Abstract
Relapsed precursor T-cell acute lymphoblastic leukemia is characterized by resistance against chemotherapy and is frequently fatal. We aimed at understanding the molecular mechanisms resulting in relapse of T-cell acute lymphoblastic leukemia and analyzed 13 patients at first diagnosis, remission and relapse by whole exome sequencing, targeted ultra-deep sequencing, multiplex ligation dependent probe amplification and DNA methylation array. Compared to primary T-cell acute lymphoblastic leukemia, in relapse the number of single nucleotide variants and small insertions and deletions approximately doubled from 11.5 to 26. Targeted ultra-deep sequencing sensitively detected subclones that were selected for in relapse. The mutational pattern defined two types of relapses. While both are characterized by selection of subclones and acquisition of novel mutations, ‘type 1’ relapse derives from the primary leukemia whereas ‘type 2’ relapse originates from a common pre-leukemic ancestor. Relapse-specific changes included activation of the nucleotidase NT5C2 resulting in resistance to chemotherapy and mutations of epigenetic modulators, exemplified by SUZ12, WHSC1 and SMARCA4. While mutations present in primary leukemia and in relapse were enriched for known drivers of leukemia, relapse-specific changes revealed an association with general cancer-promoting mechanisms. This study thus identifies mechanisms that drive progression of pediatric T-cell acute lymphoblastic leukemia to relapse and may explain the characteristic treatment resistance of this condition.Introduction
T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive malignancy of thymocytes that accounts for about 15% of pediatric acute lymphoblastic leukemias1 and can be cured in approximately 80% of affected children.2,3 However, relapsed T-ALL is highly resistant to chemotherapy4,5 and patients who experience a relapse face a dismal prognosis.5 In B-cell precursor (BCP)-ALL, integrated genomic analysis identified relapse-specific signatures of gene expression, somatic copy number alterations (CNA) and promoter methylation.6,7 Similar observations in relapsed T-ALL are scarce: gene expression analysis revealed that relapsed T-ALL is characterized by a gene response pattern that is also found in early, but not in late relapse of BCP-ALL.8 Genome-wide analysis in relapsed T-ALL, in contrast to relapsed BCP-ALL, found similar numbers of CNA as those in matched samples taken during the primary disease.9 While DNA methylation in T-ALL has been shown to differ from that in BCP-ALL,10 systematic comparisons of DNA methylation between primary and relapsed T-ALL are not available. Evolution into relapsed TALL has been attributed to clonal selection, because relapse-specific rearrangements of the T-cell receptor were backtracked to primary disease11 and because leukemic cells from primary disease can develop into a clone resembling relapsed disease after transfer into immunodeficient mice.12 Rarely, late relapse of T-ALL represents a true secondary leukemia, whereby an underlying genetic predisposition could be documented in at least one patient.13
To obtain a more comprehensive picture of the evolution of T-ALL from primary disease to relapse we performed an integrated genomic analysis of 13 patients with T-ALL, who were analyzed at the time of primary diagnosis, during remission and at relapse.
Methods
Patients’ clinical characteristics
Patients were treated according to ALL-BFM 2000 or related frontline protocols (INS-89, INS-98,14 IC15). One patient was aged 18 at diagnosis, all others were children or adolescents. The 13 patients (Table 1) were recruited between 1993 and 2007 from the ALL-REZ BFM 2002 trials (patients T-ALL-H-A61, -E114, -F110, -KI17, -MD40, -T92, -T128) or from Schneider Children’s Medical Center of Israel, Petah Tikva, Israel (patients T-ALL-H-S00169, -S00207, -S00285, -S00438, -S00456, -S00472) and selected on the basis of sufficient material being available from the time points of first diagnosis, remission and relapse. Minimal residual disease (MRD) response was assessed as described previously2,16 (Online Supplementary Table S3).
This study was approved by the institutional review boards of the Charité Universitätsmedizin Berlin and the Medical Faculty Heidelberg. Informed consent was obtained in accordance with the Declaration of Helsinki.
Exome capture, target capture and Illumina sequencing
The Agilent SureSelect Target Enrichment Kit (Agilent, Santa Clara, CA, USA; vendor’s protocol version 2.0.1) was used to capture all human exons for sequencing. The HaloPlex Target Enrichment Kit (Agilent, Santa Clara, CA, USA; vendor’s protocol version D.5, May 2013) was used according to the manufacturer’s instructions. The starting material consisted of 225 ng of genomic DNA. The captured fragments were sequenced as 100 bp paired reads using an Illumina HiSeq instrument (Illumina, San Diego, CA, USA).
Analysis of whole exome sequencing and targeted ultra-deep sequencing data
The analysis of sequencing data is detailed in the Online Supplementary Methods.
DNA methylation analysis using 450k BeadChip arrays
Genomic DNA (200 ng) was bisulfite-converted using the EZ DNA Methylation Gold Kit (Zymo Research, Irvine, CA, USA). The Infinium methylation assay (Illumina) was carried out as previously described.17 Data from the 450k Human Methylation Array were normalized by the Beta Mixture Quantile (BMIQ) method18 using the RnBeads analysis software package.19
Multiplex ligation dependent probe amplification
Multiplex ligation dependent probe amplification (MLPA) was done using the MRC Holland (Amsterdam, The Netherlands) SALSA MLPA probe mix P383-A1 TALL according to the manufacturer’s instructions. Polymerase chain reaction products were separated by capillary electrophoresis on an ABI-3130XL device; the size standard was GeneScan 500–250 (Applied Biosystems). Coffalyser software, available at http://www.mlpa.com, was used for the analyses.
