Human MLL/KMT2A gene exhibits a second breakpoint cluster region for recurrent MLL–USP2 fusions

For nearly 3 decades, the human MLL (KMT2A) gene and its rearrangements have been investigated in many different laboratories around the world. At our diagnostic center (DCAL Frankfurt), our standard strategy for the identification of MLL-r is based on two independent approaches, namely “Multiplex” (MP)-polymerase chain reaction (PCR) and “Long distance inverse” (LDI)-PCR approach [1]. The MP-PCR approach is used to rapidly identify the eight most frequent MLL fusions (AF4, AF6, AF9, AF10, ENL, ELL, EPS15, and PTDs) which encompass ~90% of all diagnosed MLL-r leukemia patients, while LDI-PCR is used for all other patients (~10%). By applying both technologies, we have accumulated 94 direct MLL-gene fusions and 247 reciprocal fusion partner genes [2]. Nearly, all breakpoints have been identified in the major breakpoint cluster region (BCR) of the MLL gene (MLL exons 8–14). However, some of the patients remained negative, although they were positively prescreened by various methods. In order to diagnose MLL breakpoints in every patient, a total of 2688 overlapping Illumina capture probes covering the whole-MLL gene were designed and used to analyze a cohort of AL patients (n= 109) where we had either limited (n= 4; PCR positive but not sequenced) or no information (n= 105) on their molecular status. As depicted in Fig. 1a, we identified chromosomal rearrangements in 93 out of 109 patient cases. Sixteen patients remained MLL-r negative and were therefore assigned as patients with “unknown status”. The data analyses of the remaining 93 patients revealed the following distribution: for 67 patients (72%) a breakpoint could be analyzed in the major BCR; 5 patients (5%) displayed only the reciprocal der(TP) with breakpoints in exon 11 (putative CEP83-MLL spliced fusion), intron 11 (n= 3; putative FKBP8-MLL spliced fusion, AF9-MLL, RELAMLL) and intron 27 (IFT46-MLL), respectively. Surprisingly, an additional 21 patients (23%) had their breakpoints outside of the major BCR, but inside a novel, minor BCR. This novel BCR is localizing between MLL intron 19 and exon 24 (with a clear preference for MLL intron 21–23). Most of the new BCR cases represented MLL–USP2 gene fusions (n= 17). USP2 is localized about 1 Mbp telomer to MLL at 11q23.3 and transcriptionally orientated in direction of the centromere of chromosome 11, classifying all these fusions as intrachromosomal inversions (see Fig. 1b). In addition, we identified four balanced translocations in the minor BCR: one patient with an USP8 fusion (see Fig. 2 and Suppl. Figure S1), two with AF4 and one with AF9. MLL–USP2 and MLL–USP8 alleles seem to be restricted to the minor BCR (see Fig. 2), because they were never diagnosed in association with the major BCR. Most of the reciprocal USP2–MLL fusions were scattered over a larger region at 11q23.3 (see Fig. 2), involving also upstream (C2CD2L) and downstream genes (USP2-AS1). Our analysis revealed also five patients with 3′-MLL deletions that were caused by microdeletions (<200 bp), larger deletions (up to 34 kbp), or complex rearrangements including other These authors contributed equally: Claus Meyer, Bruno A. Lopes


To the Editor:
For nearly 3 decades, the human MLL (KMT2A) gene and its rearrangements have been investigated in many different laboratories around the world. At our diagnostic center (DCAL Frankfurt), our standard strategy for the identification of MLL-r is based on two independent approaches, namely "Multiplex" (MP)-polymerase chain reaction (PCR) and "Long distance inverse" (LDI)-PCR approach [1]. The MP-PCR approach is used to rapidly identify the eight most frequent MLL fusions (AF4, AF6, AF9, AF10, ENL, ELL, EPS15, and PTDs) which encompass~90% of all diagnosed MLL-r leukemia patients, while LDI-PCR is used for all other patients (~10%). By applying both technologies, we have accumulated 94 direct MLL-gene fusions and 247 reciprocal fusion partner genes [2]. Nearly, all breakpoints have been identified in the major breakpoint cluster region (BCR) of the MLL gene (MLL exons [8][9][10][11][12][13][14]. However, some of the patients remained negative, although they were positively prescreened by various methods. In order to diagnose MLL breakpoints in every patient, a total of 2688 overlapping Illumina capture probes covering the whole-MLL gene were designed and used to analyze a cohort of AL patients (n = 109) where we had either limited (n = 4; PCR positive but not sequenced) or no information (n = 105) on their molecular status. As depicted in Fig. 1a, we identified chromosomal rearrangements in 93 out of 109 patient cases. Sixteen patients remained MLL-r negative and were therefore assigned as patients with "unknown status". The data analyses of the remaining 93 patients revealed the following distribution: for 67 patients (72%) a breakpoint could be analyzed in the major BCR; 5 patients (5%) displayed only the reciprocal der(TP) with breakpoints in exon 11 (putative CEP83-MLL spliced fusion), intron 11 (n = 3; putative FKBP8-MLL spliced fusion, AF9-MLL, RELA-MLL) and intron 27 (IFT46-MLL), respectively. Surprisingly, an additional 21 patients (23%) had their breakpoints outside of the major BCR, but inside a novel, minor BCR. This novel BCR is localizing between MLL intron 19 and exon 24 (with a clear preference for MLL intron 21-23).
Most of the new BCR cases represented MLL-USP2 gene fusions (n = 17). USP2 is localized about 1 Mbp telomer to MLL at 11q23.3 and transcriptionally orientated in direction of the centromere of chromosome 11, classifying all these fusions as intrachromosomal inversions (see Fig. 1b). In addition, we identified four balanced translocations in the minor BCR: one patient with an USP8 fusion (see Fig. 2 and Suppl. Figure S1), two with AF4 and one with AF9.
MLL-USP2 and MLL-USP8 alleles seem to be restricted to the minor BCR (see Fig. 2), because they were never diagnosed in association with the major BCR. Most of the reciprocal USP2-MLL fusions were scattered over a larger region at 11q23.3 (see Fig. 2), involving also upstream (C2CD2L) and downstream genes (USP2-AS1). Our analysis revealed also five patients with 3′-MLL deletions that were caused by microdeletions (<200 bp), larger deletions (up to 34 kbp), or complex rearrangements including other chromosomes as well (n = 4; chromosome regions 2p21, 4q13.1, 12p13.33, and 18p11.32). A detailed picture of the investigated MLL-USP2 and MLL-USP8 and their reciprocal fusions is shown in the Suppl. Fig. S2A-D.
All patients with a rearrangement of USP2 or USP8 fused the conserved "UCH-domain" to an extended 5′-MLL portion (see Suppl. Fig. S1A). This may indicate that the UCH domain has a functional importance for the resulting MLL fusion protein. USP genes belong to a large group of deubiquitinating proteins binding to specific target proteins [3][4][5]. The USP family exhibits a ubiquitin-specific protease (UCH domain) that is characterized by several conserved amino acids that are summarized as CYS-and ASP-box (see Suppl. Fig. 1B). USP2 protein deubiquinates and stabilizes MDM2, leading to an enhanced degradation of p53 [6]. This in turn activates MYC, because active p53 induces the transcription of several microRNAs that target MYC mRNA.
MLL fusions with the conserved 3′-UCH domain of USP2 and USP8 may change profoundly the functions of these novel MLL fusion proteins. It has already been shown that PHD2 [7] and PHD3/BD [8] both bind to proteins (CDC34 and ESC ASB2 ) that mediate the destruction of MLL by poly-ubiquitination and proteasomal degradation. Fusing single or all PHD domains to a der(11) product (MLL-AF9 and MLL-ENL) caused even a strong drop of their transforming potential [9,10]. This well-described degradation mechanism of MLL may now be counteracted by the UCH domain of MLL-USP2 or MLL-USP8, and thus, restoring their oncogenic transformation capacity.
In our cohort, we also identified new MLL fusion partner genes (n = 3). These novel fusion genes were SNX9 (6q25.3), USP8 (15q21.2), and SEPT3 (22q13.2). SNX9 encodes a protein known to be a member of the sorting nexin family which contain a phosphoinositide binding domain and are involved in intracellular trafficking. The SNX9 protein has a variety of interaction partners, including an adapter protein 2, dynamin, tyrosine kinase nonreceptor 2, Wiskott-Aldrich syndrome-like, and ARP3 actin-related protein 3. USP8 has diverse functions, being Fig. 1 Overview about all analyzed patients, their molecular information and breakpoint distribution. a Data from 109 patients which were analyzed with NGS in percentages. Their breakpoint distribution is displayed left (major BCR; n = 67) and right (minor BCR; n = 21). Five patients displayed only a reciprocal fusion, while 16 cases displayed no MLL rearrangement. b Top: chromosome 11 is depicted with highlighting of the MLL (red) and USP2 (green) genes. Below: all the genes between MLL and USP; blue marked genes: additional genes found in this study to be rearranged with MLL; orange marked genes: genes that have been earlier described to be rearranged with MLL. Recombinations between MLL and USP2 are caused by an inversion, with reciprocal alleles that carry additional deletions or complex rearrangements required for the internalization of liganded receptor tyrosine kinases and stabilization of ESCRT components. The USP8 protein is thought to regulate the morphology of the endosome by ubiquitination of proteins on this organelle and is involved in cargo sorting and membrane trafficking at the early endosome stage. SEPT3 is the seventh member of the septin family of GTPases that is fused to MLL. Members of this family are required for cytokinesis.
