Introduction

Acute myeloid leukemia (AML) is the second most common blood cancer in children and remains a challenging disease with suboptimal outcomes [1, 2]. AML develops as a result of a series of genetic alterations in a hematopoietic progenitor cell. Recent advances in genetic characterization have improved the understanding of the biology of pediatric AML and led to the development of novel therapeutic strategies. We have previously identified potentially disease-causing genetic alterations in nearly all pediatric AML patients enrolled in the Japanese Pediatric Leukemia/Lymphoma Study Group AML-05 trial. These include two rare NPM1 rearrangements: NPM1::MLF1 and NPM1::CCDC28A [3].

NPM1 is a nucleolar protein with multiple functions in ribosome biogenesis, mRNA processing and chromatin remodeling [4, 5]. Mutations of the NPM1 gene are the most common genetic alteration in AML, detected in approximately 30% of adult AML and 7% of pediatric AML [6, 7]. NPM1 mutations are almost exclusively found in exon 12 and usually consist of 4-bp frameshift insertions that lead to the generation of an additional nuclear export signal (NES) motif at the C-terminus and loss of the nucleolar localization signal (NLS) [8]. Consequently, NPM1-mutated AML is characterized by the aberrant cytoplasmic localization of NPM1 (NPM1c) [9]. In addition to the NPM1 mutations, several NPM1 rearrangements, such as NPM1::MLF1/t(3;5)(q25;q34), NPM1::RARA/t(5;17)(q35;q21), NPM1::HAUS1 and NPM1::CCDC28A, are associated with myeloid neoplasms [10, 11]. Most of these NPM1-fusion proteins also show cytoplasmic relocalization [12]. Since the NPM1c and NPM1-fusion proteins retain the ability to interact with wild-type NPM1, they could induce cytoplasmic relocalization of wild-type NPM1 through formation of heterodimers to disrupt its normal function. Loss of NPM1 nuclear function imposed by the NPM1c or NPM1-fusion proteins has been implicated in myeloid leukemogenesis [13].

Another unique feature of NPM1-mutated AML is the aberrant activation of HOX genes [14, 15]. HOX genes encode a highly conserved family of homeodomain-containing transcription factors that specify cell identity in early development and adult tissues [16]. Aberrant overexpression of HOX genes is associated with a variety of malignancies including AML [17]. Recent studies have shown that both the nuclear export of NPM1c and the upregulation of HOX genes depend on the interaction between NPM1c and the chromatin-bound nuclear exporter XPO1 [18,19,20]. The NPM1c-XPO1 interaction could be disrupted by the selective nuclear export inhibitor selinexor, which covalently binds to XPO1. Preclinical studies with selinexor have shown that XPO1 inhibition induces NPM1c chromatin release and loss of HOX expression, thereby inhibiting the growth of NPM1-mutated AML cells. Thus, while studies have gradually revealed the molecular pathogenesis of NPM1c AML, it remained largely unknown how NPM1-fusion proteins cause AML.

In this study, we examined the leukemogenic potential of the two rare NPM1-fusion proteins, NPM1::MLF1 and NPM1::CCDC28A, that are found in pediatric AML patients. NPM1::MLF1 and NPM1::CCDC28A showed modest and strong leukemogenicity, respectively, in the mouse bone marrow transformation assays. We also found the critical role of XPO1 to sustain HOX gene expression and clonogenicity in the NPM1-fusion AMLs.

Method

Mice

C57BL/6 J mice were purchased from SLC Japan and maintained in a specific pathogen-free environment at the University of Tokyo animal center. All animal experiments were performed according to the institutional guidelines and protocol (PA21-67) approved by the Laboratory Animal Research Center of the Institute of Medical Science at the University of Tokyo.

Patient samples

The pediatric patients with NPM1::MLF1 fusion or NPM1::CCDC28A were reported by Shiba et al. [3]. The patients’ leukemic samples were obtained from the bone marrow at the time of diagnosis. RNA extraction and cDNA synthesis were performed as previously described [3]. NPM1::MLF1 and NPM1::CCDC28A were amplified by polymerase chain reaction (PCR) using the following primers:

NPM1 Fw: GCTCGAGGCCACCATGGACTACAAGGACGACGATGACAAAGAAGATTC

GATGGACATG,

NPM1::MLF1 Rv: GCGCGGCCGCCAGTTATTTTTTGTTGCTTTTCAC,

NPM1::CCDC28A Rv: TGTCAATGCCAAAAAAAATGCCATTCCAGTGAG.

PCR cycling conditions were as follows: 2 min at 94 °C for one cycle, followed by 25 cycles of 10 s at 98 °C, 30 s at 55 °C, 1 min 30 s at 68 °C, followed by a final extension of 10 min at 72 °C.

