Background

Acute myeloid leukemia (AML) is characterized by accumulation of aberrantly differentiated hematopoietic myeloid progenitor cells [1]. Chromosomal translocations involving the Nucleoporin 98 gene (NUP98) result in > 30 distinct gene fusions found in different hematological malignancies including AML [2, 3]. Functional studies have shown that NUP98 fusion proteins promote myeloid progenitor self-renewal and prevent their differentiation through mechanisms that are heavily influenced by the fusion partner [2, 3]. Moreover, several recent studies suggested that transformation of myeloid progenitors directly depends on phase transition capacity of NUP98 fusion proteins, mediated by the disordered N-terminal NUP98 moiety which is shared among all fusions [4,5,6].

The karyoty**-silent NUP98-NSD1 fusion is a molecular hallmark of pediatric AML present with high white blood cell counts, and is generally associated with additional mutations of which the activating FLT3-ITD mutation is found in > 70% of the cases [7, 8]. The NUP98-NSD1 fusion carries the N-terminus of NUP98, including its disordered phenylalanine-glycine (FG) repeats and GLEBS domains, and the C-terminus of NSD1 bearing five plant homeodomains (PHD) and one PHD finger-like Cys-His-rich domain (C5HCH or PHD6), one proline-tryptophan-proline domain (PWWP), and the lysine methyltransferase SET domain [9]. Retroviral NUP98-NSD1 promotes myeloid progenitor self-renewal in mice by maintaining the expression of HoxA7, HoxA9, HoxA10, HoxB4, HoxB5, and Meis1 genes via a mechanism requiring both the NUP98 and the NSD1 moieties [10, 11]. It has been shown that the NUP98 moiety interacts with transcriptional coactivators EP300 and KMT2A (also known as MLL1) [10, 11]. PHD5 and C5HCH domains from the NSD1 moiety mediate binding of genomic loci, while their transcriptionally active state is ensured by SET domain-catalyzed H3K36me2 [10]. Despite these insights, the molecular mechanism of NUP98-NSD1 driven leukemogenesis remains poorly characterized.

Herein, we determined the NUP98-NSD1 nuclear interactome, and examined its dependency on specific domains of the fusion protein. We found that core interactome binding was largely dependent on the FG repeat domains which mediate formation of liquid-like phase-separated NUP98-NSD1 nuclear condensates. We identified condensate constituents including imitation switch (ISWI) family member SMARCA5 and BPTF (bromodomain PHD finger transcription factor), both members of the nucleosome remodeling factor complex (NURF), and validated the interaction with SMARCA5 remodeler in NUP98-NSD1+ patient cells. Furthermore, we demonstrated an important role of SMARCA5 in self-renewal of NUP98-NSD1 immortalized murine hematopoietic cells and regulation of HoxA9 and Meis1 proto-oncogene expression. As SMARCA5 knockdown/pharmacologic targeting did not affect formation of the NUP98-NSD1 nuclear condensates, we propose that a fully functional interactome is necessary to maintain the transformed state.

Materials & methods

Cell lines and lysate preparations

Cell lines were acquired from the American Type Culture Collection (ATCC) and cell culture media were obtained from Lonza. All media were supplemented with 10% fetal bovine serum (FBS). HEK-293 cells were grown in DMEM, while 32Dcl3 cells were grown in RPMI containing 10 ng/ml of murine IL-3. The cells were grown at 37 °C with 5% of CO2 and maintained according to manufacturer instructions. For transfection experiments, HEK293 cells were plated at 70% confluency. We utilized the calcium-phosphate method [12] with 10 μg of corresponding plasmids. The medium was changed after 8 h and the cells were incubated for 36 h. Nuclear extracts were made using Dignam’s protocol [13]. For stable transfection of 32Dcl3 cells the NEPA21 Electroporator was used (NEPA GENE). 32Dcl3 cells and immortalized NUP98-NSD1/FLT3-ITD primary murine cells were lysed using 4X Laemmli buffer, after which the lysate was boiled at 95 °C for 5 min. Patient cells were obtained from the lab of Prof. Olaf Heidenreich at Princes Maxima Center for Pediatric Oncology in Utrecht, The Netherlands. NUP98-NSD1+ patient cells were grown in StemSpan™ (STEMCELL Technologies) supplemented with 10 ng/mL IL-3, 10 ng/mL FLT3 ligand, 10 ng/mL GM-CSF, 150 ng/mL SCF, 100 ng/mL TPO.

