A polybasic motif in ErbB3-binding protein 1 (EBP1) has key functions in nucleolar localization and polyphosphoinositide interaction

We reveal the identification of a polybasic motif necessary for polyphosphoinositide interaction and nucleolar targeting of ErbB3 binding protein 1 (EBP1). EBP1 interacts directly with phosphatidylinositol(3,4,5)-triphosphate and their association is detected in the nucleolus, implying regulatory roles of nucleolar processes.

Human EBP1 was originally identified as a binding partner of the tyrosine kinase receptor ErbB3 [50] and as the homologue of murine p38-2AG4 [47]. PA2G4 encodes two splice variants, a long and predominant form p48, and a shorter and minor form p42 (homologue to p38-2G4) which differs in its N-terminus by the lack of the first 54 amino acids [51]. The two isoforms have distinct sub-cellular localizations, p48-EBP1 is present in the cytoplasm, nucleus as well as the nucleolus [52,53], whereas p42-EBP1 is restricted to the cytoplasm, where it is targeted for degradation via its ubiquitination [51,54]. p48-EBP1 has the ability to translocate from the cytoplasm to the nucleus upon the activation of ErbB3 [50] or upon high cell density in oral squamous carcinoma cells [55].
In the present study, we show that EBP1 binds directly to several PPIn species via two different PPIn interaction sites consisting of lysine-rich PBRs located in the two termini of the protein. The two PBRs have a different PPIn-interaction profile and contribute to EBP1 nucleolar localization, albeit differently. EBP1 interacts particularly with PtdIns(3,4,5)P 3 via its C-terminal PBR and this association is localized in the nucleolus. The Cterminal PBR is mutated in endometrial cancer and we showed that this partially prevents PtdIns(3,4,5)P 3 -interaction as well as its nucleolar localization. These data demonstrate that the EBP1 PBRs have a dual function as a PPIn interaction motif and nucleolar localization signal, and imply that the regulation of EBP1-mediated nucleolar processes is potentially regulated by PtdIns(3,4,5)P 3 .

Plasmids, cloning and site-directed mutagenesis
The phospholipase Cδ1 pleckstrin homology domain (PLCδ1-PH) cloned in pGEX-4T was obtained from Dr A.Z. Gray (University of Dundee, UK) and the general receptor for phosphoinositides-1 PH domain (GST-GRP1-PH) cloned in pGEX-4T3 was from Dr J. Hastie (MRC, University of Dundee, UK). pGEX-4T2-hEBP1 and pEGFP-C2-hEBP1 were from M. Squatrito [53]. The N-and C-terminal EBP1 fragments were amplified by PCR from pGEX-4T2-hEBP1 using primers flanked by EcoRI restriction sites and cloned into pGEX-4T1. All mutants were generated by QuickChange sitedirected mutagenesis (Agilent Technologies) according to the manufacturer's instructions and verified by sequencing using ABI Prism BigDye Terminator version 3.1 cycle sequencing kit (Applied Biosystems). GRP1-PH was amplified by PCR from pGEX-4T3-GRP1-PH using primers flanked by BglII and SalI restriction sites and subcloned into pEGFP-C2-NLS. All primers are listed in Supplementary Table S1.

GST-tagged protein expression and purification
GST-PLCδ1-PH and GST-GRP1-PH were expressed and purified as described previously [24]. GST-EBP1, full length (FL), Nand C-terminal fragments, WT and mutants were transformed into Escherichia coli BL21-RIL DE3 and bacterial cultures were grown at 37 • C and further induced with 0.5 mM isopropyl-E-D-thiogalactopyranoside for 3 h at 37 • C. Bacterial pellets were resuspended in 50 mM Tris pH 7.5, 2 mM EDTA, 1 mM DTT and 1x Sigma protease inhibitor cocktail, sonicated three times for 30 s at 4 • C and centrifuged at 4400 g for 10 min at 4 • C. GSTtagged proteins were purified with glutathione-agarose 4B beads from an overnight pull down, eluted with 50 mM Tris pH 8.0, 100 mM NaCl, 0.5 mM DTT and 10 mM reduced glutathione, and analysed by SDS/PAGE and Coomassie staining for purity. For NMR studies, E.coli were grown in M9 minimal medium, supplemented with 6 g/l Na 2 HPO 4 , 3 g/l KH 2 PO 4 , 0.5 g/l NaCl, 0.25 g/l MgSO 4 , 1 g/l 98%-enriched ( 15 NH 4 ) 2 SO 4 and/or 4 g/l 13 C6-glucose to produce 15 N-and/or 13 C-uniformly labelled GST-C-terminal domain of EBP1. Further protein expression and purification was performed using the same protocol as described above.