Plasmids and retrovirus infection

Flag-tagged forms of the NPM1::MLF1 fusion and NPM1::CCDC28A, wild-type NPM1, NPM1c, MLL::ENL were cloned into the pMYs-IRES-GFP retroviral vector. We generated ectopic retrovirus using Plat-E packaging cells [21] with the calcium-phosphate coprecipitation method. The mouse bone marrow cells were incubated with ectopic retrovirus for 48 h using retronectin (Takara Bio Inc., Otsu, Shiga, Japan)).

Western blotting

HEK293T cells were transiently transfected with the indicated plasmids using polyethyleneimine (PEI). Forty-eight hours after transfection, cells were lysed with cell lysis buffer (Cell Signaling Technology, Danvers, MA, USA; #9803). Lysates were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to a polyvinylidene fluoride membrane (Bio-Rad). The blot was incubated with anti-Flag (Sigma-Aldrich, catalog F1804, clone M2, 1:200) and anti-GAPDH (Cell Signaling Technology, catalog #5174, clone D16H11, 1:1000) antibodies. Signals were detected with ECL Western Blotting Substrate (Promega, Madison, WI, USA) and visualized with Amersham Imager 600 (GE Healthcare).

Immunostaining

HEK293T cells and K562 cells were transduced with the indicated plasmids using PEI. Twenty-four hours after transduction, the cells were fixed with 4% paraformaldehyde for 15 min at room temperature. The cells were then permeabilized with 0.2% Triton X-100 for 5 min and blocked with 5% goat serum for 1 h. The HEK293T cells and K562 cells were then stained with anti-flag antibody (Fujifilm, catalog 014-22383), followed by AlexaFluor633-mouse antibody (Invitrogen, catalog A-21052. Cell nuclei were stained with DAPI (BioLegend, catalog 422801). Fluorescence images were captured with an N-SIM super-resolution microscope (Nikon) and were analyzed with NIS-Elements software (Nikon).

Colony replating assay and bone marrow transplantation assay

Eight to ten weeks old mice were treated with 5FU (150 mg/kg, intraperitoneal administration). Bone marrow cells were harvested 4 days after 5FU injection, and were transduced with pMY-IRES-GFP vector, NPM1::MLF1 or NPM1::CCDC28A in RPMI supplemented with 10% fetal bovine serum, mouse SCF 50 ng/μl (R& D, catalog 455-MC), human TPO 50 ng/μl (R& D, catalog 288-TP), mouse FLT3 50 ng/μl (R& D, catalog 427-FL) for 3 days.

For the colony replating assay, 1 × 104 GFP-positive cells were sorted by FACS Aria III (BD Biosciences, San Jose, CA, USA) and plated in M3234 (STEMCELL Technologies) methylcellulose containing 10 ng/mL mouse SCF (R& D, catalog 455-MC), 10 ng/mL mouse IL-3 (R& D, catalog 403-ML), mouse IL-6 (R& D, catalog 406-ML), and 10 ng/μl GM-CSF (R& D, catalog 415-ML). For each round of plating, 1 × 104 cells were plated. Colonies were counted and replated every 7 days. A colony was defined as a cluster of at least 50 cells.

For the bone marrow transplantation assay, 1 × 106 GFP+ cells were injected into 8–12 weeks old C57BL/6 J mice after sublethal irradiation (4.5 Gy). GFP+ leukemia cells were harvested from bone marrows of moribund mice, and were serially transplanted into sublethally irradiated secondary recipient mice by intravenous injection.

Flow cytometry analysis

Bone marrow cells and peripheral blood cells were obtained from leukemic mice. After removal of red blood cells using RBC lysis buffer, the cells were stained with the following antibodies: CD11b-PE-cy7 (BioLegend, 101215), B220-APC-cy7 (BioLegend, 103223), CD3-APC (BioLegend, 100235) and c-kit-PE-Cy7 (BioLegend, 105814). For analysis of mouse lineage- c-Kit+ (LK) cells, the cells were incubated with a cocktail of biotinylated monoclonal antibodies against lineage markers (CD5, B220, CD11b, Gr-1, and Ter119). Cells were then stained with c-kit-PE-Cy7 (BioLegend, 105814), Sca1-APC (BioLegend, 108112), and Streptavidin-Brilliant Violet 605 (BioLegend, 405229) CD34-Alexa flora647 (BioLegend, 119314), CD16/32-PE (BioLegend, 101307). Propidium iodide (PI) and 4’, 6-diamidino-2-phenylindole (DAPI) were used to exclude dead cells. All data were collected using FACS Verse or FACS AriaIII (BD Biosciences, San Jose, CA, USA) and were analyzed with FlowJo software.

Morphological analysis

Cytospin preparations were stained with May-Giemsa staining solution. Images were captured using a BX51 microscope and a DP12 camera (Olympus).