A few cases of MLL-USP2 fusions have already been described. However, these were single patient cases and they were classified as exceptional rearrangements [11][12][13]. Our NGS approach allowed for the first time the recurrent characterization of breakpoints in this novel minor BCR region of the MLL gene. Moreover, our targeted NGS approach enabled us to overcome the technical limitations associated with LDI-PCR and MP-PCR approaches.
Another advantage of the targeted NGS approach is the simultaneous identification of 3′ MLL deletions or copy number variations. In the current study, 23 of the patients (out of 88: 26%) had a 3′ MLL deletion. According to our data, 3′-MLL deletions were present in both breakpoint groups (major and minor) to a similar extent with 26.9% and 23.8%, respectively. This seems to be much higher than previously described (Andersson et al. [12]: 13%; Peterson et al. [14]: 7%).
In diagnostic fluorescence in situ hybridization analyses, these MLL-USP2 cases revealed two major patterns: (1) loss of the 3′-MLL probe signal, and (2) a normal pattern typical for MLL wild-type (Suppl. Table S1b, Suppl. Fig. S3). Considering the clinical data (Suppl. Table S1a-c), our 17 patients with MLL-USP2 were divided into 8 males and 9 females. All of them were children, and the median age at diagnosis was 17 months (range: 3-120 months). The median leukocyte count was 30.4 × 10 9 /L (range: 3.4-324.0 × 10 9 /L), and the disease phenotype was predominantly B-ALL (n = 12), followed by mixed-phenotype acute leukemia (MPAL) (n = 4) and acute myeloid leukemia (n = 1). The MPAL cases all had mixed myeloid and B-cell phenotype. The patients were treated with diverse therapy protocols. Five patients (29%) presented with central nervous system disease, and 13 patients (76%) had positive-minimal residual disease (MRD) levels at day 33. Prednisone response was measured in 12 patients with a poor response in 5 patients (42%). The median follow-up of the patients was 1.2 years (range: 0.1-11.1 years), and 2 cases died after 5 and 9 months following The MLL gene is depicted from exon 4 to the end. The major BCR is marked in green, the minor BCR in red. Main breakpoint regions are depicted in dark green/red while regions with fewer breakpoints are depicted in light green/red. The fusions sites and the fusion partners are shown. Information about the 5 cases with no der (11) or the 23 cases with 3′-MLL deletions are given in the box at the right bottom diagnosis. The remaining patients are still at first clinical remission.
In conclusion, we have identified a minor BCR within the human MLL gene that is recurrently associated in acute leukemia patients with MLL-USP2 fusion alleles as well as MLL fusion partnerships with USP8, AF4, and AF9. However, with 17 cases out of~2500 analyzed patients the incidence is less than 1% while still ranking fourteenth of our updated fusion gene list (see Table 1 of reference 1). The discovery of a second, minor BCR extends our knowledge of the MLL-recombinome and MLL-r oncogenesis. Moreover, these findings will enable many labs to make changes in their diagnostic set-up for MLL-MRD diagnostics to ensure the best medical treatment for a group of patients that is still very hard to cure. 1 To the Editor: Activating signaling mutations are common in acute leukemia with KMT2A (previously MLL) rearrangements (KMT2A-R) [1]. When defining the genetic landscape of infant KMT2A-R acute lymphoblastic leukemia (ALL), we identified a novel FLT3 N676K mutation in both infant ALL and non-infant acute myeloid leukemia (AML) [1]. FLT3 N676K was the most common FLT3 mutation in our cohort and we recently showed that it cooperates with KMT2A-MLLT3 in a syngeneic mouse model [1,2]. To study the ability of FLT3 N676K to cooperate with KMT2A-MLLT3 in human leukemogenesis, we transduced human CD34 + -enriched cord blood (CB) cells and followed leukemia development immunophenotypically and molecularly in NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ (NSG) mice.
To determine the evolution of phenotypically distinct leukemia cells in secondary recipients, BM cells from six primary KMT2A-MLLT3 + FLT3 N676K leukemias (three each with > 60% or 20-32% of FLT3 N676K -expressing cells) and from four primary KMT2A-MLLT3 leukemias, were retransplanted. All leukemias gave rise to secondary malignancies and recipients that received BM from AML (n = 2) and ALL (n = 1) with > 60% of KMT2A-MLLT3 + FLT3 N676K cells had an accelerated disease onset and maintained leukemia immunophenotype (median latency of 117 and 41 days, for the primary and secondary recipients, respectively) (Supplementary Fig. 2A-E and Supplementary Data 2). Thus, FLT3 N676K circumvented the cytokine dependence normally required for myeloid cells in immunodeficient mice [3]. By contrast, all secondary recipients that received BM with 20-32% KMT2A-MLLT3 + FLT3 N676K cells succumbed to ALL, irrespective of disease phenotype in the primary recipients (one AML and two BLL). Thus, the myeloid FLT3 N676K -expressing cells unexpectedly decreased in size, while the KMT2A-MLLT3expressing lymphoid cells increased to clonal dominance ( Supplementary Fig. 2F, G and Supplementary Data 2). This suggests that the FLT3 N676K -containing myeloid leukemia population needs to be sufficiently large to expand in secondary recipients, either because they otherwise are outcompeted by the larger population of more easily engrafted ALL cells, or because they themselves need to mediate the permissive microenvironment that allows myeloid cells to engraft. Similarly, in all but one of the secondary recipients that received BM from leukemias expressing only KMT2A-MLLT3 (two AML, one ALL, and one BLL), the disease phenotype changed and myeloid cells did not engraft ( Supplementary Fig. 2H-J and Supplementary Data 2).
In the secondary recipient with maintained immunophenotype (a BLL, h11.13-1), the myeloid cells unexpectedly expanded from 38% to close to 50% ( Fig. 1d and Supplementary Data 2). This suggested that the myeloid cells had acquired a de novo mutation that allowed serial transplantation, similar to what was observed for FLT3 N676K . Strikingly, targeted sequencing of AMLassociated genes on hCD45 + BM from this mouse identified a KRAS G13D in 34% of the cells. Resequencing of hCD45 + CD19 − CD33 + and hCD45 + CD19 + CD33 − BM showed that KRAS G13D was present exclusively in the myeloid population and based on the variant allele frequency of 51%, that all cells carried the mutation ( Fig. 1d and Supplementary Table 1, 2). Further, KRAS G13D likely arose independently in h11.13-1 as no mutation was identified, at the level of our detection, in the primary (h11. 13) or in a separate secondary recipient (h11.13-2) from the same primary mouse (h11.13) that developed ALL (Supplementary Table 2 and Supplementary Data 2).
Gene expression profiling (GEP) followed by principal component analysis (PCA) showed that the leukemias segregated based on their immunophenotype, with an evident separation between leukemias and normal hematopoietic cells (Fig. 2a, Supplementary Fig. 3A, B, and Supplementary Table 3). All leukemias expressed high levels of known KMT2A-R target genes and showed enrichment of gene signatures associated with primary KMT2A-R leukemia, indicating that they maintain a GEP representative of human disease ( By correlating the GEPs of the xenograft leukemias to those of normal hematopoietic cells [11], DPL and ALL were found to resemble normal common lymphoid progenitors (CLPs) and AML cells normal GMPs (Fig. 2b). Further, ALL and DPL cells had significantly higher expression of the transcription factor CEBPA as compared with normal lymphoid cells which lacked CEBPA expression (Fig. 2c). CEBPA drives myeloid programs and is expressed in most hematopoietic progenitors, in particular in GMPs, with CLPs lacking CEBPA expression ( Fig. 2c) [11]. Finally, to link the GEPs of our xenograft leukemias to those of pediatric KMT2A-R leukemia, we utilized a dataset of pediatric B-cell precursor ALL (BCP-ALL) [12]. Multigroup comparison visualized by PCA showed that KMT2A-MLLT3 DPL and ALL mainly resembled KMT2A-R BCP-ALL (Fig. 2d), again highlighting that the xenograft leukemias resemble human leukemia.
Herein, we demonstrate that co-expression of KMT2A-MLLT3 and FLT3 N676K in human CB cells primarily causes AML and thus alters the lineage distribution of KMT2A-MLLT3-driven leukemia. AML could only be serially transplanted with maintained immunophenotype in the presence of FLT3 N676K . This is consistent with the idea that activated signaling allows myeloid cells to more efficiently engraft and maintain their self-renewal. In agreement, we identified a de novo KRAS G13D in myeloid KMT2A-MLLT3expressing cells that had expanded upon secondary transplantation. Altogether, this shows that constitutively active signaling mutations can substitute for external factors and influence the phenotype of the developing KMT2A-R leukemia, at least in xenograft models.

GSE127492.
Acknowledgements Author contributions AHW and AKHA designed the study and experiments; AHW, MP, AFC, ME, HS, PS, PW, CGR, JL, and HÅ performed experiments; AHW, MP, HL, and AKHA analyzed sequencing data; AHW, AKHA, AH, MJ, and RWS interpreted data; AHW and AKHA wrote the paper. MJ and RWS performed critical reading and contributed to the writing of the paper.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.  To the Editor:

Unique CDR3 epitope targeting by CAR-T cells is a viable approach for treating T-cell malignancies
Efficient and specific removal of malignant cells is the ultimate goal of cancer therapy. The current rapid development of chimeric antigen receptor T-cell (CAR-T-cell or CART) therapy potentially provides high efficiency and allows long-term surveillance, which have greatly extended the frontier of leukemia treatment.
