Characterization of leukemias with ETV6-ABL1 fusion

Abstract

To characterize the incidence, clinical features and genetics of ETV6-ABL1 leukemias, representing targetable kinase-activating lesions, we analyzed 44 new and published cases of ETV6-ABL1-positive hematologic malignancies [22 cases of acute lymphoblastic leukemia (13 children, 9 adults) and 22 myeloid malignancies (18 myeloproliferative neoplasms, 4 acute myeloid leukemias)]. The presence of the ETV6-ABL1 fusion was ascertained by cytogenetics, fluorescence in-situ hybridization, reverse transcriptase-polymerase chain reaction and RNA sequencing. Genomic and gene expression profiling was performed by single nucleotide polymorphism and expression arrays. Systematic screening of more than 4,500 cases revealed that in acute lymphoblastic leukemia ETV6-ABL1 is rare in childhood (0.17% cases) and slightly more common in adults (0.38%). There is no systematic screening of myeloproliferative neoplasms; however, the number of ETV6-ABL1-positive cases and the relative incidence of acute lymphoblastic leukemia and myeloproliferative neoplasms suggest that in adulthood ETV6-ABL1 is more common in BCR-ABL1-negative chronic myeloid leukemia-like myeloproliferations than in acute lymphoblastic leukemia. The genomic profile of ETV6-ABL1 acute lymphoblastic leukemia resembled that of BCR-ABL1 and BCR-ABL1-like cases with 80% of patients having concurrent CDKN2A/B and IKZF1 deletions. In the gene expression profiling all the ETV6-ABL1-positive samples clustered in close vicinity to BCR-ABL1 cases. All but one of the cases of ETV6-ABL1 acute lymphoblastic leukemia were classified as BCR-ABL1-like by a standardized assay. Over 60% of patients died, irrespectively of the disease or age subgroup examined. In conclusion, ETV6-ABL1 fusion occurs in both lymphoid and myeloid leukemias; the genomic profile and clinical behavior resemble BCR-ABL1-positive malignancies, including the unfavorable prognosis, particularly of acute leukemias. The poor outcome suggests that treatment with tyrosine kinase inhibitors should be considered for patients with this fusion.

Introduction

ETV6-ABL1 (TEL-ABL) fusion is a rare but recurrent genetic aberration found in hematologic malignancies. Given the orientation of ETV6 (12p13) and ABL1 (9q34) an in-frame fusion cannot be produced by a simple balanced translocation. In fact, the fusion results from a complex rearrangement involving a translocation and inversion or an insertion of ETV6 into chromosomal band 9q34 or ABL1 into 12p13. Alternative splicing generates two fusion transcripts - type A and B without and with ETV6 exon 5, respectively. Both result in constitutive chimeric tyrosine kinase activity analogous to BCR-ABL1 fusion.1 Likewise the effect of ETV6-ABL1 on cellular proliferation, cell survival and transforming capacity mirrors that seen in cases with BCR-ABL1.32 However, unlike BCR-ABL1, ETV6-ABL1 does not cause acute leukemia in mice; instead it induces a chronic myeloproliferation similar to BCR-ABL1-induced chronic myeloid leukemia.4 In humans, the oncogenic potential of ETV6-ABL1 is similar to that of BCR-ABL1 and both fusions can be found in a spectrum of malignancies including myeloproliferative neoplasms (MPN), acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML).

There is renewed interest in ETV6-ABL1 since the discovery of a “BCR-ABL1-like” (or “Ph-like”) gene expression profile (GEP) among B-cell precursor (BCP) ALL cases lacking an established chromosomal abnormality (so-called B-other ALL). The BCR-ABL1-like GEP resembles the BCR-ABL1 GEP because both are driven by the activation of kinase signaling.85 In vitro studies suggest that many of these kinase activating aberrations, including ETV6-ABL1 fusion, are sensitive to specific tyrosine kinase inhibitors (TKI).1097 In addition, there are case reports of patients responding to TKI treatment indicating that these aberrations represent a promising and relevant therapeutic target especially given the reported unfavorable prognosis of BCR-ABL1-like ALL.141175 However, the biology of this heterogeneous group of abnormalities is not fully understood and there is evidence that the prognosis of patients depends on the type of kinase activating lesion and the presence of cooperating aberrations such as IKZF1 deletions and JAK2 mutations.7 Further studies, based on patients with specific gene fusions, are therefore warranted.