Lipid overlay assays
Lipid overlay assay was carried out using PIP Strips TM (Echelon Biosciences) spotted with 100 pmol of each of the 7 PPIns in addition to other phospholipids, and PIP Arrays TM spotted with 1.56-100 pmol of PtdIns or each of the 7 PPIns. PIP strips TM and arrays were incubated with blocking buffer (3 % fatty acid-free BSA (Sigma A6003) in TBS-T (50 mM Tris pH 7.5, 150 mM NaCl, 0.1 % Tween-20) for 1 h at room temperature. PIP Strips TM were incubated with 1.5 μg/ml GST-tagged protein or dialysed neomycin extracts in the same buffer overnight at 4 • C. Detection of GST-tagged proteins and EBP1 (from neomycin extracts) was performed with an anti-GST-HRP conjugated antibody (Abcam, ab3416, 1:50000) and an anti-EBP1 antibody (M. Squatrito, 1:800) respectively, both diluted in blocking buffer. PIP arrays TM were incubated with anti-PtdIns(3,4,5)P 3 (Echelon, #Z-P345b, 1:10000) followed by anti-mouse IgG-HRP (1:10000), both diluted in blocking buffer made in PBS-T (137 mM NaCl, 2.68 mM KCl, 8 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 0.1% Tween-20). Six washes of 5 min each with TBS-T or PBS-T were performed after incubations with protein and antibody. The protein-lipid interactions were visualized using a west pico or femto chemiluminescent substrate and a Bio-Rad ChemiDoc TM XRS + Imaging System from Bio-Rad and the ImageLab TM Software Version 3.0.
Relative binding of EBP1 to the phospholipids was quantified by densitometry using ImageJ software (http://rsb.info. nih.gov/ij). The data were normalized to background signals for each blot.

NMR spectroscopy
The NMR sample contained 0.14 mM uniformly 15 N-and/or 13 C-labelled GST-CTD-EBP1 in 50 mM Tris buffer S + 5.5 containing 100 mM NaCl, 0.5 mM 2-mercaptoethanol and 90 % H 2 O/10 % D 2 O. NMR spectra were acquired at 298 K on a Bruker Avance spectrometer operating at proton frequency of 600.13 MHz using the acquisition parameters provided in Supplementary  Table S2. The spectrometer was equipped with a TCI 5-mm triple resonance cryo-probe with pulse field gradients along the z-axis. Spectra were recorded and processed in TopSpin 2.1 (Bruker Biospin). 1 H, 13 C and 15 N backbone resonance assignments for the protein were determined using CARA (Computer Aided Resonance Assignment) version 1.8.4.2 [56]. Secondary structure propensities (SSP) were calculated with the Cα and Cβ chemical shifts as input into the SSP algorithm [57]. To monitor protein-lipid binding 1 H-15 N heteronuclear single quantum coherence (HSQC) spectra acquired in the absence and presence of 0.08 mM diC16-PtdIns(3,4,5)P 3 or 0.33 mM diC8-PtdIns(3,4,5)P 3 (Echelon Biosciences) were subsequently compared and analysed. The peak intensities were measured in CARA using peak-fitting algorithm [56]. The robustness of the intensity values was tested by using different fit parameters and the variation in the output did not exceed 2 %.

Cell culture and transfection
MEL cells were cultured with DMEM supplemented with 10 % foetal bovine serum (FBS) and antibiotics (50 000 units of both penicillin and streptomycin) at 37 • C with 5 % CO 2 . AU565 cells were cultured under the same conditions but in RPMI-1640 medium. For transfections, cells were plated in six-well plates and transfected with 1-2 μg DNA and XtremeGene 9 (Roche) at 3:1 ratio for 24 h. Treatment with the pan-PI3K inhibitor LY294002 (10 μM) or DMSO (0.2 % (v/v)) was performed 4 h post-transfection.

Nuclear fractionation and neomycin extraction
Nuclei were isolated according to a method by Mukai et al. [58] with some modifications. Cells were washed two times in PBS and once briefly in buffer A (10 mM Hepes pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 340 mM sucrose, 10 % glycerol). Cells were resuspended in buffer A containing 0.1 % Triton X-100, 1 mM DTT, 5 μg/ml leupeptin and 5 μg/ml aprotinin, left to swell for 5 min on ice and centrifuged at 1300 g for 5 min at 4 • C. Nuclei were washed quickly with retention buffer (20 mM Tris pH 7.5, 70 mM NaCl, 20 mM KCl, 5 mM MgCl 2 and 3 mM CaCl 2 [59]). Nuclei were incubated twice in retention buffer for 30 min at room temperature, split into two equal fractions and further incubated in the presence or absence of 5 mM neomycin (trisulfate salt, Sigma N6386) for 30 min at room temperature. Samples were centrifuged at 9600 g for 5 min at 4 • C and supernatants were collected. For lipid overlay assays, neomycin supernatants were dialysed twice against 20 mM HEPES pH 7.5, 150 mM NaCl, 5 mM EDTA and 0.1 % NP-40.