T-cell leukemia/lymphoma accounts for 15-25% of the incidence of malignant lymphoid diseases and represents 22 clinicopathologic entities in the most recent classification [1][2][3]. Recently, researchers have started to investigate approaches using chimeric antigen receptor T-cells (CAR-T-cells or CARTs) to treat T-cell leukemia/lymphoma by targeting common antigens such as CD4, CD5, TRBC1/2, and the chemokine receptor CCR4 [4][5][6][7]. The major hallmark of T-cell leukemia is the clonal expansion of malignant lymphocytes [8][9][10], which is both a characteristic and diagnostic marker [11,12]. Given the vast diversity of the immune repertoire, leukemia cell germlines from each patient have a unique T-cell receptor (TCR) on the cell surface that distinguishes these cells from normal T-cells. We hypothesized that targeting complementarity-determining region 3 (CDR3) on TCRs on leukemia cells by personalized tumor-specific CARTs would be a valid and superior approach that may offer higher precision and lower off-tumor toxicity than current treatment approaches. Recent advancements in nextgeneration sequencing (NGS) technology have enabled researchers to investigate TCR diversity with unprecedented depth and efficiency. Here, we propose a novel pipeline based on epitope screening technology aimed at isolating unique CDR3 regions from patients with leukemia/lymphoma for the subsequent redirection of human Tcells toward malignant epitopes. We assumed that the screening of combinatorial single-chain variable fragment These authors contributed equally: Jinqi Huang, Stepanov Alexey, Jian Li * Xiao Lei Chen Cxl2015@xmu.edu.cn * Guo Fu guofu@xmu.edu.cn * Gabibov Alexander gabibov@mx.ibch.ru * Chi-Meng Tzeng cmtzeng@xmu.edu.cn (scFv) libraries derived from the immune repertoire of cancer patients would allow us to select malignancyspecific epitopes that could facilitate the generation of redirected tumor-specific T-cells. We established a platform spanning from clinical pathology to antigen identification and from targeted scFv screening to in vitro/ in vivo validation, and the entire process can be completed in a relatively short time frame (Fig. 1a). We believe that clonal T-cell CDR3-selective CART therapy is a viable approach for further clinical development, with the potential for extrapolation to B-cell malignancies. We collected pathological samples from nine leukemia patients (four with T-cell acute lymphoblastic leukemia (T-ALL) and 5 with T-cell non-Hodgkin lymphoma (T-NHL), Supplementary Table S1) pathologically confirmed by histology and flow cytometry, and four healthy donors (normal controls) to study TCR diversity. Isolated cells (as well as Jurkat and MOLT-4 cells as controls) were used for gDNA extraction, amplification of the CDR3 region of TCRβ chain sequences and NGS analysis ( Supplementary Fig. S1, Table S2). Consistent with previous reports [13], the TCR diversity of T-cells from leukemia patients showed an abnormal predominance of CDR3 sequences (the frequency ranged from 7 to 94%), while no single clone exhibited a frequency of higher than 2% in T-cells from healthy donors (Fig. 1b). Identified dominant, unique CDR3 sequences from the leukemia, Jurkat and MOLT-4 cells were chemically synthesized and used for further applications.
Seminal studies imply that a patient's immune surveillance system may harbor antibodies against neoepitopes on pathogens or malignant cells [14]. We hypothesized that the antibody library derived from a leukemia patient could be uniquely useful for the selection of antibodies against the patient's malignant TCRs. Therefore, we constructed a phage combinatorial antibody scFv library from 56 patients and 20 healthy donors. Furthermore, we performed phage   Table S3). The results of previous structural studies suggest that some CDR3 fragments form rather sophisticated tertiary conformations, which may not be fully represented by chemically synthesized linear peptides. To validate whether our selected antibodies could recognize TCRs harboring the CDR3 sequences, selected scFv-GFP fusion proteins were used to stain the biopsy samples from the patients. The scFv fusion proteins specifically labeled both dispersed cancer cells (Jurkat and patient #1-T-ALL) (Fig. 1c) and lymphoma biopsy (patient #6-peripheral T-cell lymphoma (PTCL) and patient #8-angioimmunoblastic T-cell lymphoma (AITL)) (Fig. 1d, e). To analyze off-target specificity, clinical bone marrow, liver, spleen, and lymph node tissue specimens were stained with selected scFv-GFP fusion proteins. Immunohistochemical examination did not show any detectable offtarget binding of the CDR3-specific scFvs ( Supplementary  Fig. S2). Thus, we concluded that patient B cell repertoireaided combinatorial antibody library screening is a valid approach to identify antibodies against TCR CDR3 regions on malignant cells for subsequent therapeutic applications.
To explore the possibility of personalized immunotherapy with the CAR approach, the library-selected CDR3specific scFvs against TCRs on MOLT-4 cells, Jurkat cells, and cells from patient #1 were cloned into lentiviral vectors encoding second-generation CARs and named J-CAR, MOLT-4-CAR and P#1-CAR, respectively. The corresponding lentiviruses were produced and used for transduction of activated primary human T-cells.
We next determined whether T-cells modified with the P#1-CAR and J-CAR constructs exhibited specific cytotoxicity in vitro. We incubated CARTs with biopsy cells from patient #1, Jurkat cells or modified Jurkat cells expressing TCRs with the replaced CDR3 from the tumor cells from patient #1 (P1-CDR3-Jurkat). After 24 h of coculture, activated human CDR3-specific T-cells responded to their corresponding Jurkat or patient malignant Tcells by secreting IL-2 and IFN-γ, as well as undergoing highly efficient cell lysis (Fig. 2a).
To validate the specificity of the CARTs targeting CDR3 epitopes on malignant cells, we analyzed the TCR repertoire diversity before and after coincubation of the CDR3-selective CARTs or control CART-cells with biopsy cells from patient #1 or Jurkat cells. By ranking the frequency of the V and J regions of the TCRs, we found that control CARTs did not alter the frequency of TCR sequences on malignant cells (Fig. 2b, left panel). However, both J-CARTs and P#1-CARTs drastically reduced the percentage of malignant T-cell clones (Fig. 2b, right panel), suggesting the safety and substantial therapeutic potential of TCR targeting.
The encouraging results described above prompted us to investigate whether this approach could be further validated in vivo. We tested the efficacy of J-CARTs, MOLT-4-CARTs, and P#1-CARTs in a disseminated leukemia model by i.v. engrafting the corresponding target cell lines (Jurkat, MOLT-4 or P1-CDR3-Jurkat cells expressing luciferase) into immunodeficient NOD mice. Two days after tumor inoculation, animals from all experimental groups were treated with CARTs specific to the TCRs on the injected cells or control CARTs with irrelevant CDR3 specificity.
Injection of J-CARTs, MOLT-4-CARTs, and P#1-CARTs significantly reduced the tumor burden in animals and led to cancer cell elimination, while animals injected with control CARTs hosted a progressively increasing number of malignant cells (Fig. 2c). Indeed, compared with animals in the control CART-treated groups, animals in the CDR3-specific groups exhibited improved body weight and extended survival (Fig. 2d).
We demonstrated for the first time that targeting the CDR3 regions of malignant T-cell clones by cell therapy is a viable approach to eliminate leukemia cells. Due to its intrinsic uniqueness, the CDR3 region has historically been an intriguing target with much historical discussion. A few attempts have been made to use this region as the antigen for a cancer vaccine [15]. Due to the clonal nature of leukemia cells and the uniqueness of a given CDR, targeting CDR3 should offer a few advantages not offered by targeting common antigens: (1) Lower "on-target, off-tumor" effect; (2) Minimized impact on a patient's immune system during treatment; and (3) Great capacity for development as a personalized treatment that can be tailored to individual needs, as a percentage of patients do not respond well to existing therapies with common targets. In the current study, we validated the idea of using CARTs as targeting agents and observed excellent continuous efficacy as well as specificity. Combining CDR3 targeting with the CART approach provides a solution for a substantial portion of patients with T-cell leukemia and lymphoma, with supposedly minimized side effects. The potential problems to be solved in the future include establishing efficient and streamlined good laboratory practice (GLP)-level CDR3 binder discovery and good manufacturing practice (GMP)level personalized CART manufacturing and decreasing the financial burden for individual patients. However, these issues may be short-lived as technologies develop rapidly.
Nevertheless, after validation of this strategy to eliminate pathological T-cells ex vivo and in vivo, we envisage this approach as a generally useful alternative and supplement to the popular approach of common antigen targeting to treat T-cell malignancies, especially considering its safety.
Author Contributions JH, SA, JL, GF, XC, and CMT designed the study; JH, SA, JL, JXJZ, and XX performed and analyzed the experiments; JH, DC, and XW contributed to patient clinical care and data collection; JX and GF provided scientific advice; TJ, GG, and LD proof read the manuscript, JH, SA, JL, GF, MM, AG, JX, and CMT wrote the paper; and all authors read and approved the final version of the manuscript.

Compliance with ethical standards
Conflict of interest: The authors declare that they have no conflict of interest.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. To the Editor: The t(9;22) translocation results in the expression of the constitutively active BCR-ABL1 tyrosine kinase. It is detected in chronic myelogenous leukemia (CML) and iñ 30% of adult acute lymphoblastic leukemia (ALL) patients [1]. Thus, Ph + ALL is not only the largest genetically defined subgroup of ALL, but also characterized by a poor prognosis [2]. The two major protein isoforms of Bcr-Abl are p210 and p190. Whereas the shorter p190 isoform is specific for Ph + ALL, the longer p210 isoform causes CML, but is also present in~30% of Ph + ALL patients [1,3]. BCR-ABL1 was the first oncogene targeted successfully with the tyrosine kinase inhibitor (TKI) imatinib, which results in durable remissions in most CML patients and increased remission rates and survival in Ph + ALL patients. Still, resistance to imatinib occurs particularly frequently in Ph + ALL. Several next-generation TKIs were developed to address TKI resistance and intolerance [4]. Various TKI resistance mechanisms, including dozens of Bcr-Abl mutations, were described, but causes for resistance are still elusive in a significant portion of patients [5].