Immunoprecipitation

Nuclear extracts were resuspended in IP buffer (10 mM Tris HCL pH 7.6, 150 mM NaCl, 0.4% NP-40, 1X EDTA-free Roche protease inhibitors) to a final concentration of 1 mg/ml. Protein concentration was determined using Bradford assay. FLAG-tagged proteins were immunoprecipitated using anti-FLAG M2-affinity gel (Sigma-Aldrich). For co-immunoprecipitation experiments that were followed by Western blotting, 40 μl of M2 affinity gel was added to 5 mg of nuclear extracts previously resuspended in IP buffer. If the experiment was followed by mass-spectrometry analysis, the same amount of M2 affinity gel was added to 10 mg of nuclear extracts/IP buffer. The reaction was incubated for 10 min at 4 °C, after which the beads were washed two times with “washing buffer-1” (10 mM Tris HCl pH 7.6, 500 mM NaCl, 0.5% NP-40, 1X EDTA-free Roche protease inhibitors), two times with “washing buffer-2” (PBS, 0.5% NP-40, 1X EDTA-free Roche protease inhibitors) and once with PBS. The elution was done using 3XFLAG peptide (#F4799, Sigma-Aldrich) which was resuspended in elution buffer (50 mM Tris HCl pH 7.6, 300 mM NaCl, 1% glycerol,1X EDTA-free Roche protease inhibitors) to a final concentration of 50 ng/ml. Eluted proteins were reduced using 4X Laemmli buffer and boiled at 95 °C.

Sample preparation and mass spectrometry analysis

Immunoprecipitated samples were run on a 4-12% Bis-Tris Gel (Thermo Scientific) and subjected to Coomassie staining. Protein bands were excised from the SDS-gel. Excised bands were chopped into pieces of approximately 1x1mM. Gel particles were transferred into a 1.5 ml Eppendorf tube where the reduction of proteins’ disulfide bridges was performed using 10 mM DTT in 100 mM NH4HCO3 at 56 °C for 55 min, while protein alkylation was performed using 55 mM iodoacetamide in 100 mM NH4HCO3 for 20 min at the room temperature. Proteins were digested with trypsin (0.1 μg/μl in 100 mM NH4HCO3) and incubated at 37 °C overnight. After digestion, peptides were extracted from the gel pieces using acetonitrile and 5% formic acid. Peptide extracts were then purified using the StageTip procedure [14], dried in a SpeedVac and resuspended in 1% trifluoroacetic acid (TFA) before mass spectrometry analysis. Five microliter of purified peptides were injected into the chromatographic system (Thermo Scientific, EASY-nLC 1200 Liquid Chromatography system), and separated on the self-made capillary column (ReproSil-Pur 120 C18-AQ, 1.9 μm, Dr. Maisch GmbH, 360 × 0.075 mM). Peptide separation was achieved on a linear gradient from 95% solvent A (2% acetonitrile, 0.1% formic acid) to 55% solvent B (80% acetonitrile, 0.1% formic acid) over 75 min and from 55 to 100% solvent B in 3 min at a constant flow rate of 0.25 μl/min on UHPLC Easy-nLC 1000 (Thermo Scientific) where the LC system was connected to a 23-cm fused-silica emitter of 75 μm inner diameter (New Objective, Inc. Woburn, MA, USA), packed in-house with ReproSil-Pur C18-AQ 1.9 μm beads (Dr Maisch Gmbh, Ammerbuch, Germany) using a high-pressure bomb loader (Proxeon, Odense, Denmark).