Here we present new data from 20 ETV6-ABL1 patients and thoroughly scrutinized existing data from other patients with this fusion to evaluate different detection methodologies, estimate frequency, describe clinical features and outcome, and characterize the profile of copy number aberrations and gene expression of this distinct molecular subtype.

Methods

Patients

The cohort consisted of 44 patients with an ETV6-ABL1 fusion and comprised newly identified cases (n=9), published cases with additional new data (n=11)201587 and cases with re-examined published data (n=24).3921181 Standard diagnostics, including molecular genetics, karyotyping and fluorescence in-situ hybridization (FISH) were performed according to the standard practice of the local diagnostic laboratories. Basic clinical/outcome data were collected from treating centers. Detailed diagnostic procedures have been published for existing cases. For the newly described patients, diagnostic and treatment procedures and protocols were approved by local Institutional Review Boards. Informed consent was obtained in accordance with the Declaration of Helsinki.

Primary samples from the first and second MPN in lymphoid blast crisis of case 34-a-MPN used in this study were provided by the Institute Paoli Calmettes Tumor Bank (Marseille, France). The primary childhood leukemia sample of case 09-ch-ALL was provided by the Leukaemia and Lymphoma Research Childhood Leukaemia Cell Bank working with teams in the Bristol Genetics Laboratory (Southmead Hospital, Bristol), Molecular Biology Laboratory (Royal Hospital for Sick Children, Glasgow), Molecular Haematology Laboratory (Royal London Hospital, London), and Molecular Genetics Service and Sheffield Children’s Hospital, Sheffield) in the UK.

Detection of transcript variants

The expression of alternative splice variants (type A/B) was detected by end-point reverse transcriptase polymerase chain reaction (RT-PCR)20 and quantitative RT-PCR (qRT-PCR) analyses, utilizing forward primers annealing to ETV6 exon 5 (type B-specific: 5′- GCCCATCAACCTCTCTCATCG -3′) or ETV6 exon 4/ABL1 exon 2 junction (type A-specific: 5′- CAGAACCATGAAGAAGAAGCCC -3′). For the schematic representation of transcript variants and position of primers see Online Supplementary Figure S1.

Single nucleotide polymorphism array

Copy number alterations (CNA) were determined by single nucleotide polymorphism array (HumanOmni Express BeadChip, Illumina, San Diego, USA). DNA from bone marrow or peripheral blood cells was extracted using standard methods. DNA was labeled and hybridized according to the recommended Infinium HD assay Ultra protocol from Illumina; scanning was performed using an Illumina iScan System. GenomeStudio Software v2011.1 (Illumina) was used for genotype calling and quality control. DNA CNA and regions of uniparental disomy (UPD) were analyzed using the CNV Partition 2.4.4 algorithm plug-in within the GenomeStudio followed by visual inspection in the Illumina Chromosome Browser (display option of the GenomeStudio) and manual correction. Identified CNA/UPD were mapped against human genome assembly GRCh37/hg19. Deletions corresponding to somatic rearrangements of the immunoglobulin and T-cell receptor gene loci and CNA/UPD seen in remission samples (or those that overlapped with copy number polymorphisms listed in the Database of Genomic Variants) were excluded from further analysis.

To identify recurrently affected regions, the genome was segmented into the minimal segments using genomic coordinates for the start and end of all detected CNA/UPD from all analyzed samples. Segments affected by the same type of aberration (loss or gain or UPD) in more than two cases were called recurrently affected regions (Online Supplementary Figure S2). For this analysis continuous CNA/UPD in a particular sample could be segmented into several neighboring regions when they overlapped in the different sets of samples.