Table 1.Clinical characteristics of the 13 T-ALL patients analyzed.
Integrated analysis
In order to evaluate functions that were altered by mutation or DNA methylation at relapse, Ingenuity Pathway Analysis (IPA, Version 21249400; Qiagen, Redwood City, CA, USA) was used. For details refer to the Online Supplementary Methods.
Results
Relapsed T-cell acute lymphoblastic leukemia acquires single nucleotide variants and small insertions and deletions
We performed whole exome sampling (WES) of DNA samples obtained from 13 patients (see Table 1 for clinical characteristics) at the time of first diagnosis, during remission and at relapse, and identified a total of 340 somatic single nucleotide variants (SNV) and 53 small insertions and deletions (InDels; Online Supplementary Table S1). On average, each leukemia specimen carried 11.5 (SD 7.0) somatic SNV and InDels at first diagnosis and 26.0 (SD 26.6) somatic SNV and InDels at the time of relapse – suggesting ongoing evolution of the major clone or of minor subclones. Only 85 of 126 SNV and 14 of 23 InDels present at the time of first diagnosis could also be detected in the corresponding relapse sample, indicating that the (treatment-related) eradication of leukemic clones contributes to the evolution of relapse. We defined a mutation as “relapse-specific” if it was detected by WES or MLPA in a relapse sample, but not in the corresponding primary leukemia sample from the same patient. This could include mutations that were detected in the primary leukemia of other patients.
Comparing all mutations in primary disease to those in relapse, no significant difference in the types of single nucleotide exchanges and in the ratio of transversions and transitions was identified (Online Supplementary Figure S1A). However, mutations at the central position of the trinucleotides TCA and GCT were significantly more frequent among relapse-specific SNV than in primary leukemia (Online Supplementary Figure S1B,C), consistent with a contribution of cytarabine to the acquisition of mutations during treatment.20
Somatic copy number alterations can represent subclonal, “late” events
We used MLPA in order to identify CNA in genes commonly altered in T-ALL. On average, 3.5 CNA were detected in each primary leukemia sample and 3.2 in each relapse sample, whereas 2.7 CNA were detected in both samples from the same patient (Online Supplementary Table S2). Most commonly, we found deletions of the CDKN2A/B locus21 on chromosome 9q (all patients), followed by microdeletions within the LEF1 gene22 (6/13 patients), amplification of the MYB gene23,24 (4/13 patients), deletions of the PTEN gene25,26 (3/13 patients) and homozygous deletions of the PTPN2 gene27 (2/13 patients). Thirty-five of the 45 CNA that were identified in primary disease were preserved in relapse, a proportion comparable to the corresponding numbers of SNV and InDels. In four patients, CNA found to be lost in relapse affected the CDKN2A, LEF1 or PTEN genes, indicating that the deletion of these tumor suppressors can be a late event during leukemogenesis. Although several CNA (deletions of SUZ12, LEF1, PTEN, CASP8AP2; amplification of MYB) were specific for relapse in individual patients, none of these was specific for relapse in more than one patient or significantly enriched in either primary disease or relapse.
Targeted ultra-deep sequencing detects rare subclones
In order to track the clonal evolution of the relapses, we performed targeted ultra-deep sequencing of all mutations that had previously been identified by WES using the HaloPlex target capture (Agilent). Allele frequencies inferred by HaloPlex were highly reproducible (Online Supplementary Figure S2A) and corresponded well to allele frequencies inferred by WES (Online Supplementary Figure S2B). We analyzed whether HaloPlex reflected the frequencies of rare alleles by measuring the signals in serial dilutions. This analysis indicated that the allele frequency is highly correlated with the dilution (r=0.91). With increasing dilution and decreasing allele frequencies the accuracy of quantification of allele frequencies dropped because mutant reads were overrepresented (Online Supplementary Figure S2C). Rare alleles were identified by comparing allele frequencies in the sample of interest with a reference pool consisting of healthy control DNA and unrelated leukemia samples. This technique sensitively detected alleles with a frequency in the range of 0.01 and below: 91% of leukemia-specific SNV were detected after diluting leukemia DNA in control DNA at a ratio of 10, 26% were detected in a dilution of 10 and 3% in a dilution of 10. The rate of SNV that were detected as false positives in samples from patients who never carried the respective SNV was 1.8% (Online Supplementary Figure S2D), indicating that the detection of a significant proportion of mutant reads by HaloPlex is a reliable qualitative indicator of a small subclone carrying the mutation of interest. The greatest sensitivity of detection was achieved for insertions of more than one nucleotide. For example, in patient’s S00169 remission sample two reads carrying a specific 9 bp insertion in the NOTCH1 gene among 43,567 total reads covering this region indicated persistent MRD in the order of 10 (Online Supplementary Table S6).
In ten of 13 patients we found at least one SNV that had developed from a rare (<0.05) subclonal allele in primary disease to a major clonal allele in relapse (Online Supplementary Table S5). Of the 211 SNV and 30 InDels which were found in the major clone exclusively in relapse by WES, 45 and nine, respectively, were detected at a low allele frequency already in primary disease (Online Supplementary Table S7). The average allele frequency for such relapse-specific mutations found to be subclonal in primary leukemia was 0.006 (range, 0.0002–0.036). Fifty-five SNV and InDels present in the major clone at primary disease had been lost during the transition to relapse. Fifty-four of these 55 mutations were not detected even in minor clones at the time of relapse, indicating that these clones were eradicated fully or at least to a level below the sensitivity of the HaloPlex analysis.