The LC system was coupled to the Thermo Scientific™ Q Exactive™ HF hybrid quadrupole-Orbitrap mass spectrometer. The total run time including sample loading and column reconditioning was 60 min. The Q Exactive HF was operated in a DDA top 20 method with an MS survey scan resolution setting of 60,000 and MS/MS resolution setting of 17,500. Peptide fragmentation was performed with an NCE (normalized collision energy) of 25 and an isolation window of 2.0 m/z. Automatic gain control target value and maximum ion injection times were 3 × 105 and 60 ms for MS, and 105 and 60 ms for MS/MS. Dynamic exclusion was enabled with an exclusion duration of 20s. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD026020 [15].

Proteomics analysis

All raw files were processed using MaxQuant [16] (Version 1.5.2.8) against a UniProtKB/Swiss-Prot human database containing 85.336 entries (downloaded on 02.02.2017). Carbamidomethylation was set as fixed modification while methionine oxidation and N-terminal acetylation were searched as variable modifications. Statistical analysis was done in Perseus [17] using two-sample t-tests with Benjamini-Hochberg correction set at FDR = 0.05. Volcano plots were created to display the results of t-testing, with FDR value of 0.05, and S0 value set at 5. For the visualization of this comparison, volcano plots and heatmap were produced using the in-house written R script on which x-axis shows the ratio (fold change) between log2 transformed LFQ values of the proteins bound by NUP98-NSD1 and mutated forms, while on the y-axis -log10 transformed p values (obtained in t-test) were plotted.

ChIP-Seq read alignment

FLAG-NUP98-NSD1 genome binding sites were obtained from a previously published ChiP-seq dataset (GSE112928), expressed in primary murine hematopoietic cells. Reads were aligned to the reference genome (mm10) using bowtie2 algorithm, and peak calling was performed after duplicate filtering using MACS2. To identify NUP98-NSD1 target genes, peaks were intersected with an interval of +/− 1.5 kb around the genomic transcription start sites, which yielded 553 gene targets of the fusion protein. Similarly, SMARCA5 peaks were obtained from the public ChIP-seq dataset acquired in human leukemic K562 cell line (replicates GSM2424122 and GSM2424123), mapped using hg38 human reference genome and intersected with an interval of +/− 1.5 kb around the TSS of the human genome. Acquired SMARCA5 target genes (10,377 genes) were converted into corresponding murine orthologs using an ad-hoc script in R. Gene Set Enrichment Analysis was performed on shared target genes using enrichr package in R assessing the enrichment for the (GO) Biological Processes.

Immunoblot analysis

Western blotting was performed using standard protocols. Primary antibodies were diluted in 5% milk solution, according to manufacturer’s instructions. We used anti-FLAG (#F7425, Sigma-Aldrich), anti-CHD4 (#ab72418, Abcam), anti-BPTF # A300-973A, Fortis Life Sciences), anti-SMARCA5 (#A301-017A-T, Fortis Life Sciences), anti-KDM1 (#ab37165, Abcam), anti-HDAC2 (#ab12169, Abcam). Primary antibodies were incubated for 2 h at RT with the nitrocellulose membranes. Washing of the nitrocellulose membranes was done 3 times for 5 min, after which secondary antibodies (anti-mouse or anti-rabbit, Bio-Rad) were incubated for 1 h. After three more washes, membranes were developed using the ECL-Plus Western Blotting Reagent (Amersham Biosciences).

