic events defining the established subtypes are assumed to be initiating events that require additional events for leukemia to develop. This is supported by various levels of evidence for the different aberrations. For example, studies of concordant childhood ALL in monozygotic monochorionic twins have revealed that such ALLs share primary genetic changes such as high hyperdiploidy,¹¹ ETV6-RUNX1,^12,13 BCR-ABL1,¹⁴  or MLL fusions.¹⁵  The primary genetic changes were concordant with shared genomic breakpoints or shared hyperdiploid chromosomal profile between the twin ALLs, whereas copy number changes and mutations presumed to be secondary events were discordant.^11-14  These findings are consistent with a model in which the leukemia originates from an early clone containing the primary (subtype-defining) genetic change that has been transferred from 1 twin to the other, in utero, through the shared placenta. Subsequently, this preleukemic clone has then developed into full-blown leukemia independently in the twins by acquisition of additional genetic changes after birth.

As initiating events, the genetic aberrations defining the established subtypes also appear to deregulate the transcriptional state to a higher degree than other genetic aberrations because the subtypes are associated with distinct gene expression patterns that are maintained by primary BCP-ALLs as well as BCP-ALL cell lines after passage.^16-18  Substantial efforts are under way to establish individualized therapies that target the biological processes altered in cancers.¹⁹  The importance of the subtype-defining alterations in the leukemic process therefore makes these alterations and their downstream pathways particularly attractive targets, as exemplified by tyrosine kinase inhibitors such as imatinib and dasatinib that target BCR-ABL1 tyrosine kinase activity; the inclusion of such inhibitors in current treatment regimens has significantly improved the event-free survival for BCR-ABL1–positive ALL.^20,21  Targeted treatment is also being evaluated for other fusions. For example, EPZ-5676, a drug that inhibits DOT1L (a methyltransferase required for MLL-mediated leukemogenesis), is evaluated for treatment of MLL-rearranged leukemias, and ruxolitinib, which inhibits JAK2 (a tyrosine kinase required for CRLF2 signaling), is evaluated for treatment of BCR-ABL1–like ALL with CRLF2 and/or JAK2 rearrangements (www.clinicaltrials.gov #NCT02141828 and #NCT02723994).

Recently, we and others have been able to delineate the genetic makeup of a large fraction of the remaining 20% to 30% of BCP-ALL cases not belonging to the established genetic subtypes currently used for risk stratification. Analyzing these cases with modern high-resolution techniques such as RNA-sequencing (RNA-seq), whole exome sequencing, and whole genome sequencing has revealed additional groups with distinct gene expression profiles, in most cases associated with recurrent nonoverlapping rearrangements, analogous to the established subtypes. In this review, we will summarize these recent discoveries of new oncogenic subtypes within BCP-ALL and their underlying somatic mutations.

We and others recently discovered that DUX4 rearrangements identify a distinct subtype of adult and pediatric BCP-ALL,^22-25  accounting for 4% to 5% of the childhood cases (Figure 1).^23,25  However, the first hint of this subtype’s existence was noted in microarray-based gene expression studies of childhood BCP-ALL in which a group of cases outside of the established subtypes displayed a distinct uniform gene expression profile.¹⁶  Further genomic studies revealed that ∼50% to 70% of these cases displayed intragenic ERG deletions, a genomic aberration that was practically nonexistent in other BCP-ALL cases.^26,27  This subtype was found to be associated with high CD2 expression and a favorable prognosis, even together with IKZF1 deletions, which are otherwise associated with a poor prognosis.^28,29  Although ERG deletions are common in this subtype, a number of features associated with the ERG deletions suggested that they were not the primary genetic abnormality in these cases: (1) they were not present in all cases with the distinct gene expression profile, (2) the ERG deletions were sometimes found to be subclonal at diagnosis, and (3) the ERG deletions were either lost at relapse or did not maintain the genomic breakpoints from diagnosis.^28-30  Instead, all cases with this distinct gene expression pattern were found to harbor rearrangements involving the transcription factor DUX4.^22-25  At least 1 truncated copy of the DUX4 gene, normally located within subtelomeric D4Z4 repeat regions on 4q and 10q, was found to be inserted into the IGH locus in a large majority of the cases. A less common variant involved an insertion of DUX4 into an intron of ERG.²³ 