In ten of 13 of the first disease/relapse pairs a DNA sample obtained at the time of remission (for 7 patients, MRD level from the same sample was available and ≤ 10, Table 1) was also available for ultra-deep sequencing. In these samples we searched for the presence of SNV that had been detected in the same patient at the time of either primary disease or relapse. In six patients (T-ALL-H-E114, -MD40, -KI17, -S00169, -S00285, -S00472) we detected a total of 52 (range, 1 – 21) SNV or InDels, which were significantly more abundant in the remission samples than in the reference consisting of the average of all samples from healthy controls and from patients who never carried the respective SNV of interest (Online Supplementary Table S6). Allele frequencies in remission samples ranged between 0.00005 and 0.0095. By comparison with a dilution series of leukemia samples in healthy DNA we estimate that these allele frequencies correspond to a frequency of leukemic cells of 0.0001 to 0.01 (1:10,000 – 1:100), consistent with the MRD levels that were available (Table 1). All of the mutations detected in remission samples were also found in the corresponding relapse sample with an allele frequency in the range of 0.13 to 0.91. Thirteen mutations that were detectable in remission were already found in a major clone (allele frequency 0.22–0.93) at primary disease, and an additional ten SNV were detected in primary disease samples at low allele frequencies (range, 0.0003 – 0.0120). Twenty-six SNV that were detected in remission were not detected in the corresponding primary disease sample with a sensitivity of 0.01 or higher. We propose that these 26 SNV may have originated from a mutational event during treatment, although clonal selection starting from a very small subclone present at initial diagnosis cannot be ruled out.
Figure 1.Ultra-deep sequence analysis of primary and relapsed T-ALL distinguishes two types of relapse. (A, C) Variant allele frequencies in relapse were plotted over the variant allele frequency in the corresponding primary leukemia sample. Allele frequencies were determined by HaloPlex sequencing, only in the rare case that a certain allele was not covered by HaloPlex, allele frequencies from WES were used. (B, D) The simplified models show each mutation as a single symbol (circle, square, pentagon, ellipse). Blue symbols denote mutations that were detected in the major clone of primary leukemia, red symbols denote mutations that were specific for relapse. (A, B) Type 1 relapse: all mutations present in the major clone from primary disease were also present in relapse. The clone giving rise to relapse carried all mutations that were detected in the major clone of primary leukemia. (C, D) Type 2 relapse: the major clone from primary leukemia was lost in relapse, as indicated by the mutations that were present in the major clone in primary leukemia but absent in relapse. The clone giving rise to relapse shared some but not all mutations with the major clone of primary leukemia and was derived from a common ancestor, but evolved independently already before the initial diagnosis.
In all 13 patients a minimal set of common genetic changes (at least one concordant MRD marker, Online Supplementary Table S3, or at least six concordant SNV and InDels, Figure 1) were found in both the primary leukemia and relapse samples, which indicated that all relapses were bona fide recurrences of the leukemia and not a second, unrelated neoplasm (as has previously been described in a small proportion of relapsed patients based on the identification of discordant MRD markers13). By analyzing allele frequency plots, we can distinguish two types of relapse: type 1 and type 2. Type 1 relapse, observed in six of 13 patients (Figure 1A,B), contained all mutations that were already detectable at the time of primary leukemia. This type of relapse developed either from a major sub-clone or from a smaller subclone that had acquired additional mutations late in the process of leukemogenesis. In type 2 relapse, observed in the remaining seven patients, mutations that had been present in the major clone in primary leukemia were lost at relapse (Figure 1C,D). Here, relapse developed from an ancestral pre-leukemic clone that had already diverged into distinguishable subclones at an early time point prior to the initial diagnosis. In both types of relapse, clonal selection and acquisition of novel mutations contributed to the mutational load. Type 1 showed a trend to be more frequent in early relapses (time to relapse <24 months; P=0.07, see Online Supplementary Figure S3 for a logistic regression model) and in Israeli and Palestinian patients (P=0.029, Table 2).
The example of mutations in the nucleotidase NT5C2 illustrates the genetic plasticity of T-ALL. NT5C2 mutations were identified in five of 13 relapse samples (R367Q in patients A61, S00207, S00285, T92; D407Y in patient T92; P414S in S00456; Online Supplementary Tables S1 and S5). R367Q has been shown to activate the nucleotidase activity of NT5C2 and to confer resistance against nucleoside analogs.28,29 In two patients, an NT5C2 mutation was already detected in primary disease samples at low allele frequency (A61: D407Y, allele frequency 0.3%; S00456: P414S, allele frequency 0.1%). While patient S00456 carried the same mutation in the corresponding relapse sample, patient A61 lost the D407Y mutation and acquired the R367Q mutation. NT5C2 mutations were clonal in three relapses, but subclonal in two other relapse samples (S00207: allele frequency 0.1; T92: allele frequencies 0.41 for R367Q and 0.09 for D407Y). This is compatible with the notion that acquisition of resistance to chemotherapy by NT5C2 activation can be a late, not-initiating event on the way to relapse.
