Abstract

   The interaction of oncogenes with cellular proteins is a major
   determinant of cellular transformation. The NUP98-HOXA9 and SET-NUP214
   chimeras result from recurrent chromosomal translocations in acute
   leukemia. Functionally, the two fusion proteins inhibit nuclear export
   and interact with epigenetic regulators. The full interactome of
   NUP98-HOXA9 and SET-NUP214 is currently unknown. We used
   proximity-dependent biotin identification (BioID) to study the
   landscape of the NUP98-HOXA9 and SET-NUP214 environments. Our results
   suggest that both fusion proteins interact with major regulators of RNA
   processing, with translation-associated proteins, and that both
   chimeras perturb the transcriptional program of the tumor suppressor
   p53. Other cellular processes appear to be distinctively affected by
   the particular fusion protein. NUP98-HOXA9 likely perturbs Wnt, MAPK,
   and estrogen receptor (ER) signaling pathways, as well as the
   cytoskeleton, the latter likely due to its interaction with the nuclear
   export receptor CRM1. Conversely, mitochondrial proteins and metabolic
   regulators are significantly overrepresented in the SET-NUP214 proximal
   interactome. Our study provides new clues on the mechanistic actions of
   nucleoporin fusion proteins and might be of particular relevance in the
   search for new druggable targets for the treatment of
   nucleoporin-related leukemia.

   Keywords: SET-NUP214, NUP98-HOXA9, BioID, interactome, gene ontology,
   leukemia

1. Introduction

   The nucleoporins NUP98 and NUP214 are components of the nuclear pore
   complex (NPC) and belong to the group of so-called
   phenylalanine-glycine (FG) rich nucleoporins, which are essential for
   nucleocytoplasmic transport. Via their FG domains, NUP98 and NUP214
   contribute to the selective and semi-permeable NPC barrier, and
   interact with nuclear transport receptors (NTRs) of the β-karyopherin
   family, thereby promoting the fast exchange of cargoes between the
   nucleus and the cytoplasm [[30]1,[31]2,[32]3].

   A number of chromosomal translocations involving the NUP98 and NUP214
   loci are reported in acute myeloid and lymphoblastic leukemias (AML and
   ALL, respectively). NUP98- and NUP214-related leukemia are associated
   with poor overall survival [[33]4,[34]5,[35]6,[36]7], and no specific
   or targeted therapies are as yet available to improve prognosis. The
   chromosomal rearrangements of NUP98 and NUP214 result in their fusion
   with a large range of gene partners, all of which retain the FG domain
   of the respective nucleoporin [[37]7,[38]8]. The fusion of NUP98 with
   the homeobox protein Hox-A9 (HOXA9), NUP98-HOXA9, that results from
   t(7;11)(p15;p15), has been studied as the prototype for the oncogenic
   mechanisms governing the actions of NUP98 fusions with homeodomain (HD)
   proteins in AML [[39]9]. HOXA9 is a transcription factor that regulates
   hematopoietic stem cell expansion and is abundantly expressed in
   hematopoietic precursor cells, while being progressively silenced
   during differentiation [[40]10,[41]11]. NUP214 is frequently found in
   conjunction with the oncogene SET, resulting from either
   t(9;9)(q34;q34) or an interstitial deletion at 9q34
   [[42]12,[43]13,[44]14]. SET-NUP214 is typically linked to ALL, and less
   frequently to AML [[45]15,[46]16]. SET is a chromatin-binding protein
   and an epigenetic regulator as part of the inhibitor of
   acetyltransferases (INHAT) complex [[47]17,[48]18]. Due to its role as
   an epigenetic modifier, SET is involved in a multitude of cellular
   functions, including regulation of the cell cycle, gene expression, and
   apoptosis [[49]19,[50]20,[51]21]. NUP98-HOXA9 and SET-NUP214 share
   several characteristics: both form nuclear foci that accumulate
   endogenous proteins [[52]22,[53]23]; both interact with the NTR
   chromosome region maintenance 1 (CRM1), or exportin 1 (XPO1), and
   sequester cargo-loaded CRM1-nuclear export complexes to inhibit their
   translocation to the cytoplasm [[54]23,[55]24,[56]25]. Moreover,
   NUP98-HOXA9 and SET-NUP214 interact with chromatin-binding proteins,
   such as the histone methyltransferases mixed lineage leukemia 1 (MLL1)
   and the disruptor of telomeric silencing 1-like (DOT1L)
   [[57]26,[58]27,[59]28]. Association of NUP98-HOXA9 and SET-NUP214 with
   chromatin-bound CRM1 induces over-expression of HOX genes, a hallmark
   of unfavorable prognosis in leukemia [[60]10,[61]29,[62]30]. The full
   landscape of NUP98-HOXA9 and SET-NUP214 interactors, however, has not
   yet been determined.

   In recent years, advances in enzyme-mediated protein labelling became a
   powerful approach to study specific protein–protein interactions (PPIs)
   [[63]31,[64]32,[65]33]. Proximity-dependent protein biotinylation
   (BioID) is an enzyme-mediated protein labelling approach that uses a
   modified version of the Escherichia coli (E. coli) biotin ligase BirA
   (BirA^R118G), which has a lower affinity than wild type BirA for
   biotinoyl-adenylate (bio-AMP), the active form of biotin that can bind
   lysine residues [[66]31,[67]34]. Thus, when expressed in-frame with a
   protein of interest, BirA^R118G biotinylates proximal proteins, which
   can then be purified by streptavidin pulldown and further identified by
   mass spectrometry [[68]32]. In contrast to other interaction assays,
   which may be regarded as a snapshot of PPIs at the moment of cell
   lysis, BioID interrogates both stable and transient PPIs in living
   cells, thus providing a broader picture of protein interactors. Here,
   we used a modified BioID approach to study the proximal interactome of
   NUP98-HOXA9 and SET-NUP214. We identified further common associated
   partner as well as discrete fusion protein-specific interactors in the
   environment of NUP98-HOXA9 and SET-NUP214.