Immunofluorescence

Transfected HEK293 cells were seeded on glass coverslips that were previously placed in 24well plates. 32D cl3, NUP98-NSD1/FLT3-ITD, and patient NUP98-NSD1+ cells were attached onto fibronectin-coated glass coverslips (13 mm in diameter). Cells were fixed with 4% paraformaldehyde (Sigma-Aldrich) and permeabilized with 0.2% TritonX-100 (in PBS). Blocking was done using 10% horse serum. Cells were incubated for 1 h at room temperature with primary antibodies, diluted at optimized concentrations in 5% horse serum. Following primary antibodies’ incubation (Anti-FLAG M2 mouse (#F1804, Sigma), 1:2000), the appropriate secondary antibodies were used and incubated with cells for 45 min (Alexa Fluor- coupled antibodies from Jackson ImmunoResearch. Finally, cells were counterstained with DAPI to visualize the nuclei and then mounted on microscope slides using glycerol. The samples were investigated by confocal microscopy performed on a Leica TCS SP5 confocal laser scanning microscope (Leica Microsystem, Heidelberg, Germany). The images were acquired with an HCX PL APO 63X/1.4NA oil immersion objective. The software used for all acquisitions was LAS AF (Leica). Raw images were then analyzed using Fiji software (NIH, Bethesda, USA). The figures were assembled using Adobe Illustrator.

Proximity ligation assay (PLA)

The Duolink In Situ Far Red Starter Kit Mouse/Rabbit (Sigma-Aldrich) was used for the detection of examined interactions. Cells were attached onto the fibronectin-coated coverslips during overnight incubation at 37 °C with 5% of CO2. The following day, the cells were gently washed in PBS and fixed using 4% paraformaldehyde (Sigma-Aldrich). Permeabilization was performed using 0.2% TritonX-100 in PBS at room temperature for 10 min. Then, the cells were gently washed in PBS at room temperature twice. Cells were placed in an in-house made dark and humid chamber. Duolink Blocking solution was used for the blocking step, which was performed at 37 °C for 60 min. Anti-FLAG, SMARCA5 and BPTF antibodies were diluted in Duolink Antibody diluent solution (both in same Eppendorf tube) at the concentration 1:2000, NSD1 (C-term) antibody (#N312/10, NeuroMab) was similarly diluted in Duolink Antibody diluent at the concentration 1:1000, and NUP98 (N-term) antibody (#M1-26400, antibodies-online.com) at the concentration 1:500. After the blocking, incubation with primary antibodies was performed for 60 min at room temperature. Subsequent steps were performed following manufacturer’s instructions. Finally, imaging of these samples was performed on Leica TCS SP5 confocal laser scanning microscope with an HCX PL APO 63X/1.4NA oil immersion objective, and Nikon Crest V3 Spinning Disc Confocal microscope. Images were acquired with LAS AF software (SP5) and NFI software (Nikon) analyzed using Fiji software.

Fluorescent recovery after photobleaching (FRAP analysis)

FRAP experiments were performed on a Leica TCS SP8 confocal microscope, using HC PL APO CS2 63X/1,40 objective, managed by Leica LasX software. 10 Pre-bleach images were acquired at maximum speed using the white light laser set at 488 nm and 2% power. Then, after bleaching the fluorescence signal at background level using the same laser at maximum power, post-bleach images were acquired according to the following scheme: 30 images at maximum speed, 30 images every 1 s, 50 images every 2 s. Only complete recovery curves were used for the analysis. All of the steps in the analysis were performed using the Leica LasX software, including background subtraction, correction for imaging photobleaching, normalization, curve fitting (single exponential) and mathematical data collection (recovery half-time).

Biotinylated isoxazole-mediated precipitation

Biotinylated isoxazole (b-isox, Sigma-Aldrich), was resuspended in DMSO. Briefly, HEK293 cells were lysed in lysis buffer (20 mM Tris HCl pH 7.4, 150 mM NaCl, 5 mM MgCl2, 0.5% NP-40 and 10% glycerol, supplemented with 1X EDTA-free Roche protease inhibitors for 1 h at 4 °C) [4,5,6]. However, none of the studies examined any molecular players that cooperate with NUP98 fusions in installing the leukemogenic gene regulatory network.