Patient E114 demonstrated that the evolution of the relapse-specific clone from a pre-leukemic ancestor may be facilitated by intensive induction treatment of the primary leukemia. In this patient, two preserved MRD markers confirmed the relationship between primary leukemia and relapse. In addition, targeted ultra-deep sequencing identified five mutations that had been present at a subclonal level in primary disease, persisted in remission and became predominant in relapse (Figure 1C, Online Supplementary Tables S6 and S7). Already in the remission sample, which was taken immediately after induction treatment had been completed, 16 newly acquired mutations were detected and later predominated at relapse (Online Supplementary Table S6). During the 2 years of remission before relapse occurred, only five more mutations accumulated. Interestingly, although MLPA identified deletion of CDKN2A/B and amplification of MYB in the primary disease sample of this patient (Online Supplementary Table S2), none of them was conserved between primary disease and relapse, indicating that in this specific leukemia these copy number variants were probably not the initiating event in leukemogenesis.
Relapse-specific alterations in T-acute lymphoblastic leukemia do not show association with leukemogenesis, but with cancerogenesis in general
In nine of 13 patients we identified relapse-specific mutations that, based on current knowledge, are likely to contribute to the evolution of relapse (Table 3). Besides known mechanisms of leukemogenesis in T-ALL such as NOTCH-activation or MYB-amplification, the most common alterations were mutations activating the nucleotidase NT5C2 and mutations in epigenetic modifiers, such as WHSC1, SUZ12, SMARCA4, ARID4B and USP7.
Table 2.Type of relapse is associated with treatment and with time to relapse.
In order to obtain an unsupervised view of the contribution of mutations to relapse we investigated whether the genetic alterations that are specific for relapse can be linked to certain biological functions. To this end, we grouped genetic alterations, which had been detected either by WES or by MLPA, according to the time points at which they were found. Using Ingenuity Pathway Analysis (IPA) software, genes that were mutated or deleted in the major clones at both times, primary leukemia and relapse, were compared to those genes that were found to be mutated or deleted in the major clone at relapse but not in the corresponding primary leukemia (Online Supplementary Table S8).
Using genes altered both in primary leukemia and in relapse, IPA constructed a dense network involving the nodes NOTCH, IL7R, MTOR, GATA3 and AKT, reflecting the frequent occurrence of somatic DNA alterations in established leukemia drivers in this gene set (Online Supplementary Figure S4A; Table 4). In contrast, the genes that were mutated in a relapse-specific way showed similarities to gene lists typically carrying somatic mutations in gastrointestinal and other cancers, but not leukemia (Online Supplementary Figure S4B; Table 4). These findings suggest that relapsed T-ALL share a profile of mutations with non-hematologic cancers. Among genes that were mutated in a relapse-specific manner, we did not observe a strong enrichment of canonical pathways (minimal P=0.003 for actin cytoskeleton signaling).
Hypermutation caused by somatic DNA repair deficiency can contribute to genetic instability in relapsed T-cell acute lymphoblastic leukemia
Patient S00285 carried a moderate number of eight mutations in primary disease, but an extraordinarily high number of 106 mutations in relapse, many of which were subclonal (Online Supplementary Tables S1 and S5). One of the 102 relapse-specific mutations affected the BLM gene. While this mutation was not present at the time of primary disease, it was the most abundant newly acquired mutation in remission and present in the main clone at relapse. BLM codes for a RecQ DNA helicase and is required for DNA replication and DNA repair. Inactivating germline mutations in BLM cause Bloom syndrome, a recessively inherited tumor predisposition syndrome (OMIM #210900). The K39N mutation observed in this patient is predicted to be damaging by PolyPhen30 and likely caused a somatic DNA-repair defect, resulting in the accumulation of a large number of mostly subclonal mutations during treatment.
Hypomethylated promoters in relapsed T-cell acute lymphoblastic leukemia do not show association with leukemogenesis, but with cancerogenesis in general
We used Illumina 450k arrays to compare DNA methylation in relapse samples to that in the corresponding primary T-ALL samples. In contrast to the situation in relapsed BCP-ALL,6 the average degree of methylation in relapse did not differ from average methylation in primary leukemia (Figure 2A, r=0.996). When primary disease and relapse samples from individual patients were compared, both hypo- and hypermethylation of promoters in relapse could be observed (Figure 2B).
Table 3.Mutational mechanisms likely contributing to relapse.
In order to identify promoters that may undergo differential methylation in relapse, we filtered for promoters that: (i) were represented on the 450k array by at least three different probes; (ii) had a gene symbol assigned; and (iii) had a decrease or an increase of the β-value of at least 0.2 in absolute numbers in at least three different patients. According to these criteria, a total of 239 promoters were recurrently hypermethylated and 579 promoters recurrently hypomethylated in relapse. The lists of hyper- and hypomethylated promoters (Online Supplementary Table S8) were subjected to IPA. The association of hypermethylated promoters with diseases and functions reached marginal significance for one functional category only (P=6.62 for “development of neurons”). In contrast, hypomethylated promoters, similar to the mutations specifically found in relapse, showed highly significant associations with different types of cancer, namely melanoma (P=6.89×10) and endometrial carcinoma (P=1.92×10; Table 4). We did not observe a strong enrichment of canonical pathways (minimal P=0.003 for glutamate receptor signaling) among promoters that were hypomethyled in a relapse-specific manner. Although the lack of RNA samples precludes an analysis of a correlation between methylation and gene expression data, our results show that DNA methylation changes may cooperate with mutational events in driving T-ALL relapse.