Deregulation of protein tyrosine phosphatases (PTP) plays an important role in maintaining a wide range of cancers. The ability of tyrosine phosphatases to antagonize oncogenic tyrosine kinases makes them candidate tumor suppressors. We previously showed that deregulation of PTP1B causes resistance in Ph + leukemias [6]. The phosphatase Sts-1 (suppressor of T-cell receptor signaling 1, encoded by the human UBASH3B gene) was found to be transcriptionally upregulated in Ph + ALL as compared with Ph − ALL patients [7]. Notably, Sts-1 also is one of the most prominent interactors of Bcr-Abl as determined by a systematic interaction proteomics screen [8]. In two recent independent studies, Sts-1 was found to interact more strongly with the Bcr-Abl p210 isoform than with p190 and to be phosphorylated in Bcr-Abl expressing cells [9,10]. Sts-1 and its only human and mouse paralogue, Sts-2 (UBASH3A), comprise an N-terminal ubiquitin-associated (UBA) domain, an Src homology 3 (SH3) domain and a C-terminal phosphoglycerate mutase (PGM) domain, which has structural homology with the histidine phosphatase superfamily. It was demonstrated that Sts-1 (and to a lesser extent Sts-2) possesses tyrosine phosphatase activity [11]. Strikingly, Sts-1 is a negative regulator of several tyrosine kinase pathways, including not only EGFR and PDGFR, but also ZAP-70 and SYK, thereby antagonizing T-and B-cell receptor signaling, respectively [12,13]. As genetic and functional perturbation of kinasephosphatase networks have been implicated in oncogenesis and based on our previous expression and proteomics data, we reasoned that the interaction between the Bcr-Abl kinase and the Sts-1 phosphatase may contribute to leukemogenesis. We therefore investigated the functional relationship between these two proteins, in particular the ability of Sts-1 to dephosphorylate Bcr-Abl and how it may contribute to TKI resistance in Ph + ALL patients.
To study the interaction of Bcr-Abl p190 with Sts-1 and its dependence on kinase activity in Ph + ALL cells, we performed coimmunoprecipitation (co-IP) assays of the endogenous p190 and Sts-1 in Sup-B15 cells either in the absence or presence of imatinib. Sts-1 binding to p190 was largely independent of the Bcr-Abl activation status, as their interaction was only mildly reduced in cells following exposure to imatinib (Fig. 1a). These data were confirmed by co-IP assays using murine Ba/F3 cells stably expressing p190 or Sts-1 as an independent cell-line model. p190 interacted with endogenous or overexpressed Sts-1 regardless of the activation status of Bcr-Abl (Fig. 1b, c).
In order to map the interaction mode of the two proteins, we first determined if Sts-1 binds to the Bcr-or Abl-portion of Bcr-Abl p190. These experiments were performed in Ba/ F3 cells upon coexpression of Sts-1 with full-length Bcr-Abl, the Abl-portion (#ABL1) only, the Bcr oligomerization domain directly fused to the Abl-portion (BCC-ABL1) or only the Bcr-portion (BCR; Fig. 1d). These experiments showed that Sts-1 binding was mediated by the Abl-portion of Bcr-Abl, as all Abl-containing constructs, but not the Bcr-portion alone, bound Sts-1 (Fig. 1d). To map the Sts-1 domains that are required for binding, we performed co-IP assays using loss-of-function point mutations in all domains of Sts-1, including the UBA (W72E), SH3 (W295A), and PGM (H391A) domains (Fig. 1e). To further delineate the requirements for binding to Bcr-Abl, we also assayed these mutants in combination with Bcr-Abl mutants with either abolished kinase activity (ΔK; D382N mutation) or a deletion in the C-terminal last exon region (ΔLE). Sts-1 interacted with the N-terminal part of the Abl-portion of Bcr-Abl encompassing the SH3-SH2-kinase domain unit (Fig. 1e). This interaction was independent of the activation status and the various protein-protein interaction motifs in the C-terminal last exon region of Bcr-Abl. Furthermore, the interaction does not require a functional UBA-, SH3-, or PGM-domain of Sts-1 (Fig. 1e).
To study a possible functional interdependence of the Bcr-Abl/Sts-1 kinase-phosphatase interaction, we first investigated how Sts-1 may regulate Bcr-Abl kinase activity and autophosphorylation at different tyrosine (Y) residues: Y177 in the Bcr-portion is critical for Ras-MAPK signaling, whereas Y245 and Y412 in the Abl-portion are important markers for kinase activation [14]. Thus, we examined the autophosphorylation of Bcr-Abl in Ba/F3 cells in the presence and absence of Sts-1. In line with its binding properties, Sts-1 caused strong dephosphorylation of Bcr-Abl at Abl-Y245 and Abl-Y412, whereas Bcr-Y177 was only mildly dephosphorylated (Fig. 1f). In a second step, we cotransfected HEK293 cells with either Bcr-Abl or an oligomerization-deficient mutant (ΔCC-Bcr-Abl) together with either wild-type or a phosphatase-dead (H391A) Sts-1 [11]. We found that Sts-1 dephosphorylates both Bcr-Abl and itself, and that Sts-1 is a kinase substrate of Bcr-Abl. In fact, Sts-1 dephosphorylated Bcr-Abl and ΔCC-Bcr-Abl strongly and equally well (Fig. 1g). Conversely, only phosphatase-dead Sts-1, but not wild-type Sts-1, was strongly phosphorylated in the presence of Bcr-Abl, demonstrating that Sts-1 is able to dephosphorylate itself (Fig. 1g).
Given that Sts-1 may regulate kinase activity of Bcr-Abl by modulating its autophosphorylation, we next investigated whether Sts-1 impacts on cell proliferation in IL-3independent Ba/F3 cells expressing wild-type (wt) Bcr-Abl or the gatekeeper mutation T315I, which conveys resistance to multiple TKIs. These cells were retrovirally transduced with GFP or Sts-1-GFP, and proliferation competition assays were performed over the course of 12 days. The expression of GFP alone did not alter the proliferation of BCR-ABL expressing Ba/F3 cells as revealed by the constant percentage of GFP positive cells (Fig. 2a). In contrast, expression of Sts-1-GFP reduced the proliferation of Ba/F3 cells expressing Bcr-Abl wt and T315I. Concomitant treatment with 1 µM imatinib further decreased proliferation of Bcr-Abl wt, but, as expected, not Bcr-Abl-T315I cells (Fig. 2a). These results indicate that Sts-1 activity negatively regulates cell proliferation induced by Bcr-Abl wt and T315I.
To assess the role of Sts-1 in Bcr-Abl-driven leukemogenesis, we examined the induction of Bcr-Abl p210-induced CML-like disease in wt vs. Sts-1/Sts-2 doubleknockout bone marrow cells using a transduction/transplantation model. The absence of Sts-1/Sts-2 decreased the survival of recipient mice significantly and further aggravated the pronounced splenomegaly observed in mice transplanted with Bcr-Abl-expressing wt cells (Fig. 2b). Fig. 1 The Sts-1 phosphatase interacts with and dephosphorylates Bcr-Abl. a SupB15 was treated with imatinib (2 µM for 6 h). ABL1 immunoprecipitates (left panel) from total cell lysates and 5% input fraction (right panel) were immunoblotted with the indicated antibodies. b Ba/F3 cells transduced with STS-1 were additionally transduced with BCR-ABL1 and treated with imatinib (1 µM for 6 h). STS-1 immunoprecipitates (right panel) from total cell lysates and 5% input fraction (left panel) were immunoblotted with the indicated antibodies. c Ba/F3 cells transduced with BCR-ABL1 were additionally transduced with STS-1 and treated with imatinib (1 µM for 6 h). ABL1 immunoprecipitates (right panel) from total cell lysates and 5% input fraction (left panel) were immunoblotted with the indicated antibodies. d Ba/F3 cells transduced with full-length Bcr-Abl, the Abl-portion (#ABL1) only, the Bcr oligomerization domain directly fused to the Abl-portion (BCC-ABL1) or only the Bcr-portion (BCR) was additionally transduced with STS-1. ABL1 (upper panels) and STS-1 (lower panels) immunoprecipitates from total cell lysates were immunoblotted with the indicated antibodies. e HEK293 cells were cotransfected with BCR-ABL1 wt, Bcr-Abl kinase-dead (D382N; BCR-ABLΔK) or BCR-ABL1 lacking the C-terminal last exon region (BCR-ABLΔLE) with STS-1 wt and point mutations in functional domains (UBA (W72E), SH3 (W295A), and PGM (H391A)). ABL1 immunoprecipitates were analyzed by immunoblotting using the indicated antibodies. f Ba/F3 cells retrovirally expressing BCR-ABL1 were transfected with STS-1 or empty vector. Whole-cell lysates were then analyzed with the indicated antibodies. g HEK293 cells were cotransfected with BCR-ABL1 wt or an oligomerization-defective BCR-ABL1 lacking the coiled-coil domain (ΔCC-BCR-ABL1) with either STS-1 wt or a phosphatase-dead STS-1 mutant (H391A). Total cell lysates were analyzed by immunoblotting using the indicated antibodies These results indicate that Sts-1/Sts-2 are functionally relevant negative regulators of Bcr-Abl-dependent leukemogenesis in a CML mouse model.