RNA Isolation and RNA sequence

Murine bone marrow cells were transduced with vector, NPM1::MLF1 or NPM1::CCDC28A, and were transplanted into sub-lethally irradiated mice. Ten weeks after transplantation, bone marrow cells were collected from mice and GFP+ cells were sorted by FACS AriaIII (BD Biosciences, San Jose, CA, USA). Total RNA was extracted from GFP+ cells using the RNeasy Mini kit (QIAGEN, 74004), and mRNA was purified from total RNA using poly‐T oligo‐attached magnetic beads. Pair‐end sequencing FASTQ files were aligned to the mouse reference genome (mm10) using HISAT2 30 on the Galaxy platform (https://usegalaxy.org). Raw gene counts were obtained from the read alignments using Subread 31 (v2.4.3) and converted to counts per million (CPM) by edgeR 32 (v3.32.1). After filtering out low‐expressed genes with CPM less than 1, all CPM values were log2 transformed to generate unsupervised clustering dendrograms and heat maps. Differential expression was analyzed with the linear model using limma 33 (v3.46.0). Genes with false discovery rate (FDR) < 0.05 adjusted by the Benjamini‐Hochberg method were considered significant differentially expressed genes (DEGs). For MSigDB gene pathway overlap analysis, DEGs with FDR < 0.05 were examined at http://www.gsea‐msigdb.org/gsea/msigdb/annotate.jsp.

CUT&RUN assay

1 × 105 NPM1::CCDC28A cells were collected per sample to perform the CUT&RUN assay using CUT&RUN kit (Cell Signaling, #86652). The CUT&RUN assay was performed according to the protocol provided by Cell Signaling (https://www.cellsignal.com/learn-and-support/protocols/cut-and-run-protocol) with an anti-flag antibody (FUJIFILM catalog 014-22383) and Rabbit (DA1E) mAb IgG XP® Isotype control antibody (Cell Signaling#66362) as a negative control.

Quantitative RT-PCR

Total RNA was extracted using the RNeasy Mini kit (QIAGEN) and reverse transcribed (RT) using High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems, 4368814). The cDNA was then subjected to quantitative RT-PCR using a SYBR Select Master Mix (Applied Biosystems). The sequences of the 5ʹ to 3ʹ primers used for RT-PCR in this study, from 5ʹ to 3ʹ are as follows:

mouse Hoxa3 Forward CTCATTTTGCGGTCGCTATCC

mouse Hoxa3 Reverse ATCCATGCCATTGTAGCCGTA

mouse Hoxa9 Forward ACAATGCCGAGAATGAGAGC

mouse Hoxa9 Reverse GTTCCAGCGTCTGGTGTTTT

mouse Hoxa10 Forward CAGCCCCTTCAGAAAACAGT

mouse Hoxa10 Reverse TCTTTGCTGTGAGCCAGTTG

mouse Hoxa11 Forward GGCCACACTGAGGACAAGG

mouse Hoxa11 Reverse GAACTCTCGCTCCAGCTCTC

mouse Gapdh Forward TTGATGGCAACAATCTCCAC,

mouse Gapdh Reverse CGTCCCGTAGACAAAATGGT

ChIP-qPCR

Chromatin immunoprecipitation (ChIP) was performed using Simple chip kit (Cell Signaling Technology, #9002) with antibodies against Flag (Sigma-Aldrich, catalog F1804, clone M2, 1:200) according to the manufacturer’s protocol. The purified DNA was then used for qPCR using a SYBR Select Master Mix (Applied Biosystems). The sequences of the 5‘ to 3‘ primers used for ChIP-qPCR in this study are as follows:

mouse Hoxb9 Forward CCACCGACTGGCTTCCTCGC

mouse Hoxb9 Reverse CCAGGGGTCACACCTCCCCA

mouse Hoxa9 Forward AGGAGTCGCTGCTTTCTGTT

mouse Hoxa9 Reverse ATTAGAACGGGGAGGGGTAA

mouse Hoxa10 Forward TAGATGCTTGCAGAAGGAAAGG),

mouse Hoxa10 Reverse CCATATGGCAAGAGGCAAAGA

Dual luciferase Assay

HEK293T cells were seeded in 24-well plates and co-transfected with 560 ng of each plasmid (pMYs-IG vector, NPM1::MLF1, NPM1::CCDC28A, wild-type NPM1, NPM1c or MLL::ENL), 3.8 ng pGL4.74 vector and 190 ng HoxA9 reporter [22] using PEI. Forty-eight hours after transfection, cells were lysed and firefly and renilla luciferase activities were measured with the Dual-Luciferase Assay System (Promega) according to the manufacturer’s protocol on a FLUOstar Optima.

Selinexor treatment

Selinexor (KPT-330) was purchased from Selleckchem (catalog# S7252). For the colony forming assay, 1 × 104 NPM1::MLF1 or NPM1::CCDC28A expressing AML cells were cultured in M3234 (STEMCELL Technologies) methylcellulose containing 10 ng/μl mouse SCF (R& D, catalog 455-MC), 10 ng/μl mouse IL-3(R& D, catalog 403-ML), 10 ng/μl mouse GM-CSF (R& D, catalog 415-ML) together with the indicated concentrations of selinexor (10 nM, 25 nM, 50 nM, and 100 nM). The number of colonies was counted after seven days of treatment with selinexor.

