Nuclear speckles and nucleoli targeting by PIP₂–PDZ domain interactions

Abstract

PDZ (Postsynaptic density protein, Disc large, Zona occludens) domains are protein–protein interaction modules that predominate in submembranous scaffolding proteins. Recently, we showed that the PDZ domains of syntenin-1 also interact with phosphatidylinositol 4,5-bisphosphate (PIP₂) and that this interaction controls the recruitment of the protein to the plasma membrane. Here we evaluate the general importance of PIP₂–PDZ domain interactions. We report that most PDZ proteins bind weakly to PIP₂, but that syntenin-2, the closest homolog of syntenin-1, binds with high affinity to PIP₂ via its PDZ domains. Surprisingly, these domains target syntenin-2 to nuclear PIP₂ pools, in nuclear speckles and nucleoli. Targeting to these sites is abolished by treatments known to affect these PIP₂ pools. Mutational and domain-swapping experiments indicate that high-affinity binding to PIP₂ requires both PDZ domains of syntenin-2, but that its first PDZ domain contains the nuclear PIP₂ targeting determinants. Depletion of syntenin-2 disrupts the nuclear speckles–PIP₂ pattern and affects cell survival and cell division. These findings show that PIP₂–PDZ domain interactions can directly contribute to subnuclear assembly processes.

Introduction

PDZ (Postsynaptic density protein, Disc large, Zona occludens) domains are protein–protein interaction modules of 90 amino acids (aa), containing six β strands and two α helices. It is well established that PDZ domains generally interact with the C-terminal end of the cytoplasmic tail of transmembrane proteins. The C-terminus of the interacting peptide docks in a cavity formed by the βB strand, the αB helix and the carboxylate binding loop of the PDZ domain. The first residue of the αB helix (αB₁ residue) confers specificity for the peptide ligand and forms the basis for PDZ domain classification (Sheng and Sala, 2001; Hung and Sheng, 2002). The PDZ protein syntenin-1 was originally identified as a ligand for the syndecans (Grootjans et al, 1997). Recently, we reported that the PDZ domains of syntenin-1 bind to phosphatidylinositol 4,5-bisphosphate (PIP₂) and that PIP₂, as well as syndecans, regulates the targeting of syntenin-1 to the plasma membrane (Zimmermann et al, 2002). As most proteins with PDZ domains act as submembranous scaffolding proteins, controlling the correct subcellular targeting of supramolecular signaling complexes (Nourry et al, 2003), this raises the question of whether PDZ domains in general function as phosphoinositide-binding modules.

Phosphoinositides are phosphorylated forms of phosphatidylinositol (PI), present at relatively low levels within cells compared to other phospholipids. They have emerged as essential players in a variety of cellular processes, ranging from cell surface receptor signaling, actin cytoskeleton remodeling and compartmentalization of vesicular trafficking to cell growth and proliferation. Phosphoinositides can function as precursors of second messengers or directly as docking sites for target proteins (Toker, 2002; Czech, 2003; De Matteis and Godi, 2004). Protein domains known to bind phosphoinositides are FYVE, PX, PH, ENTH, ANTH, Tubby and FERM domains (Lemmon, 2003), and were identified in proteins that are principally cytoplasmic or membrane bound (Gozani et al, 2003).

It is now established that phosphoinositides are also present in the nucleus, but their function there remains poorly understood (Irvine, 2003). PIP_2, in particular, concentrates in nuclear speckles and nucleoli (Osborne et al, 2001). Nuclear speckles are dynamic punctate subnuclear structures that are enriched in pre-messenger RNA (pre-mRNA) splicing factors and are located in the interchromatin regions of the nucleoplasm (Lamond and Spector, 2003). Nuclear PIP₂ regulates chromatin remodeling and gene transcription, and in vitro studies indicate that nuclear PIP₂ might modulate pre-mRNA splicing (Yu et al, 1998; Zhao et al, 1998; Osborne et al, 2001; Jones and Divecha, 2004). Gozani et al (2003) showed that the chromatin-associated protein ING2 binds nuclear phosphatidylinositol 5-phosphate (PI(5)P), via its PHD finger, and suggested that PHD fingers, domains present in a large number of chromatin regulatory factors, function as nuclear phosphoinositide receptors. No other protein domains recognizing nuclear phosphoinositides have been identified so far.

Here, further investigating the importance of phosphoinositide–PDZ interactions, we found that the PDZ protein syntenin-2, a protein closely related to syntenin-1, also directly interacts with PIP₂ in a PDZ-dependent mode. This interaction specifically targets the protein to nuclear PIP₂ pools. Depletion of syntenin-2, by siRNA, disrupts the nuclear speckles–PIP₂ pattern and affects cell survival and cell division.

Results

Most PDZ proteins bind weakly to PIP₂-containing lipid layers

To evaluate whether binding to PIP₂ might be a general property of PDZ domains, we investigated whether the PDZ domains of nine different proteins could directly and specifically interact with lipid layers containing PIP₂. We selected PDZ proteins containing multiple PDZ domains and PDZ domains organized in tandem as in syntenin-1, and fused these to GST. GST was included as negative control, while GST-tagged full-length (FL) syntenin-1 (human form) and GST fused to the PH domain of phospholipase C (PLC)δ1 (GST-PH-PLCδ) were used as positive controls (Zimmermann et al, 2002). In surface plasmon resonance (SPR) assays (Figure 1A), a segment of PSD95 containing the three PDZ domains of the protein (GST-PDZ1-3-PSD95, rat form aa 65–393) or a segment of nNOS containing the single PDZ domain of the protein (GST-PDZ1-nNOS, rat form aa 1–299) did not interact with PIP₂. The tandem PDZ domains of EBP50 (GST-PDZ1-2-EBP50, human form aa 1–245), the tandem PDZ domains of Mint3 (GST-PDZ1-2-Mint3, mouse form aa 391–556), a segment of Bazooka containing the three PDZ domains of the protein (GST-PDZ1-3-Bazooka, aa 291–737), and the FL PDZ proteins Magi-3 (GST-FL-Magi-3, human form), Dlt (GST-FL-Dlt) and PAR-6 (GST-FL-PAR-6, human form) bound with low affinity to PIP₂-containing lipid layers, as the sensogram showed a clear and specific association and dissociation phase, and the binding was concentration dependent; however, these interactions were at the limits of the sensitivity of the assay (data not shown). Based on the characteristics of our biosensor, we conclude that the K_D of these interactions is around or above 10⁻⁴ M.

