The characterization of the nuclear dynamics of syntenin-2, a PIP2 binding PDZ protein

Authors

  • Annelies Geeraerts,

    1. Department of Chemistry, Faculty of Science, University of Leuven, Leuven, Belgium
    2. Department of Human Genetics, Faculty of Medicine, University of Leuven, Leuven, Belgium
    Search for more papers by this author
  • Fan Hsiu-Fang,

    1. Department of Chemistry, Faculty of Science, University of Leuven, Leuven, Belgium
    Search for more papers by this author
  • Pascale Zimmermann,

    Corresponding author
    1. Department of Human Genetics, Faculty of Medicine, University of Leuven, Leuven, Belgium
    • Department of Chemistry, Faculty of Science, University of Leuven, Leuven, Belgium
    Search for more papers by this author
  • Yves Engelborghs

    Corresponding author
    1. Department of Chemistry, Faculty of Science, University of Leuven, Leuven, Belgium
    • Department of Human Genetics, Faculty of Medicine, University of Leuven, Leuven, Belgium
    Search for more papers by this author

Correspondence to: Pascale Zimmermann, Department of Human Genetics, O&N 1, Herestraat 49 bus 602, B3000 Leuven, Belgium. E-mail: Pascale.Zimmermann@med.kuleuven.be or to Yves Engelborghs, Moleculaire en Structurele Biologie, Celestijnenlaan 200G bus 2403, 3001 Heverlee, Belgium. E-mail: Yves.Engelborghs@fys.kuleuven.be

Abstract

Cellular signaling is largely dependent on the presence, that is, assembly/disassembly, of supramolecular complexes. Postsynaptic density protein, Discs-large, Zona occludens (PDZ) domains play important roles in the assembly of various signaling complexes. Syntenin-2 (S2) is a PDZ protein that interacts with nuclear phosphatidylinositol 4,5-bisphosphate (PIP2). Although nuclear lipids emerge as key players in nuclear processes, the global significance of nuclear phosphoinositide-protein interactions is still poorly understood. Those phosphoinositide-protein interactions that have been studied in detail appear to have profound physiological effects. To our knowledge none of these were investigated by dynamic studies such as Fluorescence Correlation Spectroscopy (FCS), Fluorescence Cross-Correlation Spectroscopy (FCCS), or Fluorescence Recovery After Photobleaching (FRAP). Although the exact function of S2 is unknown, siRNA experiments suggest that this PDZ protein plays a role in the organization of nuclear PIP2, cell division, and cell survival. As a consequence of its PIP2 interaction, its reported self-association in a yeast two-hybrid study and its speculated interaction with many, yet unidentified, proteins one can hypothesize that S2 plays an important role in cell signaling. Therefore, we studied the dynamics of S2 using FCS, FCCS, and FRAP, utilizing an active truncated form deleted for the first 94 amino acids (S2-ΔN). We showed that S2-ΔN self-associates and is distributed in three groups. One immobile group, one slow diffusing group, which interacts with the nuclear environment and one fast diffusing group, which is not incorporated in high molecular weight complexes. In addition, our FCS and FRAP measurements on S2-ΔN mutants affected in their PIP2 binding showed that PIP2 plays an important role in the distribution of S2-ΔN among these groups, and favors the enrichment of S2-ΔN in the slow diffusing and immobile group. This work indicates that S2 relies on nuclear PIP2 to interact with practically immobile structures, possibly chromatin. © 2013 International Society for Advancement of Cytometry

Cellular signaling is largely dependent on the presence, that is, assembly/disassembly, of supramolecular complexes. In the dynamic process of (re)arranging multimeric protein complexes, the actions of a broad variety of protein modules are crucial. PDZ domains are among the most common protein domains and are present in organisms as diverse as bacteria, yeast, plants, invertebrates, and vertebrates [1]. Their name refers to the first three proteins in which they were originally identified (PSD-95/SAP90, Discs-large and Zona occludens 1) [2].

In the canonical mode PDZ domains recognize motifs of 3–7 amino acids that are present at the C-terminal end of membrane proteins [3]. In 2005 it was shown that the PDZ domains of syntenin-2 (S2) bind phosphatidylinositol 4,5-bisphosphate (PIP2) in surface plasmon resonance experiments with reconstituted liposomes [4]. The PIP2 concentrates at the plasma membrane, in nuclear speckles and nucleoli [5]. Nuclear speckles are dynamic punctuate subnuclear structures localized in the interchromatin regions of the nucleoplasm and enriched in pre-messenger RNA splicing factors [6]. Endogenous S2 is present in the nucleoplasm, nucleoli, and at the plasma membrane of paraformaldehyde fixed MCF-7 cells. Surprisingly, when fused to eGFP or Myc, overexpressed full length S2 is only present in the cytoplasm and excluded from the nucleus, whereas nontagged overexpressed full length S2 was detected with anti-S2 antibodies in the nucleus of MCF-7 cells. This indicated that tagging full length S2 interferes with its nuclear localization and that this interference is independent on the type of the tag. Interestingly, it was shown that the tandem of PDZ domains is necessary and sufficient to target S2 to the plasma membrane and subnuclear structures. Consequently, S2 deleted for the first 94 amino acids (S2-ΔN) is present in the nucleoplasm, nucleoli, and at the plasma membrane of living MCF-7 cells similar to endogenous S2 [4]. Additionally, in paraformaldehyde fixed cells, overexpressed eGFP-S2-ΔN is clearly enriched in nuclear speckles but anti-S2 antibodies fail to detect overexpressed eGFP-S2-ΔN in these nuclear compartments [4].