2. Materials and Methods

   All experiments were carried out at room temperature (RT) unless
   otherwise specified.

2.1. Plasmids

   pcDNA3.1 MCS-BirA(R118G)-HA was a gift from Dr. Kyle Roux (Addgene
   plasmid # 36047; [[69]31]) and BirA(R118G)-HA destination vector from
   Dr. Karl Kramer (Addgene plasmid # 53581). For the cloning of
   SET-NUP214, total RNA was extracted from LOUCY cells, which carry
   del(9)(q34.11q34.13) resulting in the SET-NUP214 fusion transcript
   [[70]35]. The coding sequence of SET-NUP214 was cloned into the
   pcDNA3.1 MCS-BirA(R118G)-HA vector, as described in [71]Appendix A. The
   NUP98-HOXA9-BirA^R118G construct was generated by Gateway^® cloning.
   The coding sequence of NUP98-HOXA9 was first subcloned from
   pEGFP-NUP98-HOXA9 [[72]22] into the pENTR/TOPO vector using the TOPO^®
   TA Cloning Kit (Invitrogen, Merelbeke, Belgium) to generate the
   pENTR/NUP98-HOXA9 Gateway^® entry vector. The NUP98-HOXA9 sequence was
   then subcloned into the BirA(R118G)-HA destination vector using the
   Gateway ™ LR Clonase™ enzyme mix (Invitrogen).

2.2. Cell Lines and Transfections

   HCT-116 cells were a gift from Dr. Denis Lafontaine (Institute of
   Molecular Biology and Medicine, Université Libre de Bruxelles,
   Charleroi, Belgium). HCT-116 cells were cultured in McCoy’s 5A medium
   (LONZA^TM BioWhittaker^TM, Verviers, Belgium), supplemented with 10%
   FBS and 1% penicillin/streptomycin (P/S, GIBCO, Invitrogen), and
   cultivated in a humidified incubator at 37 °C with 5% CO[2] atmosphere.
   HCT-116 cells were transfected using the jetPRIME^® transfection
   reagent (Polyplus transfection^®, Illkirch, France). Briefly, plasmids
   were mixed with transfection reagent at a 1:2 (w/v) ratio and incubated
   for 40 min. Transfection mixes were added to the cell culture and
   incubated for 24 h. Next, the culture medium was replaced by fresh
   medium containing 50 µM biotin (Sigma–Aldrich, Overijse, Belgium) and
   biotinylation was induced for an additional 24 h. Cells were tested for
   mycoplasma contamination on a regular basis.

2.3. Immunofluorescence

   Cells were grown on polylysine-coated glass coverslips and fixed in 2%
   formaldehyde for 15 min, washed three times for 10 min with PBS, and
   permeabilized with PBS/2% BSA/0.1% Triton X-100 for 10 min. Cells were
   washed twice for 10 min in PBS/2% BSA, incubated with
   Streptavidin-Alexa Fluor ™ 488 conjugate (dilution 1:1000; Invitrogen)
   or anti-HA antibody (clone 12CA5, dilution 1/50, supernatant of a mouse
   hybridoma cell line) for 1 h and washed twice in PBS/2% BSA/0.1% Triton
   X-100. Cells incubated with anti-HA were then incubated with goat
   anti-mouse IgG Alexa Fluor™ 488 (dilution 1:1000; Invitrogen) for 1 h
   and washed twice for 10 min with PBS. All cells were mounted with
   Mowiol-4088 containing DAPI (1 μg/ml) and stored at 4 °C until viewed.
   Cells were imaged using a Zeiss LSM-710 confocal laser-scanning
   microscope (Zeiss, Oberkochen, Germany). Images were recorded using the
   microscope system software and processed using ImageJ v.1.52t
   ([73]http://imagej.nih.gov) and Inkscape 0.92 Software
   ([74]http://www.inkscape.org).

2.4. Pulldown of Biotinylated Proteins

   2 × 10^6 cells were plated in a 10 cm^2 cell-culture dish, grown for 24
   h and subsequently processed for transfection and biotinylation
   induction as described above. Cells were lysed in lysis buffer (50 mM
   Tris-HCl, pH 7.8, 150 mM NaCl, 1mM EGTA, 1.5 mM MgCl[2], 0.4% sodium
   dodecyl sulfate (SDS), 1 µl/ml benzonaze [25 U/ml], 1% Nonidet-P40, and
   protease inhibitor cocktail tablets (Roche, Basel, Switzerland).
   Bradford assay was used to determine protein concentration and 500 μg
   of protein were incubated with 50 µl of SeraMag™ magnetic
   Streptavidin-coated beads ([10 mg/ml]; GE Healthcare, Chicago,
   Illinois, USA). Protein-beads incubation and recovery of biotinylated
   proteins were carried out as detailed in [75]Appendix A. The entire
   eluate containing biotinylated proteins was subjected to sodium
   dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and
   Western blotting. Proteins were detected using horseradish-conjugated
   streptavidin (HRP-Strep; Thermo Fischer Scientific, Merelbeke,
   Belgium). For protein identification by mass spectrometry, the same
   amount of protein was used. After incubation of whole protein extract
   with streptavidin-coated magnetic beads, samples were resuspended in
   on-bead tryptic digestion buffer (20 mM Tris-HCl, pH 8.0, 2 mM CaCl[2])
   and processed for large-scale analysis by tandem mass spectrometry
   (LC-MS/MS).