Figure 1.

PIP₂ binding of selected PDZ proteins. (A) Binding of PDZ proteins or PDZ domains to PIP₂ in SPR analysis. Purified recombinant proteins were perfused at 2 μM over a sensorchip coated with vesicles containing 10% PIP₂. Values correspond to response units (RU) measured 4 min after perfusion of the proteins. (B) SPR sensograms showing the interaction of syntenin-2 with lipid layers containing 10% PIP₂. The syntenin-2 sensogram is compared to the sensogram for GST-PH-PLCδ (positive control for PIP₂ binding). The purified recombinant proteins were perfused at 1 μM. (C) SPR sensogram showing that the binding of syntenin-2 to PIP₂ increases as the PIP₂ concentration in the lipid layers increases. (D) PIP₂ binding of syntenin-2 in overlay assay. Increasing amounts of PIP₂ (as indicated) were spotted on a nitrocellulose membrane and incubated with GST-Myc as a negative control (upper panel) or Myc-tagged syntenin-2 (lower panel).

Syntenin-2 binds with high affinity to PIP₂

In contrast, the PDZ protein syntenin-2 bound strongly to lipid layers containing 10% PIP₂ (Figure 1A and B). To evaluate the PIP₂ binding of syntenin-2, we used a nontagged form of the protein, produced by self-splicing from an intein fusion protein. Decreasing the concentration of PIP₂ in the lipid layers decreased the binding of syntenin-2, illustrating that the binding was dependent on PIP₂ (Figure 1C). Apparent K_D values for PIP₂ binding were calculated for FL syntenin-2 and for GST-PH-PLCδ, and determined to be 1.5 × 10⁻⁶ and 0.2 × 10⁻⁶ M, respectively. Dot-blot assays, whereby increasing amounts of PIP₂ were immobilized before overlay with syntenin-2, indicated that PIP₂ does not need to be incorporated in lipid layers for binding to syntenin-2 (Figure 1D). We also investigated whether syntenin-2 significantly bound to phosphoinositides other than PIP₂ (Supplementary data 1). Binding of syntenin-2 to other phosphoinositides was at the limit of biosensor sensitivity, except for PI(3,4,5)P₃ for which the apparent K_D was determined to be 2.0 × 10⁻⁶ M.

Syntenin-2 is targeted to the plasma membrane, nucleoli and nuclear speckles via its PDZ domains

Specific anti-syntenin-2 antibodies were raised to study the expression and subcellular distribution of endogenous syntenin-2. Western blots of total protein extracts showed that MCF-7 and Caco-2 cells express syntenin-2 (Figure 2A). In MCF-7 cells, the main signal distributed to subnuclear structures and within the cytoplasm (Figure 2Ba and b). In Caco-2 cells, the syntenin-2 signal was mainly observed within the cytoplasm and at the plasma membrane (Figure 2Bc), but a weak signal was also present in subnuclear structures (Figure 2Bd). The endogenous syntenin-2 signal in the nucleus codistributed with nucleoli (Figure 2Ca, b and a-b), a subnuclear compartment that contains PIP₂ according to electron microscopy studies (Osborne et al, 2001), but not with PIP₂ in nuclear speckles (Figure 2Ca, c and a-c), as traced by the antibody 2C11 (Osborne et al, 2001). To specify the respective roles of the different syntenin-2 domains in this subcellular targeting, we transfected MCF-7 cells with expression vectors encoding eGFP fusions to the different domains of syntenin-2 and analyzed the subcellular locations of these proteins by confocal microscopy (Figure 2D). eGFP itself, included as a control, was diffusely distributed over the cell. When the first or the second PDZ domain of syntenin-2 was fused to eGFP in isolation, the fluorescence was distributed over the cytosol and the nucleosol, with some enrichment in the nucleosol for PDZ1. When the tandem of PDZ1 and PDZ2 was fused to eGFP, the fluorescence was concentrated in the nucleus, in discrete subnuclear structures, and at the plasma membrane. Similar distributions were observed for constructs where either the N-terminal or the C-terminal domain was added to the PDZ tandem. Nuclear enrichment predominated over plasma membrane and cytosolic distribution, with 78% (±3%) of the cells displaying fluorescence exclusively in the nucleus. Surprisingly, when FL syntenin-2 was fused to eGFP, the fluorescence was primarily localized in the cytoplasm and excluded from the nucleus. A similar exclusion was observed for a Myc-tagged FL syntenin-2 (data not shown), suggesting that nuclear exclusion was not due to the size of the eGFP fusion. However, overexpressed nontagged syntenin-2 (detected with anti-syntenin-2 antibodies, and scoring expression above endogenous levels) was detected in the nucleus of MCF-7 cells (Supplementary data 2A), indicating that tagging the FL protein interferes with its nuclear concentration. The reason for this interference is unknown, but it could be alleviated by treating the cells with the nuclear export inhibitor, leptomycin B (Supplementary data 2B). We concluded that, although other domains might influence the subcellular distribution of syntenin-2, the PDZ domains together are necessary and sufficient for targeting syntenin-2 to the plasma membrane and subnuclear structures. In contrast to endogenous syntenin-2, all tagged constructs of syntenin-2 containing both PDZ domains were concentrated in nuclear speckles, in addition to nucleoli, as illustrated for the eYFP-PDZ1-2-Cter construct (Figure 2E). Investigating possible reasons for this discrepancy revealed that our anti-syntenin-2 antibodies are not suited for detecting the protein in nuclear speckles (Supplementary data 3). We concluded that syntenin-2 concentrates at three PIP₂-rich regions—the plasma membrane, in nucleoli and nuclear speckles—by virtue of its PDZ domains.