Although the exact function of S2 is unknown, siRNA experiments suggest that this PDZ protein plays a role in the organization of nuclear PIP2, cell division, and cell survival [4]. As a consequence of its PIP2 interaction, its reported self-association in a yeast two-hybrid study and its speculated interaction with many, yet unidentified, proteins one can hypothesize that S2 plays an important role in cell signaling. Using Fluorescence Correlation Spectroscopy (FCS) and Fluorescence Recovery After Photobleaching (FRAP), we studied the diffusion of S2-ΔN and S2-ΔN mutants affected in their PIP2 binding to characterize the influence of PIP2 on the nuclear binding behavior of S2 in living cells, for example, its presence in high molecular weight complexes, its long term binding or free diffusion.

MATERIALS AND METHODS

Plasmids

For the expression of eYFP in MCF-7 cells, we used a peYFP-C1 construct (Clontech, Laboratories, Mountain View, CA, VS). eGFP-S2-ΔN, eGFP-S2-ΔN-δ1,2, eYFP-S2-ΔN, eYFP-S2-ΔN-δ1, eYFP-S2-ΔN-δ2, and eYFP-S2-ΔN-δ1,2 were constructed as earlier described [4]. To construct mCherry-S2-ΔN, S2-ΔN was cut from eGFP-S2-ΔN and ligated in mCherry using BsrGI and BamHI as restriction enzymes. To construct mCherry-S2-ΔN-δ1,2, S2-ΔN-δ1,2 was cut from eYFP-S2-ΔN-δ1,2 and ligated in mCherry-S2-ΔN using BamHI and EcoRI as restriction enzymes and replacing S2-ΔN. SC35 was obtained from ImaGenes (Image ID 5296371; Source BioScience, Nottingham, UK) and PCR-amplified using primers: sense: GGAAGGCAACTGCCTGAATTCGAGAGG and antisense: GTCTCGGATCCTTCCCAGACATTACC (engineered EcoRI and BamHI sites underlined). SC35 was ligated downstream and in frame with mCherry in mCherry-S2-ΔN using EcoRI and BamHI as restriction sites replacing S2-ΔN. SF3b155 was a kind gift from Dr. M. S. Schmidt-Zachmann (German Cancer Research Center, Heidelberg, Germany) and PCR-amplified using primers: GGTAAGCCCGGGTCTAAA CGCAAACGTAGATGGGATC sense: and GAACGCTCTAGAACCTCCCTATGGTGTTGGAG TTGC antisense: (engineered SmaI and XbaI sites are underlined). Amino acids 194 to 440 of SF3b155 were ligated downstream of mCherry in mCherry-S2-ΔN replacing S2-ΔN. eGFP-mCherry and mCherry-ASF/SF2 were a kind gift from Dr. J. Hendrix (K.U. Leuven, Leuven, Belgium).

Cell Culture and Transient Transfection Assays

MCF-7 cells were cultured in Dulbecco's Modified Eagle medium high glucose (D-MEM) without phenol red (Gibco BRL, Belgium) and supplemented with 10% fetal bovine serum (Sigma-Aldrich gmbh, Taufkrichen, Germany) and GlutaMax-I (Gibco). Cells were transfected using Lipofectamine 2000 (Invitrogen™, Belgium) according to manufacturer's guidelines. To inhibit specifically RNA polymerase II MCF-7 cells were incubated for 6 hours with 5 μg/mL α-amanitin (Roche, Belgium) at 37°C. To stimulate PLC activity in MCF-7 cells 100 μM m-3M3 FBS (Sigma-Aldrich, Bornem, Belgium) was added. FCS measurements were started 1 hour after adding m-3M3 FBS and ended when the MCF-7 cells were incubated with m-3M3 FBS for 2 hours. Before confocal imaging, FRAP or FCS measurements, the medium was changed to DMEM without phenol red and supplemented with 25 mM Hepes buffer (Gibco). For in vivo measurements, MCF-7 cells were kept at 37°C using a home-made heating stage and an objective heater (Bioptechs, Butler, PA, VS).

Paraformaldehyde Fixation

MCF-7 cells were rinsed with PBS buffer (Gibco) and fixed with paraformaldehyde (Formalin Solution, neutral buffered 10%, Sigma-Aldrich) at room temperature for 20 min. After fixation the cells were washed three times using PBS. To avoid artifacts caused by different batches of MCF-7 cells, cells measured for in vivo fluorescence imaging were afterward used for paraformaldehyde fixation.

Fluorescence Microscopy and Fluorescence Recovery After Photobleaching Experiments

Fluorescence microscopy and FRAP measurements were performed on a LSM510 ConfoCor2 microscope system (Carl Zeiss, Jena, Germany) using the 488 nm line of the argon-ion laser to excite eGFP and eYFP and the 543 nm line of the HeNe laser to excite mCherry. The fluorescence was collected with a 40x/1,2 Plan-Apochromat water immersion objective. The fluorescence emission light passed through a 505–530 nm bandpass filter for eGFP and eYFP fluorescence and a 565–615 nm bandpass filter for mCherry fluorescence.