For the luciferase assay, HEK293T cells were seeded in 24-well plates and co-transfected with 560 ng of each plasmid (pMYs-IG vector, NPM1::MLF1, NPM1::CCDC28A, wild-type NPM1, NPM1c), 3.8 ng pGL4.74 vector and 190 ng HoxA9 reporter using PEI. Selinexor (100 nM) was added to some cultures 24 h after transfection. Reporter activity was measured 48 h after transfection.

Cell viability assay

Cells were plated into flat-bottomed 96-well plates at a density of 5 × 104 for NPM1::CCDC28A cells, 1 × 105 for cord blood CD34+ cells, or 1 × 105 for mouse c-kit+ bone marrow cells per well. Cells were incubated for 72 h with Selinexor (Selleckchem catalog#S7252) at 0, 0.025, 0.05, 0.1, 0.25, 0.5, and 1 μM, or VTP50469 (MedChemExpress catalog# HY-114162) at 0, 0.1, 0.3, 1, 3, 10, and 30 μM. The growth inhibitory effect of the drugs was evaluated both individually and in combination. Following the treatment, 8 µL of Cell Counting Kit 8 (Dojindo Molecular Technologies) was added to each well, and the plates were incubated at 37 °C for an additional 1 h. The absorbance of each sample was then measured using a CLARIOstar Plus Microplate Reader (BMG LABTECH) at 450 nm.

Statistical analysis

GraphPad Prism 9 was used for statistical analysis. Two-tailed unpaired t test was used for pairwise comparisons, and one-way ANOVA was used for multiple comparisons of significance. The p value (<0.05, <0.01, <0.001) is indicated by one to three asterisks (*, **, ***). p values higher than 0.05 were considered not significant (ns).

Ethics approval and consent to participate

The present study was conducted in accordance with the Declaration of Helsinki and approved by the institutional review boards of Gunma Children’s Medical Center and the participating institutes and the ethical review board of the JPLSG AML-05 trial. Written informed consent was provided by all patients or their parents/guardians.

Results

NPM1 fusion proteins localize to the cytoplasm and promote myeloid leukemogenesis

We first generated full-length NPM1::MLF1 and NPM1::CCDC28A cDNAs with N-terminal Flag-tag using mRNAs isolated from AML patients harboring NPM1::MLF1 and NPM1::CCDC28A. Characteristics of these patients and structures of NPM1-related proteins are shown in Table 1 and Fig. 1A, respectively. The expression of these NPM1-fusions in HEK293T cells was confirmed by Western blotting (Fig. 1B). We then examined the cellular localization of wild-type NPM1, NPM1c, NPM1::MLF1 and NPM1::CCDC28A in a transient transfection assay in HEK293T and K562 cells. As expected, wild-type NPM1 localized exclusively to the nucleolus, whereas NPM1c localized mainly to the cytoplasm in HEK293T cells and to both the nucleus and cytoplasm in K562 cells. NPM1::MLF1 was present in both the nucleus and cytoplasm, and NPM1::CCDC28A localized more to the cytoplasm in HEK293T and K562 cells. (Fig. 1C, Supplementary Figs. 1A, B). These data well agree with the previous reports [12, 23] and suggest that cytoplasmic mislocalization is a key feature of the NPM1-fusion proteins.

Table 1 Patient characteristics.
Full size table
Fig. 1: Cytoplasmic localization of the NPM1-fusion proteins.
figure 1

NES nuclear export signal. AS acidic domains. NLS nuclear localization signal. A Schematic presentation of wild-type NPM1, NPM1::MLF1, NPM1::CCDC28A. Numbers indicate positions of amino acid residues from the N-terminus. B HEK293T cells were transduced with the flag-tagged NPM1::MLF1 and NPM1::CCDC28A. The cell lysates were stained with anti-Flag antibody. C K562 cells were transduced with vector, flag-tagged wild-type NPM1, NPM1c, NPM1::MLF1 or NPM1::CCDC28A, and were then stained with anti-Flag. Cell Nuclei were visualized with DAPI. Scale bar, 20 μm. WT wild-type.

Full size image

Next, we evaluated the leukemogenic activity of the NPM1-fusions. We transduced the vector, NPM1::MLF1 or NPM1::CCDC28A into primary mouse bone marrow cells and cultured the cells in semisolid media. Vector and NPM1::MLF1-transduced cells did not produce colonies beyond the second passage, whereas NPM1::CCDC28A-transduced cells acquired extensive serial replating capacity (Fig. 2A). Furthermore, NPM1::CCDC28A-expressing colonies contained many myeloblasts at the third passage, whereas vector and NPM1::MLF1-transduced colonies consisted mainly of differentiated macrophages (Fig. 2B). Thus, NPM1::CCDC28A has strong activity to enhance self-renewal of mouse bone marrow progenitors in vitro.