2.5. Mass Spectrometry

   Proteins were digested on the beads with 1 µg of trypsin (Promega,
   Leiden, Netherlands) at 37 °C for 4 h while spinning at 161× g. Beads
   were removed, an additional 1 µg of trypsin was added, and proteins
   were further digested at 37 °C, overnight. The resulting peptide
   mixture was purified using OMIX C18 pipette tips (Agilent, Santa Clara,
   California, USA). The purified peptides were dried completely and
   re-suspended in 20 µl loading solvent (0.1% TFA in water/ acetonitrile,
   2/98 (v/v)) of which 5 µl were injected for LC-MS/MS analysis on an
   Ultimate 3000 RSLCnano ProFLow system connected online to a Q Exactive
   HF mass spectrometer (Thermo, Waltham, MA, USA). Peptide trapping and
   elution and mass spectrometer operation details are described in
   [76]Appendix A (Sample injection and mass spectrometer operation).

2.6. Data Analysis

   Data analysis was performed with MaxQuant (version 1.6.3.4; Max Planck
   Institute of Biochemistry, Germany [[77]36]) using the Andromeda search
   engine with default search settings including a false discovery rate
   (FDR) set at 1% on both the peptide and protein level. Spectra were
   searched against the human UniProt Tax ID: 9606 proteins in the
   UniProt/Swiss-Prot reference database (database release version of
   January 2019, [78]www.uniprot.org) supplemented with the BirA fusion
   proteins, as detailed in [79]Appendix A (Data Analysis). Further data
   analysis was performed using the Perseus software (version 1.6.2.1, Max
   Plank Institute of Biochemistry, Germany) after loading the protein
   groups file from MaxQuant [[80]37]. First, proteins identified by site
   and reverse database hits and potential contaminants were removed.
   Label-free quantification (LFQ) values were then used to normalize
   protein abundance among NUP98-HOXA9-BirA^R118G (NHA9-BioID) and
   SET-NUP214-BirA^R118G (SN214-BioID) proximal interactors, relative to
   BirA^R118G alone (control). Given the potential differences in the
   expression of the BioID proteins (due to transient transfection), we
   performed an internal normalization to calculate the LFQ value of each
   individual protein that results from differences in NHA9-BioID and
   SN214-BioID protein expression. Subsequently, an external normalization
   to calculate the LFQ value of each individual protein relative to the
   control BirA^R118G (NHA9-BioID/BirA^R118G and SN214-BioID/BirA^R118G)
   was carried out. The results were then converted to log2 and expressed
   as fold change (F.C.) relative to the control ([81]Figure S1).

2.7. Gene Ontology and Pathway Analysis

   Gene ontology (GO) of NHA9-BioID and SN214-BioID proximal interactors
   was performed by the GO consortium-associated Protein Analysis Through
   Evolutionary Relationships (PANTHER) Classification System (version
   14.1.), available online ([82]www.pantherdb.org) using the default
   parameters: Fisher’s exact test, with the Benjamini–Hochberg FDR
   correction for multiple testing and the background reference list: Homo
   sapiens whole genome [[83]38,[84]39,[85]40]. Clustered enrichment
   analysis was performed with the Cytoscape software (v3.7.1) plugin
   ClueGO (v2.5.5), to calculate enrichment of terms as right-sided tests
   based on the hypergeometric distribution [[86]41]. ClueGO uses Cohen’s
   kappa statistics to link the terms in the network and to determine the
   association strength between the terms, which is an indication of term
   overlapping to define functional clusters [[87]41,[88]42]. For the
   analysis of proximal interactors of NUP98-HOXA9 and SET-NUP214, the
   Kyoto Encyclopedia of Genes and Genomes (KEGG) and REACTOME Pathways
   together with GO vocabularies (GO Biological Processes (GOBP), GO
   Molecular Function (GOMF), and GO Cellular Compartments (GOCC)) were
   used, and functional representations of non-redundant and
   over-represented terms within the input protein sets were generated.
   Human Ensemble Gene ID identifiers (ENSG IDs) were mapped to the
   selected annotations (GO, updated on 27th February, 2019, KEGG, updated
   on 19th December, 2019, and REACTOME pathways, updated on 19th December
   2019). The following settings were used during enrichment analysis:
   p-value cut off 0.01, and Bonferroni step-down correction for multiple
   comparisons.

2.8. Screening for Nuclear Export Signals

   The presence of classical NESs was assessed using the online available
   software NES finder 0.2
   ([89]http://research.nki.nl/fornerodlab/NES-Finder.htm) and LocNES
   ([90]http://prodata.swmed.edu/LocNES/LocNES.php), with default software
   parameters [[91]43,[92]44].

3. Results

3.1. Subcellular Distribution of BioID Fusion Proteins and Biotinylation
Induction

   To validate the subcellular localization of the BioID fusion proteins,
   we first examined the localization of NHA9-BioID, SN214-BioID, and
   control BirA^R118G, after biotinylation was induced. As shown in
   [93]Figure 1A, NHA9-BioID localized to the nucleus in a distinctive
   punctate pattern, whereas SN214-BioID formed intranuclear foci and
   localized to the nuclear rim. The localization of the two BioID fusion
   proteins is similar to the respective GFP-fusion proteins of NHA9 and
   SN214 ([94]Figure 1B; [[95]22,[96]45]). BirA^R118G was found throughout
   the entire cell ([97]Figure 1A). Streptavidin labeling revealed the
   same distribution pattern as HA- and GFP-NHA9 and -SN214, indicating
   the presence of biotin ([98]Figure 1C). Western blot analysis
   furthermore showed that endogenous proteins were biotinylated in whole
   cellular extracts of NHA9-BioID and SN214-BioID transfected cells
   ([99]Figure 1D, bound fraction), in contrast to BirA^R118G transfected
   cells. Of note, the NHA9-BioID fraction showed a strong enrichment of a
   specific band above 100 kDa, likely corresponding to the fusion protein
   itself. Due to the lower intensity of the SN214-BioID fraction,
   biotinylated SET-NUP214 cannot reliably be allocated, but might
   correspond to the band appearing below the 250-kDa band. For BirA^R118G
   no specific enrichment was observed, given the unspecific nature of
   biotinylation mediated by the biotin ligase alone.