Figure 2.

Contribution of the syntenin-2 PDZ domains in subcellular targeting. (A) Overexpressed nontagged (OE) and endogenous (E) syntenin-2 levels were analyzed by Western blotting of total cell extracts originating from MCF-7 or Caco-2 cells as indicated. Purified recombinant nontagged syntenin-2 produced in bacteria (R) was included as a positive control. (B) Immunofluorescence micrographs of MCF-7 (a, b) and Caco-2 cells (c, d) showing the distribution of endogenous syntenin-2. (C) Confocal immunofluorescence micrographs of the nucleus of an MCF-7 cell showing the distribution of endogenous syntenin-2 (a), nucleolar antigen (b) and nuclear speckles/PIP₂ (c) as detected with the 2C11 antibody that does not detect PIP₂ in nucleoli (Osborne et al, 2001). In the merge panels, syntenin-2 is in green and nucleolar antigen (a-b) or nuclear speckles/PIP₂ (a-c) are in red. (D) Schematic representation of the different domains of syntenin-2 that were transiently overexpressed as eGFP fusion proteins in MCF-7 cells. The limit of each domain is indicated by the amino-acid number. The subcellular distribution of the different eGFP fusion proteins was assessed by confocal microscopy, as illustrated. (E) Confocal immunofluorescence micrographs of the nucleus of an MCF-7 cell showing the distribution of eYFP fused to the segment of syntenin-2 that contains the two PDZ domains plus the C-terminal domain (a), nucleolar antigen (b) and nuclear speckles/PIP₂ (c). In the merge panels, syntenin-2 is in green and nucleolar antigen (a-b) or nuclear speckles/PIP₂ (a-c) are in red. Size bars are 10 μm.

Plasma membrane, nuclear speckles and nucleoli targeting of syntenin-2 PDZ domains relies on PIP₂

To evaluate whether PIP₂–PDZ domain interactions govern these subcellular distributions of syntenin-2, we performed colocalization studies with probes for PIP₂ and tested whether treatments known to affect PIP₂ levels also influence the distribution of syntenin-2. For these studies, we used the segment of syntenin-2 containing both PDZ domains plus the C-terminal domain, for simplicity further referred to as ‘tandem PDZ'.

In the small percentage of transfected cells where the eYFP-tandem PDZ was detected at the plasma membrane, it significantly colocalized with eCFP fused to the PH domain of PLCδ, a probe for plasma membrane PIP₂ pools (Figure 3A, compare plasma membrane staining in a and d). As revealed by time-lapse video microscopy, plasma membrane PIP₂ hydrolysis induced by ionomycin treatment (Varnai and Balla, 1998) caused simultaneous delocalization of eYFP-tandem PDZ and eCFP-PH-PLCδ (Figure 3A). Moreover, increasing the plasma membrane PIP₂ pools, by overexpressing a phosphatidylinositol 4-phosphate 5 kinase, or by overexpressing Arf6, a small GTPase that recruits and activates phosphatidylinositol 4-phosphate 5 kinase at the plasma membrane (Honda et al, 1999), caused enrichment of the tandem PDZ at the plasma membrane (data not shown). These results support a role for PIP₂ in recruiting the tandem PDZ domains of syntenin-2 to the plasma membrane. Since syntenin-2 interacts also significantly with PI(3,4,5)P₃ (see above), we wondered whether PI(3,4,5)P₃ plays a role in the plasma membrane targeting of the tandem PDZ domains. Neither serum stimulation, a treatment that produces a local enrichment of PI(3,4,5)P₃ at the plasma membrane (as indicated by membrane recruitment of the PH domain of Akt, a PI(3,4,5)P₃ probe), nor addition of wortmannin, which prevents synthesis of PI(3,4,5)P₃, influenced the localization of the tandem PDZ (data not shown). Therefore, we conclude that PI(3,4,5)P₃ does not significantly contribute to the plasma membrane localization of syntenin-2. This is consistent with the fact that PIP₂ is 25-fold more abundant in cells compared to other phosphoinositides and with the view that a protein will not target rare phosphoinositides unless it displays stringent affinity for this particular phosphoinositide (Lemmon, 2003).

Figure 3.

