IPO3-mediated Nonclassical Nuclear Import of NF-κB Essential Modulator (NEMO) Drives DNA Damage-dependent NF-κB Activation*

Background: Nuclear import of NEMO is critical for DNA damage-dependent NF-κB signaling, but the mechanism remains unknown. Results: IPO3 binds NEMO, promotes its nuclear import, and is critical for DNA damage-dependent NF-κB activation. Conclusion: IPO3 is a nonclassical nuclear import receptor for NEMO. Significance: IPO3 is a new player in DNA damage-dependent NF-κB signaling with implications in cancer therapy. Activation of IκB kinase (IKK) and NF-κB by genotoxic stresses modulates apoptotic responses and production of inflammatory mediators, thereby contributing to therapy resistance and premature aging. We previously reported that genotoxic agents induce nuclear localization of NF-κB essential modulator (NEMO) via an undefined mechanism to arbitrate subsequent DNA damage-dependent IKK/NF-κB signaling. Here we show that a nonclassical nuclear import pathway via IPO3 (importin 3, transportin 2) mediates stress-induced NEMO nuclear translocation. We found putative nuclear localization signals in NEMO whose mutations disrupted stress-inducible nuclear translocation of NEMO and IKK/NF-κB activation in stably reconstituted NEMO-deficient cells. RNAi screening of both importin α and β family members, as well as co-immunoprecipitation analyses, revealed that a nonclassical importin β family member, IPO3, was the only importin that was able to associate with NEMO and whose reduced expression prevented genotoxic stress-induced NEMO nuclear translocation, IKK/NF-κB activation, and inflammatory cytokine transcription. Recombinant IPO3 interacted with recombinant NEMO but not the nuclear localization signal mutant version and induced nuclear import of NEMO in digitonin-permeabilized cells. We also provide evidence that NEMO is disengaged from IKK complex following genotoxic stress induction. Thus, the IPO3 nuclear import pathway is an early and crucial determinant of the IKK/NF-κB signaling arm of the mammalian DNA damage response.

The nuclear factor B (NF-B) 2 family of transcription factors regulates multiple biological and pathological processes, including immunity and oncogenesis (1)(2)(3)(4). In resting cells, an inactive NF-B dimer is localized in the cytoplasm in association with a member of the inhibitor of NF-B (IB) family. Cell stimulation can lead to the activation of the IB kinase complex containing IKK␣ and IKK␤ catalytic subunits and a regulatory subunit, NF-B essential modulator (NEMO), resulting in phosphorylation and proteasome-mediated degradation of IB to release NF-B to the nucleus to regulate transcription of target genes. This "canonical" pathway is widely engaged by different stress, inflammatory, and microbe-associated signals (5)(6)(7)(8).
Among the stress conditions that lead to NF-B activation are DNA-damaging, anticancer agents, such as ionizing radiation (IR), camptothecin, doxorubicin (DOX), and etoposide (VP16) (9 -11). This class of activators is unique among NF-B activators due to a signal initiation event being "DNA damage" in the nucleus, which results in the activation of cytoplasmically localized NF-B, a nuclear-to-cytoplasmic signaling pathway (12). Much progress has been made regarding the activation of the NF-B signaling pathway in response to multiple genotoxic agents (10,13). In particular, NEMO was found to perform a unique and early IKK-independent role in the nucleus, thus distinguishing the genotoxic stress signaling pathway from known canonical NF-B signaling pathways often triggered by the engagement of cognate cell surface receptors (14). In this model, genotoxic agents induce increases in nuclear localization of NEMO in the absence of the catalytic IKK subunits ("IKK-free" NEMO) via an undefined mechanism. NEMO is then modified by small ubiquitin-like modifier 1 (SUMO-1) via protein inhibitor of activated STAT y (PIASy), a nuclear resident SUMO ligase, and subsequently phosphorylated by ataxia telangiectasia mutated (ATM) (15,16). p53-induced death domain protein (PIDD) and RIP1 (receptor-interacting protein 1) have been proposed to increase NEMO SUMOylation and phosphorylation (17). Similarly, DNA damage-activated PARP1 (poly-(ADP-ribose) polymerase 1) assembles a nuclear signalosome containing NEMO, PIASy, and ATM, but not PIDD or RIP1, to promote NEMO SUMOylation and possibly phosphorylation (18). NEMO is also monoubiquitinated (15,16), which is associated with subsequent activation of cytoplasmic TAK1 (TGF␤-activated kinase 1) (19), an upstream kinase for IKK␤, in a manner dependent on Lys-63-or Met-1-linked polyubiquitination of different adapter proteins, including TRAF6 (TNF receptor-associated factor 6), RIP1, and ELKS (a protein rich in glutamate, leucine, lysine, and serine) (20 -23). Once the IKK complex is active, IB is phosphorylated and degraded to release NF-B to the nucleus via the canonical signaling cascade. Under genotoxic and oxidative stress circumstances, activated IKK and/or NF-B have been shown to modulate apoptosis (16 -18, 23-27), DNA repair (28,29), cell cycle arrest (30,31), and synthesis of inflammatory mediators (32,33), thereby contributing to oncogenesis (34,35), drug resistance (36), and premature aging (37,38).
The nuclear envelope in eukaryotic cells creates a barrier to factors accessing or exiting the nucleus; therefore, regulatory mechanisms exist to control trafficking of factors across this barrier (39). These mechanisms are particularly critical for regulation of signal transduction and transcription factors that dictate gene expression programs and therefore cellular phenotypes and functions. The major mechanism of nuclear import and export regulation depends on nuclear localization signal (NLS) sequences and nuclear export sequences in the cargo proteins and a family of ϳ20 karyopherins (nuclear import and export receptors) (40,41). The classical NLS sequence is composed of a short stretch of basic amino acid residues (lysines and arginines), either consecutively or in a bipartite manner, which is directly recognized by one of seven importin ␣ isoforms (KPNA1-7) via its armadillo repeat motifs (42,80). Co-crystal structure and biochemical analyses have provided essential characteristics of classical NLS sequences, and multiple predictive models have been developed for their identification (43). Importin ␣ is insufficient to mediate nuclear import of its cargos and requires the participation of importin ␤1 (KPNB1) through interaction with the importin ␤1 binding domain (44). The cargo-importin ␣-importin ␤1 complex is then transported through the nuclear pore complex into the nucleus, where the cargo and importin ␣ are dissociated from importin ␤1 by the binding of a small G protein, Ran, in its GTP-bound form, to complete the import process.