Figure 1.

   [100]Figure 1
   [101]Open in a new tab

   Proximity-dependent biotin identification (BioID) fusion protein
   expression and biotinylation of endogenous proteins. (A) Localization
   of NHA9-BioID, SN214-BioID, and BirA^R118G was evaluated by
   immunostaining with anti-HA antibody. (B) Localization of GFP-tagged
   NUP98-HOXA9 and SET-NUP214 was evaluated by green fluorescent protein
   (GFP) fluorescence. NHA9-BioID and SN214-BioID exhibit the same
   distribution pattern in nuclear foci and at the nuclear envelope as
   their GFP-tagged counterparts. (C) Detection of protein biotinylation
   by Streptavidin-488. Cells were transfected with NHA9-BioID,
   SN214-BioID, and BirA^R118G and probed with Streptavidin-Alexa Fluor™
   488 conjugate. Shown are representative confocal images. DNA was
   visualized with DAPI (blue). Scale bars, 10 µm. (D) For detection of
   protein biotinylation by NHA9-BioID, SN214-BioID, and BirA^R118G,
   corresponding cell lysates were enriched on Streptavidin-coated
   magnetic beads and whole protein lysates and the bound fractions were
   analyzed by immunoblotting. Note, virtually no specific bands were
   detected in the bound fraction of BirA^R118G, in contrast to the bound
   fractions of NHA9-BioID and SN214-BioID, which exhibited patterns of
   differentially biotinylated proteins.

3.2. Identification of Known Proximal Interactors of NUP98-HOXA9 and
SET-NUP214

   Having validated the correct subcellular localization and the
   reliability of NHA9-BioID and SN214-BioID, we then carried out a
   mass-spectrometric analysis of the biotinylated proteins co-purifying
   with the respective BioID fusion protein. We first screened for
   candidate proximal interactors and identified several known binding
   partners of NUP98-HOXA9 and SET-NUP214. These were the protein nuclear
   export factor CRM1 [[102]24,[103]29], the mRNA export factor RAE1
   [[104]46], the transcription factor NFAT5 [[105]24], the histone
   methyltransferase MLL1 [[106]28], and the histone deacetylase HDAC1
   [[107]47] for NUP98-HOXA9 ([108]Table 1). We did not find the
   acetyltransferase CREB binding protein/p300 (CBP/p300; [[109]48]) nor
   WDR5, a component of the non-lethal specific (NSL)-histone modifying
   complex [[110]27], co-purifying with NHA9-BioID, which might be due to
   the fact that both genes are mutated in HCT-116 cells
   [[111]49,[112]50]. Among known SET-NUP214 interactors ([113]Table 2),
   we identified the nuclear RNA export factor 1 (NXF1/TAP; [[114]25]),
   CRM1 [[115]23,[116]25], and nucleoporin NUP62 [[117]23,[118]45].

Table 1.

   List of validated NUP98-HOXA9 interaction partners found in the
   biotinylated fraction of NHA9-BioID.
   Protein Gene NHA9-BioID F.C. Ref.
   Nucleoporin NUP98 NUP98 +/+ −1.84 [[119]46]
   Chromosome region maintenance 1/Exportin 1 CRM1/XPO1 +/+ −2.13
   [[120]24,[121]29]
   mRNA export factor 1 RAE1 +/+ 6.40 [[122]46]
   Nuclear factor of activated T cells NFAT5 +/- 24.81 [[123]24]
   Histone-lysine N-methyltransferase 2A/Mixed lineage leukemia 1 KMT2A/
   MLL1 +/+ −0.91 [[124]28]
   Host cell factor 1 HCFC1 +/+ −1.86 [[125]27]
   Histone deacetylase 1 HDAC1 +/+ −2.05 [[126]47]
   O-GlcNac transferase subunit p110 OGT -/- [[127]27]
   CREB binding protein CREBBP/p300 -/- [[128]47,[129]48]
   [130]Open in a new tab

   F.C.—Fold change; +/+, present in both replicates; +/−, present in one
   of the replicates; −/−, not detected in any of in NHA9-BioID nor
   BirA^R118G replicates.

Table 2.

   List of validated SET-NUP214 interaction partners found in the
   biotinylated fraction of SN214-BioID.
                  Protein                     Gene    SN214-BioID F.C.   Ref.
        Nuclear RNA export factor 1         NXF1/ TAP     +/+     1.37 [[131]25]
 Chromosome region maintenance 1/Exportin 1 CRM1/XPO1     +/+     0.16
   [[132]23,[133]25]
               Nucleoporin 62                 NUP62       +/+     1.79 [[134]23]
   [135]Open in a new tab

   F.C., Fold change; +/+, present in both replicates.