Fig. 2: NPM1::MLF1 and NPM1::CCDC28A show leukemogenic activity.
figure 2

A Colony numbers of vector, NPM1::MLF1 or NPM1::CCDC28A-transduced mouse bone marrow c-Kit+ cells. Weekly colony count per 1 × 104 replated cells are shown. Data are shown as mean ± SEM. B Wright-Giemsa staining of cells at the third round. All vector or NPM1::MLF1-transduced cells were differentiated macrophages or mastocytes, whereas NPM1::CCDC28A-transduced colonies contained myeloblasts. Scale bar, 20 μm. C Frequency of GFP+ cells in peripheral blood of all recipient mice bearing cells expressing vector (n = 6), NPM1::MLF1 (n = 12) or NPM1::CCDC28A (n = 6). PB peripheral blood. D White blood cells, hemoglobin, platelet and mean corpuscular volume levels of the recipient mice 6, 10 or 14 weeks after transplantation are shown. Data from all recipient mice with vector-transduced cells, and mice with NPM1::MLF1 or NPM1::CCDC28A-expressing cells in which the GFP+ cells increased after transplantation were used (vector, n = 6, NPM1::MLF1, n = 2, NPM1::CCDC28A, n = 6). Data are shown as mean ± SEM. One-way ANOVA was used for multiple comparisons. WBC white blood cells, Hb hemoglobin, Plt platelet, MCV mean corpuscular volume. E Kaplan-Meier survival curve of mice transplanted with bone marrow progenitor cells transduced with NPM1::MLF1 (n = 12), NPM1::CCDC28A (n = 6), or the vector (n = 6).

Full size image

We then transplanted bone marrow progenitor cells transduced with the retroviruses vector harboring NPM1::MLF1 or NPM1::CCDC28A co-expressing GFP, into sub-lethally irradiated recipient mice. The NPM1::MLF1 or NPM1::CCDC28A-transduced GFP+ cells gradually increased in the peripheral blood of some or all of the recipient mice, respectively (Fig. 2C). At 14 weeks after transplantation, the mice bearing NPM1::MLF1 or NPM1::CCDC28A-expressing cells showed lower hemoglobin and higher MCV compared to the control mice (Fig. 2D). Importantly, five out of six mice transplanted with NPM1::CCDC28A-expressing cells and two out of twelve mice transplanted with NPM1::MLF1-expressing cells developed leukemia within 6 months after transplantation (Fig. 2E). Their bone marrow was dominated by GFP+CD11b+ myeloblasts (Fig. 3A), indicating that both NPM1::MLF1 and NPM1::CCDC28A produced AML. NPM1::CCDC28A-expressing cells were enriched in a ckit+ granulocytic-monocytic progenitor (GMP)-like population known to contain leukemia stem cells, whereas NPM1::MLF1-expressing cells produced diverse cell types, including Gr1+ mature myeloid cells (Fig. 3B, C). Finally, we transplanted the NPM1::MLF1- or NPM1::CCDC28A-expressing AML cells into secondary recipient mice. In these serial transplantation assays, all recipient mice rapidly developed AML within 2 months after transplantation (Fig. 3D).

Fig. 3: NPM1::MLF1 and NPM1::CCDC28A induce the development of transplantable AML in vivo.
figure 3

A–C Bone marrow cells were collected from the moribund mice bearing NPM1::MLF1 or NPM1::CCDC28A-expressing cells. BM bone marrow. A Representative flow cytometry profiles (right) and a bar graph showing the frequency of GFP+CD11b+ cells (left). Data are shown as means ± SEM. (NPM1::MLF1: n = 2, NPM1::CCDC28A: n = 6). B Representative flow cytometry profiles (right) and a bar graph showing the frequency of GFP+c-kit+ cells and GFP+ granulocytic-monocytic progenitors are shown. Data are shown as means ± SEM. Numbers indicate the frequency of each population. GMP granulocytic-monocytic progenitors. CMP common myeloid progenitors. MEP megakaryocyte/erythrocyte progenitors. C Wright-Giemsa staining of bone marrow cells collected from the recipient mice bearing NPM1::MLF1 or NPM1::CCDC28A-expressing cells. Scale bar, 20 μm. D Kaplan-Meier survival curve of secondary recipient mice transplanted with the NPM1::MLF1 or NPM1::CCDC28A-expressing cells (NPM1::MLF1, n = 6, NPM1::CCDC28A, n = 6).

Full size image

Taken together, we concluded that both NPM1::MLF1 and NPM1::CCDC28A have leukemogenic activities. NPM1::CCDC28A efficiently drives AML development with a strong transforming capacity. NPM1::MLF1 is weakly leukemogenic and likely requires additional mutations for the development of AML.

