A somatic UBA2 variant preceded ETV6-RUNX1 in the concordant BCP-ALL of monozygotic twins

Childhood acute lymphoblastic leukemia (ALL) is a genetically heterogenous disease,^1-3  largely affecting B-lymphoid cells (85%).^4,5  A variety of recurrent genetic aberrations, predominantly chromosomal translocations or nonrandom loss or gain of entire chromosomes, are currently considered the initiating events and drivers of the disease.⁶  These aberrations are also the basis for subtype classification, the 2 most common being high hyperdiploidy and t(12;21)(p13;q22)/ETV6-RUNX1.⁶  The etiology of childhood ALL remains unknown in the vast majority of cases. Nevertheless, compelling evidence of constitutional predisposition to ALL has emerged in recent years.⁷ 

Monozygotic (mz) twins with concordant B-cell precursor ALL (BCP-ALL) have played a central role in exploring the timing of disease initiation. Molecular studies of initiating translocations have shown identical breakpoints and chimeric fusion sequences in leukemias of twins, providing solid evidence of a common clonal origin.^8-12  In addition, fusion sequences from ETV6-RUNX1–carrying preleukemic clones have been identified in archived neonatal dried blood spots and cord blood, also in nontwin cases.^10-22  Preleukemic clones arise in 1 twin in utero and spread to the sibling by vascular transfusion through anastomoses in the shared placenta.^10,11,23  In due time, secondary genetic alterations accumulate in genes crucial for B-cell development and eventually render leukemic transformation.^19,24,25 

Further, studies of latency (time from initiation to diagnosis), mutation rates, and concordance rates of BCP-ALL in mz twins have informed us on the nature of different initiating and disease-driving genetic events. In cases where latency is short (<1 year), differs little within a twin pair (months), and concordance rates are high (close to 100%), the genetic event is suggested to have stronger oncogenicity.²³  Such events are either alone sufficient to cause leukemia or efficient in causing the additional driving events required for overt leukemia.^26,27  In contrast, initiating events such as t(12;21)(p13;q22)/ETV6-RUNX1 associated with longer latency (median 4 years, range 1-12 years^28-31 ), larger differences in latency (up to 9 years¹¹ ) within a twin pair, and lower concordance rates (10% to 15%²³ ) are considered weaker oncogenic drivers. The need for additional disease-driving genetic events is also greater as illustrated by higher mutation rates and recurrent secondary genetic events.^32,33  Also, more recent studies have used concordant ALL to closer explore the molecular phylogenetics of clonal evolution.^34,35 

Despite the above efforts, our knowledge and understanding of molecular phylogenetics and causes of mutational processes during leukemogenesis remain limited. Molecular phylogenetics of leukemia is currently restricted to distinguishing shared (early in utero) from unique (late in utero or postnatal) somatic variants. In this study, we performed a comprehensive genetic analysis of a pair of mz twins with concordant ETV6-RUNX1⁺ BCP-ALL. We evaluate constitutional variants for predisposing effects, provide molecular proof of a common clonal origin of leukemia, characterize the nature of shared (early) variants, and explore the genetic divergence of concordant BCP-ALL in each twin. Unexpectedly, we were also able to elucidate the temporal order of 2 significant shared genetic events.

The ethics review board at the Karolinska Institutet approved this study (ethics number 2015-293-31/4 and 04-638/4), and informed consent from the parents were obtained according to the Declaration of Helsinki.

Genomic DNA was extracted from samples collected from both twins at 4 different timepoints: (1) neonatal dried blood spots (and their adjacent control) from the Swedish National Phenylketonuria Register at the Karolinska University Hospital, (2) bone marrow at leukemia diagnosis, (3) peripheral blood at clinical remission, and (4) saliva samples 5 years after clinical remission (Figure 1A). We retrieved 5 punches from each neonatal dried blood spot card, each 3 mm in diameter and from different locations of the spots. We also collected saliva samples from the parents.