The small size of the rearrangements, the repeat nature of the DUX4 gene (which is present in 11-100 copies on each allele in a normal genome), and its subtelomeric localization are the most likely explanations as to why the DUX4 rearrangements evaded detection until the introduction of RNA-seq. Both IGH-DUX4 and ERG-DUX4 result in the expression of a 3′ truncated DUX4 transcript, with additional nucleotides added from noncoding regions of IGH or ERG, leading to a DUX4 protein with the C-terminal end replaced with 0 to 50 (random) amino acids from noncoding parts of the partner region (Figure 2A). The relocation of DUX4 results in: (1) DUX4 expression, most likely from nearby enhancers and lack of repression from the native locus; (2) C-terminal truncation of the DUX4 protein; and (3) increased DUX4 messenger RNA stability resulting from the presence of poly-A signals in the partner region. DUX4 is a transcription factor normally expressed in germinal tissues.³¹  Its expression is partly regulated by the repeat structure of the D4Z4 domains, where a sufficient number of repeats is required to prevent ectopic expression of DUX4.³¹  Reduction of the number of D4Z4 repeats below 11 copies leads to ectopic DUX4 expression in muscle cells and causes the disease facioscapulohumeral muscular dystrophy from toxicity of DUX4 to muscle cells.³¹ 

It is currently unclear how expression of DUX4 fusions contributes to leukemia development. Yasuda et al studied the leukemogenic potential of the fusion in vivo by retroviral expression of IGH-DUX4 and native DUX4 in mouse pro-B cells transplanted into immunocompromised mice. They found that expression of IGH-DUX4 but not native DUX4 resulted in in vivo expansion of immature pro-B cells and gave rise to pro-B cell leukemia with a relatively long median latency period of 157 days.²²  Zhang et al studied DUX4 expression in relation to the ERG deletions and reported that DUX4 binds to a cryptic promoter within the ERG gene and directly induces the expression of an alternative ERG transcript, ERGalt. They suggested that the alternative ERG transcript is important for leukemia transformation and hypothesize that increased transcriptional activity in this region makes it more prone to RAG-mediated deletion, resulting in the characteristic ERG deletions.²⁵ 

In children, 4% to 5% of all BCP-ALL harbors DUX4 rearrangements (Figure 1), making this the sixth largest subtype of childhood BCP-ALL (slightly larger than BCR-ABL1).^23,25  Besides the common ERG deletions, DUX4-rearranged cases have been found to harbor aberrations common in other forms of BCP-ALL such as deletions targeting cell-cycle regulator genes CDKN2A and CDKN2B, lymphoid transcription factor genes such as IKZF1 and PAX5, and activating mutations in NRAS and KRAS.^24,25  In addition, they harbor mutations in genes that are not commonly mutated in other types of ALL, such as the transcription factor genes MYC, MYCBP2, MGA, and ZEB2.²⁵ 