NPM1-fusions induced upregulation of HOX genes and inflammation-associated genes in vivo

To examine the molecular changes induced by NPM1 fusions, we first examined the gene expression profiles of cells expressing vector, NPM1::MLF1 or NPM1::CCDC28A. We transplanted bone marrow cells transduced with vector, NPM1::MLF1 or NPM1::CCDC28A into recipient mice, and harvested GFP+ bone marrow cells two months after transplantation. None of the mice developed AML at this time point. NPM1::MLF1 and NPM1::CCDC28A-expressing cells showed distinct gene expression profiles (Fig. 4A, Supplementary Table 1), but they also shared common features, including upregulation of HOX genes and inflammation-related genes (Fig. 4B, C, Supplementary Fig. 2A). We then compared the gene expression signatures of these NPM1::MLF1 and NPM1::CCDC28C cells with those of other mouse AML cells driven by NPM1c, MLL::AF9, and RUNX1::RUNXT1 [24,25,26]. The expression profiles of NPM1::MLF1 and NPM1::CCDC28C cells were most similar to that of NPM1c, then to MLL::AF9, and were most different from that of RUNX1::RUNXT1 (Supplementary Fig. 2B). CD34 expression was low in NPM1::MLF1 and NPM1::CCDC28A cells compared to other mouse AML cells transformed by MLL::AF9 or SETBP1/ASXL1 mutations (which we refer to as cSAM cells), similar to that observed in NPM1c-driven AML (Supplementary Fig. 2C). In addition, we found that both NPM1::MLF1 and NPM1::CCDC28A upregulated genes related to apoptosis and p53 pathway, which is likely associated with a protective proapoptotic response in hematopoietic progenitor cells (Supplementary Fig. 2D).

Fig. 4: NPM1::MLF1 and NPM1::CCDC28A induced specific and shared gene expression signatures.
figure 4

A Mouse bone marrow c-kit+ cells were transduced with vector, NPM1::MLF1 or NPM1::CCDC28A and were transplanted into recipient mice. GFP+ bone marrow cells were harvested two months after transplantation to evaluate their gene expression profiles. Principal component analysis of GFP+ cells expressing vector, NPM1::MLF1 or NPM1::CCDC28A is shown. FC fold change. B Venn diagram showing upregulated (upper panel) or downregulated (lower panel) genes in cells expressing NPM1::MLF1 or NPM1::CCDC28A. C A volcano plot showing of up- or down-regulated genes in NPM1::MLF1 or NPM1::CCDC28A-expressing cells compared to vector-transduced cells. Each dot represents one gene. D Venn diagram showing the overlapping genes between CUT & RUN signal (FDR > 1) and upregulated genes from RNA-seq in cells expressing NPM1::CCDC28A. E Integrative Genomic Viewer screenshots of FLAG-NPM1::CCDC28A and IgG control CUT&RUN signals around HOX gene clusters.

Full size image

To identify genes that are directly regulated by NPM1-fusions, we then performed CUT & RUN analysis using NPM1::CCDC28A cells, which revealed NPM1::CCDC28A specific peaks around 4725 genes (FDR < 0.1). We integrated CUT & RUN and RNA seq analysis and identified 677 genes that are directly bound and upregulated by NPM1::CCDC28A, including HOX genes (Fig. 4D, E, Supplementary Table 2). Thus, NPM1::MLF1 and NPM1::CCDC28A induced specific and shared gene expression signatures during the process of myeloid transformation. Importantly, both NPM1-fusions induced HOX gene upregulation like NPM1c, suggesting that they have similar functions to NPM1c.

Selinexor inhibits colony formation and HOX gene activation induced by the NPM1-fusions

Previous studies have shown the critical role of XPO1 in promoting NPM1c-induced leukemogenesis by either promoting nuclear export [19] or chromatin binding to HOX clusters of NPM1c [18]. We therefore examined the effect of an XPO1 inhibitor selinexor on the colony forming activity of AML cells expressing NPM1::MLF1 and NPM1::CCDC28A. Selinexor substantially suppressed the colony formation of NPM1::MLF1 and NPM1::CCDC28A-expressing cells even at low concentrations (50–100 nM), whereas it had only a marginal effect on that of normal bone marrow progenitors (Supplementary Fig. 3A, Fig. 5A, B). In addition, selinexor (100 nM) induced downregulation of HOX genes (HOXA3, HOXA9, HOXA10 and HOXA11) in NPM1::MLF1 and NPM1::CCDC28A expressing cells (Fig. 5C, D). These data suggest that NPM1-fusions, like NPM1c, require XPO1 to upregulate HOX genes and maintain the leukemic state.

