Patient data
The patient, a 66-year-old man, presented with perianal pain, fatigue, and blood abnormalities. On physical examination, he displayed pale skin, but no evidence of hepatosplenomegaly. The patient’s initial full blood count revealed that hemoglobin 71 g/L, platelets 29×109/L, white blood cell counts 26.23×109/L (differential: neutrophils 22.3%, eosinophils 0.1%, basophils 0.1%, blasts 77%). Bone marrow (BM) smear showed hypercellularity with approximately 83% myeloblasts, partially with Auer body (Fig. 1A). No signs of myelodysplasia and eosinophilia were observed. Flow cytometry (FCM) on whole lysed BM revealed a 79% blast population that showed bright expression of CD13, CD33, CD117, CD123, with partial expression of CD64 and myeloperoxidase (MPO), but without CD34, CD38, CD56, human leukocyte antigen D related (HLA-DR) and terminal deoxynucleotidyl transferase (TdT). Based on the morphology and immunophenotype, the patient was diagnosed with AML.
Due to the older age and pulmonary infection, he received a lower intensity induction therapy comprising of azacytidine and HA (homoharringtonine, cytarabine). After two cycles of induction therapy, the patient was assessed as having complete response (CR) with negative minimal residual disease (MRD) in FCM. Subsequently, the patient received two courses of sequential consolidation chemotherapy with intermediate-dose cytarabine (IDAC) regimen and imatinib as maintenance therapy. However, after 15 months from diagnosis, the patient suffered a relapse with expansion of the NPM1 ancestral clone and high allelic ratio (AR) FLT3-ITD mutation. He then was administered venetoclax-azacitidine combined with sorafenib as re-induction therapy, which resulted in CR. The patient is still on regular maintenance therapy and follow-up for ongoing monitoring of his condition.
Identification of t(5;14)(q32;q12) translocation and STRN3::PDGFRB fusion transcript in BM cells
The BM cells were analyzed for chromosomal abnormalities, revealing a clonal translocation t(5;14)(q32;q12) in 2 out of 20 metaphases prior to the second induction therapy (Fig. 1B). FISH studies found the PDGFRB rearrangement in 15% of cells examined (Fig. 1C), which was above the laboratory-determined cutoff for a separated signal of 3.37%.
Further analysis through RNA-seq identified the partner of PDGFRB rearrangement as STRN3 (Fig. 1D). To further confirm the presence of STRN3::PDGFRB fusion transcript, RT-PCR was performed using STRN3- and PDGFRB-specific primers spanning the breakpoint, which yielded a predicted 371-bp PCR product from the patient’s initial BMMNC cDNA (Fig. 2A, lane 1 and 2). Full-length STRN3::PDGFRB was amplified using overlap PCR (Fig. 2B), cloned into a blunt TOPO vector and sequenced, which revealed a 2730-bp in-frame fusion transcript that spans from the start codon to exon 7 (988 nt) of STRN3 cDNA (reference NM_001083893.2) fused to PDGFRB cDNA (NM_002609.4) from exon 11 to stop codon. Chromatograms from Sanger sequencing, showing the amplicon from 371-bp PCR product incorporating the STRN3::PDGFRB fusion site, confirmed the fusion sequence identified by RNA sequencing (Fig. 2C). The fusion protein, predicted to encode a 909-aa fusion protein, contains a coiled-coil domain (77–136 aa), a caveolin-binding site (71–79 aa) and a calmodulin-binding domain (166–183 aa) in the STRN3 portion, and a spilt tyrosine domain (403–765 aa) in the PDGFRB portion (Fig. 2D). The reciprocal PDGFRB::STRN3 mRNA was not detected by RT-PCR in our study (Fig. 2A, lane 3 and 4).
Subclonal acquisition of STRN3::PDGFRB fusion in a NPM1-mut AML
The leukemic blasts exhibited morphologically and immunophenotypically homogeneous monocytic characteristics and the FISH showed 15% of examined cells with PDGFRB rearrangement. Therefore, we suspected that STRN3::PDGFRB clone might be a subclone that emerged as a secondary event during the disease development rather than as the founding clone of AML.