In a study of gene expression–based subtypes and gene fusions in 195 BCP-ALLs, we identified 6 B-other cases with a gene expression pattern identical to ETV6-RUNX1–positive cases despite lacking this fusion gene, as confirmed by reverse transcriptase polymerase chain reaction (6/6 cases) and fluorescence in situ hybridization (4/6 cases), in addition to RNA-seq.²³  This finding was unexpected given the strong correlation between primary genetic changes and global gene expression pattern in BCP-ALL.^16,17  Cooccurring ETV6 and IKZF1 aberrations (gene fusions or deletions) were detected in all cases in which comprehensive analysis could be performed at both the DNA and RNA level (5/6 cases; Figure 2B). We could identify a similar set of 4 ETV6-RUNX1–negative cases with an ETV6-RUNX1–like gene expression pattern in an independent data set. Only RNA-seq data were available for this data set, yet gene fusions involving ETV6 could be detected in 3 of the 4 cases. No IKZF1 aberrations were identified, possibly because of a lack of single nucleotide polymorphism array data. An additional 5 ETV6-RUNX1–negative cases with an ETV6-RUNX1–like gene expression profile, all harboring ETV6 aberrations, were also recently described, further confirming the presence of this subtype.³²  In addition to sharing a gene expression profile with ETV6-RUNX1–positive cases, these cases exhibited a CD27^pos/CD44^low-neg immunophenotype, otherwise associated with ETV6-RUNX1–positive cases.^32,33  The cases were enriched for IKZF1 rearrangements (3/5 cases) as well as intragenic ARPP21 deletions (3/5 cases).³²  In addition, several previously published BCP-ALL data sets contain potential ETV6-RUNX1–like cases that exhibit gene expression or methylation patterns identical to ETV6-RUNX1–positive cases,^8,34,35  although it is unclear to what extent canonical and noncanonical ETV6-RUNX1 fusions were excluded in those cases.

The 2 studies describing ETV6-RUNX1–like ALL so far indicate that 1% to 3% of all childhood BCP-ALL cases belong to this subtype.^23,32  There are no data available on ETV6-RUNX1–like cases in adult BCP-ALL. Although very few relapses have been reported among the ETV6-RUNX1–like cases, studies of larger uniformly treated cohorts are required to elucidate the prognostic significance of this subtype.

The first ZNF384 (also known as CIZ and NMP4) fusions were described in BCP-ALL in 2002,³⁶  but it was not recognized until recently that as much as 1% to 6% of childhood BCP-ALL cases and 5% to 15% of adult BCP-ALL cases harbor ZNF384 fusions.^22,24,37-39  So far, 9 different 5′ fusion partners to ZNF384 have been identified in BCP-ALL: ARID1B, BMP2K, CREBBP, EP300, EWSR1, SMARCA2, SYNRG, TAF15, and TCF3 (Figure 2C-D).^24,36-42  Despite the high number of fusion partners, cases with ZNF384 rearrangements have a distinct gene expression pattern and a characteristic immunophenotype with low CD10 expression and expression of the myeloid markers CD13 and/or CD33.³⁸  Further studies are needed on the prognostic relevance of ZNF384 rearrangements in children, although an intermediate to favorable prognosis has been described in small cohorts.^24,38  Epigenetic regulators appear to have an important cooperative role in ZNF384-rearranged ALL because the genes CREBBP, EP300, KDM6A, CHD4, or CHD8 are altered in 82% of the cases, either by sequence mutations or by being a fusion partner to ZNF384.³⁹  Mutations in RAS signaling pathway genes such as NRAS, KRAS, PIK3CD, and PTPN11 are also common in this subtype.³⁹ 

Fusion genes involving MEF2D (myocyte enhancer factor 2D) have recently been reported to be present in 1% to 4% of childhood BCP-ALL and 7% of adult BCP-ALL. MEF2D is the 5′ partner in all described fusions, and a total of 6 3′ partners have been described (BCL9, CSF1R, DAZAP1, FOXJ2, HNRNPUL1, and SS18;Figure 2E-F).^22-24,42-44 

MEF2D-rearranged cases have an inferior outcome.⁴²  Apart from MEF2D-CSF1R, which confers a BCR-ABL1–like gene expression profile, the rearranged cases also share a distinct gene expression profile that includes deregulation of MEF2D targets.⁴² 

Leukemic cells from MEF2D-rearranged cases have showed in vitro sensitivity to HDAC inhibitors, possibly from inhibition of HDAC9, which is a specifically overexpressed target of MEF2D. Hence, the addition of HDAC inhibitors could improve treatment regimens for this subtype.⁴²  Mutations in the RAS signaling pathway (NRAS, KRAS, NF1, PTPN11) are common in MEF2D-rearranged cases.⁴² 