   A major limitation of our BioID approach is the use of transient
   transfection instead of stable, inducible expression of the BioID
   fusion proteins. Due to its smaller size, the transfection efficiency
   of BirA^R118G is much higher as for the larger NHA9-BioID and
   SN214-BioID constructs. We therefore considered the rate of
   biotinylation by BirA^R118G higher than by the fusion proteins.
   Moreover, BirA^R118G is a promiscuous biotin ligase, which also likely
   results in higher biotinylation levels in the control. To control for
   these differences in biotinylation, we employed a two-step
   normalization strategy that accounted for differences between
   BirA^R118G, NHA9-BioID, and SN214-BioID protein expression ([136]Figure
   S1). To further validate our BioID approach, next we performed Gene
   ontology (GO) analysis of proteins exclusively co-purifying with the
   BirA^R118G control to determine unspecific interactors of
   BirA^R118G-mediated biotinylation [[137]29], as described in the
   Methods section. [138]Table S1 summarizes the list of proteins that
   were exclusively biotinylated by BirA^R118G and their respective
   subcellular localization. No significant enrichment in any of the GO
   categories (i) biological processes (GOBP), (ii) molecular function
   (GOMF), and (iii) cellular components (GOCC) was found, suggesting that
   biotinylation of these proteins is of an unspecific nature, as
   previously reported [[139]29].

   Moreover, given the differences in transfection efficiency between the
   BirA^R118G, NHA9-BioID, and SN214-BioID constructs, some potential
   candidates might have disappeared, because known interactors, such as
   CRM1 and others, were more strongly biotinylated by BirA^R118G as
   compared to NHA9-BioID and SN214-BioID. This in consequence explains
   the negative fold change (F.C.) values ([140]Table 1). High normalized
   label-free quantification values (LFQ[NORMALIZED]; see Data Analysis,
   [141]Section 2.6) indicated that these proteins, after pull-down, are
   more enriched with NHA9-BioID or SN214-BioID, albeit being biotinylated
   by BirA^R118G, meaning that their interaction with the fusion proteins
   is of a rather specific nature.

3.3. Identification of Novel Proximal Interactors of NUP98-HOXA9

   Next, we employed a two-step normalization strategy to evaluate
   NHA9-BioID and SN214-BioID proximal interactors relative to BirA^R118G
   to account for differences in protein expression levels due to the
   transient expression of the BioID fusion proteins ([142]Figure S1). In
   doing so, we obtained 131 proteins that were at least 50% more abundant
   for NHA9-BioID when compared to BirA^R118G ([143]Table S2, fold change
   ≥0.6). GOCC analysis revealed that these NHA9-BioID proximal
   interactors located to both cytoplasmic and nuclear compartments with a
   general enrichment of chromatin-associated proteins and particularly
   the transcription factor AP-1 complex ([144]Figure 2A).
   Cytoskeleton-related proteins were also significantly overrepresented
   in the pool of NHA9-BioID proximal interactors, namely proteins from
   the spindle pole, actin stress fibers, and centrosome ([145]Figure
   2A,B). GOMF analysis showed that NUP98-HOXA9 proximal interactors are
   significantly enriched in DNA binding proteins that target the
   promoters of genes transcribed by RNA polymerase II (RNAPII;
   [146]Figure 2C).

Figure 2.

   [147]Figure 2
   [148]Open in a new tab

   Gene ontology (GO) of NHA9-BioID proximal interactors. Most represented
   (A) cellular compartments (GOCC), (B) biological processes (GOBP), and
   (C) molecular functions (GOMF) among NHA9-BioID proximal interactors.
   Statistical analysis of the overrepresented proteins in the NHA9-BioID
   fraction (total 131 proteins) was performed with the PANTHER
   classification online software (v14.1) using Fisher’s exact test.
   Results are displayed for FDR p < 0.05.

   For a systematic analysis of the landscape of the NUP98-HOXA9 proximal
   interactome, next we performed clustered pathway analysis ([149]Figure
   3). Here, the most significantly enriched functional clusters included
   proteins involved in RNA processing and the RNAPII transcriptional
   machinery ([150]Figure 3, [151]Figure S2). Moreover, proximal
   NHA9-BioID interactors clustered in functional groups associated with
   the expression of genes from key signaling pathways frequently
   dysregulated in cancer. These included estrogen receptor (ER) and MAPK
   signaling pathways, and p53-mediated transcription of DNA damage repair
   genes ([152]Figure 3A, [153]Figure S2; [[154]51,[155]52,[156]53]).
   Further, β-catenin-mediated transcription may be negatively regulated
   by NUP98-HOXA9, suggesting a dysregulation of the Wnt signaling
   pathway. Interestingly, some members of the Wnt signaling pathway,
   i.e., the segment polarity protein disheveled homolog DVL-1, the
   trinucleotide repeat-containing gene 6A protein (TNRC6A), and the
   transducing-like enhancer protein, isoforms 1-4 (TLE 1-4), were
   exclusively biotinylated by NHA9-BioID, further supporting the
   hypothesis that NUP98-HOXA9 affects this signaling pathway. Altogether,
   GO and clustered pathway analyses of NUP98-HOXA9 suggest that
   transcription dysregulation is a major defect in NUP98-HOXA9 driven
   leukemia.

Figure 3.

   [157]Figure 3
   [158]Open in a new tab

   Clustered pathway analysis of NHA9-BioID proximal interactors. (A)
   Functionally grouped network of NHA9-BioID proximal interactors. Nodes
   correspond to functional clusters, resulting from term grouping based
   on their overlapping level. (B) Overview pie-chart showing functional
   groups. Statistical analysis was performed using the Cytoscape plugin
   ClueGo (v.2.5.5) using the hypergeometric test and the following
   parameters: p < 0.01, kappa-score (k) = 0.4; (min/%) genes = 3/4%, GO
   tree levels: 3-11. Ontology databases: GOBP, GOCC, and GOMF, Reactome
   Pathways and KEGG.