In each FRAP experiment the recovery of fluorescence was recorded in a region of interest in the nucleoplasm or nucleoli and background subtracted by the fluorescence recorded in a region outside the cell. The fluorescence from the reference spot positioned at a similar nuclear environment in the same cell, was used to account for acquisition bleaching. Each FRAP measurement was started by 10 images by 3% AOTF of 16.5 mW at 488 nm before bleaching. Afterward, a bleaching period of ∼66 ms on a circular spot with radius of 10 pixels (∼0.90 μm) was performed by 100% AOTF of 16.5 mW at 488 nm. The complete FRAP measurement was composed of 200 image frames with intervals of 100 ms for eYFP-S2-ΔN, eYFP-S2-ΔN-δ1, eYFP-S2-ΔN-δ2, and eYFP-S2-ΔN-δ1,2 in the nucleoplasm and with intervals of 200ms for eYFP-S2, eYFP-S2-ΔN-δ1, and eYFP-S2-ΔN-δ2 in nucleoli. The average fluorescence intensity in the bleaching spot was normalized before analyzing according to

display math

with math formula and math formula the background-subtracted intensities in respectively the bleach and reference spot and < math formula> and < math formula> the average of the background-subtracted intensities before bleaching in, respectively, the bleach and reference spots. Analysis of fluorescence recovery in the nucleus was done according to a diffusion model based on a circular bleaching profile [7].

Fluorescence Correlation Spectroscopy and Fluorescence Cross-Correlation Spectroscopy Experiments

FCS and FCCS measurements were performed essentially as described in [8] on a LSM510 ConfoCor2 microscope system (Carl Zeiss, Jena, Germany).

In brief, in FCS eYFP was excited using the 488 nm line of argon-ion laser and the emission light was filtered by a 505–550 bandpass filter. The fluctuation data were correlated over 15s and 15 data sets in one specific region of one MCF-7 cell. Only cells with low to medium expression level were chosen on average 20 cells were measured under one condition. The autocorrelation function (ACF) was imported into Igor Pro 5 (Wavemetrics, Lake Oswego, OR, VS) and analyzed by means of global analysis.

When the molecular detection function is approximated by a Gaussian in axial and lateral directions, and the medium is homogeneous, the ACF can be described by an analytical model. The overall autocorrelation curve contains two factors, a first factor describing fluctuations during the residence time of the molecule in the detected volume due to photophysical phenomena (GT), and a factor due to diffusion (GD):

display math

The fast fluctuation term is described as a sum of exponentials:

display math

With Fx and τx the fraction and time of the fast process and n the number of processes. These fast processes are not relevant for our analysis.

The different models used to fit the diffusion behavior have the following general forms:

display math
display math
display math

math formula describes the autocorrelation at slower timescale, representing diffusion; math formula the average number of particles in the confocal volume element, math formula the correlation time; math formula the diffusion time; S is the structure parameter, that is, math formula with math formula and math formula, respectively, the radial and axial radii of the confocal volume element, which are determined by a calibration with Rhodamine 6G [D = 635 μm2/s for diffusion at 37°C in water as recalculated from its diffusion coefficient of 426 μm2/s at room temperature [12]] and are fixed throughout the fitting routine; math formula and math formula the fractions of molecules diffusing with diffusion times math formula and math formula, respectively; math formula the anomaly factor, which is equal to 1 for free (Brownian) diffusion and smaller than 1 for restricted diffusion. Two-component diffusion describes the heterogeneity of the species, whereas anomalous diffusion is a general way to take into account the heterogeneity of the medium (see discussion) and is often used to describe diffusion in the living cell [13].

In FCCS eGFP was used as a fluorescent tag instead of eYFP to minimize bleeding through. eGFP was excited with the 488 nm line of argon-ion and mCherry with the 543 nm line of HeNe. The emission light was filtered by two filter sets (500–530 nm bandpass and 600–650 nm bandpass). Collection and analysis procedures are similar to the FCS measurements. The cross-correlation function was normalized to the green ACF.

RESULTS

S2-ΔN Subnuclear Localization in Living Cells

S2, by virtue of its PDZ domains, localizes at three PIP2 rich cellular locations in paraformaldehyde fixed MCF-7 cells, that is, the plasma membrane, nucleoli and nuclear speckles [4, 5]. Unfortunately, anti-S2 antibodies fail to detect overexpressed eGFP-S2-ΔN in nuclear speckles [4]. Here, we investigated the nuclear localization of S2-ΔN in living cells. Therefore, we fused S2-ΔN N-terminally with eGFP (eGFP-S2-ΔN). Nuclear speckles were identified in living cells using three different speckle markers fused with mCherry, that is, mCherry-SC35, mCherry-ASF/SF2, and mCherry-ΔSF3b155 [14] (Fig. 1A, Supporting Information Figs. S1A and S2A). First, the functionality of these nuclear speckle markers was tested using α-amanitin. This RNA polymerase II inhibitor causes the nuclear speckles to increase in size and change their shape after incubation with 5 μg/ml for 6 hours at 37°C [6, 15]. In agreement, all three nuclear speckle markers showed such behavior (Fig. 1D, Supporting Information Figs. S1D and S2D). When cotransfecting eGFP-S2-ΔN with mCherry-SC35, mCherry-ASF/SF2 or mCherry-ΔSF3b155, we found that eGFP-S2-ΔN concentration was lower in the nuclear speckles than in the nucleoplasm (Figs. 1A–1C, Supporting Information Figs. S1A–S1C and S2A–S2C). Moreover, after adding α-amanitin, the distribution pattern of eGFP-S2-ΔN did not change (Figs. 1D–1F, Supporting Information Figs. S1D–S1F, and S2D–S2F). Interestingly, when these cells were fixed using paraformaldehyde, eGFP-S2-ΔN was translocated to the nuclear speckles in both non-treated as α-amanitin treated cells (Figs. 1G–1L, Supporting Information Figs. S1G–S1L and S2G–S2L), indicating that this translocation is fixation dependent.

Figure 1.