Multiplex ligation-dependent probe amplification

Multiplex ligation-dependent probe amplification assays (P335-A3 ALL-IKZF1, MRC-Holland, The Netherlands) were performed according to the manufacturer’s instruction. Fragments were analyzed on an Applied Biosystems 3730 DNA Analyzer (Applied Biosystems, Carlsbad, CA, USA) and data analyzed using Peak Scanner v1.0 and Coffalyser v9.4 software. Relative copy number was calculated after intra-sample normalization against control fragments and inter-sample normalization against control samples (healthy blood donor DNA). Probes with a ratio of 1±0.3 were assigned a normal copy number of 2, probes with a ratio of less than 0.7 and less than 0.3 were considered heterozygously and homozygously deleted, respectively, while probes with a ratio of more than 1.3 were considered amplified.

Whole genome gene expression profiling and the BCR-ABL1-like signature

RNA was isolated from pretreatment diagnostic ALL samples and analyzed on Affymetrix Human Genome GeneChip U133_Plus_2.0 arrays. CEL files from the previously published 1181 cases7 (including 2 ETV6-ABL1-positive patients reported in this study: 06-ch-ALL and 14-a-ALL) were normalized together with another ten ETV6-ABL1-positive samples described here. Besides the 12 ETV6-ABL1-positive samples the cohort for gene expression profiling consisted of 664 children with National Cancer Institute–classified high-risk BCP-ALL, 348 adolescents (16 to 20 years old), and 167 young adults (21 to 39 years old). The cohort included 84 BCR-ABL1-positive ALL and 209 BCR-ABL1-like ALL. Expression data were generated using the default RMA conditions of the Expression Console 1.4.1 software (Affymetrix). The clustering was performed using MATLAB R2015a (MathWorks) with Euclidean distance and weighted linkage using 257 probe sets predicting BCR-ABL1-like status according to predictive analysis of microarrays.7 Maximum ranges for coloring high (red) and low (green) expression were set to 10-fold the median. Information for BCR-ABL1-like and BCR-ABL1 status was obtained from previously published data. The BCR-ABL1-like cluster was deduced from the expression patterns of its signature genes.7

The BCR-ABL1-like signature was determined using customized Taqman low density arrays at the South Australian Health & Medical Research Institute.19

Results

Disease subtype, frequency, demographics and clinical features

Of the 44 ETV6-ABL1-positive patients considered in this study, half were diagnosed with ALL [n=22; including one manifesting as B-lymphoblastic lymphoma (13-ch-LBL)], while the other half had chronic myeloid leukemia-like MPN (n=18) or AML (n=4) (Figure 1; Table 1). A high diagnostic white blood cell count (>50×10/L) was observed in 10/12 children and 3/7 adults with ALL. Interestingly, MPN cases presented with lower white blood cell counts (median 36×10/L with only 1/16 patients having a count >100×10/L) than those reported for patients with chronic myeloid leukemia.40 Eosinophilia, suggested as a hallmark of ETV6-ABL1 leukemia,28 was present in all MPN and AML cases but only 4/13 ALL cases had an elevated absolute eosinophil count (>0.5×10/L).

Figure 1.Malignancy subtypes, age and gender distribution of the ETV6-ABL1-positive cases. In childhood, all cases were diagnosed as ALL (albeit one originally manifested as B-lymphoblastic lymphoma). Nine childhood ALL were of BCP and two of T-cell origin. Notably, there were two infants in this group; however, most of the childhood ALL were diagnosed in the preschool/early school age (5–10 years). ALL is dominant until the fourth decade and then it is replaced by MPN, which account for more than 70% of cases in the age categories over 35 years. Overall, the male to female ratio is 1.9: 1 (28:15). While gender distribution is almost even in ALL (M:F 12:10), patients with MPN were predominantly male (M:F 12:5) and all four AML cases were diagnosed in males. *The gender of one case of MPN was not reported.

Table 1.Methods and results of the performed diagnostic/research genetic tests.