3.4. NUP98-HOXA9 Proximal Interactors are Enriched in Nuclear Export Signal
(NES) Containing Proteins

   NUP98 is essential for CRM1-mediated nuclear export and NUP98 chimeras
   have been shown to affect CRM1-mediated nuclear export, to an extent,
   however, that is not fully understood [[159]7,[160]54]. To obtain a
   deeper insight into the impact of NUP98-HOXA9 on CRM1-mediated nuclear
   export, we screened the list of NUP98-HOXA9 proximal interactors for
   the presence of classical NESs using two algorithms: NES finder 0.2 and
   LocNES [[161]43,[162]44]. Both algorithms revealed that the majority of
   NHA9-BioID proximal interactors exhibit at least one classical NES
   motif (NES+ proteins, [163]Figure 4A, [164]Table S3). Given that the
   LocNES software has previously been shown to identify a higher rate of
   false positive NESs, we applied GO analysis for predicted NES+
   interactors obtained by the NES Finder 0.2 software [[165]55]. As shown
   in [166]Figure 4B, this revealed a significant enrichment in proteins
   from cytoplasmic ribonuclear granules and in proteins from the
   microtubule organizing center. Conversely, NES- proximal interactors of
   NUP98-HOXA9 were significantly enriched in proteins implicated in
   transcription regulation, namely the ß-catenin/TCF complex and the
   transcription factor AP-1 complex ([167]Figure 4B).

Figure 4.

   [168]Figure 4
   [169]Open in a new tab

   Screening of nuclear export signals (NESs) in NHA9-BioID proximal
   interactors. The amino acid sequences of the 131 NHA9-BioID proximal
   interactors were analyzed by two different algorithms (NES Finder 0.2
   and LocNES) to evaluate the presence of classical NES. (A) The results
   from both algorithms show that the majority of NHA9-BioID proximal
   interactors have at least one putative classical NES, with a higher
   percentage of NES+ proteins determined by the LocNES algorithm. (B) GO
   analysis of NES+ and NES- proteins identified by NES Finder 0.2 shows
   an overrepresentation of proteins from cytoplasmic RNP granules and the
   microtubule organizing center. NES- proteins are mostly nucleoplasmic
   and are associated with the ß-catenin and the AP-1 transcription factor
   complexes. Statistical analysis was performed with the PANTHER
   classification online software (v14.1), using Fisher’s exact test.
   Results are displayed for FDR p < 0.05.

3.5. Identification of Novel Proximal Interactors of SET-NUP214

   Our LFQ normalization strategy produced a list of 1125 proteins
   enriched in SN214-BioID relative to the BirA^R118G control ([170]Table
   S4). To disclose the SET-NUP214 proximal interactome, we performed GO
   enrichment analysis of the SN214-BioID fraction using the same
   fold-change threshold of ≥0.6 and statistical parameters as for
   NHA9-BioID ([171]Figure 5, [172]Figures S3 and S4). For clustered
   pathway analysis, we applied a conservative approach to reduce
   redundancy of the functional network given the elevated number of
   potential SN214-BioID interactors. [173]Figure 6 shows the functional
   groups that were significantly represented, obtained with the following
   statistical parameters: p <0.01, GO levels: 13-15 (ClueGO default:
   3-8), threshold kappa score of 0.5 (ClueGO default 0.4), thus
   increasing the threshold for group overlapping, and the minimum gene
   number corresponding to percentage of genes of 40/4% (ClueGO default
   3/4%).

Figure 5.

   [174]Figure 5
   [175]Open in a new tab

   Gene ontology (GO) of SN214-BioID proximal interactors. Most
   represented (A) cellular compartments (GOCC), (B) biological processes
   (GOBP), and (C) molecular functions (GOMF) among SN214-BioID proximal
   interactors. Statistical analysis of the overrepresented proteins in
   the SN214-BioID fraction (total 1125 proteins) was performed with the
   PANTHER classification online software (v14.1) using Fisher’s exact
   test. Results are displayed for FDR p < 0.05. Summarized graphic
   representation of the significantly enriched GO terms among SN214-BioID
   proximal interactors. The detailed results of GOBP and GOCC analysis
   can be consulted in [176]Figures S3 and S4.

Figure 6.

   [177]Figure 6
   [178]Open in a new tab

   Clustered pathway analysis of SN214-BioID proximal interactors. (A)
   Functionally grouped network of SN214-BioID proximal interactors. Nodes
   correspond to functional clusters resulting from term grouping based on
   their overlapping level. (B) Overview pie-chart with functional groups.
   Statistical analysis was performed with the Cytoscape plugin ClueGo
   (v2.5.5) using the hypergeometric test and the following parameters: p
   < 0.01, kappa-score (k) = 0.5; (min/%) genes = 40/4%, GO tree levels:
   13-15. Ontology databases: GOBP, GOCC, and GOMF, Reactome Pathways and
   KEGG.

   GOCC ([179]Figure 5A and [180]Figure S3) and GOBP ([181]Figure 5B and
   [182]Figure S4) analysis produced a list of cytoplasmic and nuclear
   structures and processes likely to be affected by SET-NUP214, such as
   mRNA processing, intracellular transport, viral transport, and
   transcription ([183]Figure 5B, [184]Figure S4)
   [[185]48,[186]50,[187]51,[188]52,[189]53]. Consistently GOMF analysis
   ([190]Figure 5C) unveiled an enrichment in proteins with GTP- and
   GDP-binding activity, as well as mRNA and DNA binding proteins.
   Moreover, proteins involved in neutrophil degranulation were enriched
   with SET-NUP214 ([191]Figure 5B), establishing a direct link between
   SET-NUP214 and immune regulation. GO also showed an enrichment in
   nucleolar and ribosomal proteins ([192]Figure 5A and [193]Figure S3),
   suggesting an effect of the fusion protein on translation. It further
   revealed that mitochondrial proteins were enriched with SN214-BioID,
   especially proteins from the respiratory chain complexes and proteins
   involved in mitochondrial translation ([194]Figure 5A, [195]Figures S3
   and S4), suggesting involvement of the fusion protein in cell
   metabolism through an effect on mitochondria. In line with the GO
   results, clustered pathway analysis of proximal SET-NUP214 interactors
   revealed an enrichment in proteins involved in amino acid metabolism,
   translation and infection, and proteins involved in virus biology, such
   as transcription, transport, and interaction with host cells
   ([196]Figure 6, [197]Figures S4 and S6, Table S5). Clustered pathway
   analysis further supports GO findings of an interplay between
   SET-NUP214 and mitochondrial proteins ([198]Figure 6B, [199]Figures S3
   and S4, Table S5). Moreover, the results suggest an association of
   SET-NUP214 with several transcription factors and point to a possible
   effect on TP53-mediated transcription ([200]Figure 6A,B, [201]Table S5,
   Figure S5). Finally, our clustered pathway analysis reinforces the link
   between SET-NUP214 and immunity, with two distinct, yet overlapping,
   immunity-related clusters that link the fusion protein to immune system
   regulation, and more specifically to neutrophil activation ([202]Figure
   6A, [203]Table S5).