S2-ΔN is not enriched in nuclear speckles in living cells. (AC) Confocal fluorescence micrographs of living MCF-7 cells overexpressing mCherry-SC35 (A) with eGFP-S2-ΔN (syntenin-2 deleted for the first 94 amino acids) (B). In the merge (C) SC35 is in red and S2-ΔN is in green. Note that eGFP-S2-ΔN is not enriched in nuclear speckles in living cells. (D-F) Confocal fluorescence micrographs of living MCF-7 cells as described in (A-C) but after incubation for 6 hours with 5 μg/ml of α-amanitin. Note that the inhibition of transcription by α-amanitin affects the shape of nuclear speckles (compare A to D) and that eGFP-S2-ΔN is not recruited to nuclear speckles after α-amanitin treatment. (G-I) Confocal fluorescence micrographs of MCF-7 cells as described in (A-C) after paraformaldehyde fixation. Note the clear enrichment of eGFP-S2-ΔN in nuclear speckles after fixation. (J-L) Confocal fluorescence micrographs of MCF-7 cells as described in (D-F) after paraformaldehyde fixation. Note again the change in shape of nuclear speckles after α-amanitin treatment (compare J with G). Note eGFP-S2-ΔN enrichment in nuclear speckles. Scale bars 10 μm.

Dynamics of eYFP-S2-ΔN in the Nucleoplasm

Knowing the exact nuclear localization of S2-ΔN in living cells, we studied the mobility of eYFP-S2-ΔN in the nucleoplasm of MCF-7 cells. FCS is particularly suited to measure diffusion of proteins in living cells as it is a noninvasive technique, which can be used to measure the mobility and interactions of proteins inside a living functional cell at spatial and temporal scales [16, 17]. However, FCS in a living cell is subject to many sources of potential artefacts, as extensively described by Enderlein et al. [18]. By controlling as many optical parameters as possible and by measuring at a low penetration depth their impact on the diffusion coefficients from the ACF can be minimized [18, 19].

We compared the FCS measurements of eYFP-S2-ΔN with those of eYFP in different spots in the nucleoplasm (Fig. 2). The ACFs of eYFP-S2-ΔN and eYFP could best be fitted with the two-component diffusion model (Supporting Information Figs. S3A–S3B), which defines the existence of one component with a high and one with a low diffusion coefficient. From the fit, the diffusion coefficients together with the percentage of each component were calculated (Table 1). For eYFP, the slow component contributed only 3 ± 3 %.

Figure 2.

S2-ΔN fusion slows down the diffusion of eYFP in the nucleoplasm and in nucleoli. Normalized autocorrelation curves obtained from FCS measurements in the nucleoplasm and nucleoli of MCF-7 cells overexpressing different constructs as indicated. Note the important shift to longer timescale of eYFP-S2-ΔN compared with eYFP in the nucleoplasm. Additionally, note that eYFP-S2-ΔN is even more slowed down in nucleoli compared with the nucleoplasm.

Table 1. FCS experiments performed on MCF-7 cells transiently transfected with different constructs and treated with α-amanitin or m-3M3 FBS as indicated
 Dfast, 37°C ± s.d. (μm2/s)Dslow, 37°C ± s.d. (μm2/s)Ffast, 37°C ± s.d.χ2 ( math formula)n
  1. FCS experiments were performed on MCF-7 cells transiently transfected with the different constructs as indicated. Additionally, eYFP-S2-ΔN transfected MCF-7 cells were incubated with 5 μg/ml α-amanitin for 6 hours or with 100 μM m-3M3 FBS for minimum 1 and maximum 2 hours as indicated. All autocorrelation curves obtained were fit with the two-component diffusion model to calculate the diffusion coefficients (Dfast, 37°C and Dslow, 37°C, μm2/s), the fraction of the fast component (Ffast, 37°C) and the goodness of fit (χ2). n = number of independent cellular measurements, s.d. = standard deviation.
Nucleoplasm     
eYFP36.4 ± 3.80.6 ± 0.80.97 ± 0.031.3719
eYFP-S2-ΔN22.0 ± 6.51.8 ± 0.60.48 ± 0.084.4127
eYFP-S2-ΔN + α-amanitin17.7 ± 6.81.6 ± 0.60.53 ± 0.118.4332
eYFP-S2-ΔN-δ118.5 ± 6.51.7 ± 0.90.63 ± 0.113.7923
eYFP-S2-ΔN-δ217.5 ± 6.81.4 ± 0.70.59 ± 0.115.1045
eYFP-S2-ΔN-δ1,216.3 ± 5.50.7 ± 0.30.70 ± 0.103.2920
eYFP-S2-ΔN + m-3M3 FBS16.4 ± 8.61.0 ± 0.30.55 ± 0.094.8134
Nucleoli     
eYFP-S2-ΔN14.1 ± 6.71.0 ± 0.20.20 ± 0.092.0512
eYFP-S2-ΔN-δ117.3 ± 9.91.4 ± 0.50.52 ± 0.204.3020
eYFP-S2-ΔN-δ214.0 ± 4.60.8 ± 0.20.29 ± 0.054.6724

As shown in Figure 2, the ACF of eYFP-S2-ΔN strongly shifted to the longer time scale compared with eYFP. A diffusion coefficient of the fast component of eYFP-S2-ΔN (D = 22.0 ± 6.5 μm2/s) is obtained which is only ∼1.6 times smaller than that of eYFP (D = 36.4 ± 3.8 μm2/s). This factor is only slightly higher than the factor math formula= 1.26 which is expected for a spherical molecule with double molecular mass (Mw eYFP = 29.42 kDa; Mw eYFP-S2-ΔN = 51.06 kDa). This suggests that eYFP-S2-ΔN diffuses as a monomer or maximally as a dimer, in view of the limited accuracy of FCS. The slow component however, has to be attributed to a species interacting with elements from the nucleoplasm.