The glucocorticoid dexamethasone and Sts-1 seem to regulate several common signaling pathways: Both inhibit T-cell receptor (TCR) signaling by regulating TCR expression and Sts-1 additionally inhibits certain downstream effectors [15]. In order to harness the therapeutic potential of Sts-1's ability to inhibit growth of Bcr-Ablpositive cells, we explored whether dexamethasone may alter Sts-1 expression and activity. Exposure of Sup-B15 and Sup-B15RT cells to clinically relevant concentrations of dexamethasone increased Sts-1 expression over time, accompanied by decreased Bcr-Abl autophosphorylation in Sup-B15 cells (Fig. 2c, d), indicating that it increases the sensitivity of Bcr-Abl-transformed cells to TKIs. In cell proliferation assays, concomitant treatment with dexamethasone and imatinib showed stronger inhibition than each drug alone in both Sup-B15 and Sup-B15RT cells (Fig. 2e).
Deregulation of the tyrosine phosphatase Sts-1 may be an important and pharmacologically targetable mechanism for Bcr-Abl mutation-independent resistance. Upregulation of Sts-1 in Ph + ALL together with its direct interaction with both p190 and p210 Bcr-Abl strongly suggests a functional deregulation of protein phosphorylation similar to that we previously showed for PTP1B [6]. The fact that its normal function can be restored by ectopic overexpression not only indicates a central role for Sts-1 in the regulation of Bcr-Abl but also that the deregulation of Sts-1 is based on a loss of balance between Bcr-Abl kinase and Sts-1 phosphatase activity. This establishes the upregulation of Sts-1 by drugs, such as dexamethasone, as a valid therapeutic approach for increasing the sensitivity to TKIs.
In conclusion, we delineated the molecular interaction mode of the Sts-1 phosphatase with the Bcr-Abl kinase and provide strong evidence that Sts-1 is a negative regulator of Bcr-Abl signaling, cell proliferation, and leukemogenesis. In addition, the parallel study by Udainiya et al. (cosubmitted) shows a broad impact of Sts-1 on the Bcr-Abl b Equal amounts of primary murine Lin − c-Kit + Sca-1 + cells from wt or Sts-1/Sts-2 knockout animals expressing BCR-ABL1 were injected in lethally irradiated recipient mice (n = 8 for each group). Overall survival of transplanted mice was monitored over 60 days. P-value = 0.014 was calculated using a logrank (Mantel-Cox) test. Representative spleens from control mice, BCR-ABL1/Sts-1/2 wt and BCR-ABL1/Sts-1/2 KO mice are shown. c SupB15 Ph + ALL cells and their imatinib-resistant subline SupB15RT were exposed to 10 −7 M dexamethasone, and STS-1 expression was investigated at the given time points. α-Tubulin was used as loading control. d SupB15 Ph + ALL cells and their imatinib-resistant subline SupB15RT were exposed to 10 −7 M dexamethasone, and the effect of increasing expression of STS-1 on BCR-ABL1 phosphorylation was investigated by an anti-p-Tyr antibody. e SupB15 and SupB15RT were exposed to imatinib (1 µM) and 10 −7 M dexamethasone alone or in combination and proliferation was analyzed by the XTT assay. The bars represent the mean (±SEM) of three independent experiments, each performed in triplicates. Statistical significance was calculated using student's t test. *** p ≤ 0.001 phosphoproteome network and precisely delineated the Sts-1 interactome using quantitative functional proteomics techniques. Furthermore, we show that modulation of Sts-1 expression by dexamethasone influences TKI sensitivity of Ph + ALL cells. Therefore, the inclusion of dexamethasone for therapy regimens in Ph + ALL for may increase sensitivity to TKIs by upregulating Sts-1.
Acknowledgments This work was supported by the Swiss National Science Foundation (grant 31003A_140913; OH) and the Cancer Research UK Experimental Cancer Medicine Centre Network, Cardiff ECMCI, grant C7838/A15733. We thank N. Carpino for the Sts-1/2 double-KO mice.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.

To the Editor:
Multiple myeloma (MM) is the third most common hematological malignancy, after Non-Hodgkin Lymphoma and Leukemia. MM is generally preceded by Monoclonal Gammopathy of Undetermined Significance (MGUS) [1], and epidemiological studies have identified older age, male gender, family history, and MGUS as risk factors for developing MM [2].
The somatic mutational landscape of sporadic MM has been increasingly investigated, aiming to identify recurrent genetic events involved in myelomagenesis. Whole exome and whole genome sequencing studies have shown that MM is a genetically heterogeneous disease that evolves through accumulation of both clonal and subclonal driver mutations [3] and identified recurrently somatically mutated genes, including KRAS, NRAS, FAM46C, TP53, DIS3, BRAF, TRAF3, CYLD, RB1 and PRDM1 [3][4][5].
Despite the fact that family-based studies have provided data consistent with an inherited genetic susceptibility to MM compatible with Mendelian transmission [6], the molecular basis of inherited MM predisposition is only partly understood. Genome-Wide Association (GWAS) studies have identified and validated 23 loci significantly associated with an increased risk of developing MM that explain~16% of heritability [7] and only a subset of familial cases are thought to have a polygenic background [8]. Recent studies have identified rare germline variants predisposing to MM in KDM1A [9], ARID1A and USP45 [10], and the implementation of nextgeneration sequencing technology will allow the characterization of more such rare variants.
In this study, we sought to explore the involvement of rare germline genetic variants in susceptibility to MM.
Within our discovery cohort of peripheral blood samples (see Supplementary Methods) from 66 individuals from 23 unrelated families analyzed by WES, DIS3 (NM_014953) was the only gene in which putative loss-of-function variants were observed in at least two families. An additional cohort of 937 individuals (148 MM, 139 MGUS, 642 unaffected relatives and eight individuals with another hematological condition) from 154 unrelated families (including the individuals in the discovery cohort) were screened for germline variants in DIS3 using targeted sequencing (Supplementary  Table S1). In total, we detected DIS3 germline putative lossof-function variants in four unrelated families. The DIS3 genotypes for the identified variants were concordant between WES and targeted sequencing (where available) and independently confirmed by Sanger sequencing on DNA extracted from uncultured whole blood. The variant allele frequencies (VAF) were close to 50%, as expected of a germline variant (Supplementary figure S1).
The DIS3 gene, located in 13q22.1, encodes for the catalytic subunit of the human exosome complex, and is recurrently somatically mutated in MM patients [4,5,11,12]. The somatic variants are predominantly missense variants localized in the RNB domain mainly abolishing the exoribonucleolytic activity [4,13], and are often accompanied by LOH or biallelic inactivation due to 13q14 deletion, implying a tumor suppressor role for DIS3 in MM [5,12,13].
The first DIS3 variant, observed in 2 affected siblings (1 MGUS and 1 MM case) from family B (Fig. 1a), was located in the splice donor site of exon 13 (c.1755+1G>T; chr13: 73,345,041; GRCh37/hg19, rs769194741) (Supplementary Figure S1a). It is predicted to abolish the splice donor site and cause skipping of exon 13, introducing a premature termination codon (p.Arg557Argfs*3) and result in a truncated DIS3 protein that lacks part of the exonucleolytic active RNB and S1 domains (Fig. 1b, c). The presence of this variant in two siblings, implying Mendelian segregation, is consistent with a germline, rather than somatic, origin. We investigated whether a DIS3 transcript from the variant allele is generated but is subsequently eliminated by Nonsense Mediated Decay (NMD) by incubating Lymphoblastoid Cell Lines (LCLs) derived from the two c.1755+1G>T allele carriers with and without puromycin, which suppresses NMD. The mRNA transcript corresponding to the variant allele was clearly present in LCLs treated with puromycin in both carriers, whereas not detectable in untreated LCLs (Fig. 2a), consistent with the variant allele being transcribed but subsequently degraded via the NMD pathway. In line with this observation, analysis of DIS3 mRNA expression by qRT-PCR showed an average 50% reduced expression in the c.1755+1G>T carriers (range 40.7−61.4%) as compared to non-carriers (Fig. 2b). A second splicing variant (c.1883 +1G>C; chr13: 73,342,922; GRCh37/hg19) located in the splice donor site of exon 14 within the RNB domain was identified in a MM case from family D (Fig. 1a, b, Supplementary figure S1c). However, the individual's mother (Q59), affected with amyloidosis, did not carry the variant, implying that MM in the allele carriers' maternal uncles is unlikely to be explained by this DIS3 variant. Whether the mRNA transcript encoded by this germline variant undergoes NMD could not be explored due to lack of appropriate material (LCLs, RNA).
A third DIS3 variant disrupting the wild-type termination codon (stop-loss) (c.2875T>C; p.*959Gln; chr13:73,333,935; GRCh37/hg19, rs141067458) (Fig. 1b, Supplementary Figure S1b) was identified in two unrelated families (A and C, Fig. 1a). This variant is expected to result in a putative readthrough variant and a DIS3 protein with an additional 13 amino acids in the C-terminus (p.*959Glnext*14). It was detected in 3 out of 4 affected siblings (2 MGUS (M63, O53) and 1 MM case (O29)), as well as 5 unaffected relatives (N14, N13, L41, M33 M50) from family A. The Mendelian segregation of this variant in this pedigree is also consistent with germline origin. An additional MM case from family C carried the variant, while we were unable to assess the other MM-afflicted family member (Fig. 1a). As expected of a stop-loss variant, NMD was not observed (data not shown), and gene expression analysis showed no effect on DIS3 mRNA levels (Fig. 2b). However, western blot analysis demonstrated that DIS3 protein levels were markedly lower (~50%) in the p.*959Glnext*14 carrier (O53, family A) compared to non-carriers (Fig. 2b, c).