Modulation of cellular PIP₂ levels affects the distribution of syntenin-2 PDZ domains. (A) Time-lapse microscopy of MCF-7 cells expressing the tandem PDZ domains of syntenin-2 (tandem PDZ) fused to eYFP (a–c), and the PH domain of PLCδ fused to eCFP (d–f) before (a, d), 7 min after (b, e) or 8 min after (c, f) ionomycin-induced hydrolysis of plasma membrane PIP₂. (B) MCF-7 cells expressing the tandem PDZ fused to eYFP were treated with DNase I (a, b), RNase A (c, d) or α-amanitin (e, f). The subnuclear distributions of eYFP-tandem PDZ (a, c, e) and nuclear speckles/PIP₂ (b, d, f) were analyzed by confocal microscopy. In the merge panels, eYFP-tandem PDZ is in green and nuclear speckles/PIP₂ are in red. The structure below the asterisk in panels e and e-f most likely corresponds to a nucleolar fragment. (C) Distribution of the tandem PDZ after activation of nuclear PLCs. MCF-7 cells expressing the eGFP-tandem PDZ were scored for nuclear enrichment of the fluorescence. Diagram, left part: Swiss 3T3 cells were treated with IGF-I for 20 min and compared to nontreated cells (N). Diagram, right part: MCF-7 cells were treated with the PLC activator m-3M3 FBS for different periods, as indicated, and compared to nontreated cells (N), or cells that were allowed to recover in serum-supplemented medium for 1 h after removal of the activator (recov). The distribution of the fluorescence was scored on three separate experiments, with 100 cells counted for each condition. Bars indicate standard deviations. Pictures: Time-lapse micrographs of MCF-7 cells before and after 25 min incubation with m-3M3 FBS. Note that the treatment induces cytosolic distribution of the fluorescence. (D) Confocal micrograph of MCF-7 cells expressing the eYFP-yeast PLC1 with a mutated NES ‘eYFP-yPLC1-NES' (a) and a Myc fusion to the tandem PDZ (b). eYFP-yPLC1-NES was detected by eYFP fluorescence and the Myc-tandem PDZ using anti-Myc antibodies (9E10). In the merge (a-b), eYFP-yPLC1-NES is in green and the Myc-tandem PDZ is in red. Note the nuclear concentration of the Myc-tandem PDZ in the cell on the right side (below the asterisk) and the cytosolic/plasma membrane distribution in the cell expressing yPLC1-NES, on the left side. Size bars are 10 μm.

RNA, but not DNA, is known to be essential for the association of PIP₂ with nuclear speckles (Osborne et al, 2001). RNase A treatment (Figure 3Bc-d) reduced the nuclear localization of the eYFP-tandem PDZ, while DNase I treatment (Figure 3Ba-b) had no effect. Speckles are dynamic structures, and their size, shape and number varies in response to environmental signals that influence the available pools of active splicing and transcription factors. When transcription is inhibited, splicing factors and PIP₂ accumulate in enlarged and rounded speckles (Boronenkov et al, 1998). Treating cells with the transcriptional inhibitor α-amanitin, at doses that allow specific inhibition of RNA polymerase II, caused reorganization of the speckles/PIP₂ structures (Figure 3Bf, compare the number and the size of the speckles in f and b) and of the distribution of eYFP-tandem PDZ (Figure 3Be). This indicates that the tandem PDZ and PIP₂ are physically and dynamically associated in nuclear speckles. A pool of the eYFP-tandem PDZ did not colocalize with the speckles/PIP₂ in these experiments (asterisk in Figure 3Be and e-f), and most probably represents a nucleolar fragment as inhibition of transcription can cause breakdown or loss of the nucleolar structure (Haaf and Ward, 1996). To determine whether the nuclear localization of the tandem PDZ can be modulated by nuclear PIP₂ breakdown, we induced cells to activate nuclear phospholipases. In Swiss-3T3 cells, IGF-I treatment activates nuclear PIP₂ breakdown via activation of PLCβ1 (Cocco et al, 1988; Divecha et al, 1991). Treating Swiss-3T3 cells that expressed the eGFP-tandem PDZ with 80 ng/ml of IGF-I for 20 min caused an enrichment of the fluorescence in the cytoplasm and at the plasma membrane (Figure 3C, left part of the diagram). The eGFP-tandem PDZ was also translocated from the nucleus to the cytoplasm and to the plasma membrane in MCF-7 cells when the cells were exposed to 100 μM m-3M3 FBS, a general PLC activator (Bae et al, 2003) (Figure 3C, right part of the diagram). Time-lapse microscopy revealed a clear translocation after 25 min (Figure 3C, micrographs). A complete release from the nuclear organelles was observed after 2 h treatment, but by that time nucleoli were not distinguishable anymore by differential interference contrast (DIC) (data not shown). This effect was reversible, as removal of the PLC activator and incubation of the cells in serum-rich medium for 1 h led to recovery of nuclear enrichment (Figure 3C, recov). Finally, expression of the yeast phospholipase PLC1 (mutated in its nuclear export signal (NES) to allow nuclear localization) shifted the distribution of the tandem PDZ from nucleoli/speckles toward the cytoplasm and the plasma membrane (Figure 3D, compare the distribution of the tandem PDZ in the left cell, positive for PLC1, with the distribution in the right cell, negative for PLC1). In these experiments, nucleoli could still be traced by antinucleolar antigen staining (Supplementary data 4). Nevertheless, this staining was weaker than in non-PLC1-transfected cells, suggesting that PLC overexpression affects nucleolar structure. Altogether, these experiments indicate that nuclear PIP₂ is required for the subnuclear enrichment of syntenin-2 tandem PDZ.

Both PDZ domains of syntenin-2 contribute to PIP₂ binding, but PDZ1 plays a crucial role in PIP₂ subnuclear targeting