The abundance of nuclear speckles in the nucleoplasm (∼20–50 per cell) together with their small size [0.8–1.8 μm diameter [20]] made it impossible to exclude these nuclear compartments from our nucleoplasmic FCS measurements, even with the coexpression of a nuclear speckle marker. When eYFP-S2-ΔN was measured in the nucleoplasm of α-amanitin treated MCF-7 cells the ACFs could again best be fitted with the two-component diffusion model (Supporting Information Fig. S3C). Only a limited but significant decrease was observed for the fast diffusion coefficient of α-amanitin treated compared with untreated MCF-7 cells. This decrease could be explained by additional nucleoplasmic interactions of eYFP-S2-ΔN caused by the inhibition of transcription. In addition, we cannot conclude from our confocal microscopy studies that eYFP-S2-ΔN is completely excluded from nuclear speckles. However, the slow diffusion coefficient and the amplitude fraction (Ffast 37°C) of α-amanitin treated and untreated MCF-7 cells were not significantly different (Table 1 and Supporting Information Table S1).

Dynamics of eYFP-S2-ΔN in Nucleoli

Comparable to the nucleoplasm, in nucleoli, the two-component diffusion model also gives the best description of the ACF of eYFP-S2-ΔN (Supporting Information Fig. S4A). A pronounced decrease in the fraction of the fast component was noted in nucleoli compared with the nucleoplasm. In addition, both diffusion coefficients were somewhat decreased, most likely due to an increased “viscosity” of the nucleolus (Table 1 and Fig. 2).

We also used FRAP to detect potentially slow or immobile groups of eYFP-S2-ΔN in cells, which could be missed by FCS. As the diffusion of eYFP-S2-ΔN was relatively fast, live cells were imaged at room temperature (∼21°C) to slow down these processes. Additionally, at this lower temperature displacements of the cells during the measurements were reduced as well. The recovery curves obtained in minimum nine independent measurements were normalized and fitted using the diffusion model proposed by Soumpasis [7]. A small immobile fraction of 8% (Fimm = 0.08 ± 0.05) was obtained in the nucleoli in a 18s measurement. Bleaching could deplete the nucleolar fluorescent S2 pool, and therefore, its immobile fraction would be overestimated. However, in fluorescent loss of photobleaching experiments we observed a fast exchange of S2 between the nucleoplasm and nucleoli (results not shown). Therefore, the depletion of the fluorescent S2 pool in nucleoli can be neglected. In the nucleoplasm, already after 10s, an almost 100% recovery was noted (Fimm = 0.05 ± 0.03) (Table 2).

Table 2. FRAP experiments performed on MCF-7 cells transiently transfected with the different constructs as indicated
 Fimm±s.dn
  1. All FRAP curves were normalized and fitted as described in materials and methods. s.d. is the standard deviation and n the number of independent cellular measurements.
Nucleoplasm  
eYFP-S2-ΔN0.05 ± 0.039
eYFP-S2-ΔN-δ10.04 ± 0.046
eYFP-S2-ΔN-δ20.04 ± 0.039
eYFP-S2-ΔN-δ1,20.02 ± 0.0410
Nucleoli  
eYFP-S2-ΔN0.08 ± 0.0511
eYFP-S2-ΔN-δ10.03 ± 0.057
eYFP-S2-ΔN-δ20.03 ± 0.0420

In the Nucleus, PIP2 Favors the Enrichment of S2-ΔN in the Slow Diffusing Components

Nucleoplasm

Using previously described S2-ΔN mutants [4] that are defective in their PIP2 binding in the first (S2-ΔN-δ1), the second (S2-ΔN-δ2), or both (S2-ΔN-δ1,2) PDZ domains, we aimed to explore the effects of PIP2 interactions in living cells. Similar to eYFP-S2-ΔN, the best fit of the ACFs of these mutants was obtained with the two-component diffusion model for all constructs in the nucleoplasm and nucleoli (Supporting Information Figs. S3D–S3F and S4B–S4C). When S2-ΔN was unable to bind PIP2 (eYFP-S2-ΔN-δ1,2) a strong increase of the fraction of the fast component was noted compared with eYFP-S2-ΔN (full-PIP2 binding). The average diffusion coefficients of eYFP-S2-ΔN-δ1,2 were surprisingly somewhat lower than those of eYFP-S2-ΔN, although a broad distribution of diffusion coefficients is observed (see standard deviations Table 1). Second, a significant enrichment in the fast fraction was observed when only one PDZ domain was mutated for its PIP2 binding compared with eYFP-S2-ΔN in the nucleoplasm. However, this enrichment was less than with the double mutant. In addition, both the fast and slow diffusion coefficient of the single mutants was again slightly decreased. However, only eYFP-S2-ΔN-δ2 was significantly different from eYFP-S2-ΔN (Table 1 and Supporting Information Table S1).

Nucleoli

To define if nuclear PIP2 also influences the diffusion of nucleolar S2-ΔN, we measured eYFP-S2-ΔN-δ1 and eYFP-S2-ΔN-δ2 using FCS in the nucleoli. Again, a significant increase in the percentage of the fast component because of the loss of PIP2 binding was noted compared with eYFP-S2-ΔN (Table 1 and Supporting Information Table S1).