Next, we sought to determine if rare, putative deleterious variants in DIS3 were more frequent in an independent series of MM cases compared to unaffected individuals. We performed mutation burden tests between 781 MM cases and 3534 controls from the MMRF CoMMpass Study with WES data available. After testing for systemic bias in this dataset (see Supplementary Methods, Supplementary Figure S2 Figure S3a). We additionally genotyped the p.*959Glnext*14 stop-loss variant in an independent series of sporadic MM cases and controls from the IMMEnSE Consortium. While this variant was very rare in this series (8/3020 MM cases relative to 3/1786 controls), there was a consistent but non-significant association between this variant and MM (OR = 3.15 95% CI: 0.74-13.43 p = 0.122).
To explore the functional consequence of germline DIS3 variants, we compared MM tumor transcriptomes from patients harboring germline (n = 21) and somatic (n = 96) DIS3 putative functional variants to non-carriers (n = 655). Differential expression analyses showed an enrichment of pathways associated with global ncRNA processing and translational termination in germline DIS3 carriers including ncRNA processing, ncRNA metabolic process, translational termination, and RNA metabolism. Among somatic DIS3 carriers, significantly enriched pathways include interferon alpha/beta signalling, mRNA splicing, mRNA processing and transcription (Supplementary Figure S3b, Supplementary tables S3 and S4a-d). These findings are consistent with the proposed DIS3 role in regulating mRNA processing [14] and c Somatic DIS3 variants were identified in sporadic MM cases from the MMRF CoMMpass Study. We observe that in contrast to the clustering of somatic DIS3 missense variants in the RNB and PIN domains, germline variants are scattered throughout the gene and consist of splicing, stop-loss and missense variants more specifically mRNA decay, gene expression and small RNA processing [15]. We also observed that, several longintergenic non-protein coding RNAs, non-coding and antisense RNAs were significantly enriched among DIS3 carriers (Supplementary table S5a, b) supporting previous studies that demonstrate an accumulation of transcripts from non-protein coding regions, snoRNA precursors and certain lncRNAs in DIS3 mutant cells, along a general deregulation of mRNA levels probably due to the sequestration of transcriptional factors from the accumulated nuclear RNAs [16].
To our knowledge, this is the first observation of germline DIS3 likely deleterious variants in familial MM and our results suggest that the involvement of DIS3 in MM etiology may extend beyond somatic alterations to germline susceptibility. We reported rare germline DIS3 variants in~2.6% of our cohort of families with multiple cases of MM and MGUS (4/154). The germline variants described here are predicted to have loss-of-function impact on DIS3. Consistent with this, the 1755+1G>T (rs769194741) splicing variant induces NMD and results in reduced DIS3 mRNA expression, supporting the proposal that DIS3 is acting as a tumor suppressor gene in MM [13]. Moreover, the c.2875T>C (rs141067458) stop-loss variant (p.*959Glnext*14) results in reduced DIS3 protein expression suggesting that the mutant allele is translated but degraded shortly after. Notably, in contrast to the clustering of somatic DIS3 mutations in the PIN and RNB domains, germline variants identified both in familial and sporadic MM cases are scattered throughout the gene (Fig. 1b,  c). Despite the fact that these variants do not segregate perfectly with MM in the identified families and the rarity of DIS3 germline likely deleterious variants limits our statistical power, the subsequent mutation burden and transcriptome showed a mixture of the wild-type and mutant transcript lacking exon 13, which was not detected in the non-treated cells (without puromycin). Thus, the mutant transcript is degraded by NMD. b Box plot representing the relative DIS3 mRNA expression in c.1775+1G>A (n=2) and p.*959Glnext*14 (n = 1) carriers compared to non-carriers (n = 4). All reactions were performed in triplicates. c Western blot with an anti-DIS3 antibody was performed in LCLs from one p. *959Glnext*14 carrier and two wild-type individuals (anti-GAPDH antibody as internal control). The relative DIS3 expression in the p. *959Glnext*14 carrier was reduced by 50% compared to non-carriers, suggesting that the mutant allele is translated but degraded shortly after analyses provided supportive data towards DIS3 acting as an

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/. transmitted germline risk variants that can render B-cell precursors susceptible to additional somatic hits, allowing BCP-ALL development. Expanding this knowledge is a crucial step towards targeted treatments as well as precisionprevention programs.
Acute lymphoblastic leukemia (ALL) is the most common pediatric cancer (4/100,000) under the age of 15, with the majority of cases affecting B-cell precursors (BCP-ALL) [2]. Although survival rates exceed 90%, it remains a significant cause of death in young children. Thus, to develop novel individualized therapeutic approaches, or even better, to envision precision-prevention programs-in particularly for subtypes of high-risk ALL, elucidation of the tumor genetics are fundamental [3]. In this regard, inherited germline variants are of special interest. Recently, novel germline predisposition syndromes have been described [4], and current studies increasingly highlight the importance of trio-calling analyses in childhood cancer [5,6].
In this study, we performed trio-calling and describe a novel scenario for lymphoid malignancies, in which two susceptibility loci are inherited-a paternal one and a maternal one-and both act synergistically in the same signaling pathway, thereby forming a susceptible B-cell precursor compartment that is prone to secondary mutations driving BCP-ALL.
Utilizing whole-exome sequencing we identified two concomitant germline single-nucleotide variants (SNVs) affecting the JAK2/STAT3 pathway in a boy with BCP-ALL (Fig. S1A). The JAK2 variant, rs139504737 (c.1711G > A), leads to an aminoacid substitution from glycine to serine (p.G571S) and is rarely found within the general population (minor allele frequency (MAF) < 0.01). Interestingly, the same JAK2 p.G571S germline variant was identified in a Down syndrome (DS-ALL) patient from an independent family (Fig. 1a). The second variant found in the BCP-ALL patient, constitutes an extremely rare, and so far for leukemia undescribed, missense mutation in the STAT3 gene (c.1109 A > G), causing an exchange of lysine to arginine (p.K370R). While JAK2 p.G571S was transmitted from the father, STAT3 p.K370R was inherited from the mother's side (Fig. 1a). Both variants are located in functionally relevant domains, with STAT3 p.K370R being Fig. 1 a Pedigree of two independent families harboring JAK2 p. G571S. The ALL patients are marked with a red triangle, while the transmission of the identified target mutations is highlighted with red symbols (JAK2 = circle; STAT3 = +; trisomy 21 = #). b Domain organization of JAK2 and STAT3 proteins. Identified mutations (red) and the known activating JAK2 mutation p.V617F (black), which was used as a control, are highlighted. c Overview of the STAT3-WT (gray) with bound DNA (orange). The blue rectangle indicates the region with the residues of interest (sticks). A close-up depicts the mutated region in STAT3-WT (gray) and the p.K370R variant (dark green). While in the STAT3-WT, p.K370 only interacts with p.E455, p.R370 of the p.K370R variant interacts with both p.E442 and p.E455 as indicated by the yellow lines. The mutation of lysine to arginine also distorts the local backbone conformation. d Homology model of the p.G571S variant in the JAK2 protein (green). P.S571 (sticks) is located in a loop between two β-strands next to p.Y570 (sticks), which is one of the most important phosphorylation sites in JAK2 localized in the DNA-binding domain and JAK2 p.G571S in the pseudo-kinase domain, respectively (Fig. 1b).
Furthermore, 11 somatic mutations were identified in the BCP-ALL (Fig. S1B), with a missense mutation in the CEP89 gene (p.K138T) being the only tumor specific somatic variant. In addition, CytoScan ™ -HD analysis revealed loss of CDKN2A and IKZF1 loci, which groups the patient into the recently described IKZF1 plus subgroup that was shown to have a particularly poor outcome [7] (Fig. S1C). Moreover, a somatic JAK2 rearrangement was detected by cytogenetic analysis in the second JAK2 allele. While molecular genetic analysis could rule out a classical JAK2 fusion with ETV6, BCR, or PAX5, the actual fusion partner was not identified. The patient was enrolled into the AIEOP-BFM 2009 therapy protocol, responded poorly to therapy (prednisone-poor response at day 8, nonremission at day 33) and underwent hematopoietic stem cell transplantation from a matched unrelated donor after achieving first remission (MRD level < 10 −4 ).
To understand potential phenotypic influences of both mutant proteins on a structural level, we generated homology models of the mutated pseudokinase domain of JAK2 and the mutated STAT3, and compared the models to their respective wild types.
STAT3 p.K370R is located in a loop adjacent to the DNA binding site (Fig. 1c). Moreover, STAT3 p.K370 is an important site for acetylation, which enables STAT3's interaction with RELA, in turn promoting further downstream signaling [8]. In the deacetylated state, STAT3 p.K370 interacts with p.E455 in a β-sheet. Substitution of lysine to arginine at position p.370 has two implications: first, the arginine can interact with both p. E455 and p.E442 simultaneously, that way strengthening the interaction in the β-sheet. Second, in contrast to lysine, arginine cannot be acetylated. This combination leads to a constitutively non-acetylated form of STAT3 at position p.370, which was shown to have functional consequences in its interaction capacity with various signaling partners in HEK293T cells [8]. Furthermore, the STAT3 p.K370R protein did not show an activating phenotype in STAT3 reporter luciferase assays in HEK293T cells (Fig. S2A). This is in line with the structural modeling, suggesting impaired acetylation rather than phosphorylation (Fig. S2B).