Using WGS and TES, e identified several structural alterations and sequence mutations at various time points (Fig. 3A), including the DNMT3A (L905P) mutation which was present throughout the disease course and may contribute to clonal hematopoiesis. At diagnosis, we detected tow distinct subclones with STRN3::PDGFRB and FLT3-ITD mutations, which emerged after IDH2 (R140Q) and NPM1(c.863–864 insCCTG) mutations (Fig. 3B). After conventional chemotherapy and imatinib maintenance therapy, the STRN3::PDGFRB subclone disappeared, and the FLT3-ITD clone became the dominant clone at relapse (Fig. 3B). However, the FLT3-ITD (c.1751_1807dup) mutation clone at relapse was distinct from the FLT3-ITD (c.1754_1837dup) clone at diagnosis. In contrast, the IDH2 and NPM1 mutations remained stable with persistence of the diagnosis mutations at relapse. Additionally, two distinct subclones with NUP50 or ZPBP2 mutations arose from the dominant clone with IDH2/NPM1/FLT3-ITD mutations (Fig. 3B). Our findings were further supported by data obtained from ddPCR and real-time quantitative PCR (RT-qPCR), which were used to track the novel fusion gene clone and the major clone, and by examining the frequency of hotspot mutations and genomic lesions in 8 longitudinal samples obtained between diagnosis and relapse (Fig. 3C).
Oncogenic properties of STRN3::PDGFRB fusion gene and its transforming capabilities in vitro
To investigate the transforming capabilities of STRN3::PDGFRB, we designed a bicistronic MSCV-based retroviral plasmid containing STRN3::PDGFRB and eGFP, along with truncated mutation variants lacking different domains of the STRN3 portion of fusion protein (Fig. 4A). We compared these constructs to an empty vector carrying only eGFP and the previously described fusion oncogene ETV6::PDGFRB. The IL-3 dependent murine hematopoietic cell line Ba/F3 and 32D were infected with these retroviral vectors, and after IL-3 removal, cells infected with the MSCV empty vector or the coiled-coil domain truncated mutant died (Fig. 4C and 4D). In contrast, cells infected with either STRN3::PDGFRB or ETV6::PDGFRB were IL-3 independent (Fig. 4C and 4D). The proliferation rate of STRN3::PDGFRB transduced cells were faster than that of ETV6::PDGFRB transduced cells without IL-3. When IL-3 was removed, the caveolin-binding site truncated mutant transduced cells proliferated slower than STRN3::PDGFRB transduced cells (Fig. 4C and 4D). The caveolin-binding site truncation reduced the transforming capabilities of the novel fusion oncoprotein but not as significantly as the coiled-coil domain, whereas the calmodulin-binding site truncation has less effect on the fusion protein. The expression of the fusion genes was confirmed by Western blotting (Fig. 4C and 4D). The IL-3 independence of these cells suggested the novel PDGFRB fusion oncogene has transforming properties.
All previously reported PDGFRB fusion proteins have shared the ability to self-associate and form homodimers. Similarly, Strin3 forms homodimers through its coiled-coil domain6. To examine whether STRN3::PDGFRB can form homodimers, Co-IP assays were performed in 293T cells. The Flag-tagged STRN3::PDGFRB could coimmunoprecipitate with Myc-tagged STRN3::PDGFRB (Fig. 4B), indicating that STRN3::PDGFRB can self-associate and form homodimers. In Ba/F3 cells overexpressing the fusion gene, both STRN3::PDGFRB and ETV6::PDGFRB displays predominantly cytoplasmic localization (Fig. 4E). The coiled-coil domain truncated mutant also confirmed the importance of the oligomerization domain, suggesting that the novel fusion protein transforms cell lines through oligomerization and constitutive activation of the tyrosine kinase domain, similar to other tyrosine kinase fusion oncogenes.