Alternatively, a cargo is also imported via the nonclassical import pathway involving the direct binding by ϳ11 importin ␤ family members in the absence of importin ␣ participation (40,45). The nonclassical NLS sequence are currently ill defined except for a few instances (40). For example, co-crystal structure analysis demonstrated that TRN1 (transportin 1, TNPO1, IPO2) makes extensive contacts with nonclassical NLS sequences known as the proline-tyrosine-NLS (PY-NLS) spanning 30 -40 amino acids via its HEAT repeat motifs (46,47). The sequence model suggested that the PY-NLS motif is present in potentially over 80 distinct proteins (47,48). TRN1 also recognizes cargos that do not have PY-NLS motifs (48), further diversifying the number of substrates it can recognize. In addition, a recent study identified an importin-independent pathway mediated by direct binding of some ankyrin repeat-containing proteins to Ran-GDP and NTF2 (nuclear transport factor 2) (49). Thus, multiple distinct mechanisms and pathways exist to import protein cargos.
Despite the discovery of nuclear NEMO as a mediator of a NF-B activation pathway induced by multiple genotoxic agents over a decade ago (14), the molecular mechanism by which "IKKfree" NEMO is inducibly localized to the nucleus remains an important unanswered question. Here we identify NEMO sequences that are critical for its nuclear localization and subsequent NF-B activation in response to genotoxic agents. We identify IPO3 (importin 3, transportin 2, TRN2, TNPO2), a nonclassical importin ␤ family member, as a critical NEMO nuclear import receptor and a driver of IKK/NF-B signaling in the mammalian DNA damage response. Finally, we provide evidence that NEMO is disengaged from the IKK complex to liberate IKK-free NEMO during genotoxic stress signaling.
Cell Culture, Transient Transfection, and Generation of Stable Cell Lines-NEMO-deficient 1.3E2 murine pre-B cells and stable clones expressing human NEMO-WT and mutant versions were cultured in RPMI1640 medium (GE Healthcare HyClone) supplemented with 10% (v/v) fetal bovine serum (GE Healthcare HyClone), 50 M ␤-mercaptoethanol, and antibiotics (100 IU/ml penicillin and 100 g/ml streptomycin), as described previously (50). HEK293 and HeLa cells were cultured in DMEM (Corning Cellgro) supplemented with 10% fetal bovine serum and antibiotics as described above. All cells were grown at 37°C in a 5% CO 2 humidified incubator (Thermo Scientific) except for HEK293 cells, which were grown on 0.1% (w/v) gelatin-coated dishes at 37°C in a 10% CO 2 humidified incubator. All cells were passaged at least twice weekly before reaching a cell density of 1 ϫ 10 6 cells/ml for suspension cells or reaching 90% confluence for adherent cells. HEK293 cells were transiently transfected by using a standard calcium phosphate precipitation method (12). To generate 1.3E2 or HeLa stable cell lines, the cells were transfected with NEMO-WT or NEMO mutants/pcDNA3.1(ϩ) or shRNAs against IPO3/pSUPERIOR.retro.puro (VEC-IND-0010, Oligo-Engine) by electroporation or Lipofectamine 2000 transfection reagent (11668019, Invitrogen), respectively. The transfected NEMO-reconstituted 1.3E2 cells were selected with G418 as described previously (51). The transfected HeLa cells were selected with puromycin as described previously (52).
Electrophoretic Mobility Shift Assay (EMSA) and Western Blot Analysis-EMSA and Western blot analysis were performed as described previously (53).
Immunofluorescence Analysis-1.3E2 cells stably reconstituted with either NEMO-WT or NEMO-DN were seeded onto 0.01% poly-D-lysine (P0296, Sigma)-coated coverslips and cultured overnight. After the cells were treated with 10 M VP16, the cells were gently washed with PBS prior to fixation with 2% formaldehyde in PBS for 30 min at room temperature. The fixed cells were permeabilized with PBS containing 0.5% Triton X-100 for 5 min. The permeabilized cells were blocked with blocking solution (10% goat serum, 1% BSA, and 1% gelatin) for 30 min at room temperature followed by incubation with anti-NEMO antibody diluted in 3% BSA by 1:100 for 1 h at room temperature. After washing the cells three times with PBS containing 0.1% Triton X-100, the cells were incubated with FITCrabbit antibody diluted in 3% BSA by 1:1000 for 1 h at room temperature in the dark. The cells were washed three times with PBS containing 0.1% Triton X-100, stained with Hoechst (H3570, Invitrogen) for 5 min at room temperature, and washed two times with PBS containing 0.1% Triton X-100. The coverslips were mounted on slides with Vecta Shield mounting medium (NC9611543, Fisher) and sealed with nail polish. Wild type or the stably IPO3-knocked down HeLa cells grown on coverslips were rinsed in 37°C PBS and fixed in 4% formaldehyde for 7 min at 37°C. The cells were permeabilized in 0.2% Triton X-100 in PBS for 6 min. Antibodies were diluted as above. Fluorescent images were collected by a Nikon Eclipse Ti microscope with a DS-Qi1 camera using ϫ40 or ϫ100 objective lens and analyzed by Nikon Elements software.
The relative nucleus and cytoplasmic signal intensities were analyzed as described previously (54). Briefly, binary masks were created from both DAPI and NEMO fluorescence images by standard thresholding algorithms in ImageJ. The DAPI binary mask was subtracted from the NEMO binary mask to create a combined "cytoplasmic" mask used to distinguish nucleus and cytoplasmic regions. Mean nuclear and cytoplasmic intensities were calculated by dividing the total fluorescence intensities in each region by their respective areas. The ratio of the means, referred to here as the intensity ratio IR, was then calculated for each cell in the population by dividing the mean nuclear intensity by the mean cytoplasmic intensity. Thus, cell populations with increased signals in the nucleus had increased fractions of cells having a shift of the histogram to the right relative to the histogram of unstimulated cells.
Purification of Recombinant Proteins-GST-tagged or tagless recombinant proteins were purified as described previously (51).