Routine screening of newly diagnosed childhood ALL patients for ETV6-ABL1 was started in 2003 and 2007 in the Czech Republic and Italy, respectively. By the end of 2014, 730 and 2795 cases had been screened by PCR. This prospective screening coupled with two smaller studies4241 allows us to estimate the frequency of ETV6-ABL1 in childhood ALL at 0.17% (6/3610) which corresponds to ~1–2% of BCR-ABL1-like cases. Meta-analysis of the data from four studies of adult ALL cases showed a frequency of 0.38% (4/1060).42392018 One screening study in AML found a single case (1/1197) which coupled with the fact that only four ETV6-ABL1-positive AML cases have been published so far suggests that ETV6-ABL1 is very rare in AML.18 Estimating the incidence of ETV6-ABL1 in MPN is difficult because no systematic screening data are available.

Identification and location of the ETV6-ABL1 fusion

Among the 22 cases of ETV6-ABL1-positive ALL, the fusion was originally identified by PCR screening (n=10), cytogenetics/FISH (n=6), whole transcriptome sequencing (RNAseq, n=3) or another approach (n=3). Among MPN/AML cases the majority of the fusions were originally detected by cytogenetics/FISH (16/21) and only a minority by PCR (n=2) or other techniques (n=3) (Table 1). In 14 cases, we re-analyzed samples to determine ETV6-ABL1 splice variants. We detected both types (A and B, with type B expressed at significantly higher levels) in all but one case (patient 18-a-ALL in whom only type B was detected, possibly due to poor quality RNA) (data not shown).

A total of 28 cases were tested using commercial ETV6 and/or ABL1 FISH probes. Abnormal signal patterns indicative of ETV6 or ABL1 rearrangement were observed in only 19/28 (67%) cases tested with BCR-ABL1 probes and 10/19 (53%) with ETV6 probes. A further eight ETV6 or ABL1 aberrant signals were detected using specific BAC/PAC/YAC/cosmid FISH probes (7 ETV6, 1 ABL1) (Table 1 and Online Supplementary Table S1). In 30 patients the cytogenetic/FISH result enabled localization of ETV6-ABL1 to a particular chromosome (Online Supplementary Figure S3). In most cases (18/30) the fusion was present on chromosome 12 (10/10 BCP-ALL, 7/16 MPN, 1/3 AML) whereas it was located on chromosome 9 in nine cases (7/16 MPN, 1/1 T-ALL, 1/3 AML). In three cases with more complex rearrangements involving a third chromosome the fusion signal was found on chromosome 8 (AML), 16 (MPN) and 17 (MPN).

Copy number aberration profile of ETV6-ABL1-positive leukemia

We assessed the CNA profile of 22 samples from 18 ETV6-ABL1 patients by high-density single nucleotide polymorphism array further refined by multiplex ligation-dependent probe amplification analysis in seven cases. The cohort comprised 15 patients with ALL and three with MPN (Figure 2; Table 1). Three MPN samples were taken at lymphoid blast crisis (LBC) and two in chronic phase (CP). Among ALL and MPN-LBC samples, we found an average of 12 regions of CNA or UPD per case (range, 2 to 29 alterations). Deletions were more common than gains (206:19). No CNA or UPD abnormalities were detected in two MPN-CP samples (Online Supplementary Tables S2 and S3). Genome segmentation based on the overlap of all CNA identified 74 genomic regions which were affected in at least two patients. The majority of these regions are neighboring loci on chromosomes 7 and 9 (Online Supplementary Figure S1). The most prevalent aberrations detected by single nucleotide polymorphism array in ALL/MPN-LBC were deletions of CDKN2A/CDKN2B (16/17; 94%, although in one MPN case the deletion was apparent only at the second LBC) and deletions of IKZF1 (15/17; 88%), followed by PAX5 (8/17; 47%) and BTG1 (7/17; 41%) deletions (although again in the same MPN case the PAX5 deletion was detected only at the second LBC). In eight patients an aberration of the ABL1 gene was detected (gain of entire ABL1 in 4 cases, partial ABL1 gain in 3 and partial deletion in 1 case) and three patients had partial deletions of ETV6 exons 6–8 (associated with ABL1 gain in all 3 cases). Taking into account the affected exons, these partial deletions/gains seem to result from the primary unbalanced genomic rearrangement creating the fusion gene rather than being secondary aberrations. Other recurrent aberrations occurring in two to five patients included deletions of SLX4IP, ATP10A, BTLA, CD200 and RB1 genes. Overall, the frequency of the aberrations is similar to that in BCR-ABL1-positive and BCR-ABL1-like (Ph-like) ALL. Notably, while the two MPN-CP samples had no detectable CNA, the MPN cases analyzed in LBC had the same CNA pattern as ALL cases (Figure 2).