We then analyzed the relative contributions of the individual PDZ domains to the PIP₂ binding of syntenin-2. In isolation, both PDZ domains bound specifically and similarly to lipid layers containing 10% PIP₂ (Figure 4A). Yet, the binding of the wild-type form of the isolated PDZ domains was weak in comparison to the binding of the tandem PDZ (compare Figure 4B with Figure 4A and note differences in protein concentrations and RU values). We then constructed PDZ domains that are unable to interact with PIP₂, using an approach similar to that used for syntenin-1 (Zimmermann et al, 2002). Two lysine to alanine mutations, affecting a lysine in the carboxylate binding loop and a second lysine close to the αB1 residue, were introduced in the PDZ1 domain (K113A and K167A, i.e. mut PDZ1) and in the PDZ2 domain (K197A and K244A, i.e. mut PDZ2). As intended, introduction of these point mutations into the isolated PDZ domains of syntenin-2 abolished their PIP₂ binding (Figure 4A). Mutating the PDZ domains separately in the PDZ tandem of syntenin-2 reduced the binding to PIP₂, but could not abolish binding. Only the PDZ tandem of syntenin-2 in which both PDZ domains were mutated failed to bind PIP₂-containing lipid layers (Figure 4B). Interestingly, such mutant syntenin-2 still interacts with the cytoplasmic domain of tetraspanin L6A, the sole peptide ligand identified so far for the syntenin-2 PDZ domains (Borrell-Pages et al, 2000), and is still able to self-associate, as described for the wild-type protein (Koroll et al, 2001) (data not shown). Next, we explored the impact of these mutations on the subcellular localizations of the tandem PDZ in MCF-7 cells. When expressed as eYFP fusion protein, the tandem PDZ of syntenin-2 with two mutant PDZ domains (eYFP-mut PDZ1+mut PDZ2) (Figure 4Cd) was not enriched at the plasma membrane, nor in speckles or in nucleoli. This was in contrast to the wild-type tandem PDZ (eYFP-wt PDZ1+wt PDZ2) (Figure 4Ca). Mutation of only one of the PDZ domains of the tandem (eYFP-wt PDZ1+mut PDZ2 or eYFP-mut PDZ1+wt PDZ2) did not abolish PIP₂ subcellular targeting (Figure 4Cb and c).

Figure 4.

Respective contribution of syntenin-2 PDZ domains in PIP₂ binding and targeting to specific subcellular PIP₂ pools. (A) SPR sensograms showing the binding of the individual PDZ domains of syntenin-2 to 10% PIP₂-containing lipid layers. Wild-type (wt) and mutant (mut) PDZ domains were purified as GST fusion proteins and perfused at 2 μM. Mutant PDZ domains (mut PDZ1 and mut PDZ2) were made by changing two lysine residues into alanine residues. (B) SPR sensograms showing the interaction of the tandem PDZ of syntenin-2 with 10% PIP₂-containing lipid layers. Wild-type and mutant PDZ domains were purified as Myc fusion proteins and perfused at 1 μM over the sensorchip. (C) Subcellular localization of wild-type and mutant tandem PDZ of syntenin-2 in MCF-7 cells was analyzed by confocal microscopy. Proteins were expressed as eYFP fusions. (D) Confocal micrographs of MCF-7 cells expressing different combinations of the PDZ domains of syntenin-1 and syntenin-2 as indicated. Size bars are 10 μm.

Since the tandem PDZ of syntenin-1 also binds with high affinity to PIP₂ (Zimmermann et al, 2002) but never concentrates in subnuclear PIP₂-rich regions (Figure 4Da), the relative contributions of the PDZ domains to plasma membrane and nuclear PIP₂ targeting were also tested by domain-swapping experiments. We constructed chimeric proteins where the first PDZ domain of syntenin-2 was fused to the second PDZ domain of syntenin-1 (eYFP-chim-PDZ1_synt2+PDZ2_synt1; chim: chimera) or the inverse (eYFP-chim-PDZ1_synt1+PDZ2_synt2), and tested the subcellular localization of these chimeras by confocal microscopy. Although the first chimera localized to all PIP₂-rich regions (Figure 4Db), the second chimera was restricted to the plasma membrane (Figure 4Dc), suggesting that the PDZ1 of syntenin-2 determines subnuclear PIP₂ targeting. Consistently, a construct containing two syntenin-2 PDZ1 domains in tandem (eYFP-chim-PDZ1_synt2+PDZ1_synt2) and also a chimera containing the PDZ1 domain of syntenin-2 fused to the PDZ1 of syntenin-1 (eYFP-chim-PDZ1_synt2+PDZ1_synt1) were also targeted to PIP₂-rich regions in the nucleus (Figure 4Dd and e). Nevertheless, the structural combination of one PDZ1 domain with one PDZ2 domain seems important for syntenins to target PIP₂ pools. Indeed, the tandem PDZ of syntenin-2 (Figure 4Ca) represented the best combination for nuclear PIP₂ targeting, while a chimera between the first PDZ domain of syntenin-1 and the second PDZ domain of syntenin-2 (Figure 4Dc) represented the best combination for plasma membrane PIP₂ targeting. Moreover, the combination of two PDZ2 domains, one originating from syntenin-2 and the other from syntenin-1 (eYFP-chim-PDZ2_synt2+PDZ2_synt1), was not targeted to any PIP₂ pool but diffusely distributed (Figure 4Df) and diffuse distribution was also observed with a construct of two PDZ1 domains of syntenin-1 (data not shown).

Syntenin-2 loss-of-function disrupts nuclear speckles–PIP₂ pattern and affects cell division and survival

To clarify the function of syntenin-2, we investigated the effect of a syntenin-2 depletion in cultured cells. Transfection of MCF-7 or U-2 OS cells with siRNA for syntenin-2 (RNAi) strongly reduced the levels of syntenin-2 protein, compared to cell transfection with nontargeting RNA (NT) (Figure 5A). While we did not observe substantial differences in the morphology of the nucleoli between cells that were treated with NT or RNAi, loss of syntenin-2 expression had a drastic consequence on nuclear PIP₂ as observed by 2C11 staining. While in NT-treated cells, the 2C11 pattern was similar to that in nontransfected cells, 70% of the RNAi-treated cells displayed dispersed smaller speckles (Figure 5B, compare b with a and d with c). To further investigate the potential consequences of a syntenin-2 depletion for nuclear processes, U-2 OS cells, stably expressing histone 2B fused to red fluorescent protein, were transfected with RNAi or NT and followed by time-lapse video microscopy over 20 h. Cell death, accompanied by cell surface blebbing and chromatin compaction, was observed two-fold more frequently in RNAi-treated cells than in NT-treated cells (Figure 5C). Cell division and BrdU staining were reduced by a factor of 3 and 2 respectively in RNAi-treated cells (Figure 5C). These results indicate that syntenin-2 may function as an organizer of nuclear PIP₂ and that syntenin-2 expression levels can affect cell viability and the rate of cell division.