GST Pull-down and Immunoprecipitation Analyses-To test interaction of GST-NEMO with His-IPO3 or His-TRN1, human NEMO in pGEX6p-1 and IPO3 or TRN1 in pET28a plasmids were co-transformed into BL21, and the proteins were induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside at 22°C for 5 h. Bacterial lysates were prepared, and a GST pulldown assay was performed as described previously (51). To carry out immunoprecipitation analysis, His-IPO3 and tagless NEMO proteins in pRSFDuet TM -1 were expressed in BL21. An equal amount of bacterial lysates were incubated in 800 l of the IP buffer containing 2 g of anti-NEMO antibody. The mixture was then rotated for 5 h at 4°C, and then 20 l of protein G-Sepharose was added to the mixture. The binding mixture was rotated for another 2 h at 4°C. After four washes of the beads with the wash buffer, the immunoprecipitated proteins were detected by Western blot analysis.
Quantitative RT-PCR Analysis-HEK293 cells transiently transfected with either control siRNA (non-targeting siRNA) or siRNA against importin ␤ family members were left untreated or treated with VP16 or TNF␣ for the indicated times. Control HeLa cells or IPO3 knockdown stable HeLa clones were left untreated or treated with doxorubicin for 8 h. Total RNAs were prepared by an RNeasy miniprep kit (catalog no. 74104, Qiagen). cDNAs were synthesized from the total RNAs. Quantitative RT-PCR was performed and analyzed using a Bio-Rad CFX Connect real-time system. The mRNA levels of the samples were normalized to GAPDH mRNA levels and shown as -fold induction relative to untreated samples.
Preparation of Cytosolic Extracts from HeLa Cells-Cytosolic extracts from untransfected or IPO3-knocked down HeLa stable cell clones were prepared as described previously (55). Briefly, the harvested cell pellets from 4 ϫ 10 7 cells were washed twice with phosphate-buffered saline (PBS) and washed with 1 ml of wash buffer (10 mM HEPES, pH 7.3, 110 mM potassium acetate, 2 mM magnesium acetate, and 2 mM DTT). The cell pellets were resuspended in twice the volume of cell pellets using cold hypotonic lysis buffer (5 mM HEPES, pH 7.3, 10 mM potassium acetate, 2 mM magnesium acetate, 2 mM DTT, 1 mM PMSF) and then incubated on ice for 10 min. The cells were permeabilized on ice by adding 20 g/ml digitonin until ϳ90% of the cells were permeable to trypan blue (SV3008401, Fisher) in parallel aliquots. The permeabilized cells were centrifuged at 1500 ϫ g for 15 min at 4°C, and the supernatants were centrifuged again at 15,000 ϫ g for 1 h at 4°C. The final supernatants (0.5 ml) were dialyzed in 500 ml of transport buffer (20 mM HEPES, pH 7.3, 110 mM potassium acetate, 2 mM magnesium acetate, 1 mM EGTA, 2 mM DTT, 1 mM PMSF) with three changes of the buffer at 4°C with 3 h between each buffer exchange.
In Vitro Nuclear Import Assay-In vitro assay to measure nuclear transport of GFP-NEMO followed a previously detailed protocol (55) with modifications described below. Briefly, HeLa cells grown on coverslips were washed three times with ice-cold transport buffer and permeabilized with 20 g/ml digitonin at room temperature. After about 80% of the cells were permeabilized, the cells were washed three times with ice-cold transport buffer to remove digitonin. The cells were incubated at 30°C for 10 min and washed three times with transport buffer to remove the cytosol. The cytosol-depleted cells were then incubated with import reaction mix, including GFP-NEMO, 0.1 mM GTP, and ATP-regenerating system (10 mM ATP, 1 mg/ml creatine phosphate, and 15 units/ml creatine phosphate kinase) with or without cytosolic extracts. Recombinant GST-IPO3 protein was added to the import reaction for the rescue experiment. The cells with the import reaction mix were incubated at 30°C for 30 min in a humidified chamber. The import reaction was washed with transport buffer three times, and the cells were fixed with 3.7% formaldehyde in PBS for 15 min. The fixed cells were washed three times with PBS containing 0.1% Triton X-100 to remove non-imported GFP-NEMO. The nuclei were stained with Hoechst for 5 min at room temperature and washed three times with PBS containing 0.1% Triton X-100. Fluorescent images were collected by a Nikon Eclipse Ti microscope with a DS-Qi1 camera using a ϫ40 or ϫ100 objective lens and analyzed by Nikon Elements software.
The relative nuclear envelope and inside nuclear envelope signal intensities were analyzed as described previously (54) with one minor modification, where the "cytoplasmic" mask was replaced by a "nuclear envelope" mask as follows. A binary mask from the DAPI fluorescence image only was created by a standard thresholding algorithm in ImageJ. This original binary mask (DAPI-1 mask) was processed using a binary "erode" command to remove two pixels around the outer circumference of each cell, resulting in a secondary binary mask (DAPI-2). By subtracting DAPI-2 from DAPI-1, a mask was created consisting of "rings" exactly 2 pixels in thickness, representing the nuclear envelope. Subsequently, mean intensities of the nuclear envelope and the internal (non-envelope) region were calculated by dividing the total fluorescence intensities in each region by their respective areas, as before. The intensity ratio was then calculated for each cell in the population by dividing the mean internal intensity by the mean nuclear envelope intensity (similar in direction to the nuclear-cytoplasmic intensity ratio). Thus, cell populations with increased signals in the internal region showed a shift of the histogram to the right relative to the histogram of GFP-NEMO control in the absence of cytosolic extracts.