SDS/PAGE and Western immunoblotting
Proteins were resolved by SDS/PAGE and transferred to nitrocellulose membranes. Membranes were blocked with 5 % non-fat milk, incubated with primary antibodies overnight at 4 • C and with secondary antibodies conjugated to HRP for 1 h at room temperature. Protein detection was performed by chemiluminescence using the SuperSignal West Pico Chemiluminescent Substrate (Pierce) and imaged using the Molecular Imager ® ChemiDoc TM XR + Imaging System and the ImageLab TM Software Version 3.0 (Bio-Rad).

EBP1 binds directly to phosphoinositides
We have previously identified EBP1 as a potential PtdIns(4,5)P 2binding protein by combining PtdIns(4,5)P 2 pull down from neomycin-displaced nuclear proteins and quantitative MS [24]. The peptides which identified EBP1 are shown in Supplementary Figure S1 and include a peptide which only matches the long p48 variant. We first validated that p48-EBP1 could be displaced by neomycin from the nucleus by Western immunoblotting analyses ( Figure 1A). We next assessed EBP1 as a direct PPIn-interacting protein by performing lipid overlay assays on phospholipid-immobilized strips ( Figure 1B) using recombinant p48-GST-EBP1. As shown in Figure 1(C), GST-EBP1 bound to several PPIn species including the three mono-phosphorylated PPIns and PtdIns(3,5)P 2 . Binding to PtdIns(4,5)P 2 , PtdIns(3,4)P 2 , phosphatidic acid (PA) and phosphatidylserine (PS) is weaker and not always detected ( Figures 1C and 1D). GST showed no interaction by itself. Control binding experiments for PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 were also performed using specific probes, i.e. GST-fused PH domains of PLC-δ1 and GRP1 respectively. As shown in Figure 1(C), the PH domain of PLC-δ1 bound strongly to PtdIns(4,5)P 2 , whereas very weak binding was observed for PtdIns3P, Ptdins5P and PtdIns(3,5)P 2 . The PH domain of GRP1 interacted strongly with PtdIns(3,4,5)P 3 and very weakly with both PtdIns(4,5)P 2 and PtdIns(3,5)P 2 . Considering that EBP1 interacts with most anionic phospholipids, we tested the effect of increasing NaCl concentration on these interactions ( Figures 1E and 1F). Doubling the NaCl concentration had overall little effect on interaction but binding to the three mono-phosphorylated PPIns and PtdIns(3,5)P 2 was essentially the same as measured by retention of the protein to the strips, although PA and PS binding was reduced. Adding 400 mM NaCl nearly abolished all interaction detected except for PtdIns4P and PtdIns(3,5)P 2 . These results suggest that the interactions with PtdIns3P, PtdIns5P and particularly with PtdIns4P and PtdIns(3,5)P 2 have the highest affinity for the FL EBP1 protein.

EBP1 binds to PPIns via two lysine-rich PBRs
EBP1 does not harbour PPIn-binding modules, such as PH, PX or FYVE domains [60], that could account for the observed interactions. However, stretches of basic amino acids denoted as PBRs [61] or KR-motifs following the sequence K/R-(X n = 3-7 )-KXKK, have also been implicated in PPIn-binding via electrostatic interactions [62]. Such basic amino acid stretches have since then been identified in several nuclear proteins in complex with PPIn [24] including the nuclear proteins ING2 [30], Pf1 [28], SAP30L [29], and more recently in BASP1 [43], UHRF1 [63] and TAF3 [42] (Figure 2A). EBP1 has a KR-motif in the unstructured C-terminal region ( 364 RKTQKKKKKK 373 ) [24], named C-term PBR, as well as a reverse KR-motif situated on a protruding loop of the N-terminal part ( 65 KKEKEMKK 72 ), named N-term PBR ( Figures 1A and 1B). These two PBRs are highly conserved, suggesting a functional importance (Supplementary Figure S2). Point mutations of three basic residues out of five to alanines within the PBR of SAP30L ( 87 KRKRK 91 → 87 KAAAK 91 ) led to a significant decrease in binding to monophosphorylated PPIns [29]. In order to investigate if the C-terminal PBRmotif of EBP1 played a similar role in PPIn-binding, four out of six lysines were substituted to alanines in the FL GST-EBP1, resulting in the following quadruple C-term PBR mutant 368 KAAAAK 373 (FL-C-K4A). FL-C-K4A was tested for its PPIn-binding properties by lipid overlay assay in parallel with the wild-type (WT) FL GST-EBP1. This mutant did not show significant change in PPIn-binding compared with WT except for a variable decrease in binding to PtdIns(3,5)P 2 (results not shown). A mutant was also produced in the N-terminal PBR motif residing on the protruding loop ( Figure 2B), resulting in the FL-N-K2A mutant 65 AAEKEMKK 72 . Again, this mutant did not show any change in the PPIn-binding pattern compared with WT (results not shown). These results suggested that both motifs may act independently in binding to PPIns. An N-terminal fragment (amino acids 1-351), harbouring the N-term PBR motif and a C-terminal fragment (amino acids 352-394) containing the Cterm PBR, were therefore created ( Figure 2C and Supplementary Figure S3) and tested in lipid overlay assays ( Figure 2D). As shown in Figure 2(D) both fragments retained the ability to bind to PPIns, including the three mono-phosphorylated PPIns and PtdIns(3,5)P 2 for both the N-and C-terminal fragments. The C-terminal fragment was able to bind to the remaining PPIns but with variable intensity. The C-K4A mutant abolished PPIn-binding, when tested in the C-term construct ( Figure 2D). The N-K2A mutant showed a great decrease in binding when tested in the N-term construct (results not shown) and binding was completely blocked when an additional lysine at K68 was substituted to alanine ( Figure 2D). Finally, when the C-K4A and N-K3A mutants were both introduced in the FL protein, PPIn interaction was completely abolished.