With the addition of the previously mentioned novel subtypes, ∼10% of pediatric BCP-ALL cases remain (Figure 1) that do not belong to either the established subtypes (BCR-ABL1, ETV6-RUNX1, high hyperdiploid, hypodiploid, IL3-IGH, MLL-rearranged, and TCF3-PBX1), the WHO provisional entities (BCR-ABL1–like and intrachromosomal amplification of chromosome 21), or the novel subtypes (ETV6-RUNX1–like, DUX4-rearranged, MEF2D-rearranged, and ZNF384-rearranged).^23,45  For adult patients, the frequency of unclassified cases is higher.⁴⁵  The molecular data on the remaining unclassified cases are somewhat conflicting. We reported that the majority of such cases (14/17) harbor in-frame fusions in a consecutive series of pediatric BCP-ALL,²³  whereas Gu et al detected a smaller proportion of in-frame fusions (65/187 cases) in BCP-ALL cases from more heterogeneous age groups.⁴²  Presumably, the cases lacking in-frame fusions harbor other aberrations responsible for leukemia development.

The in-frame fusions among these cases include recurrent fusions, possibly indicating very rare subtypes such as TCF3-HLF, associated with a very unfavorable prognosis,⁴⁶  and fusions involving IGH and CEBP transcription factor family genes.⁴⁷  Rare fusions involving NUTM1 have been reported in several studies, but whether these cases have uniform characteristics and hence should be considered a separate subtype remains to be elucidated.^23,24,42,48 

Fusions between PAX5 and a range of partners can be found in 40% of cases with dicentric chromosomes involving 9p such as dic(7;9), dic(9;12), and dic(9;20), whereas 50% of such cases instead show deletion of PAX5.⁴⁹  However, these dicentric chromosomes sometimes occur together with established primary aberrations such as BCR-ABL1 and ETV6-RUNX1, and it is unclear whether ALL with dicentric chromosomes involving 9p should constitute a separate subtype.^49-51  Similarly, the fusions IGH-CRLF2 and P2RY8-CRLF2, resulting in upregulation of CRLF2, can occur both together with primary aberrations such as BCR-ABL1, ETV6-RUNX1, and high hyperdiploidy and as a standalone fusion gene outside of established subtypes. Hence, it is possible that these 2 gene fusions in some cases constitute primary aberrations, whereas in other cases they constitute secondary aberrations.

The recent years have witnessed a dramatic progress in BCP-ALL classification. The new oncogenic subtypes described in this review, together with the BCR-ABL1–like subtype, explain around two-thirds of previously unclassified pediatric BCP-ALL cases. We expect that this increased molecular knowledge will be incorporated in clinical risk stratification, although, as of yet, standardized assays for identifying the DUX4-rearranged, ETV6-RUNX1–like, and BCR-ABL1–like subtypes are lacking. The latter 2 subtypes, being defined by gene expression profiles, will most likely require screening by expression-based assays such as RNA-seq or specific low-density expression arrays.⁵²  In the longer perspective, we expect the increased knowledge of BCP-ALL biology will permit the development and deployment of targeted therapies and thereby further reduce the mortality and side effects caused by the intense treatment regimens currently in use for this disease.

This work was supported by the Swedish Cancer Society, the Swedish Childhood Cancer Foundation, the Swedish Research Council, the Knut and Alice Wallenberg Foundation, the Inga-Britt and Arne Lundberg Foundation, the Gunnar Nilsson Cancer Foundation, the Medical Faculty of Lund University, and Governmental Funding of Clinical Research within the National Health Service.

Contribution: H.L. and T.F. analyzed data and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Henrik Lilljebjörn, Department of Clinical Genetics, Lund University, BMC C13, 221 84 Lund, Sweden; e-mail: henrik.lilljebjorn@med.lu.se; or Thoas Fioretos, Department of Clinical Genetics, Lund University, BMC C13, 221 84 Lund, Sweden; e-mail: thoas.fioretos@med.lu.se.