Figure 5.

Suppression of syntenin-2 affects the nuclear PIP₂ staining pattern, cell viability and cell division. (A) Western blot analysis of total cell extracts from MCF-7 cells (left panels) or U-2 OS cells (right panels), treated with an NT or an siRNA targeting syntenin-2 (RNAi). Note that endogenous levels of syntenin-2 decrease in RNAi-treated cells (upper panels) while calnexin signals are not affected (lower panels). Numbers indicate molecular weight markers in kDa. (B) Confocal micrographs of MCF-7 cells (a, b) and U-2 OS cells stably expressing RFP-histone 2B (c, d), transfected with an NT (a, c) or an siRNA targeting syntenin-2 (b, d) together with an expression vector for farnesylated eGFP (GFP-F) as a marker of transfected cells (blue signal at the plasma membrane). The nuclear speckles/PIP₂ pattern was visualized using 2C11 antibody (green). RFP-histone 2B in U-2 OS is in red. Size bars are 10 μm. (C) Cell death and cell division in U-2 OS cells stably expressing RFP-histone 2B, treated with an NT or an siRNA for syntenin-2 (RNAi), were quantified in three independent experiments, by multipositioning time-lapse video microscopy, looking at the fate of a minimum of 50 transfected cells per condition in each experiment. Incorporation of BrdU by U-2 OS cells was quantified in three independent experiments, by wide-field fluorescence microscopy, on 200 cells per condition in each experiment. Bars indicate standard deviations.

Discussion

PDZ domains: phosphoinositide-binding domains?

Most PDZ proteins tested in this study displayed very weak affinity for PIP₂. Also, phosphoinositides other than PIP₂ interact only weakly with some of these PDZ proteins in vitro (data not shown). Moreover, when these weakly interacting PDZ proteins, or their PDZ domains, were overexpressed as fluorescent fusion proteins, these were neither targeted to discrete subcellular PIP₂ pools nor to compartments rich in other phosphoinositides (data not shown). This is reminiscent of what is known for most PH domains (Lemmon, 2003; Yu et al, 2004), which also tend to bind PIP₂ rather weakly, are quite nonspecific in their phosphoinositide binding and are by themselves not able to target their host proteins to phosphoinositide-rich domains. The functional importance of such weak interactions has not yet been established. In organisms, PDZ proteins generally oligomerize directly or indirectly, forming larger complexes containing multiple membrane anchors with affinity for transmembrane receptors (Bilder et al, 2003; Roh and Margolis, 2003). In such complexes, avidity effects could become important. As cultured cells and overexpression experiments might not reproduce the correct environmental contexts (cell polarity, associated subunits and presence and concentration of peptide ligands) of these PDZ proteins, one cannot exclude that in physiological situations, PIP₂–PDZ domain interactions might augment the membrane-binding capacity of these PDZ proteins and modulate their plasma membrane targeting, but further work is needed to clarify this issue.

High-affinity binding to PIP₂ by the PDZ domains of syntenins

This screen identified syntenin-2 as a second PDZ protein that binds with high affinity to PIP₂ (apparent K_D=1.5 μM) via its PDZ domains. This suggests that high-affinity binding to PIP₂ might be a unique characteristic of syntenin PDZ domains. Although syntenin-2 is highly related to syntenin-1 in its overall domain organization and sequence, the known peptide ligands for the PDZ domains of syntenin-1 do not interact with syntenin-2 (Koroll et al, 2001), and vice versa, arguing that functionally these proteins are not redundant. It is therefore remarkable that the interaction of syntenin-2 with PIP₂ is conserved. For both syntenin-1 and syntenin-2, the PDZ domains taken in isolation are not sufficient for targeting to PIP₂-rich regions in cells. The tandem is required. Consistently, the in vitro interaction with PIP₂ of the tandem PDZ domains of both syntenins is also stronger than that of the PDZ domains taken in isolation. Nevertheless, in syntenin-1, the first PDZ domain clearly displays higher affinity for PIP₂ than does the second PDZ domain (Zimmermann et al, 2002), while in syntenin-2, the relative contribution to PIP₂ binding is evenly distributed over both PDZ domains. Mutating the PDZ1 domain (introducing K119A in the carboxylate binding loop and replacing 171SDK173 by HEQ in the αB helix) in the tandem repeat already totally abolishes syntenin-1 binding to PIP₂ (Zimmermann et al, 2002) but only moderately affects syntenin-1–syndecan-2 binding (Grootjans et al, 2000). Mutation of both PDZ domains of the tandem (introducing K to A mutations in the carboxylate binding loop and close to the αB1 residue in PDZ1 and PDZ2) is required to abolish syntenin-2–PIP₂ interaction, but preserves binding to the peptide ligand tetraspanin L6A and the capacity to self-associate. This means that although some of the residues implicated in lipid and peptide binding overlap, the two activities can be dissociated. At this stage, it is difficult to delineate the structural determinants in PDZ domains that mediate high-affinity binding to PIP₂. By amino-acid alignments, comparing PDZ domains displaying high affinity for PIP₂ (the first PDZ domain of syntenin-1 and the two PDZ domains of syntenin-2) to PDZ domains displaying low or no affinity for PIP₂, we failed to identify a unifying ‘PIP₂-binding motif'. Kachel et al (2003) studied the interaction of the PDZ2b domain of PTP-Bas with PIP₂ and PI(3,4,5)P₃ by NMR, and identified 10 residues making contacts with the lipids (Kachel et al, 2003). Sequence alignments of the PDZ2b domain of PTP-Bas and the PDZ domains of the syntenins also failed to identify a consensus sequence for PIP₂ binding. It should be noted, however, that the PDZ domains of the human ortholog of PTP-Bas display only weak PIP₂ interaction in the experimental settings used in the present study (unpublished observation). Clearly, our studies indicated that positively charged residues are implicated, as in many other phosphoinositide-binding motifs. Nevertheless, competition experiments with inositol (1,4,5)-triphosphate (IP₃) and water-soluble PIP₂ also showed that the inositol phosphate headgroup is not sufficient for the syntenin-2 interaction with PIP₂ (data not shown), suggesting that hydrophobic interactions could also play a role. Moreover, our data also imply that one has to consider the tandem of the PDZ1-2 domains of the syntenins as one functional unit rather than as two independent modules interacting with PIP₂. Structural studies of syntenin–PIP₂ complexes are ongoing and will hopefully clarify which residues are implicated in the interaction with PIP₂.