Table 4.Ingenuity Pathway Analysis.
Discussion
By applying a deep coverage target enrichment technique to sensitively and quantitatively detect rare mutations in primary disease, remission and relapse of pediatric T-ALL, we distinguish between relapses arising from the major clone of the primary leukemia (type 1) and relapses arising from a pre-leukemic ancestral clone (type 2). In both types, selection of subclones and acquisition of novel mutations contributed to clonal evolution. Similar observations have been made before for BCP-leukemia6,7,9 and for acute myeloid leukemia.31 The size of the subclones that later gave rise to relapse can be estimated to range between 1:100 and 1:1,000 of the primary leukemia in our series. The association of type 1 relapse with a short duration of remission may intuitively be explained by a relative resistance of the primary leukemia to treatment. The more frequent occurrence of type 1 relapses in the Israeli/Palestinian group of patients is unlikely to be related to differences of treatment, because the protocols used in Israel and in Germany are very similar and give comparable overall results.2,14
More than three-quarters (187 of 241) of relapse-specific mutations could not be detected in primary disease samples despite a sensitivity of detection that exceeded 1:100 for most mutations. The mutational load during relapse is more than doubled compared to that during primary leukemia, suggesting that mutations are truly acquired de novo during treatment and/or during remission. Relapse-specific patterns of mutations attributed to chemotherapy have been described by others.20,31 Comparing relapse-specific mutations to mutations in primary leukemia, we found a significant increase in single nucleotide exchanges at the central position of the trinucleotide sequence TCA, a mutational signature that has been attributed to the mutagenic effects of cytarabine.20 However, the overall mutational pattern changed only slightly in relapse, indicating that the contribution of chemotherapy to the acquisition of mutations may be minor in this cohort of patients.
The leukemia that acquired a BLM mutation during treatment was of particular interest. This mutation first appeared at the time of remission and was clonal at the time of relapse. The presence of this mutation was associated with an unusually high number of mostly subclonal relapse-specific mutations, indicating that acquired mutations of DNA repair genes may result in somatic hypermutation during the clonal evolution of T-ALL. While haploinsufficiency is not evident in heterozygous carriers of germline BLM mutations,32,33 the effect of somatic BLM mutations in the context of leukemia may be amplified by mutagenic stress induced by chemotherapy.
While the genetic alterations shared by primary leukemia and relapse are, as expected, enriched in genes that are known to be implicated in leukemogenesis,3,26,27,34 the genes that we specifically found to be mutated in TALL relapse are not enriched for leukemia-specific genes. Relapse-specific genes are more closely related to other cancer types, implying that relapsed T-ALL, in addition to the genetic repertoire required to induce leukemia in T cells, have acquired properties that are linked to cancer in a less specific sense. Potentially, these genes contain a previously unrecognized category of genes that are specific for T-ALL relapses. These observations suggest that the type of malignancy is determined by the early genetic changes of the normal hematopoietic precursor and that the progression may be driven by more general abnormalities of carcinogenesis. These progression-related changes may also explain the clinical observation that relapsed T-ALL is frequently refractory to leukemia-directed chemotherapy.
An analysis of the 450k methylome data revealed that, in contrast to BCP-ALL,6 overall DNA methylation levels in T-ALL hardly differ between primary leukemia and relapse. However, a set of promoters was found to be recurrently hypomethylated in relapse compared to primary disease. This set contains many genes related to cancer, albeit not specifically involved in leukemogenesis. The correlation between DNA methylation and RNA abundance could not be analyzed due to lack of RNA samples. However, a strong inverse correlation between promoter methylation and RNA expression has been described before in ALL.6,10 Hence, both relapse-specific mutations and changes of the DNA methylome are consistent with the activation of additional oncogenic mechanisms in relapsed T-ALL.
Figure 2.Methylation analysis identifies promoter specific hypomethylation in relapsed T-ALL. β values in relapse were plotted over β values in primary leukemia in scatter plots. (A) Average β values (single sites) of all patients analyzed. (B) The β value for each promoter in relapse was compared to that in primary leukemia. Promoters with a β value in relapse that was at least 0.2 higher than in primary leukemia were considered to be hypermethylated. Promoters with a β value in relapse that was at least 0.2 lower than in primary leukemia were considered to be hypomethylated. Only promoters of those genes for which at least one gene symbol is assigned and which are represented by at least three probes on the 450k array are represented here.
The importance of epigenetic changes is specifically highlighted by the findings in patient S00169 (Figure 1A, Online Supplementary Table S1), who suffered from a very early relapse and in whom the only detectable mutational difference between primary leukemia and relapse was the acquisition of the activating mutation E1099K in the H3K36 histone methyl transferase WHSC1. This hotspot mutation has been recurrently reported in mantle cell lymphoma35 and in pediatric BCP-ALL36,37 and results in cellular transformation.36 The emergence of a clone carrying the WHSC1 E1099K mutation may have been sufficient to induce early relapse in this patient.
The only gene that was recurrently mutated in a relapse-specific manner was NT5C2,28,29 partially explaining the relative success in the use of nelarabine, which is not a substrate for NT5C2, in T-ALL relapse.38 In our small series of patients, no other mutation was recurrent and many mutations present in the primary leukemia were lost in relapse. The application of novel, targeted therapies to patients with relapsed T-ALL will, therefore, likely require a thorough genetic characterization of relapse-specific targets.