Nucleolar localization of EBP1 is mediated predominantly by the C-terminal PBR
EBP1 has previously been shown to localize in the nucleolus of HeLa and NIH-3T3 cells [53] and we confirmed this finding in the breast cancer cell line AU565 cells by co-immunostaining EBP1 with the nucleolar protein NPM. As shown in Figure 3(A), EBP1 was detected both in the cytoplasm, and in punctate foci within NPM-stained nucleoli. The first 48 amino acids in p48-EBP1 were previously shown to be necessary for its nuclear targeting whereas the region spanning amino acids 301-394 was shown to be responsible for its nucleolar localization [53]. Using the nucleolar localization sequence detector (NoD [64]), a nucleolar localization sequence (NoLS) was predicted in EBP1 in amino acids 357-385, which lies within the region previously found to be responsible for nucleolar targeting ( Figure 3B and Supplementary Figure S4A). This putative NoLS is well conserved and, interestingly, contains the C-terminal PBR ( Figure 3B and Supplementary Figure S4B). We argued therefore that this PPIn-binding motif could contribute to the localization of EBP1 in the nucleolus. AU565 cells were transfected with FL EGFP-tagged WT, C-K4A and N-K3A mutants as well as the N-K3A/C-K4A double mutant and examined by fluorescence microscopy ( Figure 3C). In contrast with EGFP alone, which was found in both the cytoplasm and nucleus, EGFP-EBP1 WT and mutants exhibited overall three different patterns of localization (Supplementary Figure S5). The different localization patterns were quantified for WT and each of the mutants ( Figure 3D). The first pattern (pattern #1) is characterized by the appearance of EBP1 in the cytoplasm, the peri-nuclear area and the nucleolus, as reported previously [51,53]. This pattern was observed in 37 % of cells expressing WT-EBP1 but did not occur in any of the mutants (Figures 3C and 3D). The second pattern (pattern #2) includes either a restricted localization in the cytoplasm with an intense peri-nuclear signal (pattern #2a, Supplementary Figure S5 (Figures 3C and 3D). Some differences were however observed among these mutants for this pattern. The C-K4A mutant was not only excluded from the nucleolus but also strongly retained in the rest of the nucleus (pattern #3c Supplementary Figure S5, Figure 3C). In contrast, the N-K3A mutant and double mutant, which were devoid of nucleolar localization, did not allow the nuclear retention of EBP1 (pattern #3d, Supplementary Figure S5, Figure 3C). The protein levels of all three mutants were lower compared with WT, and the N-K3A mutant had the lowest decrease compared with the other two EBP1 mutants ( Figure 3E). These results suggest therefore that the C-and N-term PBRs contribute to the nucleolar localization of EBP1, albeit in a different manner. The C-term PBR has in addition nuclear export properties.