Clear Native (CN)-and Blue Native (BN)-PAGE-Discontinuous Tris-glycine CN-polyacrylamide gels were cast with 4% stacking gel and 6% resolving gel (acrylamide monomer final concentration). Samples for CN-PAGE were prepared by lysing cell pellets in TOTEX buffer (20 mM HEPES, pH 7.9, 350 mM NaCl, 1 mM MgCl 2 , 0.5 mM EDTA, 0.1 mM EGTA, 20% glycerol, 1% Nonidet P-40, 0.5 mM DTT, 1ϫ Halt TM protease inhibitor mixture) on ice for 30 min. Thirty g of total protein from each sample was mixed with CN-PAGE sample buffer (62.5 mM Tris-Cl, pH 6.8, 10% glycerol, 1 mM DTT, and ϳ0.01% bromphenol blue) and run until the dye front migrated to the bottom edge of the gel. For two-dimensional CN-PAGE, samples were electrophoresed as above prior to excising gel lanes. Excised lanes were soaked in Laemmli sample buffer for 15 min and laid over a 10% SDS-PAGE resolving gel. Samples were run until the dye front migrated to the lower edge of the resolving gel. Discontinuous blue native gels were prepared as described previously (56). Gels were poured with either imidazole or BisTris as base buffer, using the compatible cathode buffer for electrophoresis as appropriate. Samples were prepared by lysing cell pellets in BN-PAGE sample buffer (50 mM NaCl, 50 mM BisTris (or imidazole), pH 7.0, 750 mM 6-aminohexanoic acid, 1 mM DTT) on ice for 40 min. Samples were centrifuged at ϳ16,000 ϫ g for 15 min at 4°C. Protein concentrations were determined by a Bradford assay (5000006, Bio-Rad), and samples of total equal protein amount were brought to equal total volume with additional BN-PAGE sample buffer. Samples were loaded directly onto empty wells without running buffer and carefully overlaid with blue cathode buffer (50 mM Tricine, 7.5 mM imidazole or BisTris, 0.2% Coomassie Blue G-250, pH 7.0). Samples were electrophoresed in the cold room at constant voltage (50 V) until migration into the resolving gel. Gels were switched to constant current (15 mA) and run until the dye front migrated about one-third of the way through the resolving gel. Gels were stopped, and blue cathode buffer was replaced with clear cathode buffer (identical except with Coomassie Blue G-250 omitted), and electrophoresed until the blue dye front migrated to the bottom edge of the gel. Following electrophoresis, gels were equilibrated in Tris-glycine transfer buffer for about 15 min at room temperature. Gels were transferred to PVDF membrane in Tris-glycine ϩ 20% methanol for 1 h at 100 V. Membranes were briefly destained with methanol and destaining solution (25% methanol, 10% acetic acid in water) prior to blocking and incubation with primary antibodies for immunoblotting.
Statistical Analysis-Values are presented as means Ϯ S.E. with the indicated number of independent experiments. The statistical significance of differences between groups was determined by the Student's t test or Welch t test. p values less than 0.05 were considered statistically significant.

Atypical Nuclear Localization Signals Modulate NF-B
Activation by Genotoxic Agents-To identify potential NLS sequence in NEMO, we manually scanned its primary sequence but were unable to identify any classical NLS sequence that satisfies the criteria established by both co-crystal structure and biochemical analyses (reviewed in Ref. 43)). We next employed online NLS sequence search programs (listed in Ref. 57) and found that although most failed to identify any NLS sequence in NEMO, NucPred identified RKRG (aa 254 -257 in human NEMO) as a possible NLS sequence with a nuclear score of 87%, and PSORT II found RKRH (aa 357-360 in human NEMO) with the nuclear score of 74% (Fig. 1A). NucPred failed to identify RKRH, and PSORT II also failed to identify RKRG as a putative NLS sequence.
To test whether these putative NLS sequences are important for NF-B activation, we mutated RKRG (NLS1) to AAAG and RKRH (NLS2) to AAAA in NEMO and stably introduced the mutant constructs into NEMO-deficient 1.3E2 murine pre-B cells to determine whether these mutations affected nuclear NEMO localization and NF-B activation in response to genotoxic stress as we have done previously for the identification of other functional NEMO mutants (14,16). Based on similar expression levels of NEMO (Fig. 1B), which are critical to ensure comparable NF-B activation potentials by genotoxic agents (51), we identified multiple independent 1.3E2 clones harboring NEMO-WT (wild type NEMO), NEMO-NLS1, NEMO-NLS2, or NEMO-DN (single or double NLS mutants, respectively) and analyzed their NF-B activation capacities by EMSA. Whereas single mutants had mild effects (data not shown), NEMO-DN clones showed consistently reduced NF-B activation in response to VP16 (Fig. 1C), camptothecin, and IR (Fig. 1D) with comparable activation following treatment with bacterial LPS (Fig. 1, D and E). NF-B activation in NEMO-DN cells was also significantly reduced over various doses and the time course of VP16 treatment relative to NEMO-WT cells (Fig. 1, F and G). Similarly, phosphorylation of IKK␣/␤ and degradation of IB␣ in response to the DNAdamaging agents were reduced in NEMO-DN clones compared with NEMO-WT clones (Fig. 1D). In contrast to human NEMO, the murine version harbors NLS2 but apparently lacks NLS1; nevertheless, multiple 1.3E2 clones stably expressing mouse NEMO-NLS2 mutant also consistently showed reduced NF-B activation by VP16 but not LPS (Fig. 1H).
Atypical Nuclear Localization Signals Modulate NEMO Nuclear Localization in Response to Genotoxic Agents-Next, to test whether NF-B activation defects in NEMO-DN clones were arising from reduced nuclear localization of NEMO, we analyzed the subcellular localization of NEMO by indirect immunofluorescence using anti-NEMO antibody ( Fig. 2A). We have recently developed a method to semiquantitatively analyze relative nuclear and cytoplasmic signal intensities of protein factors in individual cells (54). By measuring the changes in nuclear-cytoplasmic intensity ratios before and after VP16 treatment in NEMO-WT and NEMO-DN clones (ϳ1000 -3000 individual cells at each time point for each clone), we detected increases in nuclear NEMO intensities in NEMO-WT clones in response to VP16 treatment, but such increases were significantly reduced in NEMO-DN clones (Fig. 2, B and C). Thus, combined mutations of NLS1 and NLS2 led to reduced nuclear translocation of NEMO and NF-B activation by genotoxic stress inducers.
Screening of Importin ␣ and ␤ Members and Identification of IPO3 as Critical for NF-B Activation by Genotoxic Agents-To identify a specific import receptor mediating NEMO import via NLS1/2 and NF-B signaling by genotoxic agents, we next established a set of criteria to identify a nuclear import receptor for NEMO. First, deficiency of the expression of the import receptor reduces NF-B activation in response to genotoxic stimuli. Second, the import receptor associates with NEMO-WT but not NEMO-DN mutant. Third, silencing of the expression of the import receptor reduces NEMO nuclear translocation in response to genotoxic stimuli. Fourth, in vitro nuclear import of NEMO in digitonin-permeabilized cells is defective with cytosolic extracts obtained from the import receptor-deficient cells. And finally, a recombinant purified form of the import receptor complements in vitro nuclear import of NEMO in digitonin-permeabilized cells supplemented with import receptor-deficient cytosolic extracts.