In conclusion, the data presented here identify two molecularly defined types of relapse in pediatric T-ALL and implicate the selection of subclones, the acquisition of novel somatic mutations and the hypomethylation of promoters as mechanisms driving the progression of T-ALL from primary disease to relapse. It is noteworthy that the relapse-specific alterations tend to activate general mechanisms of carcinogenesis rather than known leukemia-specific drivers.
Acknowledgments
The authors would like to thank the following institutions for grants: German Consortium for Translational Cancer Research (DKTK), German Ministry of Education and Research (BMBF, NGFN Plus), “Tour der Hoffnung”, Manfred Lautenschläger Stiftung, European Commission (FP7, ERA-NET on Translational Cancer Research, TRANSCALL to MUM; Health-F2-2010-260791 and a European Research Council grant to JOK). JE and JBK received fellowships from the Heidelberg Research Center for Molecular Medicine. ORB received grants from the Excellence Initiative, Innovation Fund FRONTIER, University of Heidelberg. We thank the EMBL GeneCore sequencing team, the EMBL IT unit, the DKFZ genomics and proteomics core facility, Elena Orlova and Margit Happich for excellent technical assistance.
Footnotes
- ↵* JBK and TR contributed equally to this work.
- ↵** JOK, MUM and AEK contrubuted equally to this work.
- The online version of this article has a Supplementary Appendix.
- Authorship and DisclosuresInformation on authorship, contributions, and financial & other disclosures was provided by the authors and is available with the online version of this article at www.haematologica.org.
- Received April 24, 2015.
- Accepted August 11, 2015.
References
- Dores GM, Devesa SS, Curtis RE, Linet MS, Morton LM. Acute leukemia incidence and patient survival among children and adults in the United States, 2001–2007. Blood. 2012; 119(1):34-43. PubMedhttps://doi.org/10.1182/blood-2011-04-347872Google Scholar
- Schrappe M, Valsecchi MG, Bartram CR. Late MRD response determines relapse risk overall and in subsets of childhood T-cell ALL: results of the AIEOP-BFM-ALL 2000 study. Blood. 2011; 118(8):2077-2084. PubMedhttps://doi.org/10.1182/blood-2011-03-338707Google Scholar
- Van Vlierberghe P, Pieters R, Beverloo HB, Meijerink JP. Molecular-genetic insights in paediatric T-cell acute lymphoblastic leukaemia. Br J Haematol. 2008; 143(2):153-168. PubMedhttps://doi.org/10.1111/j.1365-2141.2008.07314.xGoogle Scholar
- Nguyen K, Devidas M, Cheng SC. Factors influencing survival after relapse from acute lymphoblastic leukemia: a Children’s Oncology Group study. Leukemia. 2008; 22(12):2142-2150. PubMedhttps://doi.org/10.1038/leu.2008.251Google Scholar
- Einsiedel HG, von Stackelberg A, Hartmann R. Long-term outcome in children with relapsed ALL by risk-stratified salvage therapy: results of trial acute lymphoblastic leukemia-relapse study of the Berlin-Frankfurt-Munster Group 87. J Clin Oncol. 2005; 23(31):7942-7950. PubMedhttps://doi.org/10.1200/JCO.2005.01.1031Google Scholar
- Hogan LE, Meyer JA, Yang J. Integrated genomic analysis of relapsed childhood acute lymphoblastic leukemia reveals therapeutic strategies. Blood. 2011; 118(19):5218-5226. PubMedhttps://doi.org/10.1182/blood-2011-04-345595Google Scholar
- Ma X, Edmonson M, Yergeau D. Rise and fall of subclones from diagnosis to relapse in pediatric B-acute lymphoblastic leukaemia. Nat Commun. 2015; 6:6604. PubMedhttps://doi.org/10.1038/ncomms7604Google Scholar
- Bhojwani D, Kang H, Moskowitz NP. Biologic pathways associated with relapse in childhood acute lymphoblastic leukemia: a Children’s Oncology Group study. Blood. 2006; 108(2):711-717. PubMedhttps://doi.org/10.1182/blood-2006-02-002824Google Scholar
- Mullighan CG, Phillips LA, Su X. Genomic analysis of the clonal origins of relapsed acute lymphoblastic leukemia. Science. 2008; 322(5906):1377-1380. PubMedhttps://doi.org/10.1126/science.1164266Google Scholar
- Nordlund J, Backlin CL, Wahlberg P. Genome-wide signatures of differential DNA methylation in pediatric acute lymphoblastic leukemia. Genome Biol. 2013; 14(9):r105. PubMedhttps://doi.org/10.1186/gb-2013-14-9-r105Google Scholar
- Eckert C, Flohr T, Koehler R. Very early/early relapses of acute lymphoblastic leukemia show unexpected changes of clonal markers and high heterogeneity in response to initial and relapse treatment. Leukemia. 2011; 25(8):1305-1313. PubMedhttps://doi.org/10.1038/leu.2011.89Google Scholar