PtdIns(3,4,5)P 3 is localized in the nucleoplasm and nucleolus
EBP1-p48 has previously been shown to bind to NPM [65]. Considering that NPM was also shown to be pulled down in a complex with PtdIns(3,4,5)P 3 from isolated nuclei [46], and that we showed that the C-terminal PBR of EBP1 could bind to PtdIns(3,4,5)P 3 , at least among other PPIns ( Figure 2D), we argued that PtdIns(3,4,5)P 3 could be found in the nucleolus, in association with EBP1. Consequently, we used the PtdIns(3,4,5)P 3 -specific probe GRP1-PH, fused it with NLS-EGFP and examined its localization by immunofluorescence. In contrast with NLS-EGFP, which was predominantly diffuse in the nucleus, the NLS-EGFP-GRP1-PH was found in 81 % of cells in the nucleus and strongly in the nucleolus together with nucleolin ( Figures 4A and 4B). Furthermore, administration of the pan-PI3K inhibitor LY 294002 significantly impaired the nucleolar localization of NLS-EGFP-GRP1-PH by more than half (32 %).
We also examined the localization of both PtdIns(3,4,5)P 3 and the nucleolar protein nucleolin by immunostaining in AU565 cells using a protocol previously established to detect detergentresistant PPIns in the nucleus [18]. PtdIns(3,4,5)P 3 was detected in the cytoplasm, nucleoplasm and nucleolus together with nucleolin ( Figure 4D). The specificity of the anti-PtdIns(3,4,5)P 3 antibody was validated firstly by lipid overlay assays using PIP arrays ( Figure 4C). Secondly, pre-incubation of the antibody with different PPIns showed that nucleoplasmic and nucleolar staining was abolished by the presence of PtdIns(3,4,5)P 3 but not by PtdIns3P and PtdIns(3,4)P 2 (Supplementary Figure S6). Furthermore, the localization of p110β, one of the class IA phosphoinositide 3-kinase (PI3K) catalytic subunits producing PtdIns(3,4,5)P 3 , has previously been reported to be present in the nucleus [66]. Following immunostaining of AU565 cells, we detected p110β in the cytoplasm and strongly in the nucleolus together with NPM ( Figure 4E). In summary, these results show the specificity of the anti-PtdIns(3,4,5)P 3 antibody utilized, and that the nucleolar sites detected by the antibody are PtdIns(3,4,5)P 3 . The presence of p110β in nuclei further substantiates the existence of nucleolar PtdIns(3,4,5)P 3 .

EBP1 co-localizes partially with PtdIns(3,4,5)P 3 in the nucleolus
Considering that EBP1 and PtdIns(3,4,5)P 3 were independently detected in nucleoli, we sought to determine if they co-localized by immunofluorescence. Using confocal microscopy, EBP1 was found to partially co-localize with PtdIns(3,4,5)P 3 within nucleoli, suggesting therefore a potential association ( Figure 5A). Binding of recombinant EBP1 to PtdIns(3,4,5)P 3 is however weaker compared with binding to other PPIns (Figures 1D and 2D) and we considered that endogenous EBP1 may bind to PPIns in a different profile. To further examine the PPIn-binding properties of nuclear EBP1, we performed lipid overlay assays using neomycin-displaced protein extracts obtained from AU565 nuclei followed by detection with an anti-EBP1 antibody ( Figure 5B). The pattern of interaction was similar to the recombinant protein, and in particular to the C-terminal fragment, and showed binding to most PPIns including PtdIns(3,4,5)P 3 . These extracts were also resolved by Western immunoblotting to demonstrate the specificity of the anti-EBP1 antibody ( Figure 5C). Taken together, these results show that endogenous EBP1 and PtdIns(3,4,5)P 3 can also associate in nucleoli.