Because the algorithms employed to detect NLS1/2 in NEMO were based on classical NLS sequences (57), we first tested the involvement of this import pathway. To test whether any of the importin ␣ (KPNA) family members can interact with NEMO, we co-transfected human embryonic kidney cells (HEK293) cells with Myc-NEMO and each of the FLAG-KPNA1-7 isoforms, treated the cells with VP16, and performed anti-Myc immunoprecipitation followed by anti-FLAG immunoblotting. We found that KPNA1, -6, and -7 were capable of interacting with NEMO above the background level (Fig. 3A). However, these interactions were not inducible by VP16 treatment (Fig. 3B). Moreover, KPNA1, -6, and -7 could not distinguish NEMO-WT from NEMO-DN by co-immunoprecipitation analysis (Fig. 3B). Finally, knockdown of the KPNAs using siRNA SmartPools against KPNA1, -6, and -7, singly or in combination, failed to reduce NF-B activation by VP16 treatment (Fig. 3C). Thus, whereas KPNA1, -6, and -7 inter-action with NEMO could be detected by co-IP analysis, importin ␣ family members failed to satisfy two of the first three criteria above. These results may not be surprising, given that sequences surrounding NLS1 and NLS2 of NEMO do not fully match the consensus sequences of importin ␣ association (43). Our results thus suggested the possibility of the involvement of a nonclassical nuclear import mechanism in the import of NEMO.
We next screened functional relevance of the nonclassical nuclear import receptors, the importin ␤ (IPO␤) family members, by transfecting HEK293 cells with siRNA SmartPool against each of the importin ␤ family members. These transfected cells were then treated with VP16 and analyzed for their impacts on NF-B activation by EMSA. We found that siRNAs against IPO3, IPO7, IPO8, and IPO11 reduced NF-B activation after VP16 treatment (Fig. 4A). Whereas the inhibition of NF-B activation in response to VP16 treatment by IPO11 (Fig.  4B) and IPO3 (see below) siRNA SmartPools was consistently reproducible, that of IPO7 and IPO8 siRNAs was not (data not shown). Knockdown efficiencies and specificities of siRNA SmartPools against individual IPO␤ members were assessed by quantitative RT-PCR (Fig. 4D), showing variable knockdown efficiencies. Although IPO11 knockdown showed significant  JULY 17, 2015 • VOLUME 290 • NUMBER 29 JOURNAL OF BIOLOGICAL CHEMISTRY 17973 reduction in NF-B activation after VP16 treatment, we failed to observe IPO11-NEMO interaction (Fig. 4C).

Regulatory Mechanism of NEMO Nuclear Import
We further validated the role of IPO3 in IB␣ degradation and NF-B activation in response to treatment with VP16 or IR but not TNF␣ (Fig. 5A). Moreover, the efficiency of IPO3 knockdown by individual siRNAs present in SmartPool as measured by Western blot analysis (Fig. 5B) correlated with inhibition of NF-B activation by VP16 treatment (Fig. 5C). The functional significance of IPO3 was further validated by significant reduction of NF-B-dependent proinflammatory cytokine induction with VP16 but not TNF␣ stimulation following siRNA-mediated knockdown of IPO3 (Fig. 5D). Although we attempted to rescue the effects of IPO3 siRNAs by transiently transfecting a siRNA-resistant IPO3 construct, we were unable to restore IPO3 function. Thus, although our shared effects of multiple siRNAs and the correlation between IPO3 knockdown efficiency and inhibition of NF-B signaling support the role for IPO3, further analyses are required to establish whether or not IPO3 siRNA effects were truly due only to IPO3 knockdown. Equal amounts of the cell lysates were immunoprecipitated using Myc antibody, and co-immunoprecipitated FLAG-KPNAs were visualized by Western blot analysis using FLAG antibody. 2% of cell lysates were used for input. C, HEK293 cells transfected with either control or siRNAs against KPNA1, -6, or -7 were left untreated or treated with 10 M VP16 for the indicated times or TNF␣ (10 ng/ml, 30 min). NF-B activation was measured by EMSA. Nonspecific binding complex (n.s.) served as a loading control.

IPO3 Directly and Inducibly Interacts with NEMO-WT, but
Not NEMO-DN, in Response to Genotoxic Agents-If IPO3 is a nonclassical nuclear import receptor for NEMO, it should directly interact with NEMO. To test this, we next performed co-expression and co-immunoprecipitation analysis as before. We found that IPO3 was able to interact with NEMO in a VP16inducible manner (Fig. 6A). In contrast, IPO3 failed to bind NEMO-DN over the background level. The VP16-inducible interaction between endogenous IPO3 and stably expressed NEMO was observed in cytoplasmic extracts, consistent with the assembly of IPO3-NEMO complex prior to nuclear import (Fig. 6B). Furthermore, recombinant NEMO-WT, but not NEMO-DN, interacted with His-IPO3 (Fig. 6C). Finally, cotransformation in E. coli and GST pull-down analysis demonstrated that IPO3, but not closely related TRN1, was able to associate with NEMO (Fig. 6D). These studies identified IPO3 as capable of interacting with NEMO in a manner dependent on NLS1/2 sequences.
IPO3 Mediates NEMO Nuclear Import-To test whether IPO3 could satisfy the remaining criteria required for being a nuclear import receptor of NEMO, we employed HeLa cells frequently used for in vitro import assays (55,58,59). We first established HeLa cell clones stably expressing shRNAs against IPO3. Two shRNA constructs corresponding to siRNAs capable of targeting two different sequences of IPO3 mRNA (siRNA #1 and #3 in Fig. 5B) were generated and transfected into HeLa cells to isolate stable clones with reduced IPO3 expression (shIPO3 #1-3 and #3-6 clones; Fig. 7A). Two independent shRNAs minimized the involvement of shared off target effects. Because HeLa cells efficiently activated NF-B with doxorubicin, but not with VP16, we employed doxorubicin in subsequent studies. The efficiency of IPO3 knockdown in independent shIPO3 clones as measured by Western blot analysis correlated with inhibition of NF-B activation by doxorubicin treatment (Fig. 7A). Functional signaling defects in these shIPO3 HeLa cell clones were demonstrated by the nearly complete inhibition of IL-8 induction in response to doxorubicin treatment (Fig. 7B). Abundant TRN1 was detected in each of these IPO3 knockdown clones but was unable to rescue IPO3 deficiency (Fig. 7, A and B), consistent with the inability of TRN1 to directly interact with NEMO (Fig. 6D).