STRN3::PDGFRB induced a murine MPN/MDS-like disease in founder and caused a T lymphoblastic lymphoma phenotype in sublethally irradiated secondary recipients
In vitro results suggested that STRN3::PDGFRB fusion gene acted as an oncogene to drive malignant transformation. We introduced the novel fusion gene, as well as empty vector, into murine c-kit BM cells using retroviral transduction, and transplanted these cells into lethally irradiated C57BL/6 mice. During a 9-month observational period, all mice in the STRN3::PDGFRB group developed MPN/MDS-like disease, whereas none of the control group developed leukemia (Fig. 5A). The diseased mice exhibited leukocytosis (WBCs mean 37.71; range, 7.95–90.43×10^9/L; n = 8), anemia (HGBs varying between 25–130 g/L, mean 75 g/L), thrombopenia (PLTs varying between 41–668×10^9/L, mean 168×10^9/L), with peripheral blood (PB), BM, spleen FCM revealing a predominance of mature neutrophils and abnormal granules, similar to MPN/MDS patient. Autopsies on the disease mice showed marked splenomegaly (Fig. 5B), with leukemia cells infiltration into spleen, liver, and lungs (Fig. 5C). The normal splenic architecture was disrupted by foci of mature neutrophils (Fig. 5C). All funder tissues analyzed contained a dominant population of B220−, CD3−, CD11b+, Gr-1+, Ter119− cells (Fig. 5D), consistent with chronic myeloid leukemia (CML).
To further understand the disease process, cells from the spleen were transplanted into sublethally irradiated secondary recipients and monitored weekly. We found that the Gr-1/ CD11b positive cells did not have a clonal advantage and disappeared gradually after transplant, but CD3/CD4/CD8 positive T cells gradually become dominant clone (Supplementary Fig. 1). These secondary recipients developed T cell lymphoblastic lymphoma with a latency period of 4 to 6 weeks. In previous CML and CMML models14,15, the disease could be serially passaged up to two rounds and in all cases of serial passage, the myeloproliferative disease transforms into a T-cell lymphomas. The secondary receipt mice were killed when symptomatic, and all lymph node groups were affected, as well as periportal tissue in the kidney, diffuse involvement of the spleen and liver, and of BM with lymphoblastic lymphoma (Fig. 6A, 6C). The Wright-Giemsa–stained PB smear showed tumor with features of lymphoblastic leukemia, including intermediate-sized lymphoid cells with scant cytoplasm, and dispersed nuclear chromatin, (Fig. 6B). Western blot of spleen cells from different mice confirmed the expression of the STRN3::PDGFRB fusion protein (Supplementary Fig. 2). FCM from BM, spleen, blood, and lymph nodes showed that the leukemia cells are CD3+, CD4+, CD8+, CD11b−, Gr-1− cells (Fig. 6D).
Ba/F3 cells with STRN3::PDGFRB were sensitive to TKIs and XPO1 inhibitor, however, only imtinib alone could improve survival of STRN3::PDGFRB tertiary recipient mice
Ba/F3 cells expressing the STRN3::PDGFRB and ETV6::PDGFRB were significantly sensitive to TKIs including imatinib, dasatinib, ponatinib, olverembatinib dimesylate, and XPO1 inhibitor selinexor (Fig. 7A-E). The efficacy of combination of imatinib and selinexor was analyzed using an interactive analysis of multidrug combination profiling with the ZIP model, which showed a marked synergistic effect in STRN3::PDGFRB and ETV6::PDGFRB transduced Ba/F3 cells (Fig. 7F-H). The highest ZIP synergy scores were observed at concentrations of 10–50 nM, indicating that even low concentrations of imatinib and selinexor are effective in combination.
Previous studies have shown that imatinib can improve the survival of mice with ETV6::PDGFRB-induced malignancies16,17. To evaluate the effect of imatinib on this model, S7A4’s spleen cells were introduced into syngeneic recipients (Supplementary Fig. 3). The tertiary recipients were divided into groups that received daily doses of vehicle, selinexor (15mg/kg, biw) alone, imatinib (100 mg/kg, qd) alone, or a combination of imatinib (75mg/kg, qd) and selinexor (10mg/kg, biw). Mice treated with imatinib alone had significantly prolonged survival over other groups, demonstrating that imatinib can inhibit the growth of tumor cells expressing STRN3::PDGFRB even after the transformation and is safe (Fig. 7J). Treatment of imatinib alone or combination can obviously reduce leukemia burden in the spleen (Fig. 7I). However, treatment of tertiary recipients with selinexor (10mg/kg twice per week or 15mg/kg twice per week) was ineffective in prolonging survival (Fig. 7J) and caused serious gastrointestinal side effects. All mice treated with selinexor were emaciated (Mean weight: 22g to 16g), many of them appeared abdominal distension and anorexia, indicating intestinal obstruction (Supplementary Fig. 4). The serious gastrointestinal side effects also significantly affected imatinib absorption, resulting in no increase in survival when the combination group compared with control group (Fig. 7J).