Genomic DNA extraction from dried blood spots was performed using QIAamp DNA Micro Kit (Qiagen, Hilden, Germany), following the manufacturer’s recommendations. Genomic DNA from bone marrow at diagnosis and blood at clinical remission had been performed prior to this study at the Department of Clinical Genetics at Karolinska University Hospital, following standard procedures. Saliva samples were collected using Oragene DNA saliva collection kit (DNA Genotek, Ontario, Canada). Genomic DNA extraction from saliva samples was performed using prepIT-L2P kit (DNA Genotek, Ontario, Canada), following the manufacturer’s recommendations. The concentrations were determined by Qubit dsDNA HS Assay Kit in Qubit 3.0 fluorometer (Life Technologies, Carlsbad, CA, USA).

Karyotyping with G-banding (Giemsa staining), interphase fluorescence in situ hybridization (FISH), and array comparative genomic hybridization had been performed prior to this study, following standard protocols at the Department of Clinical Genetics at Karolinska University Hospital. We obtained clinical and laboratory data for each twin, summarized in Table 1, from medical records.

Whole-genome sequencing (WGS) was performed on genomic DNA from diagnostic bone marrow and matched peripheral blood from remission. Libraries for sequencing on Illumina HiSeq X (Illumina Inc, San Diego, CA, USA) were prepared from genomic DNA using Illumina TruSeq polymerase chain reaction (PCR)-free kit with a mean insert size of >350 base pairs, resulting in over 700 million (range 729-915M) mapped unique sequences per sample with mean read depth 37× (range 34-39×). Alignment of reads to human reference genome (GRCh37/hg19) and variant calling was performed by Science for Life Laboratory (SciLifeLab, Stockholm, Sweden) (supplemental Table 1).

Somatic single nucleotide variants (SNVs)/indels were identified using MuTect2.³⁶  Constitutional variants were identified and processed using best practices of the Genome Analysis Toolkit.³⁷  Variants were functionally annotated using Variant Effect Predictor (version 89)³⁸  and loaded into a database using GEMINI (GEnome MINIng) (v0.20.0).³⁹  Variants were explored in GEMINI using built-in tools and visualized in Integrative Genomics Viewer.⁴⁰  Leukemia predisposition–centered constitutional variant analysis was performed applying the gene panel “Hematological malignancies cancer susceptibility,” curated by experts in Genomics England PanelApp⁴¹  (supplemental Table 2), to whole-genome data from remission.

Structural variants were detected using FindSV pipeline (https://github.com/J35P312/FindSV) merging calls from CNVnator v0.3.2 and TIDDIT.^42,43  Immunoglobulin heavy chain (IgH) and T-cell receptor rearrangements were excluded from this analysis. Immunoglobulin rearrangements were instead identified using IgCaller.⁴⁴  Somatic structural variants were analyzed as previously described.^45,46  Complex translocations were plotted using circos.⁴⁷  To detect presence of any copy number neutral structural aberrations, loss-of-heterozygosity (LOH) analysis was performed. Regions of homozygosity were called using GEMINI built-in regions-of-homozygosity function with window sizes ranging between 100 Kb and 300 Kb in both tumor-normal pairs.

Sanger sequencing was performed using ABI3730xl DNA Analyzer (Applied Biosystems, Foster City, CA, USA) and BigDye Terminator v3.1 (Applied Biosystems, Foster City, CA, USA) to validate translocation breakpoints and for segregation analysis. Primer sequences are available upon request.

Ten nanograms DNA extracted from dried blood spots generated up to 12 to 20 micrograms of DNA using illustra Ready-To-Go GenomiPhi V3 DNA Amplification Kit (GE Healthcare, Buckinghamshire, UK) according to the manufacturer’s protocol.