To test the requirement of IPO3 in nuclear translocation of NEMO, subcellular localization of NEMO in response to doxorubicin treatment was measured in shIPO3 3-6 HeLa cell clones and control cells as before. NEMO was localized in the cytoplasm in unstimulated cells (Fig. 7, C and D). Whereas NEMO or co-immunoprecipitated FLAG-IPO11 was visualized by Western blot analysis using NEMO or FLAG antibodies, respectively. 2% of cell lysates were used for input. D, knockdown efficiencies of IPO␤ family members in HEK293 cells as in A were assessed by qRT-PCR. Total RNAs were extracted and analyzed for expression of each IPO␤ using qRT-PCR. Each IPO␤ expression was normalized to GAPDH expression, and a graph representing relative -fold change of IPO␤ expression compared with untransfected HEK293 cells was shown with the error bars representing S.E. from technical duplicates.
increased nuclear NEMO signals were detected from 30 to 60 min poststimulation in control HeLa cells, NEMO nuclear translocation was much reduced in shIPO3 3-6 clones (Fig. 7,  C and D). To further test whether IPO3 is necessary to induce efficient nuclear translocation of NEMO, we next permeabilized HeLa cells with digitonin to deplete cytosolic factors (including nuclear import receptors) as described previously (55) (Fig. 8, A and B). We then assessed nuclear import of recombinant GFP-NEMO (Fig. 8C) in the absence and presence of cytosolic extracts obtained from IPO3-depleted and control HeLa cells. In the absence of cytosolic extracts, GFP-NEMO signals were detected on the nuclear envelope but not in the nucleus (Fig. 8, A and B, b). It is unclear whether NEMO detection on the nuclear envelope in this in vitro assay is due to specific or nonspecific interaction. Nevertheless, it provided a convenient signal to mark the nuclear envelope and for quantification of NEMO within the nuclear compartment (see "Experimental Procedures"). The addition of cytosolic extracts prepared from control HeLa cells induced nuclear accumulation of GFP-NEMO (Fig. 8, A and B, c). However, the addition of cytosolic extracts prepared from IPO3-deficient HeLa cells failed to do so (Fig. 8, A and B, e and g). These results further verified the necessity of IPO3 for nuclear import of NEMO. Because we were unable to rescue the effect of IPO3 knockdown by a siRNA-resistant IPO3 construct, it was crucial to determine whether IPO3 could rescue NEMO nuclear import in the nuclear import assay. To this end, we generated GST-IPO3 recombinant protein (Fig. 8C) and supplemented it into IPO3-deficient HeLa cytosolic extracts. Under these conditions, nuclear import of NEMO was restored in both shIPO3 FIGURE 5. IPO3 deficiency is associated with defective NF-B activation by genotoxic agents. A, HEK293 cells transfected with either control or IPO3 siRNA were left untreated or treated with VP16 (VP; 10 M, 2 h), IR (10 grays, 2 h), or TNF␣ (10 ng/ml, 30 min). IB␣ degradation and NF-B activation were measured by Western blot analysis and EMSA, respectively. Oct-1 binding and tubulin levels served as loading controls. B, HEK293 cells were transfected with either control or individual siRNAs against IPO3. Knockdown efficiency of IPO3 was assessed by Western blot analysis using IPO3 antibody. A tubulin blot served as loading control. n.s., nonspecific. C, HEK293 cells transfected with either control or individual siRNA 3 or 5 in B were left untreated or treated with VP16 (VP, 10 M, 2 h). NF-B activation was measured by EMSA. Oct-1 binding was used as a loading control. D, HEK293 cells transfected with either control or IPO3 siRNA were left untreated or treated with VP16 for the indicated times in triplicate. IL-8 expression was analyzed by qRT-PCR, normalized to GAPDH, and plotted on a graph representing relative -fold change of IL-8 expression. Results are presented as means Ϯ S.E. (error bars). *, p Ͻ 0.02 relative to IPO3 knockdown samples.
1-3 and 3-6 HeLa cytosolic extracts (Fig. 8, A and B, f and h). Therefore, IPO3 was able to restore the NEMO nuclear import deficiency of IPO3 knockdown cells. Thus, these results indicated that IPO3 was able to satisfy all of the criteria for being the nuclear import receptor for NEMO.
Genotoxic Stress Induces the Appearance of IKK-free NEMO-The above finding demonstrated that IPO3-mediated nuclear import of NEMO does not require genotoxic signals in digitonin-permeabilized cells. Whereas NEMO normally exists as a regulatory subunit of the cytoplasmic IKK complex (60), nuclear translocation of IKK␤ is not generally detected (14,16,18). Moreover, known nuclear NEMO interactors, including ATM, PIASy, and PARP1, all engage NEMO at the N terminus in a region overlapping/adjacent to the sequence involved in IKK␤ interaction in a manner exclusive to IKK␤ binding (15,16,18). Thus, "IKK-free" NEMO has been hypothesized to translocate into the nucleus (10,14,16). It is possible that there is a pool of preexisting IKK-free NEMO in unstimulated cells, or alternatively NEMO is disengaged from cytoplasmic IKK complexes prior to nuclear entry mediated by IPO3.
To test these possibilities, we analyzed NEMO fractions using non-denaturing native gel electrophoresis (CN-PAGE). In extracts obtained from NEMO-WT 1.3E2 cells, NEMO migrated to a prominent position ("A") (Fig. 9A). Upon stimulation with VP16 treatment, a newly migrating NEMO form ("B") was also detected. We next subjected the CN-PAGE lane to separation in a second dimension by conventional SDS-PAGE and found that IKK␤ co-migrated with the A, but not B, form of NEMO (Fig. 9B). Thus, the B form represents IKK-free NEMO, which was produced following cell stimulation with the genotoxic agent.