   We identified a total of 47 proteins that were enriched with both
   NHA9-BioID and SN214-BioID, of which a vast majority presented at least
   one putative classic NES ([204]Figure S7A and Table S6). Furthermore,
   GO analysis showed that the pool of shared proximal interactors is
   enriched in proteins associated with microtubule cytoskeleton,
   suggesting that SET-NUP214, like NUP98-HOXA9, might be involved in
   microtubule organization ([205]Figure S7B).

4. Discussion

   We used a modified BioID approach to study the proximal interactome of
   NUP98-HOXA9 and SET-NUP214. We compared the pool of proteins
   biotinylated by NHA9-BioID and SN214-BioID, respectively, with the pool
   of proteins biotinylated by BirA^R118G, using a normalization strategy
   that accounts for differences in NHA9-BioID, SN214-BioID and BirA^R118G
   expression due to their transient transfection. After validating the
   expression, cellular distribution, and the capacity of protein
   biotinylation of all three BioID proteins, we performed gene ontology
   (GO) and clustered pathway enrichment analysis of the proximal
   interactors of NHA9-BioID and SN214-BioID. As expected, we observed
   that BirA^R118G alone biotinylates endogenous proteins in an unspecific
   manner [[206]31]. The respective proximal interactomes of NHA9-BioID
   and SN214-BioID were, somewhat surprisingly, enriched in cytoplasmic
   proteins, which appears inconsistent with the nuclear localization of
   the fusion proteins. This might be explained by the fact that mitosis
   in vertebrates involves nuclear envelope breakdown and mixing of the
   cytoplasmic and nuclear protein content [[207]56]. At the end of
   mitosis, proteins must be appropriately segregated between the nucleus
   and the cytosol. This process depends on transport factors, including
   CRM1, which alone is predicted to transport one fourth of the entire
   proteome [[208]57]. As shown by us and others, NUP98-HOXA9 and
   SET-NUP214 sequester CRM1 nuclear export complexes, resulting in the
   nuclear accumulation of proteins and RNPs [[209]23,[210]24,[211]25].
   Thus, it is possible that in cells expressing NUP98-HOXA9 and
   SET-NUP214 fusion proteins protein segregation after mitosis is
   affected, and that otherwise exclusively cytoplasmic proteins are
   retained in the nucleus.

   The results from GO and pathway enrichment analysis suggest that both
   NUP98-HOXA9 and SET-NUP214 interact with major regulatory proteins,
   such as members of the RNAPII complex, ribosomal proteins, and
   transcription factor complexes, suggesting a widespread effect on gene
   expression ([212]Figure 2, [213]Figure 3, [214]Figure 4, [215]Figure 5
   and [216]Figure 6). NUP98-HOXA9 and SET-NUP214 form dynamic nuclear
   foci that accumulate endogenous proteins [[217]23,[218]46]. These
   structures may represent factories that bring transcriptional and
   epigenetic regulators into close proximity to promote gene expression
   changes [[219]29,[220]30,[221]58]. Interestingly, our analyses suggest
   that p53-mediated transcription is a common target of NUP98-HOXA9 and
   SET-NUP214. The association of the fusion proteins with p53 might occur
   via their NUP98 and SET portions, respectively. Both proteins regulate
   TP53-mediated transcription, namely of the cyclin-dependent kinase
   inhibitor p21 (p21^CDKN1A), a DNA damage response and cell cycle
   regulator [[222]59,[223]60,[224]61]. p53 dysregulation is frequent in
   both ALL and AML, and loss of p53 function was previously reported to
   promote AML progression in a NUP98-HOXD13 mouse model, suggesting that
   it might contribute to NUP98-related leukemia
   [[225]62,[226]63,[227]64]. Nevertheless, despite the direct binding
   between SET and p53, the status of p53 signaling in SET-NUP214 leukemia
   has not been studied.

   Changes in gene expression by NUP98-HOXA9 might result from its
   interaction with major transcription factors, such as TFAP2A
   ([228]Table S2) and the AP-1 transcription factor complex ([229]Figure
   2 and [230]Figure 3), and proteins from cytoplasmic RNP granules
   ([231]Figure 2 and [232]Figure 3), where mRNA is processed for
   translation or targeted for degradation
   [[233]47,[234]65,[235]66,[236]67,[237]68]. Chromatin
   immunoprecipitation (ChIP) experiments showed that NUP98-HOXA9 (but not
   NUP98 nor HOXA9) binds at genomic regions that are adjacent to the
   TFAP2A sequence recognition motif, further supporting a potential
   interaction between the fusion protein and this transcription factor
   [[238]47]. TFAP2A regulates the transcription of several developmental
   genes and has been reported to promote HOX gene upregulation of
   clustered HOX genes in AML [[239]69]. Moreover, members of Wnt, MAPK,
   and ER signaling pathways are enriched with NHA9-BioID ([240]Figure 3).
   Among them, the DVL-1, TLE, isoforms 1–4, and TNRC6A proteins, which
   are common to Wnt and ER signaling, were exclusively found in the
   NHA9-BioID pool of biotinylated proteins, reinforcing the idea of a
   specific effect of NUP98-HOXA9 in these two signaling pathways. In line
   with our findings, previous work showed that in NUP98-HOXA-transduced
   hematopoietic stem cells, Wnt and ER signaling pathways were
   dysregulated, which correlated with cellular transformation [[241]67].