Genomic DNA (15-50 ng) was amplified with 1X QuantStudio 3D Digital PCR Master mix, TaqMan assay for reference gene, and hydrolysis probes for target region according to manufacturer’s instructions. Data were analyzed using QuantStudio 3D Analysis Suite Cloud Software, version 3.1.6-PRC-build2 with default parameters (confidence level of 95%, desired precision of 10%, and Poisson plus quantification algorithm).

We have previously demonstrated the versatility of digital PCR to detect and quantify somatic structural variants and its applicability to DNA extracted from dried blood spots.⁴⁸  TaqMan probe for the shared complex rearrangement was designed manually, targeting the fusion sequence of GRM5-ETV6 (chromosome 11q to 12p). Probes for NSD2 p.E1099K and UBA2 deletion were designed by ThermoFisher Scientific’s internal bioinformatics platform. TaqMan Copy Number Reference Assay (Applied Biosystems), human, ribonuclease P RNA component H1 (RPPH1; chromosome 14(GRCh37): 20811565), labeled with VIC was used as internal control.

Genomic DNA (67 ng) from saliva was amplified in triplicates using the droplet digital PCR (ddPCR) Supermix for Probes (No dUTP) kit (BioRad, Hercules, CA, USA). The TaqMan assay for UBA2 used for chip-based dPCR was also used in this experiment. Droplets were generated on the Automated Droplet Generator (BioRad, Hercules, CA, USA), and PCR was performed according to the manufacturer’s instructions. After completion of PCR, the droplets were read QX200 on the Droplet Reader (BioRad, Hercules, CA, USA). The QuantaSoft Analysis Pro v.1.0 was used to analyze the data. The threshold for true signal positivity was adjusted based on the signal in the control samples. The ratio of the number of positive droplets to the total number of droplets was calculated for each sample using Poisson 95% confidence intervals.

The twins studied here were monozygotic, monochorionic, and diamniotic females born by acute cesarean section at gestational week 36 plus 3. Both twins developed BCP-ALL at ages 3 years 4 months (Tw1) and 3 years 10 months (Tw2), a 6-month difference in latency. All clinical data have been summarized in Table 1. Immunophenotypes displayed the classical composition of BCP-ALL and were by and large identical, differing only in expression of CD99 and CD34. Cytogenetic analysis at diagnosis identified a t(11;12)(q21;p13) as well as the classical t(12;21)(p13;q22) rearrangement in the main clone of both leukemias. In addition, Tw1 carried a subclonal deletion of ETV6 in 6% of cells, whereas Tw2 carried a subclonal deletion of ETV6 and a t(1;12)(p13;p13),+der(1)t(1;12) in 44% of cells.

Both twins were treated according to the NOPHO-ALL 2008 protocol, standard risk arm, and remain in full remission 5.5 (Tw1) and 5 (Tw2) years after diagnosis. Both twins have a normal psychomotor development, growth parameters within the normal range, and no malformations, dysmorphic features, or clinical signs of neurofibromatosis.

The increasing awareness of the contribution of constitutional predisposition to childhood leukemia urged our inclusion of a constitutional variant analysis. Both twins were found heterozygous for a constitutional missense variant in tumor suppressor gene NF1 (chr17:g.29557883C>A, NM_000267.3:c.3137C>A) NP_001035957.1:p.(Thr1046Lys). The detected variant was not present in gnomAD,⁴⁹  COSMIC,⁵⁰  or LOVD⁵¹  and had a CADD⁵²  (v1.4) c-score of 27.1, indicative of a novel, seemingly damaging, variant, further supported by a majority of in silico predictions (supplemental Table 3). Pathogenic variants in NF1 are known to cause neurofibromatosis type 1 (NF1).⁵³  However, the variant detected here is not previously described in NF1, neither did the twins display any clinical signs obligate for NF1 diagnosis. Segregation analysis showed that this variant is inherited from the healthy mother, who is also heterozygous. Hence, we classified this aberration as a variant of unknown significance. The potential impact of this variant is discussed below. No other pathogenic constitutional variants were detected.