We noted significant differences in the total amount of NEMO detected in CN-PAGE although the traditional SDS-PAGE demonstrated that the total NEMO levels were equivalent. CN-PAGE separation relies on both size and charge (61), but BN-PAGE can separate protein complexes according to size only due to the presence of anionic Coomassie Blue dye at neutral pH (56,61). As in CN-PAGE, we could detect both A and B forms of NEMO by BN-PAGE analysis using VP16-stimulated NEMO-WT cell extracts, and the overall amount of the B form was much smaller than the A form in BN-PAGE compared with that found in CN-PAGE (Fig. 9C). Again, IKK␤ did not migrate to the B position of NEMO in BN-PAGE (Fig. 9C). The use of IKK␣ Ϫ/Ϫ IKK␤ Ϫ/Ϫ double knock-out (IKK-DKO) mouse embryo fibroblasts confirmed that NEMO in the absence of IKK subunits (i.e. IKK-free) migrated faster in BN-PAGE similar to the B form (Fig. 9D). Thus, the lack of IKK-free NEMO in unstimulated cells, coupled with the appearance of the IKK-free B form of NEMO in response to VP16 treatment, indicated that a fraction of IKK-free NEMO was generated under the genotoxic stress condition. Overall, our data suggest FIGURE 6. IPO3 interacts with NEMO in a manner dependent on NLS1/2 sequences. A, HEK293 cells co-transfected with FLAG-IPO3 and either Myc-NEMO-WT or -DN were left untreated or treated with 10 M VP16 (VP) for the indicated times. Equal amounts of cell lysates were immunoprecipitated (IP) using Myc antibody and analyzed by Western blot analysis using FLAG antibody. 2% of cell lysates were used for input. B, HEK293 cells stably expressing Myc-NEMO were left untreated or treated with VP16 for 1 h. Cytoplasmic extracts was used for co-immunoprecipitation using Myc antibody followed by Western blot analysis using IPO3 and NEMO antibodies. Tubulin served as a loading control. C, His-IPO3 and either NEMO-WT or -DN were co-expressed in E. coli and immunoprecipitated using either rabbit IgG or NEMO antibodies. Co-immunoprecipitated His-IPO3 was visualized by Western blot analysis using IPO3 antibody. 10% of cell lysates were used for input. D, GST-NEMO and either His-IPO3 or -TRN1 were co-expressed in E. coli, and a GST pull-down assay was performed. Co-purified His-IPO3 or -TRN1 was visualized by Western blot analysis. 5% of bacterial lysates were used for input.
that IKK-free NEMO is imported into the nucleus by IPO3 in response to genotoxic stress induction.

Discussion
Despite the proposal in 2003 that NEMO plays an essential, early, and IKK-independent role in the nucleus to mediate NF-B activation in response to genotoxic stimuli (14) and further documentation of the critical function of nuclear NEMO by multiple groups (15-18, 20, 23, 30), how NEMO gained access to the nucleus in a signal-selective manner remained an important unanswered question. In the present study, we uncovered a nonclassical nuclear import receptor, IPO3, as a binding partner and mediator of nuclear import of NEMO under genotoxic stress conditions. The following evidence supports this conclusion. First, we identified a NEMO mutant (NEMO-DN) that is defective in nuclear import and supporting activation of IKK and NF-B in response to genotoxic agents. Second, we were unable to reveal the role for importin ␣ family members in NF-B activation by genotoxic FIGURE 7. IPO3 deficiency is associated with defective nuclear import of NEMO. A, control or HeLa cell clones stably expressing the indicated IPO3 shRNA constructs were left untreated or treated with doxorubicin (8 g/ml, 2 h). NF-B activation and the protein levels of IPO3 and TRN1 were measured by EMSA and Western blot analysis, respectively. B, HeLa cells in A were left untreated or treated with doxorubicin (4 g/ml, 8 h) in triplicate, and induction of IL-8 was analyzed as in Fig. 5D. Error bars, S.E. *, p Ͻ 0.01 relative to IPO3 knockdown stable cell clones. C, HeLa cells from A were left untreated or treated with 4 g/ml doxorubicin for the indicated times and immunostained with NEMO antibody. Nuclei were visualized by Hoechst staining. Scale bar, 5 m. D, nuclear to cytoplasmic NEMO intensity ratios from ϳ1100 cells of each condition in C were quantified as in Fig. 2B, and histograms were plotted. Percentages of the cell population that showed nuclear translocation of NEMO in each condition are indicated in blue (4 g/ml doxorubicin for 30 min) or red (4 g/ml doxorubicin for 60 min).

FIGURE 8. IPO3 induces nuclear translocation of NEMO in vitro.
A, HeLa cells were permeabilized by digitonin, and cytosolic extracts from either control HeLa cells or shIPO3 HeLa clones were added to support nuclear import of GFP-NEMO in the absence or presence of GST-IPO3 recombinant protein. Nuclei were visualized by Hoechst staining. Scale bar, 5 m. B, inside nuclear envelope to nuclear envelope NEMO intensity ratios from ϳ300 cells of each condition were quantified as in Fig. 2B, and histograms were plotted for corresponding panels in A. Gray, histogram of GFP-NEMO in the absence of cytosolic extract; brown, histogram of GFP-NEMO in the presence of cytosolic extract with or without GST-IPO3. C, Coomassie staining of GFP-NEMO and GST-IPO3 used in A with BSA standards.
agents. Third, among the importin ␤ family members, IPO3 and IPO11 knockdowns prevented NF-B activation by genotoxic agents. However, IPO3 but not IPO11 directly interacted with NEMO in vitro and inducibly in vivo in response to stimulation with genotoxic agents. Fourth, the interaction between IPO3 and NEMO was inhibited by the NEMO DN mutation. Fifth, IPO3 deficiency significantly reduced nuclear translocation of NEMO in digitonin-permeabilized cells. Finally, recombinant IPO3 induced NEMO nuclear translocation in an in vitro nuclear import assay.