NMR analyses of EBP1-PtdIns(3,4,5)P 3 interaction
To establish further the interaction of EBP1 with PtdIns(3,4,5)P 3 at the molecular level, high resolution NMR was used. The Cterm PBR was examined since we have shown that it binds better to PtdIns(3,4,5)P 3 in lipid overlay assays and it strongly affects the nucleolar localization of EBP1, where EBP1 and PtdIns(3,4,5)P 3 co-localize. Figure 6(A) shows a typical 1 H-15 N HSQC fingerprint spectrum of the 15 N isotopically labelled GST-C-terminal fragment of EBP1 (see Supplementary Figure  S7 for more details). In order to first assign each amide signal to a specific residue in the protein, standard NMR experiments  were performed to detect 13 C chemical shifts and sequential connectivities. Heteronuclear 3D NMR experiments led to the chemical shift assignment of 1 H, 15 N and 13 C nuclei for 37 out of 43 (86 %) residues in the C-terminal fragment of EBP1. Ala 352 , Glu 353 , Leu 354 , as well as the following residues from the PBR, Lys 369 , Lys 370 , Lys 371 , were not assigned due to signal broadening beyond detection, which could be attributed to the intermediate solvent or conformational exchange rates of the protein. Altogether, the number of amide proton cross-peaks, as well as nearly complete assignment of the protein, demonstrated that the observed signals originated almost exclusively from the target protein and not from GST (see also Supplementary Figure  S7). The narrow signal dispersion of the spectrum, observed for the backbone amide protons (cross-peaks positioned within 8.7-7.7 ppm range in the 1 H dimension), indicates an overall disordered state of the protein. For proteins without a stable, folded structure, chemical shifts are essential parameters encoding local conformational propensities of the protein in solution [67,68]. To predict these propensities for the C-terminal fragment of EBP1, the obtained chemical shifts were processed to evaluate its secondary structure propensity (SSP) score using the SSP program [57]. Figure 6(B) illustrates the SSP score calculated for each assigned residue of the C-terminal fragment based on the combined Cα and Cβ chemical shifts. According to the SSP data, the C-terminal fragment has a largely disordered conformation except for the N-terminal region of the fragment (aa 355-365) showing a slight propensity to form an α-helix, consistently with the X-ray data for this region [69]. We next mapped the interaction site of the C-terminal EBP1 with either diC16-( Figure 6C) or diC8-PtdIns(3,4,5)P 3 ( Figure 6E). Addition of either long or short acyl chain PtdIns(3,4,5)P3 resulted in a decrease in signal intensity for all cross-peaks ( Figures 6D and 6F), which could be due to resonance broadening caused by the presence of the lipid. Nevertheless, differences in signal ratios were observed across the C-term PBR. The most pronounced signal intensity changes were observed when diC16-PtdIns(3,4,5)P 3 was added, in particular for residues located in the N-terminus part and the PBR (Lys 355 , Ala 356 , Leu 357 , Leu 358 , Gln 359 , Ala 362 , Ser 363 , Lys 365 , Thr 366 , Gln 367 , Lys 368 , as well as Lys 372 , Lys 373 , Ala 374 , Ser 375 , Lys 376 , with a ratio <0.2, and for Ser 360 , Ser 361 and Arg 364 with a ratio < 0.4). Similar effects were observed for the PBR residues when diC8-PtdIns(3,4,5)P 3 was added, but less intensity changes were overall detected in residues in the N-terminal part. These results validate that protein-lipid contacts involve most importantly electrostatic forces between lysines and phosphates. Hydrophobic interactions between nonpolar N-terminal residues and with long acyl chains, notably in the region covered by 356 ALL 358 , may also be involved, particularly with long hydrocarbon chains.
A tumour-associated mutant within the C-terminal PBR affects EBP1 nucleolar localization EBP1 has previously been implicated to have tumour suppressor properties in different cancer types including salivary adenoid cystic and hepatocellular carcinoma (HCC), as well as prostate and breast cancer [70][71][72][73]. Consistently, the levels of EBP1 have been shown to be down-regulated in prostate cancer and HCC while its overexpression induced a decrease in cell proliferation in breast, thyroid and HCC cancer cells [70,73,74].
In contrast, recent studies have shown opposite role for EBP1 in cell proliferation since knockout mice present growth retardation [75] and overexpression induces proliferation in muscle stem cells [76]. PA2G4, the gene encoding EBP1, is mutated at a very low frequency (0.24 % in all tumours). A mutation was however identified in the C-term PBR, i.e. K372N ( Figure 7A), in an endometrial tumour sample. We therefore introduced this point mutation in EGFP-EBP FL or GST-EBP1 C-terminal constructs and examined its effect on EBP1 compartmentalization ( Figure 7B) and PPIn binding ( Figures 7C and 7D and Supplementary Figure S8). The K372N point mutant exhibited a more pronounced pattern #2 (cytoplasmic with either diffuse or no signal within the nucleus) in 93 % of K372N-expressing cells compared with 61 % in WT. Concomitantly, a decrease in pattern #1 (cytoplasmic, peri-nuclear and nucleolar) was observed with 7 % in K372N-expressing cells, compared with 38 % in WT. In contrast with the C-K4A mutant, this point mutation was however not sufficient to affect nuclear export as it did not lead to the retention of EBP1 in the nucleoplasm. In addition, PPIn interaction was overall only slightly reduced when the mutation was introduced in the C-terminal EBP1 construct but interestingly a stronger decrease in interaction was observed for PtdIns(3,4,5)P 3 , as well as for PtdIns(3,4)P 2 and PA ( Figures 7C  and 7D). In summary, the K372N point mutant has an intermediate localization and PPIn-binding properties compared with the quadruple C-K4A mutant. The occurrence of this mutation in cancer may however point to the importance of this PBR in EBP1 function.