From WGS data, we characterized a shared complex rearrangement t(11;12;21)(q23;p13;q22) (Figure 1B; Table 2) with identical fusion sequences in both twins’ leukemias (supplemental Figure 1), providing proof of a common clonal origin. The rearrangement gave rise to the classical ETV6-RUNX1 fusion gene recurrent in BCP-ALL, which was detected by FISH analysis at diagnosis, although its origin in a complex rearrangement was not. It also created an additional fusion gene, GRM5-ETV6, by the in-frame fusion of GRM5 exon 2 (chr 11) to ETV6 exon 6 (chr 12), previously unreported in Mitelman⁵⁴  and St. Jude Cloud (https://www.stjude.cloud)⁵⁵  (accessed December 2020). Sequence coding for erythroblast transformation specific-domain in ETV6 but only part of atrial natriuretic factor-receptor domain in GRM5 was retained in the fusion gene. Nevertheless, under the promoter of GRM5, expression of GRM5-ETV6 fusion in lymphocytes is likely low to nonexisting (https://www.gtexportal.org/home/) (accessed December 2020). No other shared structural variants were found.

With the aim to quantify the complex rearrangement (ie, a preleukemic clone) at birth, the unique fusion sequence of GRM5-ETV6 was used for chip-based dPCR (Figure 1C). Results stated in percent (%) refers to the fraction of mutant target DNA in the analyzed sample. As expected, the complex rearrangement was readily detected at diagnosis in Tw1 (26%) and Tw2 (29%) but not at remission. However, the rearrangement was beyond detection in both twins also at birth, indicating its copy number was beyond our detection limit (1 in 1000 copies).

In addition, 58 somatic SNVs/indels were found shared by both leukemias (Figure 2A; supplemental Table 4). Our analysis highlighted a somatic frameshift deletion in UBA2 (NM_005499.2: c.463_470del; NP_005490.1: p.(Thr156Leufs*2)), which has recently been implicated in childhood BCP-ALL as a novel driver gene (Figure 2B).⁵⁶  Variant allele frequencies from WGS at diagnosis was 40% (Tw1) and 44% (Tw2). The deletion, affecting the UBA2 exon 6 of 17, caused a frameshift, introducing a proximate premature stop.

To quantify the presence of UBA2 deletion at diagnosis and, simultaneously, repeat our effort to quantify the presence of preleukemic clones at birth, we again applied chip-based dPCR analysis (Figure 2C). As expected, the UBA2 deletion was readily detected at diagnosis in Tw1 (36%) and at birth in Tw1 (3%) and Tw2 (6%). Unfortunately, lack of further sample prevented analysis of Tw2 at diagnosis. Unexpectedly, the UBA2 deletion was detected at remission (sampling 2 years 11 months [Tw1] and 2 years 5 months [Tw2] after diagnosis) in both twins (2.6%). When reviewed manually, the variant was found in 1 out of 45 WGS reads (2.2%) at remission of Tw1. Five years after the clinical remission, the UBA2 deletion in DNA extracted from the twins’ saliva was detected 1.14% in Tw1 and 0.78% in Tw2 (data not shown). In children, cellular content of saliva consists of ∼13.5% lymphocytes and 70% epithelial cells.⁵⁷  Therefore, our finding is most likely due to the presence of lymphocytes in the saliva rather than presence of mosaicism in epithelial cells.

We identified 10 deletions unique to Tw1’s leukemia and 10 deletions, 2 translocations, and 1 complex rearrangement unique to Tw2’s leukemia from WGS data (Table 2). Six deletions were analogous to one another (ie, affecting the same chromosomal regions but with different breakpoints). Two analogous deletions encompassed the ETV6 and ATF7IP locus, respectively, whereas RAG1/RAG2 locus was affected by 2 deletions in each twin (supplemental Figures 2, 3, and 4, respectively). The ETV6 deletions were also seen at diagnosis by FISH analysis and detected here by targeted manual inspection of WGS data.