Figure 2.Overview and frequency of selected recurrent copy number aberrations in 22 samples (17 ALL, 3 MPN-LBC, 2 MPN-CP) from 18 ETV6-ABL1-positive cases. The most prevalent aberrations detected by the SNP array in ALL/MPN-LBC were deletions of CDKN2A/CDKN2B and deletions of IKZF1 followed by PAX5 (8/17; 47%) and BTG1 (7/17; 41%) deletions. CDKN2A and CDKN2B as well as BTLA and CD200 genes were co-deleted in all cases, thus the pairs are merged into one column. Deletion of ATP10A was recently shown to be a recurrent aberration in BCP-ALL, associated with decreased 10-year event-free survival⁵⁶. Gray indicates loss, black indicates a gain, deleted exons within IKZF1, ABL1 and ETV6 genes are indicated by numerals (for IKZF1 white numerals depict losses resulting in a dominant negative IKZF1 isoform or complete loss of IKZF1 expression, black numerals depict haploinsufficiency with the potential exception of combined 2–3/4–7 loss which can affect either a single allele or both gene alleles and thus can result in either haploinsufficiency or complete loss). For frequencies of the selected recurrent copy number aberrations in published ALL studies see Online Supplementary Figure S4. a: adult; ch: childhood; LBC: lymphoid blast crisis of MPN; CP: chronic phase of MPN. ^$In these cases some of the findings from the SNP array were further refined by multiplex ligation-dependent probe amplification. ^#ABL1 deletions were not found recurrently in the present study, thus, only ABL1 gains were considered for CNA frequency. * The relatively high frequency of ETV6 deletions in ETV6-ABL1-positive cases compared to that in BCR-ABL1-positive ALL probably does not reflect frequency of genuine secondary ETV6 loss but rather results from ETV6 loss during primary genomic rearrangement.

Gene expression profile and BCR/ABL1-like status

Expression profiling was performed on 12 samples from ten patients: diagnostic samples from seven ALL cases along with relapse/LBC samples from two cases of BCP-ALL and one MPN case. We compared the GEP of these ETV6-ABL1 samples with those of a previously published cohort of 1179 BCP-ALL cases tested on the same platform at the University of New Mexico.7 Overall, with the 257 probe sets predicting BCR-ABL1-like status according to predictive analysis of microarrays,7 the GEP of all the ETV6-ABL1 cases was consistent with the signature produced from BCR-ABL1 cases; furthermore, in an unsupervised analysis, all ETV6-ABL1 samples clustered within the BCR-ABL1-like cluster and in close vicinity to BCR-ABL1 cases (unlike other ABL1 and ABL2 fusion genes) (Figure 3).