JAK2 p.G571 is located in a 12-residue loop connecting 2 β-strands (Fig. 1d), adjacent to p.Y570, one of the most important phosphorylation sites in JAK2. This suggests an influence on the phosphorylation of p.Y570 either through steric hindrance, interactions of the serine sidechain or changes in the backbone conformation near p.Y570. Its unique position affecting amino acid 571, which lies adjacent to the p.Y570 residue that downregulates kinase activity via autophosphorylation hints at a potential functional mechanism of p.G571S by inhibiting p.Y570directed negative feedback.
To assess the cooperative potential of both variants, BaF3 depletion assays were carried out. In normal BaF3 cells, neither STAT3 p.K370R nor JAK2 p.G571S protein expression alone were sufficient to induce IL-3 independent growth, although immunoblot analyses confirmed increased p-STAT3 levels in cells expressing both JAK2 p.G571S and STAT3 p.K370R (Fig. 2a). Since dimerization by a cytokine receptor facilitates the constitutive activation of JAK2 mutants, we further transfected BaF3 cells which expressed human CRLF2 and the IL7R alpha chain (leading to the formation of the heterodimeric receptor for thymic stromal lymphopoietin (TSLP) [9]) with the identified target variants. In BaF3/CRLF2-IL-7RWT cells, JAK2 p.G571S protein expression conferred IL-3 independent growth. Moreover, the combination of both mutant proteins (JAK2 p.G571S + STAT3 p.K370R) showed a mild but significant growth advantage starting 2 days after IL-3 withdrawal (Fig. 2b). Immunoblot analysis revealed high-p-AKT levels in cells expressing JAK2 p.G571S, which was changed to p-STAT3 through additional expression of STAT3 p. K370R. Since hyperactivation of p-AKT negatively affects precursor B-cell survival [10], the here observed signal rewiring indicates a synergistic effect of both variants by balancing out signaling strengths.
Besides the cooperating capacity of both identified germline variants, we further observed that the STAT3 p.K370R mutant protein alone changed the phenotype of the cells in culture, with an accumulation of enlarged BaF3 cells. Surprisingly, this phenotype was reversed in the double mutant cells expressing both STAT3 p.K370R and JAK2 p.G571S (Fig. S2C). Cell cycle analysis was in line with this observation, showing a significant increase in >4n cells (p = 0.0009), while those in the G-1 phase were significantly decreased (p = 0.0031) in STAT3 p.K370R expressing BaF3 cells compared to STAT3-WT cells (Fig. 2c). Again, this phenotype was reduced in BaF3 cells transfected with both mutants simultaneously (G1 phase p = 0.0026; > 4n, p = 0.0032). In line with the reversed cell cycle phenotype in the double mutant cells, immunoblot analysis of the different conditions confirmed increased p-CDC-2, p-CyclinB1/ Cyclin-B1, and Cyclin-A2 in BaF3 cells expressing both JAK2 p.G571S and STAT3 p.K370R compared to STAT3 p. K370R alone (Fig. 2c). Taken together, STAT3 p.K370R conferred a cell cycle arrest in BaF3 cells which is consistent with a loss-of-function phenotype in the STAT3 reporter luciferase assay and the structural modeling.
These data suggest that the two mutations can act in concert to exert a germline susceptibility toward BCP-ALL development by the accumulation of a susceptible precursor compartment. This compartment might be prone to acquire additional secondary lesions in IKZF1 and CDKN2A or alterations of the JAK2 WT allele, which act as somatic oncogenic drivers leading to malignant growth.
Activation of the JAK-STAT pathway is known to be a key event in a variety of hematological malignancies. JAK2 p.V617F leads to constitutively active STAT5 signaling in about 80% of patients suffering from myeloproliferative neoplasms [11], and a high frequency of somatic rearrangements or SNVs activating JAK2 are found in Ph-like ALL [12] and DS-ALL [13,14]. However, germline JAK2 mutations particularly in BCP-ALL are rare. Here, we identified a second patient harboring germline JAK2 p. G571S in a DS-ALL cohort of 88 patients (Fig. 1a and Fig. S3) [14]. A major proportion of DS-ALL shows high expression of CRLF2 and somatic JAK/STAT pathway activation [15]. Although the analyzed DS-ALL patient did not express a P2RY8-CRLF2 transcript or show a CRLF2 p.232 mutation, he may harbor an activating translocation of CRLF2 into the IGH chain locus. However, due to the lack of patient material we could not test this hypothesis. Therefore, though CRLF2 activation in association with JAK2 p.G571S and trisomy 21 seems highly likely, we cannot confirm an active CRLF2 status in this patient. Nevertheless, the germline JAK2 p.G571S mutation in combination with trisomy 21 can be a complementary scenario of two germline variants acting in synergy to render cells susceptible to additional somatic alterations that can drive ALL development. This insight further strengthens the unique and important role of weak oncogenic germline risk variants (e.g., JAK2 p.G571S) and how they can synergize with additional low-penetrance mutations/ alterations to predispose to ALL development. To the Editor: The signal transducer and activator of transcription 5B (STAT5B), downstream of IL-15 signaling and Janus kinase (JAK)1, and 3-mediated activation, is a master regulator of development, survival, and function of innate and innatelike lymphocytes (including natural killer (NK) and NKT cells) [1][2][3]. Gain-of-function mutations in the SH2 domain of human STAT5B, especially STAT5B N642H , are associated with aggressive forms of CD56 + T cell (NKT) and NK cell lymphomas/leukemias [4][5][6]. We described a mouse model expressing human (h)STAT5B N642H under the Vav-1 promoter, which develops severe CD8 + T cell neoplasia [7]. Here, we explore the ability of hSTAT5B N642H to serve as an oncogenic driver in innate lymphocyte neoplasms.
We found an increase in absolute NK cell numbers in the spleen of T cell-diseased hSTAT5B N642H transgenic compared to wild-type (WT) mice (Fig. 1a), despite the relative decrease of the proportion of NK cells among splenic lymphocytes (Fig. S1A), while nonmutant hSTAT5B control mice showed intermediate NK cell numbers (Fig. 1a,   Fig. S1A). No significant differences in NK cell numbers were observed in the bone marrow (BM) between genotypes (Fig. S1B). In addition, both hSTAT5B and hSTAT5B N642H mice showed a similar increase in the proportion of mature NK cells (CD27 − CD11b + and KLRG1 + ) in the spleen (Fig. S1C, D). These findings suggest that the enforced expression of nonmutant hSTAT5B is sufficient to boost NK cell maturation, which is not further enhanced by introducing the activating hSTAT5B mutation. We hypothesized that any phenotypic alterations affecting innate lymphocytes might be masked by the hSTAT5B N642H -driven aggressive CD8 + T cell disease established in hSTAT5B N642H mice at the age of 6-8 weeks [7]. To explore the potential of hSTAT5B N642H to promote NK cell expansion in vivo, we transplanted CD3depleted BM from hSTAT5B N642H or hSTAT5B mice in immune-deficient Rag2 −/− γc −/− recipient mice. Using this approach, we observed an enhanced expansion of NK cells in the blood of hSTAT5B N642H compared to hSTAT5Btransplanted recipients over a time course of 4 weeks (Fig. 1b). Long-term analysis of NK cells in hSTAT5B N642Htransplanted mice was not possible due to expansion of residual CD8 + T cells (Fig. S1E, F), which forced us to terminate the experiment. However, after 4 weeks, increased numbers of NK cells were also detected in the spleen upon transplantation of CD3-depleted hSTAT5B N642H compared to hSTAT5B BM (Fig. S1G).
To investigate long-term effects of hSTAT5B N642H on innate lymphocytes, we sorted Lineage (Lin) − (CD3/B220/ Ter119/Gr1/CD11b) Sca1 − c-Kit − CD127 − CD8 − cells. This cellular fraction is devoid of hematopoietic stem cells, common lymphoid progenitors, and CD8 + T cells and was obtained from hSTAT5B N642H or hSTAT5B BM for further transplantation into Rag2 −/− γc −/− recipient mice. We monitored NK and T cell numbers in the blood over a period of 5 weeks. Again, a modest increase but subsequent drop in NK cell numbers in mice transplanted with hSTAT5B N642H compared to hSTAT5B BM cells was observed (Fig. S1H), whereas we failed to detect CD8 + T cells in the blood of the recipient animals (Fig. S1I). This indicated that our sorting strategy successfully eliminated CD8 + T cell disease, but did not allow us to uncover persistent hSTAT5B N642H -mediated NK cell expansion. Of note, after 4.5-10 months all three hSTAT5B N642H -recipient mice (#1-3) developed a rapidly progressing disease, characterized by expansion of CD3 + NK1.1 + NKT cells in blood, BM, spleen, and liver (Fig. 1c, Table S1). Disease development was restricted to transplantation of mutant hSTAT5B cells and was not detected upon transplantation of hSTAT5B BM, despite following up these animals for 10 months.