In Tw1, the 4 remaining deletions, sized 22 to 40 kb, affected chromosomes 1, 9, 19, and X (Table 2). The 34 kb deletion of CDKN2A/CDKN2B (supplemental Figure 5) on chromosome 9 was detected through manual investigation of recurrent second-hit regions in BCP-ALL. In Tw2, the 4 remaining deletions, sized 30 to 160 kb, affected chromosomes 1, 7, 14, and 22 (Table 2). The 2 translocations fused proximate regions of chromosome 1p and 12p, whereas the complex rearrangement involved chromosomes 11, 12, and 17 (Table 2; Figure 1B, red color).

In addition, 2421 (Tw1) and 2955 (Tw2) unique somatic SNVs/indels were identified, out of which 915 (Tw1) and 1040 (Tw2) mapping to protein coding genes (supplemental Table 5). An in-frame insertion in ETV6 (NM_001987.4: c.309_310insCGGCCTAGC, NP_001978.1:p.(R103_Y104insRPS)) was highlighted in Tw1. Further, a somatic SNV in NSD2 (WHSC1) p.E1099K (COSMIC ID: COSM379334)⁵⁸  (supplemental Table 5) was found in leukemias of both twins. This gene was reviewed manually due to its known recurrence in t(12;21)/ETV6-RUNX1⁺BCP-ALL.⁵⁹  Applying chip-based dPCR, the NSD2 variant was readily detected at diagnosis in Tw1 (29%) and Tw2 (15%) but not present at remission. At birth, it was undetectable in Tw1 but quantified to 0.1% in Tw2 (data not shown).

LOH analysis was also performed on WGS data for both shared and unique variants and it did not reveal any novel deletions, duplications, nor copy number–neutral LOH events.

IgCaller identified 1 IgH with shared D-J segment (IGHJ4 - IGHD3-22) but different V segments (IGHV3-69 and IGHV1-73) in the leukemias. All other IgH and immunoglobulin light chain rearrangements identified differed between the twins (supplemental Table 6).

The ETV6-RUNX1 fusion has long been considered an initiating event and disease driver in BCP-ALL.⁶  The basis for this view has been its recurrence, its identification in cord blood of healthy neonates, the detection of identical ETV6-RUNX1 fusion sequences in mz twins with concordant BCP-ALL, and the absence of any other damaging shared events detectable by WGS in concordant cases.^8-22,34,60  In twins with concordant BCP-ALL, a preleukemic clone is believed to arise in 1 twin and spread to the sibling by vascular transfusion in the shared placenta.^10-22  The detection of 58 shared SNVs/indels and, especially, a complex rearrangement t(11;12;21)(q23;p13;q22) with identical breakpoints in leukemias of both twins provided solid molecular evidence of a common in utero origin in this case also (Figure 3). Moreover, detection of a single identical IgH D-J rearrangement and otherwise only divergent IgH V(D)J and immunoglobulin light chain rearrangements in the twins’ leukemias (supplemental Table 6) suggested the shared preleukemic clone was initiated in between pro- and pre-B cell stage.

Although detected at diagnosis by FISH analysis, the origin of the ETV6-RUNX1 fusion in the shared complex rearrangement was revealed in our study by structural variant analysis of WGS data. This adds yet another example of complex rearrangements underlying the classical ETV6-RUNX1 fusion^61-65  and further emphasizes the advantage of WGS-based applications to accurately detect and in detail characterize relevant genomic aberrations at leukemia diagnosis.^61,63,65 