- Clappier E, Gerby B, Sigaux F. Clonal selection in xenografted human T cell acute lymphoblastic leukemia recapitulates gain of malignancy at relapse. J Exp Med. 2011; 208(4):653-661. PubMedhttps://doi.org/10.1084/jem.20110105Google Scholar
- Szczepanski T, van der Velden VH, Waanders E. Late recurrence of childhood T-cell acute lymphoblastic leukemia frequently represents a second leukemia rather than a relapse: first evidence for genetic predisposition. J Clin Oncol. 2011; 29(12):1643-1649. PubMedhttps://doi.org/10.1200/JCO.2010.30.2877Google Scholar
- Stark B, Nirel R, Avrahami G. Long-term results of the Israeli National Studies in childhood acute lymphoblastic leukemia: INS 84, 89 and 98. Leukemia. 2010; 24(2):419-424. PubMedhttps://doi.org/10.1038/leu.2009.254Google Scholar
- Fronkova E, Mejstrikova E, Avigad S. Minimal residual disease (MRD) analysis in the non-MRD-based ALL IC-BFM 2002 protocol for childhood ALL: is it possible to avoid MRD testing?. Leukemia. 2008; 22(5):989-997. PubMedhttps://doi.org/10.1038/leu.2008.22Google Scholar
- Breit S, Stanulla M, Flohr T. Activating NOTCH1 mutations predict favorable early treatment response and long-term outcome in childhood precursor T-cell lymphoblastic leukemia. Blood. 2006; 108(4):1151-1157. PubMedhttps://doi.org/10.1182/blood-2005-12-4956Google Scholar
- Bibikova M, Le J, Barnes B. Genome-wide DNA methylation profiling using Infinium(R) assay. Epigenomics. 2009; 1(1):177-200. PubMedhttps://doi.org/10.2217/epi.09.14Google Scholar
- Teschendorff AE, Marabita F, Lechner M. A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data. Bioinformatics. 2013; 29(2):189-196. PubMedhttps://doi.org/10.1093/bioinformatics/bts680Google Scholar
- Assenov Y, Muller F, Lutsik P, Walter J, Lengauer T, Bock C. Comprehensive analysis of DNA methylation data with RnBeads. Nat Methods. 2014; 11(11):1138-1140. PubMedhttps://doi.org/10.1038/nmeth.3115Google Scholar
- Fordham SE, Cole M, Irving JA, Allan JM. Cytarabine preferentially induces mutation at specific sequences in the genome which are identifiable in relapsed acute myeloid leukaemia. Leukemia. 2015; 29(2):491-494. PubMedhttps://doi.org/10.1038/leu.2014.284Google Scholar
- Fizzotti M, Cimino G, Pisegna S. Detection of homozygous deletions of the cyclin-dependent kinase 4 inhibitor (p16) gene in acute lymphoblastic leukemia and association with adverse prognostic features. Blood. 1995; 85(10):2685-2690. PubMedGoogle Scholar
- Gutierrez A, Sanda T, Ma W. Inactivation of LEF1 in T-cell acute lymphoblastic leukemia. Blood. 2010; 115(14):2845-2851. PubMedhttps://doi.org/10.1182/blood-2009-07-234377Google Scholar
- Lahortiga I, De Keersmaecker K, Van Vlierberghe P. Duplication of the MYB oncogene in T cell acute lymphoblastic leukemia. Nat Genet. 2007; 39(5):593-595. PubMedhttps://doi.org/10.1038/ng2025Google Scholar
- Clappier E, Cuccuini W, Kalota A. The C-MYB locus is involved in chromosomal translocation and genomic duplications in human T-cell acute leukemia (T-ALL), the translocation defining a new T-ALL subtype in very young children. Blood. 2007; 110(4):1251-1261. PubMedhttps://doi.org/10.1182/blood-2006-12-064683Google Scholar
- Palomero T, Sulis ML, Cortina M. Mutational loss of PTEN induces resistance to NOTCH1 inhibition in T-cell leukemia. Nat Med. 2007; 13(10):1203-1210. PubMedhttps://doi.org/10.1038/nm1636Google Scholar
- Zuurbier L, Petricoin EF, Vuerhard MJ. The significance of PTEN and AKT aberrations in pediatric T-cell acute lymphoblastic leukemia. Haematologica. 2012; 97(9):1405-1413. PubMedhttps://doi.org/10.3324/haematol.2011.059030Google Scholar
- Kleppe M, Lahortiga I, El Chaar T. Deletion of the protein tyrosine phosphatase gene PTPN2 in T-cell acute lymphoblastic leukemia. Nat Genet. 2010; 42(6):530-535. PubMedhttps://doi.org/10.1038/ng.587Google Scholar
- Meyer JA, Wang J, Hogan LE. Relapse-specific mutations in NT5C2 in childhood acute lymphoblastic leukemia. Nat Genet. 2013; 45(3):290-294. PubMedhttps://doi.org/10.1038/ng.2558Google Scholar
- Tzoneva G, Perez-Garcia A, Carpenter Z. Activating mutations in the NT5C2 nucleotidase gene drive chemotherapy resistance in relapsed ALL. Nat Med. 2013; 19(3):368-371. PubMedhttps://doi.org/10.1038/nm.3078Google Scholar
- Adzhubei IA, Schmidt S, Peshkin L. A method and server for predicting damaging missense mutations. Nat Methods. 2010; 7(4):248-249. PubMedhttps://doi.org/10.1038/nmeth0410-248Google Scholar