Figure 3.Unsupervised hierarchical clustering of ETV6-ABL1 cases with a large cohort of BCP-ALL. Twelve ETV6-ABL1 cases clustered with a cohort of 1179 BCP-ALL cases using data from RMA normalized U133_Plus_2.0 arrays. The 1191 samples are shown in columns, while the 257 BCR-ABL1-like predicting probe sets (described previously⁷) are in rows. The yellow box highlights the BCR-ABL1-like cluster. BCR-ABL1, ETV6-ABL1 and other ABL1 and ABL2 fusion (NUP214-ABL1, SNX2-ABL1, ZMIZ1-ABL1, RSCD1-ABL1, RSCD1-ABL2, PAG1-ABL2, ZC3HAV1-ABL2) status is shown with black, red and magenta bars, respectively, across the top of the heatmap. All other Ph-like cases are shown with blue bars using the status defined in their original report.⁷ The order of samples in the Ph-like cluster is enlarged on the top (with the same color coding). The order of the ETV6-ABL1 samples as shown in the figure, from left to right is 34-a-MPN (1^st LBC)/08-ch-ALL/10-ch-ALL/09-ch-ALL/06-ch-ALL/14-a-ALL/11-ch-ALL/12-ch-ALL/13-ch-LBL (1^st relapse)/05-ch-ALL (2^nd relapse)/05-ch-ALL (1^st relapse)/34-a-MPN (2^nd LBC).

Moreover, we analyzed ETV6-ABL1 cases by custom Taqman low-density array19 to determine whether they would be classified as BCR-ABL1-like also when using a method which, unlike GEP, is applicable in routine diagnostics. Eleven of 12 cases (9 BCP-ALL, 1 T-ALL and 1 MPN-LBC) tested positive for the BCR-ABL1-like signature indicating that these cases do indeed share considerable similarity with other BCR-ABL1-like cases. The single negative sample came from a child with BCP-ALL (11-ch-ALL) (Table 1).

Clinical characteristics, treatment and outcome

Response to treatment is the most important predictor of outcome in ALL.4543 Treatment response is determined differently per protocol but is usually measured after 1 or 2 weeks of treatment with corticosteroids (“prednisone response”) and again at the end of induction therapy. In this series, both children with a poor prednisone response died whereas three of four patients with a good prednisone response remain alive after >3 years (Table 2). Residual disease was assessed at the end of induction in ten children; four cases had no or low levels of minimal residual disease (<10) whereas six cases had higher levels of residual disease (>10). Of the four patients with low-risk minimal residual disease, three are alive >4 years after diagnosis whereas one patient died after a bone marrow relapse. One infant case with a borderline level of minimal residual disease (3×10) is alive after a central nervous system relapse. Out of the five patients with minimal residual disease levels ≥10 at the end of induction therapy, three relapsed and died while the other two are in first remission, but only 19 and 28 months after diagnosis (Table 2). Of the nine adults with ALL, seven died, one relapsed with no further outcome data and one was in short-term remission after stem cell transplantation. The survival rate of the ALL patients is 46% in children (6/13 alive) and 13% in adults (1/8 alive). Three of the four AML patients failed to achieve a complete remission and hence died within a few months. The overall survival of ETV6-ABL1 MPN cases is 50% (9/18 alive) but the risk of death is more than double for patients diagnosed in blast crisis. Importantly, of the 23 patients with ETV6-ABL1-positive malignancies in whom the cause of death could be evaluated, 19 (>80%) died of disease progression or relapse.

Table 2.Demographics, clinical characteristics and outcome of ETV6/ABL1-positive cases.

Although TKI were not used uniformly to treat this cohort of patients, 17 patients were given a TKI at some point during their treatment. The two children with ALL who received imatinib during their frontline treatment remain in first remission 46 and 57 months after initial diagnosis even though one experienced an isolated central nervous system relapse after 20 months. Two children were treated with TKI after relapse but both died shortly afterwards. In adult ALL, one patient (aged 81) was treated with TKI within frontline treatment and died after relapse (Table 2).

A total of five MPN patients diagnosed in CP received imatinib during frontline therapy and all are still alive. However, two cases did switch to nilotinib because of disease progression. In comparison only three out of eight MPN patients diagnosed in CP who did not receive a TKI are still alive. Two patients received a TKI at disease progression only but both subsequently died. Five of the MPN patients were diagnosed in blast crisis (Table 2). Four of these patients received a TKI during frontline treatment and the fifth patient only after relapse. Four of these MPN patients with blast crisis died, three within 1 year. One patient who received a TKI during frontline therapy remains alive after 12 months but has cytogenetically detectable residual leukemia.