Chip-based dPCR detected the deletion at 2.6% in the remission bone marrow samples of both twins, suggesting the clone involving the UBA2 deletion remained in their respective bone marrows after ALL treatment and more than 2 years in complete remission (Figure 3). The variant allele frequency of the UBA2 deletion was increased from 3% in Tw1 and 6% in Tw2% to 40% in Tw1 and 44% in Tw2 at diagnosis. The expansion of the clone during leukemia strongly supports that the leukemic clone arose in 1 of the cells of this clone and expanded over time. The complex rearrangement, harboring ETV6-RUNX1, was undetectable at birth using dPCR. We conclude that the level of this complex rearrangement at birth is below 0.1%, which is the limit of detection for the employed method. As the level of the UBA2 deletion is at least 30 times more abundant than that of the complex rearrangement at birth, we infer that the UBA2 variant precedes the complex chromosomal rearrangement. Thus, we suggest the shared complex rearrangement emerged in a delUBA2⁺ cell in 1 twin and then established in the bone marrows of both twins’ following a second event of vascular transfusion. Considering the ETV6-RUNX1 fusion’s well-established role as a leukemia initiating event, discovering that the UBA2 deletion preceded the shared complex rearrangement compelled us to reevaluate the assumption of ETV6-RUNX1 as the initiating event in our cases.

UBA2 is predicted to be haploinsufficient and thus likely causative of a dominant disease when mutated. It is essential to and highly conserved across many organisms (https://varsome.com/gene/uba2) (accessed January 2021).⁶⁶  The UBA2 frameshift deletion reported here introduced an early stop codon which most likely led to protein truncation and nonsense-mediated decay of any transcribed product, resulting in heterozygous loss of UBA2. Constitutional loss-of-function variants and deletions affecting UBA2 have recently been associated with congenital malformations.^67-74  No associated cancer predisposition has been reported; however, cases are few, and reduced penetrance of cancer phenotype is likely in dominant disease. Altogether, we assessed our reported deletion as likely pathogenic.

Interestingly, UBA2 was recently uncovered as a recurrent target gene for somatic aberrations in BCP-ALL.^56,75  In line with our findings, UBA2 variants have also been reported enriched in ETV6-RUNX1⁺ cases.⁷⁶  However, the temporal relation between UBA2 variants and the ETV6-RUNX1 fusion has not been determined previously. To our knowledge, this is the first report of a UBA2 variant preceding the ETV6-RUNX1 fusion or any other known recurrent aberration associated to BCP-ALL.

Moreover, as reviewed by Han et al, the UBA2 protein is an essential component in posttranslational protein modification by small ubiquitin-like modifier (SUMO) attachment (SUMOylation).⁷⁷  As part of the SUMO E1-activating enzyme heterodimer, UBA2 activates the highly conserved SUMO proteins to exercise their effects on target proteins. SUMOylation regulates crucial cellular processes such as gene expression, cell signaling, DNA damage repair, cell cycle progression, apoptosis, etc, and is known to be exploited by viruses as a result interfering with diverse cellular mechanisms.⁷⁸  Consequently, it affects the function and activity of most intracellular pathways and thus nearly all physiological and pathological processes.⁷⁷  Importantly, a disrupted balance of SUMOylation has been implicated in cancer, including leukemia.^77,79  Given the somatic heterozygous loss of UBA2 predicted in our cases, downregulation of SUMO E1-activating enzyme and thus of SUMOylation appeared the most likely functional outcome. Conversely, SUMOylation has been found upregulated in malignant cells of many cancer types,⁷⁷  speaking against a cancer-predisposing effect of our detected variant.

Nevertheless, SUMOylation is part of an intricate regulatory network, also cross-talking with other regulatory systems such as acetylation, phosphorylation, and ubiquitination.⁸⁰  Also, the contribution to carcinogenesis by dysregulated SUMOylation takes place in a context-dependent manner.⁸⁰  For example, functional effects of dysregulated SUMOylation in a cell before and after malignant transformation may differ. In that respect, cancer predisposing and/or promoting effects of dysregulated SUMOylation may encompass different causal relationships. Therefore, we argue that a possible association between downregulated SUMOylation and leukemia development should not yet be dismissed.