DISCUSSION
We have previously reported that short motifs consisting of basic residues or PBRs following the sequence motif K/R-(X n = 3-7 )-K-X-K/R-K/R were implicated in PPIn-binding of nuclear proteins, suggesting a mode of interaction for PPIninteracting proteins localized in the nucleus [24]. In the present study, we have identified two PBRs involved in the interaction of EBP1 with PPIns via lysine residues. The N-term PBR lies within a protruding loop although the C-term PBR is part of an unstructured sequences described by Monie et al. [77]. The two PBRs demonstrate the same profile of interaction with PPIns including the three mono-phosphorylated species and PtdIns(3,5)P 2 . In addition the C-terminal PBR was more promiscuous and interacted with all the other PPIns albeit more weakly. This pattern of interaction is consistent with other studies showing the importance of PBRs for PPIn binding in other nuclear proteins such as ING2 [30], Pf1 [28], SAP30L [29], and more recently UHRF1 [63] and TAF3 [42]. In particular, mono-phosphorylated PPIns interact with SAP30L, Pf1, TAF3 and ING2 whereas PtdIns(3,5)P 2 interacts with Pf1 and TAF3. The consistency of the pattern of interaction of these PBRs with certain PPIns is rather interesting and may hence point to specificity due to the position of the phosphate on the inositol ring for short basic motifs. Clusters of basic residues in PBRs may hence preferentially accommodate interaction with either one phosphate or PtdIns(3,5)P 2 with sufficiently spaced phosphates. Interestingly, SAP30L binds to mono-phosphorylated PPIns via a stretch of three basic residues within its PBR which also functions as an NLS adjacent to its DNA binding domain. The C-terminal PBR of EBP1, which contains six adjacent lysines, binds more extensively to PPIns and includes in addition the other bis-phosphorylated as well as PtdIns(3,4,5)P 3 . The higher number of lysines in this motif may hence offer more binding probabilities and alternatives to several PPIns with differently spaced phosphates on the inositol ring, at least in vitro.
The two PBRs may act independently and may provide different cellular functions to EBP1. Considering that PPIns are known to target proteins to different sub-cellular localizations, the two PBRs may regulate the different cellular locations of EBP1 allowing for potential roles in the nucleolus and cytoplasm bound to ribosomes [78]. We showed that the mutated N-terminal PBR restricted EBP1 mostly to the cytoplasm and exhibited either little or a diffuse nuclear signal in the majority of cells. The N-terminal PBR may hence cooperate with the first 48 amino acids, including the two lysines K20 and K22, shown previously to be necessary for the nuclear localization of EBP1 [53]. The C-terminal PBR is clearly important for the nucleolar localization of EBP1, and the sequestration of the C-terminal PBR mutant in the nucleoplasm coincident with its nucleolar exclusion suggest that this PBR also plays a role in nuclear export of EBP1. Considering that the first 48 amino acids in EBP1 are critical for nuclear targeting [53], the C-terminal PBR mutant is therefore able to reach the nucleus but is however unable to be retained in the nucleolus. Another nuclear protein, SAP30L, was shown to harbour a NoLS, located C-terminal of its NLS, and to contribute to PPIn interaction in addition to the NLS/PBR described previously [29]. The NoLS in SAP30L consists of the following polybasic sequence 120 RRYKRHYK 127 and when all these residues were mutated to alanines, SAP30L was excluded from the nucleolus but retained in the nucleus [79], which is consistent with the behaviour of the C-terminal PBR of EBP1. These studies raise the possibility of multi functions for NoLS beyond nucleolar targeting, including nuclear export as well as PPIn interaction.
The importance of the C-term PBR in the localization of EBP1 is highlighted by the occurrence of a mutation in endometrial cancer, K372N, which we have demonstrated to be important in both PPIn interaction and nucleolar localization. The importance of EBP1 compartmentalization in cancer takes precedence in salivary cancer in which EBP1 is sequestered in the cytoplasm in tumour areas whereas adjacent normal cells localize EBP1 in both the cytoplasm and nucleus [80]. Subnuclear details could however not be identified in the present study to distinguish nucleolar staining. The sub-cellular localization of EBP1 has moreover been suggested to be important for its function and its nucleolar localization correlates with its role in cell proliferation suppression [53]. Based on this, we also suggest that the C-terminal PBR is involved in regulating the cellular function of EBP1 by inducing changes in its sub-cellular localization. Cancer cells may hence benefit from targeting the C-terminal motif to alter the sub-cellular localization of EBP1 and perhaps more specifically by preventing EBP1 from entering the nucleolus. The role of EBP1 in the nucleolus remains to be identified.