Colocalization experiments and treatments known to affect cellular PIP2 levels have shown that the subnuclear localization of S2-ΔN is PIP2 dependent. In paraformaldehyde fixed MCF-7 cells [4] and in living cells (Fig. 3C), eYFP-S2-ΔN-δ1,2 is excluded from the nucleoli. We used FRAP to determine if the interaction with nuclear PIP2 leads to long-term binding of S2-ΔN in nucleoli. When the immobile fraction of eYFP-S2-ΔN (0.08 ± 0.05) was compared with those of eYFP-S2-ΔN-δ1 and eYFP-S2-ΔN-δ2 (Fimm, eYFP-S2-ΔN-δ1 = 0.03 ± 0.05 and Fimm, eYFP-S2-ΔN-δ2 = 0.03 ± 0.04) only a significant difference was measured for eYFP-S2-ΔN-δ2 (P = 0.0106) (Table 2 and Supporting Information Table S2). However, for both eYFP-S2-ΔN-δ1 and eYFP-S2-ΔN-δ2 a clear trend to a decreased long-term binding of S2-ΔN in nucleoli is observed when PIP2 binding is lost in the first or second PDZ domain (Fig. 3D and Table 2). Although in the nucleoplasm the immobile fraction of eYFP-S2-ΔN was very low (Fimm = 0.05 ± 0.03), the same trend to decrease the immobile fraction when one or both PDZ domains were mutated for their PIP2 binding is observed (Fimm, eYFP-S2-ΔN-δ1 = 0.04 ± 0.04; Fimm, eYFP-S2-ΔN-δ2 = 0.04 ± 0.03; Fimm, eYFP-S2-ΔN-δ1,2 = 0.02 ± 0.04) (Table 2).

Figure 3.

The recruitment of S2-ΔN towards the nucleoli is PIP2 dependent. (A-C) Confocal fluorescence micrographs of MCF-7 cells overexpressing eYFP-S2-ΔN mutated to abolish PIP2 binding in the first (eYFP-S2-ΔN-δ1) (A), the second (S2-ΔN-δ2) (B) or both (S2-ΔN-δ1,2) PDZ domains (C). Note that eYFP-S2-ΔN-δ1,2 is excluded from nucleoli. Scale bars denote 10 μm. (D) The averaged normalized fluorescence recovery from minimum 7 independent FRAP experiments in nucleoli of MCF-7 cells expressing different constructs as indicated are plotted in function of time. Error bars correspond to standard deviations from the mean. Note the decrease of the immobile fraction when eYFP-S2-ΔN-δ1 and S2-ΔN-δ2 are measured and compared to the fluorescence recovery of eYFP-S2-ΔN.

Because mutations were introduced in the carboxylate binding loop and the N-terminus of the αB-helix, which are both known to be important for the peptide binding of PDZ domains, an additional approach was attempted to confirm our measurements. As shown previously, nuclear PIP2 is required for subnuclear enrichment of eGFP-S2-ΔN [4]. PLC-mediated hydrolysis of PIP2 is the major catabolic pathway for PIP2 [21]. Activating PLC by incubating eYFP-S2-ΔN transfected MCF-7 cells with 100 μM m-3M3 FBS, that is, a general PLC activator, should decrease cellular PIP2 levels [22]. As previously reported, already after 25 minutes eGFP-S2-ΔN translocates from the nucleus toward the cytoplasm and the plasma membrane [4]. In agreement, in our confocal microscopy measurements, eYFP-S2-ΔN translocates from the nucleus toward the cytoplasm and the plasma membrane after incubation with 100 μM m-3M3 FBS (data not shown). This suggests that nuclear PIP2 levels are effectively decreased. Therefore, after incubation with 100 μM m-3M3 FBS for minimum 1 and maximum 2 hours, the diffusion of the remaining eYFP-S2-ΔN was measured in the nucleoplasm of MCF-7 cells using FCS. Taking the standard deviations into account a limited decrease in the diffusion coefficients and a limited increase in the percentage of the fast component was observed when eYFP-S2-ΔN measured in m-3M3 FBS incubated cells was compared with eYFP-S2-ΔN wild type. The trend in the parameters is the same as the one observed for eYFP-S2-ΔN-δ1,2 (Table 1) and are thus consistent with our mutant studies.

S2-ΔN Self-Associates in a PIP2 Independent Manner

Some PDZ proteins are reported to homodimerize and heterodimerize [23-28] and S2 was reported to self-associate in yeast two-hybrid experiments [29]. This obliged us to find out if S2-ΔN forms dimers (or oligomers) in living cells. Therefore, we used FCCS to check the codiffusion of eGFP-S2-ΔN and mCherry-S2-ΔN in the nucleoplasm of MCF-7 cells [30-32]. Control experiments were designed to calculate the false signal because of cross-talk between the green and red channel. Therefore, MCF-7 cells transfected with eGFP-S2-ΔN were measured. The counting rate ratio of red to green was around 10% on 488 nm excitation. This 10% signal in the red channel is due to bleeding through from green to red. No significantly detectable bleeding through from red to green was observed. As a negative control of codiffusion, FCCS was measured in the nucleoplasm of MCF-7 cells cotransfected with eGFP-S2-ΔN and mCherry. The transfection procedure was optimized to obtain equal to almost equal number particles for eGFP-S2-ΔN and mCherry in the green and red channel, respectively. A normalized cross-correlation amplitude of 0.17 ± 0.05 was obtained (Supporting Information Fig. S5A). Second, the fusion protein eGFP-mCherry was used as a positive control and gave a normalized cross-correlation amplitude of 0.38 ± 0.03 (Supporting Information Fig. S5B). After optimization of the transfection procedure similar to the negative control, FCCS measurements in the nucleoplasm of MCF-7 cells cotransfected with eGFP-S2-ΔN and mCherry-S2-ΔN showed a normalized cross-correlation amplitude of 0.35 ± 0.13 (Fig. 4A and Supporting Information Fig. S5C). This indicates that S2-ΔN self-associates in the nucleoplasm. Interestingly, when eGFP-S2-ΔN-δ1,2 and mCherry-S2-ΔN-δ1,2 were measured, a normalized cross-correlation value not significantly different from S2-ΔN (P = 0.1012) was calculated (S2-ΔN math formula; S2-ΔN-δ1,2 math formula) (Supporting Information Figs. 4A and S5D). This indicates that the self-association of S2-ΔN in the nucleoplasm is largely independent of nuclear PIP2 binding.