To us, the biological example posed by the twins provides the most compelling support for a possible leukemia-predisposing effect of UBA2 deletion. The likelihood of the ETV6-RUNX1 fusion arising in a delUBA2 clone by pure chance appears inarguably small, even more so if the predicted downregulated SUMOylation would exert a protective effect against leukemogenesis. Moreover, other frameshift variants in UBA2, also likely causing downregulation of SUMOylation, can be found among somatic variants in B-cell ALL.⁵⁵ 

SUMOylation has been implicated in both DNA damage response and DNA-repair pathways.⁸¹  For example, SUMOylation is essential for the function of XRCC4, Ku70, and Ku80,⁸²  all key regulatory factors of nonhomologous end joining (NHEJ), the major repair mechanism for double-stranded breaks in DNA.^83,84  NHEJ is key to V(D)J recombination, the tight regulation of which is essential for normal T- and B-cell development and prevention of oncogenic genetic events.^85,86  V(D)J recombination is mainly driven by RAG endonucleases, the off-target activity of which has been found the main driver of secondary somatic events in ETV6-RUNX1⁺ BCP-ALL.⁸⁷  However, the ETV6-RUNX1 fusions themselves bear no signs of off-target RAG activity.^60,87  Rather, illegitimate recombination of multiple simultaneous double-stranded breaks by NHEJ is considered the mechanism behind ETV6-RUNX1 fusion.⁸⁸  The above provides 1 theoretical mechanistic explanation to the effects of downregulated SUMOylation on the propensity of forming initiating and secondary genetic events leading to leukemogenesis.

In line with the established notion that additional somatic events are required for progression to overt BCP-ALL of ETV6-RUNX1⁺ preleukemic clones,^32,33  WGS identified 10 deletions and 2421 SNVs/indels unique to Tw1’s leukemia and 10 deletions, 2 translocations, 1 complex rearrangement, and 2955 SNVs/indels unique to Tw2’s leukemia. This was well within range of the reported average of 3.5 (range 0-14) copy number variations in pediatric ALL, supporting gross genomic instability is not a common trait. Latency of 3.5 and 4 years, respectively, in the twins corresponded well to overall latency among ETV6-RUNX1⁺ BCP-ALL.^25,89-91 

Six of the unique deletions in either leukemia were analogous, encompassing the same locus but with different breakpoints, targeting ETV6, ATF7IP, and RAG1/RAG2. These genes are known recurrent targets of secondary events in ETV6-RUNX1⁺ BCP-ALL^92,93  and have been shown to predominantly result from aberrant RAG-mediated recombination.^59,87  Also, RAG1/RAG2 and ATF7IP deletions have been reported to frequently cooccur in ETV6-RUNX1⁺ BCP-ALL.^56,87  Nonetheless, the independent loss of both ETV6, ATF7IP, and RAG1/RAG2 in both leukemias provides evidence of a convergent clonal evolution in the leukemias, which may only be explained by a strong selective pressure, most likely exerted by the genetic and consequential biological circumstances in the shared preleukemic clone.

Another known recurrent secondary event found in both twins was the SNV in NSD2 (WHSC1) p.E1099K (COSMIC ID: COSM379334).⁵⁸  Being the most common NSD2 variant in ETV6-RUNX1⁺ BCP-ALL,⁵⁸  these variants were most likely independently acquired. This NSD2 variant also frequently cooccurs with variants in UBA2.⁵⁶ 