The molecular mode of retention of EBP1 in the nucleolus is not clear but it is tempting to suggest that PtdIns(3,4,5)P 3 interaction could play a role. Firstly, EBP1 association with PtdIns(3,4,5)P 3 was demonstrated in several ways: (1) the Cterm PBR binds to this PPIn species in vitro, as shown by lipid overlay assays and NMR analyses, (2) endogenous nuclear EBP1 has binding capacity as shown by lipid overlay assays. Secondly, partial co-localization of endogenous EBP1 with PtdIns(3,4,5)P 3 within the nucleolus was observed, and this could be consistent with a functional association. In addition, NMR studies revealed that other residues located N-terminal of the PBR, including the nonpolar 356 ALL 358 motif, made contact with PtdIns(3,4,5)P 3 containing 16-carbon but not 8-carbon acyl chains. This would suggest that the acyl chains contribute to the interaction together with the head group. The participation of acyl chains in protein-PPIn interaction is a mechanism that has previously been suggested in several nuclear proteins identified to bind diC16-PtdIns(3,4,5)P 3 without the involvement of structured PI-binding domain [27]. Thus, the identification of nuclear proteins which can potentially accommodate the acyl chains of PPIns may provide an explanation for the presence of PPIns in the non-membranous environment of the nucleus and notably the nucleolus.
The presence of PtdIns(3,4,5)P 3 in the nucleus has been reported previously [16] and in particular in cells treated with γ -irradiation [81] but the presence of this PPIn and the PI3K catalytic subunit p110β in the nucleolus is reported for the first time in the present study. A minor pool of nuclear PtdIns(4,5)P 2 has previously been detected in the nucleolus of asynchronously growing cells [18,[82][83][84], which could hence be the source of PtdIns(3,4,5)P 3 production in this nuclear site. We have initially identified EBP1 as a potential PtdIns(4,5)P 2 -binding protein by pull down assay and mass spectrometry [24] and we have shown in the present study that the C-terminal PBR of EBP1 can bind directly to both PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 , the two PPIns found to be present in the nucleolus. EBP1 may therefore associate with PtdIns(4,5)P 2 as well as PtdIns(3,4,5)P 3 in this nuclear site. We observed that the pattern of detection of nucleolar PtdIns(3,4,5)P 3 varies in asynchronous AU565 ( Figure 4C) and HeLa (results not shown) cells and can consist of either a punctate and intense signal within nucleoli or a more diffuse distribution in both the nucleoplasm and nucleolus. We suggest therefore that EBP1 may associate differentially with PtdIns(3,4,5)P 3 or PtdIns(4,5)P 2 under certain conditions of the cell cycle. Consistently p110β has been previously reported to be active during G1 to S phase transition of the cell cycle [85]. We suggest therefore that changes in p110β activity could explain the differential pattern of PtdIns(3,4,5)P 3 in the nucleolus and nucleoplasm and regulate the function of EBP1 and we are currently pursuing this line of study.
Although the localization of EBP1 in the nucleolus can be due to the interaction of its C-terminal PBR with PtdIns(3,4,5)P 3 , another mode of retention of EBP1 is possible via rRNAinteraction. The C-terminal region of EBP1 spanning amino acids 361-394 was shown to be necessary for RNA interaction [77], including the processed rRNA species 18S, 28S and 5.8S [53]. This region comprises the C-terminal PBR described in the present study and the lysines present within the PBR shown to be involved in both nucleolar retention and PPIn interaction could be responsible for RNA binding via electrostatic interactions. Considering that this PBR is responsible for both PPIn interaction and nucleolar retention and is implicated in nucleic acid binding, we would suggest that PtdIns(3,4,5)P 3 and rRNA compete for binding to the C-terminal PBR. This scenario has indeed been reported for the HIV-1 viral protein Gag where binding of RNA to a highly basic region was demonstrated as a mechanism to prevent PtdIns(4,5)P 2 -mediated binding to the plasma membrane [86]. In addition, PtdIns(3,4,5)P 3 interacts with NPM via lysine residues [46], which are part of a basic, intrinsic disordered region involved in RNA interaction [87]. Finally, PtdIns(3,4,5)P 3 binds ALY, a protein regulating mRNA export via basic residues [35], and this interaction contributes to ALY-mediated recognition of specific mRNA transcripts for their nuclear export [37]. In light of our results obtained with the C-term PBR of EBP1 and studies in NPM and ALY, it is tempting to suggest that PtdIns(3,4,5)P 3 may regulate protein-RNA interaction utilizing basic motifs in disordered regions which have dual functions in PPIn and RNA interaction.
In conclusion, we have shown that the PBRs identified in EBP1 have a dual function as they contribute to PPIn interaction, potentially via electrostatic interactions, as well as nucleolar localization. Considering that sub-cellular localization often correlates with function, PBRs may provide a molecular mechanism allowing EBP1 to switch between different subcellular compartments and functions due to their interaction with PPIns. In addition, this is the first report providing evidence of the presence of PtdIns(3,4,5)P 3 as well as the class I PI3K catalytic subunit p110β in the nucleolus. EBP1 binds PtdIns(3,4,5)P 3 and this association is detected in the nucleolus. Our data imply novel regulation of nucleolar functions by PtdIns(3,4,5)P 3 , which are lines of research that we are currently pursuing.