Discussion

The ETV6-ABL1 fusion is rare but can occur in a range of hematologic malignancies; principally BCP-ALL (sporadically manifesting as lymphoma) and MPN but also TALL and AML. In this respect, ETV6-ABL1-positive malignancies are reminiscent of BCR-ABL1-positive neoplasms.46 An additional similarity lies in the non-random age distribution with acute leukemias dominating in children and young adults and MPN in older adults.46 However, the incidence of ETV6-ABL1 leukemia is considerably lower than that of BCR-ABL1 leukemia, accounting for <1% of cases of ALL at any age. An accurate estimate of the frequency of ETV6-ABL1 in MPN remains elusive as there has not been any systematic screening. However, all cases of BCR-ABL1-negative chronic myeloid leukemia-like MPN should be screened for this fusion as it might account for a significant proportion of this relatively rare condition.

Given that at least three chromosomal breaks are required to generate an in-frame fusion, the genomic rearrangement is not uniform across cases and typically involves cryptic insertions. Our study suggests that insertions of ABL1 into the ETV6 locus are favored in BCP-ALL (10/10 patients) whereas in other cases the location of the fusion was evenly distributed - we can only speculate about the biological basis and relevance of this finding or whether it is simply a coincidence. The complex and heterogeneous nature of the rearrangement, coupled with its rarity, makes routine detection challenging. Routine karyo typing is usually inconclusive and FISH with either BCR-ABL1 or ETV6-RUNX1 probes (including specific “break-apart” probes) can miss the fusion – the cryptic insertions are often too small to generate a visible signal split. Indeed, insertions of part of the ABL1 gene into the ETV6 locus on 12p were frequently missed with the ETV6 probes (9/13 cases), and, vice versa, insertions of the 5’ part of the ETV6 gene into the ABL1 locus on 9q can be missed with the ABL1 specific probes (6/8 cases) (Online Supplementary Figure S3). Thus, only a combination of ETV6 and ABL1 probes reliably identifies the fusion by FISH.

There are two types of ETV6-ABL1 transcript, caused by alternative splicing, but their expression was not consistently analyzed in previous studies. We detected both transcripts in all but one of the samples we analyzed, with the type B variant, including ETV6 exon 5, expressed at higher levels. These observations do not agree with those of previous studies which reported four cases expressing only the type A transcript. However, in three cases the type B fusion could have been missed due to the primers used for RT-PCR.3933 In the other case, the absence of ETV6 exon 5 in the expressed transcript was shown to result from direct disruption of this exon by a fusion-generating DNA break.28 A chimeric kinase without ETV6 exon 5 lacks the GRB2 binding site and has attenuated oncogenic potential analogous to BCR-ABL1 deficient for GRB2 interaction.484728 The predominance of the type B transcript does, therefore, have a biological basis. The recently reported complex network of kinase lesions in ALL makes RNA sequencing a useful additional diagnostic tool and in three cases in this series the ETV6-ABL1 type B fusion was identified this way. Some clinical study groups have opted to pre-select BCP-ALL for BCR-ABL1-like cases prior to screening for specific fusions.87 However, we present an ETV6-ABL1 BCP-ALL case classified as BCR-ABL1-like-negative by Taqman low density array analysis (despite clustering with other ETV6-ABL1 and BCR-ABL1 cases on GEP) and a T-ALL case with ETV6-ABL1 classified as BCR-ABL1-like positive. Thus although methods to evaluate the BCR-ABL1-like signature were developed on expression data from BCP-ALL, they might also have the potential to identify T-ALL cases with kinase-activating lesions. This potential does, however, need to be evaluated further.