To investigate, whether hSTAT5B N642H NKT cells are indeed transformed and fulfill the criteria of being leukemic, we started to perform serial whole BM transplants from the earlier diseased recipient #1 (survival: 133 d) with the presumably more aggressive disease. To do so, we transplanted BM containing 1 × 10 6 NKT cells into immune-deficient Rag2 −/− γc −/− or NSG-recipient mice. This serial transplantation approach was continued for a total of six rounds. The first two rounds were performed to maintain and amplify the NKT cell disease. We confirmed the presence of the hSTAT5B N642H transgene in a diseased mouse (Fig. S1J). Upon the 3rd to the 6th round of serial transplant, we characterized the manifested disease as an aggressive leukemia accompanied by hepatosplenomegaly (Fig. 2a), increased white blood cell (WBC) counts (Fig. S2A) and a high frequency of NKT cells in blood, BM, spleen, and liver (Fig. S2C, D) with a mean survival of 17.6 days (Fig. S2B). Infiltration of leukemic cells into various organs was confirmed by histological analysis (Fig. S2E). In the 6th round, a titration of transplanted NKT cell numbers was performed. Disease severity at the end point was comparable, while a delay of 3 days occurred between each titration step (Fig. 2a, Fig. S2A-D). Despite the increased disease latency in recipients #2 (survival: 294 d) and #3 (survival: 301 d), the disease could also be serially transplanted giving rise to a similarly aggressive disease (Fig. S3A-H) as for recipient #1.
A more detailed characterization of surface markers on transformed NKT cells from the serially transplanted disease of recipient #1 verified expression of the T cell receptor (TCR) β chain, while CD4, CD8, and TCRγδ expression was not detected. Leukemic cells also stained positive for CD122 and the NK cell markers NKp46 and DX5. No interaction with the CD1d tetramer was observed (Fig. 2b). This surface marker profile corresponds to previously described CD1d-independent NKT cells [8,9], which were recently identified as a source for CD3 + NK1.1 + T-cell large granular lymphocyte (T-LGL) leukemia in IL-15 transgenic mice [10]. In analogy to our NKT cell leukemia model, human CD56 + T-LGL leukemia cells were also shown to express NKp46 [10]. Additionally, CD3 + CD56 + blasts in human T-LGL leukemia express CD8 [4]. Interestingly, we found CD8 expressed on a subset of hSTAT5B N642H NKT leukemia cells in the two later diseased recipients #2 and #3, but not in recipient #1 (Table S1).
Furthermore, the serially transplanted leukemic NKT cells from recipient #1 were largely negative for activating and inhibitory NK cell receptors, except for Ly49G2 and KLRG1 (Fig. S4A), but stained positive for CD43, CD44, and CD69, while lacking CD62L (Fig. S4A). In summary, this surface receptor expression profile is consistent with an activated phenotype comparable to the leukemic blasts found in IL-15 transgenic mice [11]. The fact that IL-15 transgenic mice develop NK1.1 + T-LGL (NKT) or NK cell leukemia [10,11] supports the importance of the IL-15-STAT5 axis for the transformation of subsets of innate lymphocytes. The occurrence of leukemia with an NKT cell profile in our model suggests that additional IL-15-derived signals, independent of STAT5, may be required for NK cell transformation. In contrast to IL-15 transgenic mice, in which transformation is restricted to innate-like lymphocytes, NKp46 + NKT cell disease required the absence of classical CD8 + T cells bearing mutant hSTAT5B in vivo. As NKp46 + NKT cells represent a minor population [10], the higher number of CD8 + T cells may simply outcompete potentially transformed NKT cells. Alternatively, it is possible that CD8 + T cells actively suppress the development of NKT cell tumors providing a hostile tumor microenvironment. hSTAT5B N642H -driven CD8 + T cell disease is susceptible to JAK1/JAK2 inhibition by Ruxolitinib treatment [7]. As JAK inhibitors have also been suggested as therapeutic strategy for NK/T-cell lymphoma and aggressive NK cell leukemia [6,12], we transplanted NSG mice with hSTAT5B N642H NKT leukemia cells and treated them with ruxolitinib for 3 weeks. Treatment attenuated disease severity and decreased hepatosplenomegaly and WBC counts compared to the control group ( Fig. 2c and Fig. S4B).
Although initial attempts to grow transformed NKT cells in vitro failed, after 6-8 weeks we detected outgrowth of hSTAT5B N642H NKT cell lines from cultured hepatic leukocytes of three out of nine NSG mice from different serial transplant rounds from recipient #1 (Table S2, Fig. S5A). Two cell lines, derived from the same mouse cultured without or with IL-2 (4165_1 and 4165_2, respectively), were treated with Ruxolitinib and remained sensitivity towards it (IC 50 of 83 and 72.5 nM, respectively) (Fig. S5B). Furthermore, these cell lines could give rise to leukemia when injected into immune-competent Ly5.1/ CD45.1 + mice (Fig. S5C-E).
In this study, we identify STAT5B N642H as an oncogenic driver in innate-like lymphocytes. Our novel NKT leukemia Fig. 1 hSTAT5B N642H drives limited NK cell expansion and gives rise to leukemia characterized by expansion of NKT cells. a Absolute numbers (#) of splenic NK cells (CD3 − NK1.1 + NKp46 + ) were determined in WT, hSTAT5B, and hSTAT5B N642H mice by flow cytometry (n = 10 (WT), n = 8 (hSTAT5B), n = 7 (hSTAT5B N642H ) pooled from four independent experiments). Symbols represent results from individual mice, horizontal lines indicate mean ± SEM. * p < 0.05, one-way ANOVA. b CD3-depleted bone marrow (BM) from hSTAT5B or hSTAT5B N642H mice was transplanted into Rag2 −/− γc −/− recipient mice (n = 3) and NK cell numbers were monitored in the blood weekly for 4 weeks by flow cytometry. One representative experiment from two independent experiments is shown. Symbols represent results from individual mice, horizontal lines indicate mean ± SEM. * p < 0.05, unpaired t-test with Welch's correction. c Lineage (Lin) − Sca1 − c-Kit − CD127 − CD8 − cells were sorted from BM of hSTAT5B or hSTAT5B N642H mice and 1 × 10 5 cells were injected into Rag2 −/− γc −/− recipient mice (n = 3). Percentages of NKT cells (CD3 + NK1.1 + ) among living cells in blood, BM, spleen, and liver were analyzed by flow cytometry from the diseased hSTAT5B N642H -transplanted recipient mice (#1-3) (days of survival after transplant are indicated in brackets) and a healthy hSTAT5B-transplanted control mouse. Dot plots are shown. We could not analyze NKT cell frequency in the spleen of mouse #3 (marked by *), due to limited quality of spleen tissue. Analysis was performed when the diseased mice reached the humane endpoint Fig. 2 hSTAT5B N642H induces leukemia of CD1d-independent NKp46 + NKT cells, which is serially transplantable and sensitive to Ruxolitinib treatment. a BM from NKT cell-diseased recipient #1, transplanted with sorted hSTAT5B N642H BM cells, was serially transplanted into Rag2 −/− γc −/− or NSG-recipient mice for six rounds. Whole BM, containing 1 × 10 6 transformed NKT cells, was transplanted. Of note, in the 6th serial transplant (ST) round a titration of the number of transplanted NKT cells was performed, with two mice each receiving BM containing 1 × 10 6 (circle), 0.3 × 10 6 (rectangle), or 0.1 ×10 6 (triangle) NKT cells. Representative images of liver and spleen from an hSTAT5B N642H NKT cell-transplanted (6th ST) compared to a nontransplanted NSG mouse are shown; line denotes 1 cm (left panel). Relative liver and spleen to body weights are shown for the 3rd to 6th ST (n = 3 (3rd ST), n = 4 (4th ST), n = 13 (5th ST), n = 6 (6th ST)) (right panel). Symbols represent results from individual mice, horizontal lines indicate mean ± SEM. *** p < 0.001, unpaired t-test with Welch's correction. b Surface marker expression on transformed hSTAT5B N642H NKT cells was analyzed by flow cytometry. Representative histograms from BM NKT cells of one diseased NSG mouse (4th ST) are shown (unfilled histogram: negative staining control; filled histogram: surface staining). Numbers depict the mean of MFI (median fluorescence intensity, normalized to negative staining control) ± SEM from three diseased mice (4th ST). a, b Analysis was performed when the diseased mice reached the humane endpoint. c NSG mice were transplanted with BM containing 1 × 10 6 hSTAT5B N642H NKT cells (from 5th ST) and treated with Ruxolitinib (RUXO) (85 mg/kg body weight, twice daily) or vehicle control (Nutella ® ) (n = 4 per treatment), starting 1 day after NKT cell transplant for 21 days (one experiment). Representative images of liver and spleen to body weights from untreated control and RUXO-treated mice are depicted; line denotes 1 cm (left panel). Relative liver and spleen weights were analyzed in the mice after 3 weeks of treatment (right panel). Symbols represent results from individual mice, horizontal lines indicate mean ± SEM. ** p < 0.01, *** model, which is serially transplantable and inducible by frozen material or from cell lines, recapitulates human CD56 + LGL leukemia and provides new means for translational research. It will allow a deeper understanding of mechanisms driving NKT leukemia progression and maintenance as well as facilitate finding new therapeutic options. This is of vital importance as the STAT5B N642H mutation in CD56 + LGL leukemia patients is associated with a particularly aggressive chemo-refractory phenotype [4].
Acknowledgments We thank Sabine Fajmann and Philipp Jodl for great technical support and Manuela Kindl for maintaining and monitoring the mice.
Funding The work was supported by the Austrian Science Fund FWF grant SFB-F06105 to RM and SFB-F06107 to VS and FWF grant W1212 to VS.
Author contributions VS initiated and supervised the study. KK, AWS, and BM designed the experiments; KK, AWS, BM, DP, and NL performed the experiments; MPM and TS provided technical support; KK and GH analyzed the data; RM provided the mouse model. KK and VS wrote the manuscript. AWS, BM, DP, TS, and RM revised the manuscript.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.