Fig. 5: Selinexor inhibits colony formation and HOX gene activation induced by the NPM1-fusions.
figure 5

A Colony numbers of NPM1::MLF1-induced leukemia treated with either 100 nM selinexor or DMSO for 7 days. Error bars indicate the standard error (SE). Two-tailed unpaired t test was used for pairwise comparisons. B Colony numbers of NPM1::CCDC28A-induced leukemia treated with either 100 nM selinexor or DMSO for 7 days. Error bars indicate the standard error (SE). Two-tailed unpaired t test was used for pairwise comparisons. C, D HOXA gene levels in cells expressing NPM1::MLF1 (C) or NPM1::CCDC28A (D) treated with/without 100 nM selinexor for 7 days. Data are shown as means ± SEM. Two-tailed unpaired t test was used for pairwise comparisons.

Full size image

We then assessed the effect of selinexor on the subcellular localization of NPM1::MLF1 and NPM1::CCDC28. As expected, selinexor treatment suppressed nuclear export, resulting in nuclear retention of both NPM1-fusion proteins in 293 T cells (Fig. 6A, B). Since recent studies have reported the novel role of XPO1 in recruiting NPM1c to HOX clusters, we next investigated whether NPM1-fusions also bind to the HOX clusters in cooperation with XPO1. We performed chromatin immunoprecipitation quantitative PCR (ChIP-qPCR) using 293 T cells transfected with vector, NPM1::MLF1 or NPM1::CCDC28A in the presence or absence of selinexor. Selinexor significantly reduced the binding of NPM1::CCDC28A to the HOX clusters, and tended to reduce the binding between NPM1::MLF1 and the HOX clusters (Fig. 6C, D). To further investigate the association between NPM1-fusion proteins and HOX genes, we examined the transcriptional activity of NPM1::MLF1, NPM1::CCDC28A, NPM1c and wild-type NPM1 using the reporter construct containing the HOXA9 promoter [22]. MLL::ENL, an oncogenic fusion gene known to upregulate HOX genes through direct binding to their promoter, was also used as a positive control [27]. The NPM1-fusion proteins as well as NPM1c significantly increased the luciferase activity of the HOXA9 reporter as efficiently as MLL::ENL, whereas wild-type NPM1 showed only a modest activation of the HOXA9 promoter. Importantly, the HOX activation induced by the NPM1-fusions was suppressed by the selinexor treatment (Fig. 6E). Taken together, we concluded that NPM1::MLF1 and NPM1::CCDC28A cooperate with XPO1 to bind directly to the HOX gene cluster, leading to the upregulation of HOX genes.

Fig. 6: Selinexor inhibits cytoplasmic localization and chromatin binding of NPM1-fusions.
figure 6

293 T cells were transfected with Flag-tagged NPM1::MLF1 (A) or Flag-tagged NPM1::CCDC28A (B) and were treated with vehicle (DMSO) or Selinexor (100 nM) for 24 h. Cells were stained with anti-Flag antibody. Cell nuclei were visualized with DAPI. Representative images (left) are shown. Scale bar, 10 μm. Subcellular localization of NPM1::MLF1 and NPM1::CCDC28A was evaluated in 100 cells (right). Mouse AML cells expressing NPM1::MLF1 (C) or NPM1::CCDC28A (D) were treated with 100 nM selinexor for 7 days. Binding of NPM1::MLF1 (C) and NPM1::CCDC28A (D) on the HOX gene cluster was assessed by ChIP-qPCR. Data are shown as mean ± SEM from triplicate wells. Two-tailed unpaired t test was used for pairwise comparisons. E 293 T cells were cotransfected with vector, MLL::ENL, NPM1::MLF1, NPM1::CCDC28A, wild-type NPM1 or NPM1c together with pGL4.74 vector and HoxA9 reporter. Selinexor was added 24 h after transfection (right), and the HOXA9 activation was measured 48 h after transfection using the Dual-Luciferase Reporter Assay System. Data are shown as means ± SEM from triplicate wells. One-way ANOVA was used for multiple comparisons.

Full size image

Co-targeting XPO1 and menin efficiently inhibited the growth of NPM1:CCDC28A cells

Recent clinical studies have shown that NPM1c AMLs are also sensitive to menin inhibitors [28]. Therefore, we next evaluated the dose-dependent effect of selinexor and a menin inhibitor (VTP50469) on NPM1::CCDC28A cells. VTP50469 has been shown to be effective against mouse menin [29]. Both selinexor and VTP50469 inhibited the growth of NPM1::CCDC28A cells in a dose-dependent manner, and NPM1::CCDC28A cells were more sensitive to these drugs than normal cKit+ bone marrow cells (Fig. 7A). Finally, we examined the combined effect of XPO1 and menin inhibition on the growth of NPM1::CCDC28A cells. Selinexor and VTP50469 cooperatively suppressed the growth of NPM1::CCDC28A cells, whereas they showed only a marginal effect on mouse cKit+ bone marrow cells and human cord blood CD34+ cells (Fig. 7B, C). These results suggest that co-targeting XPO1/menin is a promising therapeutic strategy for NPM1-fusion AMLs.