NF1 is an autosomal-dominant neurocutaneous disorder characterized by highly variable clinical features including multiple café-au-lait spots, neurofibromas, and ocular and skeletal abnormalities.⁵³ NF1 being a tumor suppressor gene,⁹⁴  the syndrome also entails cancer predisposition to a range of tumors and malignancies, including leukemias.^95,96  In the studied twins, detection of a novel constitutional missense variant in NF1, predicted highly damaging but evaluated as a variant of unknown significance due to lack of phenotypic correlation, was intriguing. NF1 was dismissed in our cases due to complete lack of clinical symptoms obligate for this diagnosis.⁵³ NF1 LOH in malignant cells of NF1 patients who developed cancer is common, in accordance with Knudson’s 2-hit hypothesis, but not obligatory.⁵⁶  Also, the gene is considered to be haploinsufficient,⁹⁷  reflecting the cancer-predisposing effects of heterozygous loss of NF1. Cases with constitutional NF1 variants and cancer but lacking other clinical signs of NF1 have also been reported.⁹⁸  Altogether, we suggest that the absence of obligate NF1 symptoms and lack of LOH in leukemic cells does not exclude the possibility of a cancer-predisposing effect of our reported variant. Interestingly, a previous study has shown significant pairwise cooccurrence of somatic NF1 and UBA2 in childhood BCP-ALL.⁵⁶  Could this be an indication that emergence of UBA2 deletion in the presence of the constitutional NF1 variant was not a coincidence?

In summary, we report a detailed genetic characterization with partial temporal delineation of some central genetic aberrations in the concordant BCP-ALL of a mz twin pair. A common in utero origin of their preleukemic clone was supported by a shared complex rearrangement with identical breakpoint sequences in both twins and 58 shared SNV’s/indels, including a frameshift deletion in UBA2. The complex rearrangement generated the well-known recurrent ETV6-RUNX1 fusion gene, whereas the UBA2 deletion was predicted to cause heterozygous loss of function. Typical to ETV6-RUNX1⁺ BCP-ALL, additional copy number variants were required for progression to overt leukemia. Interestingly, these second hits targeted a number of recurrent second-hit genes. Surprisingly, UBA2 deletion was retained in remission of both twins, providing indirect proof of its emergence preceding the shared complex rearrangement. UBA2 variants are highly recurrent in ETV6-RUNX1⁺ BCP-ALL. To our knowledge, this is the first report of a UBA2 variant preceding the ETV6-RUNX1 fusion or any other known recurrent aberration associated to BCP-ALL.

The authors wish to thank the twins and their parents for their participation in this study. They are also very grateful to Lene Sörensen at the Swedish National Phenylketonuria Register at Center for Inherited Metabolic Diseases at Karolinska University Hospital for providing samples. The authors would like to acknowledge support from Science for Life Laboratory, the National Genomics Infrastructure (NGI), Sweden, the Knut and Alice Wallenberg Foundation, and Swedish National Infrastructure for Computing (SNIC) through Uppsala Multidisciplinary Center for Advanced Computational Science (UPPMAX) under project sens2017130 for providing assistance in massively parallel DNA sequencing and computational infrastructure.

This study was supported by grants from the Swedish Childhood Cancer Fund, the Swedish Cancer Society, the Cancer Society of Stockholm, the Swedish Research Council, Mary Béves Stiftelse för Barncancerforskning, the Swedish Rare Diseases Research foundation (Sällsyntafonden), Berth von Kantzow’s foundation, Karolinska Institutet Research Foundation, the Swedish Medical Research Council, and the Stockholm County Council.

Contribution: F.T., V.Z., B.B., and A.N. designed the study and interpreted the data; F.T. and J.E. contributed to data acquisition, and F.T. prepared the figures; F.T., B.B., and A.N. wrote the manuscript; M.H., G.B., A.H.-S., and A.N. collected patient samples; F.T., B.B., M.H., G.B., V.Z., and A.N. contributed to data interpretation; and all authors revised the manuscript and approved the final version.

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

Correspondence: Fulya Taylan, Department of Molecular Medicine and Surgery, Karolinska Institutet, BioClinicum J10:20, 171 64, Stockholm, Sweden; e-mail: fulya.taylan@ki.se.