In our study, we demonstrated that both genetic and pharmacological targeting of SMARCA5 did not affect the condensates indicating that biomolecular condensation per se is not sufficient to maintain the transformed phenotype. One recent study compared the protein interactome of five AML-associated epitope-tagged NUP98 fusions, including NUP98-NSD1, ectopically expressed in the human HL-60 AML cell line and identified by mass spectrometry 157 shared interactors [6]. Notably, in contrast to our study, Terlecki-Zaniewicz and colleagues identified SMARCA5 (also known as SNF2H) as general constituent of biomolecular condensates in human HL-60 AML cells, but not as interactor of any of the 5 NUP98-fusions studied. We believe that the reason for the observed discrepancy lies in diverse experimental approaches linked to different objectives of this recent publication. Terlecki-Zaniewicz and colleagues performed a very challenging screen for common interactions among 5 different NUP98 fusions using whole cell lysates. On the other hand, we used nuclear extracts in our mass spectrometry and Western blot analysis (as NUP98-NSD1 is expressed exclusively in the nucleus), in order to reduce the number of cytoplasmic contaminants. Furthermore, we focused on dissecting in-depth the nuclear interactome of solely NUP98-NSD1, and therefore utilized an immunoprecipitation protocol enabling the detection of high-confidence protein-protein interactions [27]. Moreover, we provided specific functional information related to the dissection of NUP98-NSD1 nuclear condensate constituents through structural domain-dependent interactome analysis. Finally, we also validated some interactions in NUP98-NSD1 immortalized primary murine hematopoietic cells and patient cells using proximity ligation assay.

Another recent study demonstrated that altered chromatin loo**, induced by the phase separation of NUP98-HOXA9, is driving the onset of the leukemogenic transcription programs [4]. Protein-mediated phase separation contributes to the three-dimensional organization of the genome and transcription regulation, thus influencing the expression of gene regulatory networks driving the cell fate [28]. SMARCA5 chromatin remodeler is one of the principal regulators of genome-wide nuclear topology, and is deeply involved in recruitment of transcription factors, such as CTCF [29]. We showed that genomic co-localization of SMARCA5 and NUP98-NSD1 occurs at the transcriptionally active genomics sites regulating stem cell differentiation, therefore suggesting that SMARCA5 may recruit NUP98-NSD1 to its target sites, thus promoting leukemic transformation.

Genetic ablation of Smarca5 in mice leads to an early lethal phenotype due to blocked maturation of erythroid and myeloid lineages underlining its role as a critical hematopoietic regulator [30]. In hematopoiesis, highest SMARCA5 levels are expressed in progenitor cells of the myeloid and erythroid lineage (http://servers.binf.ku.dk/bloodspot/). Likewise, exposure to a small molecule inhibitor (ED2-AD101) targeting the chromatin remodelers SMARCA5 and CHD4 at doses of between 1 to 10 μM induced differentiation of THP-1 leukemic cell line [25]. We observed some degree of differentiation of NUP98-NSD1+ patient cells upon treatment with 10 μM ED2-AD101, and very high rate of cell death, most probably due to the low specificity of the compound. Therefore, more selective SMARCA5-inhibiting compounds are clearly necessary to explore a therapeutic window against AML.

We, as well as Terlecki-Zaniewicz and colleagues, did not identify the previously published interaction of NUP98-fusions with KMT2A detected using whole-cell lysates and proximity dependent biotinylation (BioID) followed by Western blotting [11]. This may be explained by the biochemical complementarity of the different approaches [31, 32]. We also compared our NUP98-NSD1 interactome with the affinity-purified NUP98-HOXA9 interactome reported by Shima et al. and found a rather low overlap, suggesting that two fusion proteins may utilize different mechanisms for their molecular function [33]. Nevertheless, both datasets shared previously published NUP98 interactions with RAE1 and XPO1.

Conclusions

In conclusion, we characterized multiple nuclear protein interactions of the AML-associated NUP98-NSD1 fusion and identified SMARCA5 as important mediator to maintain the transformed phenotype of NUP98-NSD1-immortalized hematopoietic cells. Our data contain thorough functional information that may be crucial for designing specific inhibitors to be used in NUP98-NSD1 driven AML.