Fig. 7: Dose-dependent effects of XPO1 and menin inhibitors on NPM1::CCDC28A cells.
figure 7

A NPM1::CCDC28A cells and mouse cKit+ bone marrow cells were incubated with selinexor or VTP50469 at the indicated concentration for 72 h. Cell viability was assessed using the Cell Counting Kit-8. Data are shown as means ± standard deviation (SD) from three technical replicates. NPM1::CCDC28A cells and mouse cKit+ bone marrow cells (B) or human cord blood CD34+ cells (C) were incubated with selinexor + VTP50469 at the indicated concentrations for 72 h. Cell viability was assessed using the Cell Counting Kit-8. Data are shown as means ± standard deviation (SD) from three technical replicates.

Full size image

Discussion

In this study, we demonstrated that NPM1::MLF1 and NPM1::CCDC28A can induce transplantable AML in the mouse transplantation assay, providing experimental evidence that the two novel NPM1-fusions have acquired an oncogenic potential. Consistent with the previous reports [30], NPM1::MLF1 alone was not sufficient to increase the colony-forming ability of mouse bone marrow cells in vitro and to induce AML in most recipient mice, suggesting that additional genetic alterations are required for the development of AML. Indeed, at least one pathogenic mutation (WT1, FLT3-ITD, IDH2 or PTPN11 mutation) has been identified in all the reported myeloid tumors harboring NPM1::MLF1 [10]. On the other hand, NPM1::CCDC28A has a stronger leukemogenic activity with the ability to immortalize mouse bone marrow cells in vitro. Since CCDC28A also fuses with NUP98 to form the NUP98::CCDC28A fusion gene in T-cell leukemia [31], it is likely that the oncogenic activity of CCDC28A enhances the leukemic activity of NPM1::CCDC28A.

Mechanistically, both NPM1::MLF1 and NPM1::CCDC28A exhibit cytoplasmic localization and induce HOX gene activation with the help of XPO1, which are very similar to the known functions of NPM1c [12, 19]. Thus, our study provides the first experimental evidence that NPM1-fusions and NPM1c share many features to promote myeloid leukemogenesis. In addition, we showed that selinexor, a selective inhibitor of XPO1 that has been used as a pan-cancer agent in preclinical and phase I clinical trials [32, 33], effectively inhibits the growth of AML cells harboring the NPM1-fusions, as it does in NPM1c AML. How the cytoplasmic NPM1 induces HOX gene activation in cooperation with XPO1 and how the XPO1 inhibitor selinexor suppresses leukemogenesis are currently under debate. Since XPO1 is a major transport receptor responsible for the export of proteins from the nucleus to the cytoplasm [34, 35], nuclear relocalization of NPM1c was considered to be a major mechanism of action of selinexor [19]. Indeed, we found that selinexor treatment promoted the nuclear retention of NPM1::MLF1 and NPM1::CCDC28A in 293 T cells. However, this theory could not explain how the cytoplasmic NPM1 located outside the nucleus could regulate the HOX gene expression. Instead, recent studies have proposed a novel role for XPO1 in recruiting NPM1c to the HOXA cluster region, leading to increased transcription of HOX genes [18, 20]. Indeed, we found that selinexor inhibited the binding of the NPM1-fusions to the HOX cluster and suppressed the activation of the HOXA9 reporter driven by the NPM1-fusions. Thus, our data suggest that XPO1 is involved in both nuclear export and chromatin binding of the NPM1 fusion proteins. Given the strong activity of NPM1::MLF1 and NPM1::CCDC28A to directly enhance the HOXA9 reporter activity, the XPO1-mediated binding of NPM1 fusion proteins to the HOX gene cluster is likely to be critical for NPM1 fusion-driven myeloid leukemogenesis.

Menin inhibitors are promising new agents currently under clinical development that specifically target the HOX transcriptional program, which plays a critical role in the development of AML with MLL (KMT2A) rearrangements (MLL-r) or NPM1c. Menin inhibitors have been shown to disrupt the menin-MLL complex, induce downregulation of HOX genes, and promote differentiation and apoptosis of MLL-r and NPM1c AML cells. Given the potent inhibitory effect of a menin inhibitor, VTP50469, on the growth of NPM1::CCDC28A cells, it is likely that NPM1-fusion AMLs also depend on the menin-MLL complex to maintain the leukemogenic HOX transcriptional program. In addition, VTP50496 and Selinexor cooperatively suppress the growth of NPM1::CCDC28A cells, providing a rational basis for clinical testing of the XPO1/menin co-targeting approach for NPM1-fusion AML.

In summary, we demonstrated the leukemogenic activity of two novel NPM1 fusion genes, NPM1::MLF1 and NPM1::CCDC28A. Both NPM1 fusions promote myeloid leukemogenesis through HOX activation with the help of XPO1. The mouse models of NPM1-fusion leukemia established in this study will be useful tools for understanding the pathogenesis of AML with NPM1 abnormalities. Our study also highlights XPO1 and menin as promising targets for the treatment of NPM1-fusion AML.