However, the defects in NEMO nuclear translocation and NF-B activation associated with NLS1/2 mutation or IPO3 knockdowns were incomplete. The RKR motif is probably too short to confer full IPO3 binding specificity because known nonclassical NLS sequences contact IPO␤s with much longer amino acid sequences contacting multiple HEAT repeats (40,47). For example, TRN1 contains 20 HEAT repeats and makes extensive contacts with PY-NLS sequences containing 30 -40 aa residues via 13 HEAT repeats (46,47). Although IPO3 knockdowns showed incomplete inhibition of NF-B activation in HEK293 cells as measured by EMSA, NEMO nuclear translocation, NF-B-dependent IL-8 induction, and in vitro nuclear import of GFP-NEMO were almost completely prevented in multiple independent shIPO3 stable HeLa cell clones. Thus, an incomplete knockdown of IPO3 in HEK293 cells probably accounted for the partial defects in NF-B activation. However, we cannot rule out the possibility of the presence of an alternative import receptor for NEMO in a cell type-selective manner, which could possess a minor role in NEMO nuclear import and compensate for IPO3 deficiency.
IPO3 was originally cloned as transportin 2 due to its 84% amino acid sequence identity with TRN1 (62), and therefore TRN1 represented the prime candidate as a compensatory nuclear import receptor for NEMO. Indeed, both transportins have been reported to redundantly import HuR and ALS-associated FUS (fused in sarcoma) into the nucleus (63)(64)(65). However, Siomi et al. (62) reported that heterogeneous nuclear ribonucleoprotein A1 binds to TRN1 but not IPO3, although another group later reported that heterogeneous nuclear ribonucleoprotein A1 could weakly bind IPO3 in vitro (65). Apobec-1 complementation factor interacts with IPO3, but not TRN1 (66), being a rare IPO3-selective cargo. Two variable regions between TRN1 and IPO3 lie within HEAT repeats 8 and 17 of TRN1, where the latter repeat of TRN1 is involved in the detection of PY-NLS (47). In the case of NEMO, IPO3 deficiency was clearly not compensated for by the presence of TRN1 in shIPO3 stable HeLa cell clones. Moreover, TRN1 failed to interact with NEMO. Thus, the differences in the critical HEAT repeats between IPO3 and TRN1 could be responsible for the differential interaction with NEMO. Whereas well established cargos for TRN1 and IPO3 are involved in RNA metabolism (48), NEMO represents a unique cargo of IPO3 in a non-RNA processing pathway, thereby identifying genotoxic stress-induced NF-B signaling as an important biological pathway regulated by IPO3, not shared by TRN1.
We also found a strong inhibitory phenotype of NF-B activation when IPO11 siRNA SmartPool was used in our siRNA screen in HEK293 cells. IPO11 knockdown did not prevent NF-B activation by TNF␣ stimulation, showing selectivity toward the genotoxic signaling pathway. Despite this robust phenotype, we were unable to observe any evidence for IPO11-NEMO interaction with different experimental approaches (co-immunoprecipitation, GST pull-down, etc.). This suggests that IPO11 might have a role in importing some factor (or factors), distinct from NEMO to mediate NF-B activation by genotoxic agents. The potential role of off target effects of IPO11 SmartPool cannot be excluded, although multiple independent IPO11 siRNAs could also display similar phenotypes (data not shown). IPO11 is known to play a role in nuclear import of ubiquitin-charged human class III ubiquitinconjugating enzymes, such as UbcM2 (67), UbcH6, and UBE2E2 (68), as well as ribosomal protein rpL12 (69) and the gag polyprotein of Rous sarcoma virus (70). However, none of these has been previously implicated in NF-B activation by genotoxic stimuli. Although the mechanisms accounting for nuclear import of some of the players implicated in NF-B activation by genotoxic agents (e.g. ATM and PARP1) have been elucidated (71,72), those accounting for the import of other known players (e.g. PIASy, PIDD, and RIP1) remain unclear. Thus, a further investigation of the role of IPO11 could reveal an import mechanism for any of the known factors or new player(s) in this signaling pathway.
Nuclear NEMO has been generally suggested to be free of IKK subunits in the genotoxic stress-induced NF-B signaling pathway (14,16,18), although the presence of nuclear IKK␣ is well documented in cytokine signaling (73,74), and a recent study implicated the role of nuclear IKK␤ in mediating UV- induced NF-B signaling (75). While we employed multiple approaches to detect the presence of IKK-free NEMO (e.g. quantitative immunodepletion and gel filtration), the use of CN-and BN-PAGE permitted the detection of an IKK-unbound fraction of NEMO in response to genotoxic stress conditions. It is possible that some posttranslational modification(s) could be induced on one or more of the IKK subunits, including NEMO, to induce NEMO disengagement. However, it is important to note that it was technically challenging to detect IKK-free endogenous NEMO in response to VP16 treatment, which instead of being soluble seemed to have partitioned to the insoluble cell pellet fraction in the CN/BN-PAGE buffer systems (not shown). Because a 6ϫ Myc tag appended to NEMO adds 18 negative charges at neutral pH, it may be possible that we fortuitously uncovered the presence of free NEMO by CN/BN-PAGE because of the presence of this tag. The tag per se does not affect NF-B activation because other tags, such as 2ϫ HA and GFP, permit equivalent NF-B activation when reconstituted into 1.3E2 cells (data not shown). Further studies are therefore required to determine how the generation of IKKfree NEMO is regulated in genotoxic stress conditions and how charge properties may affect detection and solubility of IKKfree NEMO in the CN/BN-PAGE buffer system.
Although the role of nuclear NEMO was originally discovered in the context of genotoxic stress-induced NF-B signaling (14), recent studies have further implicated its potential role in mediating additional processes, including the pathogenesis of myeloid dysplastic syndrome and acute myeloid leukemia (35), resistance to histone deacetylase inhibitors (36), and premature aging syndromes (37,38). Adding to the classical importin ␣-dependent nuclear import pathway mediating the nuclear import of some of the NF-B family members (76 -79), coupled with the recent identification of the Ran-GDP/ankyrin repeat ("RaDAR") import pathway for the IB family of proteins (49), our current identification of the IPO3 nonclassical import pathway in regulation of NEMO nuclear import not only expands our fundamental understanding of the nuclear trafficking aspects of the NF-B signaling system but also points to new therapeutic strategies against cancer resistance as well as the aforementioned medical conditions by targeting the NF-B system-related nuclear import mechanisms.