Targeting to plasma membrane versus nuclear PIP₂ pools

Our data suggest that subcellular targeting to PIP₂ pools only occurs for constructs with an affinity for PIP₂ that is above a certain threshold, and that this threshold can only be reached when the PDZ domains of syntenins are in tandem. Nevertheless, the tandem PDZ domains of syntenin-1 are exclusively targeted to PIP₂ pools at the plasma membrane (Figure 4Da) while the tandem PDZ domains of syntenin-2 are predominantly targeted to nuclear PIP₂ pools (Figure 4Ca). Endogenous syntenin-1 and syntenin-1 tandem PDZ can also be found in the nucleus, but they never concentrate in speckles or nucleoli (Zimmermann et al, 2001). Domain-swapping experiments indicate that the first but not the second PDZ domain of syntenin-2 contains the nuclear PIP₂ targeting determinants (Figure 4Db–e). Several hypotheses can be put forward to explain this property. Most obvious is that targeting to nucleoli and nuclear speckles might necessitate an additional peptide anchorage for syntenin-2 PDZ domains in these structures, but so far we failed to identify such ligand. Alternatively, as detergent-insoluble pools of PIP₂ are found in the nucleus (Divecha et al, 1993; Maraldi et al, 1999), the first PDZ domain of syntenin-2 might be better suited to recognize non-membrane-associated nuclear PIP₂. Finally, interaction with unconventional ligands like RNA remains an unexplored possibility.

Understanding the biology of phosphoinositides has tremendously benefited from the identification of protein domains displaying specific and high-affinity binding for the different phosphoinositides and that can be used as probes in living cells (Varnai et al, 2002). The PH domain of PLCδ has been shown to be an excellent probe for recognizing PIP₂ at the plasma membrane (Varnai and Balla, 1998), but cannot target PIP₂ pools in the nucleus (Okada et al, 2002). So far, a probe detecting nuclear pools of PIP₂ was lacking. The tandem PDZ domains of syntenin-2 represent such a probe.

Syntenin-2 as an organizer of nuclear PIP₂

Understanding the role of PIP₂ in nuclear compartmentalization and as a signaling molecule in these compartments is an emerging challenge. PIP₂, by interacting with histone H1, might counteract transcriptional repression (Yu et al, 1998). By controlling the nuclear actin cytoskeleton, it might influence chromatin remodeling and pre-mRNA processing (Zhao et al, 1998; Skare et al, 2003). Moreover, PIP₂ also undergoes PLC cleavage in the nucleus (Divecha et al, 1991). Subsequent generation of DAG and activation of PKC leads to phosphorylation events, resulting for example in lamin breakdown during mitosis (Haas and Jost, 1993). On the other hand, IP₃ can serve as precursor of inositol polyphosphate, necessary for efficient mRNA export (York et al, 1999).

Multiple PIPK isoforms, diacylglycerol kinase θ and particular phospholipases are concentrated at nuclear speckles, indicating that PI signaling pathways are localized at this site (Martelli et al, 2004). Conceivably, syntenin-2 could function as a scaffold that brings PIP₂ in proximity of components of the inositol cycle machinery or other speckles components (RNA, proteins, etc.). One possibility could have been that syntenin-2 links PIP₂ with the nuclear enzyme PLCβ1, which contains a PDZ binding domain interacting with NHERF (Tang et al, 2000). Unfortunately, direct syntenin-2–PLCβ1 binding assays failed to support this idea (unpublished observation). Macromolecular interactions form the basis for speckle morphogenesis (Lamond and Spector, 2003), and when one structural component is missing, the architecture of the whole complex might be affected. Downregulation of syntenin-2 by RNAi causes a dispersion of the nuclear speckles/PIP₂ (Figure 5B). Syntenin-2 could support and stabilize speckle structure by its propensity to self-associate and to bind PIP₂. Syntenin-2 might protect PIP₂ from degradation, and in such case the dispersed PIP₂ pattern in the RNAi-treated cells could be due to reduced levels of PIP₂. Speckles are also dynamic structures and their components travel to other nuclear locations. In this context, syntenin-2 could be part of a proteolipid complex that functions as a PIP₂ transporter. Our data show that syntenin-2 is able to shuttle between several subcellular compartments. The disruption of the PIP₂–syntenin-2 complex in our RNAi experiments may disturb one or several crucial processes, and lead to cell death and reduced cell divisions under these conditions. The characterization of other components interacting with syntenin-2 in the nucleus and of the cues that influence its localization should provide further elements to understand its function.