Figure 4.

S2-ΔN forms homo-dimers in the nucleoplasm. (A) Histograms of the mean amplitudes of cross-correlation curves obtained from Fluorescence Cross-Correlation Spectroscopy (FCCS) experiments measured in the nucleoplasm of MCF-7 cells transiently expressing different constructs as indicated. Error bars correspond to standard deviations from the mean of minimum 12 independent measurements. Fusion of eGFP and mCherry was used as a positive control. Cross-correlation between mCherry and eGFP-S2-ΔN was used as a negative control. Note the codiffusion of eGFP-S2-ΔN and mCherry-S2-ΔN. Note also that eGFP-S2-ΔN-δ1,2 and mCherry-S2-ΔN-δ1,2 still co-diffuse. * denotes a significant difference (P < 0.05).

(B) Brightness-values or counts per molecules (cpm in kHz per molecule) are measured for eYFP and eYFP-S2-ΔN in the nucleoplasm of transfected MCF-7 cells and divided into different brightness-intervals. The histogram shows the percentage of events that a given brightness-value belongs to the indicated brightness-interval.

To estimate the degree of homoassociation, we compared the molecular brightness of eYFP-S2-ΔN with eYFP [33]. Therefore, the averaged fluorescent intensities of eYFP and eYFP-S2−ΔN were divided by their respective average number particles to obtain the apparent brightness or counts per molecules (cpm) of eYFP and eYFP-S2-ΔN. Afterward, these brightness values were normalized to the averaged cpm of eYFP in the nucleoplasm and plotted into histograms (Fig. 4B). The distribution of the brightness values of eYFP-S2-ΔN was increased compared with eYFP and a second peak with an increased observed brightness of 1.5 ± 0.2 kHz is visible. The presence of two peaks indicates that in part of the nucleoplasmic measurements only monomeric eYFP-S2-ΔN is observed, whereas in others a mixture of monomeric and probably dimeric eYFP-S2-ΔN is detected. However, in the ideal case of no quenching, bleaching, or binding to endogenous S2 the homodimerization of eYFP-S2-ΔN should lead to a brightness value of 2.0 kHz. In the two-components diffusion model the total apparent brightness is dependent on the number particles and brightness of each component in a nonlinear Equation [34]. Using this dependence and assuming a brightness value of 2.0 kHz for the dimer, we calculate that our observed brightness of 1.5 ± 0.2 kHz corresponds to a fraction of protein in dimers of 50%. It should be realized that this equilibrium position is valid at the micromolar concentrations of fluorescent S2 observed in our selected cells. It is possible that the endogenous concentration is much lower, and therefore, the degree of association of the endogenous S2 might be lower.

DISCUSSION

S2-ΔN is Not Enriched in Nuclear Speckles in Living Cells

As a PDZ protein, S2 most likely interacts with many partners but only a few are identified today. Interestingly, in 2005, it was shown that nuclear PIP2 binds to the PDZ domains of S2 [4]. Unfortunately, for S2 only limited information is available and although much is known about the phosphoinositide signaling at the cytoplasm, only in the last decades evidence accumulated suggesting its presence in the nucleus. However, the PDZ-PIP2 interaction of S2 together with its reported self-association in yeast two-hybrid experiments suggest that S2 plays an important role in cell signaling. Therefore, we studied the influence of PIP2 on the binding behavior of S2 in the nucleus of living cells. To study this, FCS, FCCS, and FRAP are excellent tools as they are noninvasive techniques, which can be used to measure the mobility and interactions of proteins inside a living functional cell at spatial and temporal scales [16, 17].

According to the major view in the field at present, nuclear speckles function as storage/assembly/modification compartments that can supply splicing factors to active transcription sites [35]. Depletion of S2 by siRNA disrupts the nuclear speckles-PIP2 pattern in fixed MCF-7 and U-2 OS cells, which suggests a role for S2 in the nuclear speckles formation [4]. In paraformaldehyde fixed cells, eYFP-S2-ΔN is present at the plasma membrane, in the nucleoplasm and highly enriched in nucleoli and nuclear speckles [4]. Interestingly, however, in our colocalization studies with different nuclear speckle markers we showed that the enrichment of S2-ΔN in nuclear speckles is fixation dependent. Redistribution toward nuclear speckles after fixation is also described in the literature for RNA polymerase II [36]. Although we could not conclude that S2 is excluded from nuclear speckles, this finding was rather surprising and stresses the risk of biological misinterpretations after fixation. However, S2 diffusion is influenced when α-amanitin, known to inhibit RNA polymerase II and enlarge nuclear speckles, is added to the cell. Therefore, we speculate that a low percentage of S2 is still present in these nuclear compartments, although its function remains unknown.