   Most of the NHA9-BioID proximal interactors are predicted to have at
   least one NES, supporting the idea that the biological consequences of
   CRM1 inhibition by the fusion protein might depend on the type of
   cargoes that become trapped in the nucleus. NES+ NHA9-BioID proximal
   interactors were enriched in cytoplasmic RNP granules and microtubule
   organizing center (MTOC)-associated proteins ([242]Figure 4B). As a
   MTOC docking protein, CRM1 facilitates microtubule nucleation at the
   NPC periphery [[243]70]. Several cell-death-related processes, such as
   autophagy and apoptosis, which are activated under stress, rely on the
   microtubule system [[244]71,[245]72]. Still, the biological
   implications of the nuclear retention of MTOC-related proteins, which
   has been reported in cancer cells, have remained unclear
   [[246]73,[247]74].

   Among SET-NUP214 proximal interactors, our results revealed an
   unexpected enrichment of mitochondrial proteins. This enrichment might
   be explained by changes in the communication between the mitochondria
   and the nucleus (anterograde/retrograde transport) that may be imposed
   by SET-NUP214 at the NPC. The molecular determinants of anterograde and
   retrograde transport are still unclear, but current evidence shows that
   some mitochondrial proteins also have functions in the nucleus, and
   that nuclear translocation of mitochondrial proteins is an emerging
   mitochondrial signaling pathway [[248]75,[249]76]. This system relies
   on the microtubule network that proteins and even entire organelles use
   to move from and towards the vicinity of the NE, respectively
   [[250]77]. Given the recently reported MTOC function of CRM1, it is
   tempting to hypothesize that SET-NUP214 might dock CRM1 at the
   cytoplasmic side of the NPC and support its MTOC function, thereby
   promoting the accumulation of mitochondrial proteins (or even entire
   mitochondria) in the periphery of the NPC. The possibility that
   SET-NUP214 is related to mitochondria dysregulation is supported by the
   observation that patients with SET-NUP214 T-related ALL are resistant
   to glucocorticoid (GC) therapy, which induces a metabolic shift from
   glycolysis to oxidative phosphorylation (OXPHOS; [[251]20,[252]78]).
   The AML-associated DEK-NUP214 fusion protein, which has the same NUP214
   portion as SET-NUP214, also promotes a shift from glycolysis to OXPHOS
   [[253]79]. Yet, in a recent proteomics report, no mitochondrial
   proteins were reported in the DEK-NUP214 interactome and dysregulation
   of the mammalian target of rapamycin (mTOR) was assumed as the main
   cause of metabolic shift [[254]80].

   Among SN214-BioID interactors, eight proteins (i.e., RAB6A, PCBD1,
   RPTOR, RIN1, NRAS, GCC2, LYN, and CRKL) were common to the DEK-NUP214
   interactome ([255]Table S4; [[256]80]). The association of these
   proteins with DEK-NUP214 was proposed to promote activation of several
   cancer associated pathways, such as AKT/mTOR, Src family kinase (SFK),
   ABL1, and c-MYC pathways [[257]81,[258]82,[259]83,[260]84]. Further
   studies will be necessary to confirm the same pathway profile
   activation by SET-NUP214. Nevertheless, given the structural
   similarities of the two chimeras, it is reasonable to expect a
   significant overlap in the biological processes that are affected by
   both fusion proteins [[261]12,[262]14,[263]85].

   Despite the identification of the above-mentioned numerous NUP98-HOXA9
   SN214-BioID interactors, our approach possibly falls short of others
   due to the promiscuous biotinylation activity of BirA^R118G. Although
   our normalization strategy accounts for differences in the expression
   of the BioID fusion proteins, we observed that some known NUP98-HOXA9
   and SET-NUP214 binding partners, such as CRM1 (for both fusion
   proteins) and MLL1 (for NUP98-HOXA9), were not strongly enriched in the
   NHA9-BioID and SN214-BioID fractions relative to BirA^R118G alone. We
   hypothesize that the same may occur with other NUP98-HOXA9 and
   SET-NUP214 interactors and for that reason we recognize that our BioID
   approach might miss relevant candidate binding partners. Nevertheless,
   the unspecific protein biotinylation by BirA^R118G alone, and the
   identification of several known binding partners enriched with
   NHA9-BioID and SN214-BioID, reinforce our approach in the study of
   NUP98-HOXA9 and SET-NUP214 proximal interactome.

5. Conclusions

   Overall, our report provides new data on the landscape of potential
   binding partners of nucleoporin fusion proteins, and suggests novel
   cellular processes and signaling pathways that may be affected by
   NUP98-HOXA9 and SET-NUP214. Although we identified several previously
   validated binding partners of both fusion proteins, experimental
   validation by gold-standard experimental methods, such as protein
   immunoprecipitation, is necessary to confirm the predicted interaction
   candidates. Our work unveils, for the first time, putative new players
   in nucleoporin-related leukemia, and provides the basis for a new
   understanding of the biological actions of nucleoporin fusion proteins.
   Our findings may be of particular relevance in the search for new
   druggable targets, such as the MAPK and ER pathways, that might be
   explored in the development of specific therapies for NUP98 and NUP214
   leukemia.

Acknowledgments