- Ding L, Ley TJ, Larson DE. Clonal evolution in relapsed acute myeloid leukaemia revealed by whole-genome sequencing. Nature. 2012; 481(7382):506-510. PubMedhttps://doi.org/10.1038/nature10738Google Scholar
- Gruber SB, Ellis NA, Scott KK. BLM heterozygosity and the risk of colorectal cancer. Science. 2002; 297(5589):2013. PubMedhttps://doi.org/10.1126/science.1074399Google Scholar
- Zauber NP, Sabbath-Solitare M, Marotta S. Clinical and genetic findings in an Ashkenazi Jewish population with colorectal neoplasms. Cancer. 2005; 104(4):719-729. PubMedhttps://doi.org/10.1002/cncr.21230Google Scholar
- Van Vlierberghe P, Palomero T, Khiabanian H. PHF6 mutations in T-cell acute lymphoblastic leukemia. Nat Genet. 2010; 42(4):338-342. PubMedhttps://doi.org/10.1038/ng.542Google Scholar
- Bea S, Valdes-Mas R, Navarro A. Landscape of somatic mutations and clonal evolution in mantle cell lymphoma. Proc Natl Acad Sci USA. 2013; 110(45):18250-18255. PubMedhttps://doi.org/10.1073/pnas.1314608110Google Scholar
- Jaffe JD, Wang Y, Chan HM. Global chromatin profiling reveals NSD2 mutations in pediatric acute lymphoblastic leukemia. Nat Genet. 2013; 45(11):1386-1391. PubMedhttps://doi.org/10.1038/ng.2777Google Scholar
- Oyer JA, Huang X, Zheng Y. Point mutation E1099K in MMSET/NSD2 enhances its methyltranferase activity and leads to altered global chromatin methylation in lymphoid malignancies. Leukemia. 2014; 28(1):198-201. PubMedhttps://doi.org/10.1038/leu.2013.204Google Scholar
- Gokbuget N, Basara N, Baurmann H. High single-drug activity of nelarabine in relapsed T-lymphoblastic leukemia/lymphoma offers curative option with subsequent stem cell transplantation. Blood. 2011; 118(13):3504-3511. PubMedhttps://doi.org/10.1182/blood-2011-01-329441Google Scholar
- Schrappe M, Reiter A, Ludwig WD. Improved outcome in childhood acute lymphoblastic leukemia despite reduced use of anthracyclines and cranial radiotherapy: results of trial ALL-BFM 90. German-Austrian-Swiss ALL-BFM Study Group. Blood. 2000; 95(11):3310-3322. PubMedGoogle Scholar
- Moricke A, Reiter A, Zimmermann M. Risk-adjusted therapy of acute lymphoblastic leukemia can decrease treatment burden and improve survival: treatment results of 2169 unselected pediatric and adolescent patients enrolled in the trial ALL-BFM 95. Blood. 2008; 111(9):4477-4489. PubMedhttps://doi.org/10.1182/blood-2007-09-112920Google Scholar
- O’Neil J, Grim J, Strack P. FBW7 mutations in leukemic cells mediate NOTCH pathway activation and resistance to gamma-secretase inhibitors. J Exp Med. 2007; 204(8):1813-1824. PubMedhttps://doi.org/10.1084/jem.20070876Google Scholar
- Thompson BJ, Buonamici S, Sulis ML. The SCFFBW7 ubiquitin ligase complex as a tumor suppressor in T cell leukemia. J Exp Med. 2007; 204(8):1825-1835. PubMedhttps://doi.org/10.1084/jem.20070872Google Scholar
- Weng AP, Ferrando AA, Lee W. Activating mutations of NOTCH1 in human T cell acute lymphoblastic leukemia. Science. 2004; 306(5694):269-271. PubMedhttps://doi.org/10.1126/science.1102160Google Scholar
- Tosello V, Mansour MR, Barnes K. WT1 mutations in T-ALL. Blood. 2009; 114(5):1038-1045. PubMedhttps://doi.org/10.1182/blood-2008-12-192039Google Scholar
- Ntziachristos P, Tsirigos A, Van Vlierberghe P. Genetic inactivation of the polycomb repressive complex 2 in T cell acute lymphoblastic leukemia. Nat Med. 2012; 18(2):298-301. PubMedhttps://doi.org/10.1038/nm.2651Google Scholar
- Huether R, Dong L, Chen X. The landscape of somatic mutations in epigenetic regulators across 1,000 paediatric cancer genomes. Nat Commun. 2014; 5:3630. PubMedhttps://doi.org/10.1038/ncomms4630Google Scholar
- Goss KH, Risinger MA, Kordich JJ. Enhanced tumor formation in mice heterozygous for Blm mutation. Science. 2002; 297(5589):2051-2053. PubMedhttps://doi.org/10.1126/science.1074340Google Scholar
- Holmfeldt L, Wei L, Diaz-Flores E. The genomic landscape of hypodiploid acute lymphoblastic leukemia. Nat Genet. 2013; 45(3):242-252. PubMedhttps://doi.org/10.1038/ng.2532Google Scholar
- Wu MY, Eldin KW, Beaudet AL. Identification of chromatin remodeling genes Arid4a and Arid4b as leukemia suppressor genes. J Natl Cancer Inst. 2008; 100(17):1247-1259. PubMedhttps://doi.org/10.1093/jnci/djn253Google Scholar
- Bains T, Heinrich MC, Loriaux MM. Newly described activating JAK3 mutations in T-cell acute lymphoblastic leukemia. Leukemia. 2012; 26(9):2144-2146. PubMedhttps://doi.org/10.1038/leu.2012.74Google Scholar