Genomic and gene expression profiling demonstrates the similarity of ETV6-ABL1 and BCR-ABL1 cases. Besides the involvement of ABL1, these cases have a similar profile of secondary aberrations including frequent deletions of IKZF1, CDKN2A/B, PAX5, BTG1 and RB1 which distinguish them from other ALL.5149 We have previously shown the prenatal origin of ETV6-ABL1 in a 5-year old child with ALL20 demonstrating the need for cooperating mutations to induce an overt disease. The two most frequent secondary aberrations in ETV6-ABL1 ALL (CDKN2A/B and IKZF1 loss) have both been shown to cooperate with BCR-ABL1 during leukemogenesis in mice and likely represent cooperating lesions also in ETV6-ABL1 leukemia.5352 These and other CNA recurrently found in our cohort (ATP10, BTLA/CD200) have been associated with an unfavorable prognosis in ALL overall and specifically in BCR-ABL1 and BCR-ABL1-like ALL.5754136 Interestingly, we found no secondary aberrations in MPN-CP samples but progression to LBC was accompanied by the gain of a profile similar to that of ETV6-ABL1 ALL.

Most studies show that the BCR-ABL1-like subgroup has an inferior prognosis.75 The final outcome was shown to depend significantly on the particular primary lesion (with childhood ALL bearing ABL-class aberrations having 5-year event-free survival and overall survival rates of 50% and 68%, respectively) and also on the presence of secondary aberrations (e.g. IKZF1 deletions).7 A recent study showed that, on current minimal residual disease-based protocols, the prognosis of BCR-ABL1-like patients treated with standard chemotherapy and stem cell transplantation is not inferior to that of patients with other ALL; however, only one out of 40 analyzed BCR-ABL1-like patients in this study bore an ABL gene fusion (NUP214-ABL1) and thus the results are not directly applicable to ETV6-ABL1 cases.58 Our study confirms the poor outcome associated with ETV6-ABL1 in ALL and AML; however, we did observe long-term survivors in childhood ALL among patients with a good initial response. The number of patients in this study was small but there is corroborating evidence in the literature from patients with other kinase fusions58 to support the notion that a TKI can be useful for treating slow-responding or high-risk ALL. Again among a small number of cases of MPN there was a trend towards improved outcome for CP patients treated upfront with a TKI (5/5 patients alive while 5/8 MPN-CP patients without TKI treatment died). In line with increased genomic complexity during LBC, patients who presented at this advanced stage did not survive despite TKI therapy. These observations are consistent with recent data in adult BCR-ABL1 ALL which showed that the effect of a TKI is optimal when it is administered early during frontline therapy.59

In conclusion, the ETV6-ABL1 fusion is a rare and complex rearrangement occurring in a variety of hematologic malignancies but especially ALL and MPN. The diagnosis is often incidental and only targeted RT-PCR screening (e.g. in a multiplex setting with primers for ETV6-RUNX1 and BCR-ABL1 fusions, also detecting ETV6-ABL1 cases), a combination of both ETV6 and ABL1 FISH probes or RNAseq will reliably detect all cases. ETV6-ABL1 leukemias resemble BCR-ABL1 leukemias in many aspects including expression profile and secondary genetic aberrations. The outcome of ETV6-ABL1 ALL is often poor but in childhood ALL with an excellent early treatment response, continuous remissions seem to be achievable with current therapy without TKI. There is increasing evidence that among poor responding childhood ALL cases, as well as in adult acute leukemias and in MPN, the poor outcome can be abrogated by the use of a TKI which should be administered early, before progression of the disease, and thus depends on timely diagnostics. Future screening algorithms should include appropriate measures to detect ETV6-ABL1 so that the most appropriate treatment strategy can be determined.

Acknowledgments

The authors would like to thank the Institute Paoli Calmettes Tumor Bank (Marseille, France) and Leukaemia and Lymphoma Research Childhood Leukaemia Cell Bank for providing cell samples. The work was supported by grants IGA-MZ NT/13170-4, GAUK 694414 and AZV 15-30626A and by grants from the National Institutes of Health (USA) CA98543 (COG Chair’s grant), CA98413 (COG Statistical Center), and CA114766 (COG Specimen Banking).

Footnotes

• Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/101/9/1082

• Received February 12, 2016.
• Accepted May 18, 2016.