S2 is Distributed in Three Diffusing Groups in the Nucleus

Studying the diffusion of S2-ΔN in the nucleus, we could gain insight in its binding behavior in living cells since, for example, high molecular weight complexes, long term binding or free diffusion can be determined. In the nucleoplasm and nucleoli the ACFs of eYFP-S2-ΔN had to be fitted by a two-component diffusion model indicating two pools of the protein that are not in fast exchange. In the nucleoplasm, we noticed a ∼1.5 decrease in the fast diffusion coefficient of eYFP-S2-ΔN compared with eYFP. This factor is only slightly higher than what can be expected from its molecular weight relative to eYFP and suggests the presence of a monomeric protein. However, using FCCS, we could show that S2-ΔN does (in) directly self-associate in the nucleoplasm. Interestingly, when the slow component of eYFP-S2-ΔN is compared with its fast component, the decrease in the diffusion coefficient is remarkable (about 10-fold). This factor is much too high to represent an increased molecular mass. The slow diffusion of eYFP-S2-ΔN is therefore likely due to interactions with immobile nuclear structures, possibly chromatin.

In nucleoli a further decrease in the high diffusion coefficients was noted compared with the nucleoplasm. This decrease is most likely caused by an increased “viscosity” [37]. Additionally, an enrichment in the slow component is observed. Therefore, also in nucleoli two fractions of eYFP-S2-ΔN are detected by FCS: a minor fraction which is (almost) freely diffusing and a major fraction that is interacting with practically immobile structures, possibly nucleolar chromatin. Identification of partners of eGFP-S2-ΔN in the nucleolus will clarify the nature of its slow diffusing components.

Measuring FRAP we observed a 5% and 8% immobile fraction in, respectively, the nucleoplasm and nucleoli. This immobile fraction indicates long term binding and is, in addition to the two components measured by FCS, the third group of S2-ΔN present in the nucleus.

The PDZ-PIP2 Interaction Plays an Important Role in the Nucleoplasm and in Nucleoli

Although nuclear PIP2 is implicated in a diverse set of nuclear activities such as gene transcription, chromatin remodeling, cell proliferation, cell cycle regulation, and mRNA metabolism [5, 38-43], the global significance of nuclear phosphoinositide-protein interactions is still poorly understood. This is largely due to the fact that only few nuclear effector proteins are known. Interestingly, those phosphoinositide-protein interactions that have been studied in detail appear to have profound physiological effects [43-45]. To our knowledge none of these were investigated by dynamic studies such as FCS, FCCS, or FRAP. Nevertheless, efforts are taken in the literature to explore phosphoinositide-protein interactions and recently a proteomic screen identified 28 nuclear PIP2 interacting proteins that could provide an important resource on which to base further investigations [46].

To get insight into the influence of nuclear PIP2 on the binding behavior of S2, we studied the diffusion of S2-ΔN PIP2 mutants in the nucleus of living MCF-7 cells. In both the nucleoplasm and nucleoli, an increased percentage of the fast component was measured by FCS and a decreased immobile fraction by FRAP when these S2-ΔN PIP2 mutants were compared to eYFP-S2-ΔN. This indicates that nuclear PIP2 plays an important role in the distribution among the three pools of eYFP-S2-ΔN and favors the enrichment of this PDZ protein in the slower diffusing pool. Therefore, this suggests that when eYFP-S2-ΔN looses the ability to bind nuclear PIP2, the expected scaffolding properties of this PDZ protein decrease. However, even without any PIP2 binding, as in the case of eYFP-S2-ΔN-δ1,2, a slow diffusing group is still noted that indicates that also PIP2-independent interactions contribute. Interestingly, when eGFP-S2-ΔN-δ1,2 and mCherry-S2-ΔN-δ1,2 were measured in the nucleoplasm using FCCS, no significant decrease or increase in the cross-correlation value was noted compared to intact S2-ΔN. Therefore, the self-association of S2-ΔN seemed to be independent of PIP2. In addition, limited decreased diffusion coefficients measured for both the slow and fast component of eYFP-S2-ΔN PIP2 mutants compared with eYFP-S2-ΔN make the story even more complex. This decrease suggests that the loss of PIP2 binding enables S2-ΔN to associate with additional interactors, not yet identified for S2-ΔN. All together, the different combinations of S2 binding with nuclear structures, PIP2 and not yet identified interactors fit well with the expected function of S2 in cell signaling. Moreover, cellular differences in PIP2 concentration can regulate S2 interactions with nuclear structures and other binding partners not identified yet.

When measuring mutant forms of eYFP-S2-ΔN defective for PIP2 binding, results must be interpreted with caution, as these mutations were made in structural features known to be important for the peptide binding of PDZ domains (i.e., the carboxylate binding loop and the N-terminus of the αB-helix). Interestingly, mutating the PIP2 binding in one or both PDZ domains of S2 does not destroy the peptide binding of tetraspanin L6A, that is, the only known peptide ligand of S2 until now [4, 47]. A possible way to confirm that the effects were related to PIP2, was to decrease nuclear PIP2 levels by adding drugs. Therefore, we incubated MCF-7 cells with a general PLC activator, that is, m-3M3 FBS, which decreases cellular PIP2 as reported previously [22]. As a result, the percentage of the fast component of eYFP-S2-ΔN increased, although not to the same extent as of eYFP-S2-ΔN PIP2 mutants. However, it is to be expected that the diffusion of eYFP-S2-ΔN PIP2 mutants, which are unable to bind PIP2 in one or both PDZ domains, will be more affected than the diffusion of eYFP-S2-ΔN in an environment with decreased but not fully depleted PIP2 levels.

The interactions we have identified were all based on the diffusion or absence of diffusion of the fluorescent proteins under study. In this way, it was possible to identify interactions with unknown nonfluorescent partners. Also, FCCS measurements of homo-oligomerization are based on codiffusion. This is in contrast with other techniques, for example, based on FRET signals, where specific interactions are measured independent on diffusion and more dependent on a specific short distance interaction. For these type of studies the partners have to be known, and the degree of association and the complex stoichiometry can in this case be calculated provided the intrinsic FRET efficiency of the complex can be determined [48, 49].

Ancillary