Materials and methods

Molecular biology

Untagged and Myc-tagged (EQKLISEEDL) forms of syntenin-2 cDNAs were subcloned in the pTYB1 vector (New England Biolabs, USA). The various phosphoinositide probes, the PDZ domains and the PDZ protein cDNAs were cloned in pGEX vectors. All proteins were expressed as intein fusion or GST fusion proteins in ER2566 cells and purified as described before (Grootjans et al, 2000). The purified proteins were dialyzed in HBS buffer (20 mM Hepes pH 7.4, 150 mM NaCl and 1 mM EDTA). Eukaryotic expression vectors encoding eGFP, eCFP and eYFP fluorescent proteins were obtained from Clontech (BD Biosciences, USA). The eukaryotic expression vector pCDNA3.1/Zeo⁽⁺⁾ for expression of untagged and Myc-tagged proteins was obtained from Invitrogen (CA, USA). The yeast PLC1 open reading frame was amplified from yeast genomic DNA and TA cloned into pGEM-T-EASY (Promega, USA). The nuclear export sequence 188-KLKAH-192 was changed into KRKRH. Mutated yeast PLC1 was subcloned in the mammalian expression vector peYFP-C. All constructs were certified by sequencing. More detailed information is available upon request.

Lipid-binding assays

Lipid manufacturers, PIP₂ overlays, preparation of vesicles and detailed protocols for SPR experiments can be found in Supplementary data 5. Briefly, reconstituted composite vesicles containing PC, PE, PS and PI-PIPs prepared in appropriate molar ratios (30:40:20:10) were coated on L1-chips. Coated chips were perfused with purified proteins in HBS for association. Sensograms were corrected for background association to 10% PI vesicles. For apparent K_D measurements, various protein concentrations, ranging from 250 to 3000 nM, were perfused on the chips and signals obtained at the equilibrium (Req) were plotted as a function of protein concentration. Apparent K_D values were deduced from these plots, as the concentration corresponding to Req(max)/2.

Anti-syntenin-2 antibodies

Two rabbits were immunized (Eurogentec, Belgium) with purified recombinant human syntenin-2 produced in the impact T7 system according to the instructions of the manufacturer (New England Biolabs, UK). Anti-syntenin-2 polyclonal antibodies were isolated from the sera from these animals by affinity purification on a syntenin-2-Sepharose 4B column (Amersham Biosciences, UK). Characterization of these antibodies indicated that they do not crossreact with syntenin-1 and recognize epitopes in the N-terminal, PDZ-1, PDZ-2 and C-terminal domains of syntenin-2, in Western blot and in immunocytochemistry. However, they fail to recognize the nuclear speckles-associated form of syntenin-2.

Western blotting

Cells were lysed in DIP buffer (1% Triton X-100, 0.1% SDS and 1% NP-40 in PBS). Western blotting was performed as described before (Zimmermann et al, 2001). The blots were incubated with anti-syntenin-2 (1 μg/ml) or anti-calnexin (StressGen Biotech. Corp., Canada) antibodies and appropriate secondary antibodies coupled to horseradish peroxidase (Amersham Biosciences, UK). Signals were visualized by enhanced chemiluminescence.

Cell biology and microscopy

Cell cultures, transfections and stainings were performed as described before (Zimmermann et al, 2001). For microscopical analysis of fixed samples, cells were plated on uncoated eight-well chamber slides (Nalgene Nunc International, USA). For live cell imaging, cells were plated on two-well chamber borosilicate coverglass (Nalgene Nunc International, USA). For plasma membrane PIP₂ breakdown experiments, cells were washed in Krebs–Ringer buffer (Varnai and Balla, 1998) before incubation with 10 μg/μl of ionomycin (Sigma, USA) in the same buffer. For inhibition of transcription, MCF-7 cells were treated with 5 μg/ml of α-amanitin (Roche, UK) for 6 h. For PLC activation assays, IGF-I was purchased from Invitrogen (USA) and m-3M3 FBS (Bae et al, 2003) from Calbiochem EMD Biosciences (USA). Anti-syntenin-2 antibodies were used at 10 μg/ml to detect endogenous syntenin-2 and at 1 μg/ml to detect overexpressed syntenin-2. Anti-PIP₂ antibodies (2C11) were applied as described elsewhere (Osborne et al, 2001). Antinucleolar antigen antibody was purchased from Biomeda (Foster City, CA). RNase A (Roche UK; 0.5 mg/ml, 5 min) and DNase I (Roche UK; 100 μg/ml, 1 h) treatments were performed as described elsewhere (Osborne et al, 2001). Incorporation of BrdU was detected with anti-BrdU antibody (Becton Dickinson, San Jose, CA) in cells that were pulsed for 15 min with 20 μM BrdU (Calbiochem, EMD Biosciences, USA). Images were obtained by digital imaging fluorescence microscopy using a cooled charge-coupled device camera (Photometrics, Tucson, AZ) or by confocal microscopy (MRC-1024 Laser Scanning Confocal Imaging System, Bio-Rad, CA). Live cell imaging experiments were performed with a Leica AS MDW workstation (Leica Microsystems, Wetzlar, Germany).

siRNA

siRNAs, specifically targeting syntenin-2, were purchased from Dharmacon Inc. (USA). They consist of four RNA duplexes (sigenome-on-target™ duplexes for human-SDCBP2). Duplex 2 was proven to be the best to downregulate endogenous syntenin-2 and was used in all our assays. NT (siSTABLE #1, Dharmacon Inc.) was used as a control. Cells were transfected at 40–50% density with 100 nM siRNA or NT together with an expression vector for farnesylated eGFP to recognize targeted cells and were analyzed after 48 h. Transfection was achieved with oligofectamine (Invitrogen, USA), according to the recommendations of the manufacturer. For live experiments, cells were grown in two-well chamber borosilicate coverglass to allow a parallel study of siRNA- (well 1) or NT- (well 2) treated cells. Cell fate was recorded every 3 min by fluorescence and DIC time-lapse microscopy.

Supplementary Material

Supplementary Information 1

Supplementary Information 2

Supplementary Information 3

Supplementary Information 4

Supplementary Information 5