The centrosome-nucleus attachment is a prerequisite for faithful chromosome segregation during mitosis. We addressed the function of the nuclear envelope (NE) protein Sun-1 in centrosome-nucleus connection and the maintenance of genome stability in Dictyostelium discoideum. We provide evidence that Sun-1 requires direct chromatin binding for its inner nuclear membrane targeting. Truncation of the cryptic N-terminal chromatin-binding domain of Sun-1 induces dramatic separation of the inner from the outer nuclear membrane and deformations in nuclear morphology, which are also observed using a Sun-1 RNAi construct. Thus, chromatin binding of Sun-1 defines the integrity of the nuclear architecture. In addition to its role as a NE scaffold, we find that abrogation of the chromatin binding of Sun-1 dissociates the centrosome-nucleus connection, demonstrating that Sun-1 provides an essential link between the chromatin and the centrosome. Moreover, loss of the centrosome-nucleus connection causes severe centrosome hyperamplification and defective spindle formation, which enhances aneuploidy and cell death significantly. We highlight an important new aspect for Sun-1 in coupling the centrosome and nuclear division during mitosis to ensure faithful chromosome segregation.
The nuclear envelope (NE) separates the nuclear compartment from the cytoplasm. It is composed of two membranes: the outer nuclear membrane (ONM) and the inner nuclear membrane (INM). The lumen between the two membranes is the perinuclear space (PNS). The ONM is continuous with the endoplasmic reticulum (ER), whereas the INM harbors a unique set of proteins. INM and ONM proteins can interact within the PNS. Underneath the INM, the nuclear lamina is located, which is formed by intermediate filament (IF) proteins and associated proteins. The lamina forms the nucleoskeleton and associates with the INM, chromatin and nuclear pore complexes. Proteins of the NE have important roles. They are involved in nuclear migration and positioning and are essential for many processes such as mitosis, meiosis, differentiation and cell migration. Furthermore, several of the NE proteins have been associated with inherited diseases (1,2).
Research in mammalian cells and in Caenorhabditis elegans has identified conserved components of the NE that link the nucleoskeleton to the cytoskeleton. In C. elegans, two putative INM proteins, matefin/SUN-1 and UNC-84, bind to the nuclear lamina and extend their C-terminus into the PNS where they interact with the C-termini of KASH domain proteins (Klarsicht/Anc-1/Syne homology, designated KASH domain). Matefin/SUN-1 and UNC-84 belong to the SUN family of proteins based on the presence of the conserved SUN (Sad1/UNC-84 homology) domain at their C-terminus. KASH domain proteins are type II transmembrane proteins of the NE and have been identified as molecular linkers connecting the nucleus to actin filaments [filamentous actin (F-actin)], IFs and microtubules (MTs) (3). C. elegans harbors three KASH domain proteins, Anc-1, a huge ONM protein with an F-actin-binding domain (ABD) of the α-actinin type at its N-terminus with which it can link to the actin cytoskeleton, UNC-83 that associates with MTs, and ZYG-12, which provides a link to the centrosome. Anc-1 and UNC-83 interact with UNC-84, ZYG-12 with matefin/SUN-1 (1).
In mammalian cells, Nesprin-1/Enaptin and Nesprin-2/NUANCE are homologues of Anc-1, and their largest isoforms also have ABDs for association with F-actin; Nesprin-3 binds to the ABD of plectin, which itself can bind to the IF cytoskeleton. SUN-1 and SUN-2 are the best characterized SUN/UNC-84 homologues in mammalian cells and, like the C. elegans proteins, interact with the C-terminus of the Nesprins in the PNS (3). Taken together, by combination of the SUN domain proteins with diverse KASH domain proteins, the nucleus can be linked simultaneously to different cytoskeletal elements and to the centrosome.
In yeasts, recent studies revealed that SUN domain proteins organize the chromatin during homologous chromosome pairing in meiosis. In Saccharomyces cerevisiae, the SUN domain protein Mps3 that is involved in karyogamy and sister chromatid cohesion interacts with the meiotic telomere protein Ndj1 to cluster the telomeres to the NE, which is known as the bouquet formation (4–6). In Schizosaccharomyces pombe, the bouquet formation is mediated by the complex containing the SUN domain protein Sad1 and the linker proteins Bqt1 and Bqt2. Sad1 is associated with the spindle pole body that is the centrosome equivalent in fission yeast. The movement of the chromosomes toward their appropriate homologues occurs through the interaction of Sad1 with the KASH domain protein Kms1 that engages a cytoplasmic dynein motor complex (7). Similarly, in C. elegans, the complex containing matefin/Sun-1 and Zyg-12 tethers and moves the chromosomal pairing centers along the NE to allow homologue recognition and clustering. Accordingly, mammalian Sun-1 attaches the telomeres to the NE to promote homologous chromosome pairing and synapsis formation in meiotic prophase I (8–10).
To date, little is known about the NE and NE proteins in Dictyostelium discoideum. Like yeast, D. discoideum does not have lamins and undergoes a closed mitosis. The first NE protein described so far, interaptin (11), may function as a KASH domain protein in analogy to the findings in higher eukaryotes. The prediction of SUN domain proteins in the D. discoideum genome prompted us to investigate the function of Sun-1 and its interplay with interaptin. Here, we report that Sun-1 is retained in the INM through binding to chromatin and defines the spacing of the NE lumen as well as the centrosome-nucleus juxtaposition. Abrogation of the Sun-1 binding to chromatin induces dramatic separation of the two nuclear membranes and loss of the centrosome-nucleus connection that ultimately results in centrosome hyperamplification causing defective mitotic spindle formation that induces chromosome missegregation and genome instability.
Sun-1 is an INM protein and binds to chromatin
The 105 kDa D. discoideum Sun-1 (DDB0219949) homologue is composed of an N-terminal coiled-coil domain (amino acids 170–221), a single transmembrane domain (aa 291–313), a pair of coiled-coil domains (aa 412–457 and 507–571) and a C-terminal SUN domain (aa 712–859) that terminates in a putative ER retention signal SDEL that is unique for D. discoideum Sun-1 (Figure 1A). Sun-1 has features found in different SUN domain protein homologues: (i) it shares the single transmembrane domain with the C. elegans UNC-84 and the S. pombe Sad1. (ii) Like the mammalian SUN domain proteins, Sun-1 possesses coiled-coil domains that are absent in Ce UNC-84 and Sp Sad1 (12,13). The SUN domain of Dd Sun-1 shows highest homology to the one of vertebrate Sun proteins (Figure 1A). Sun-2 (DDB0186751) represents the second SUN domain protein in D. discoideum. It has a different domain structure with a centrally located SUN domain and belongs to the SUN-like proteins, which are also present in the proteome of yeast and flies (14).
As the SUN domain proteins that are involved in nuclear positioning and migration possess a C-terminal SUN domain, we focused on the analysis of Sun-1 in D. discoideum that may represent an orthologue of the classical SUN domain proteins.
In wild-type AX2 cells, Sun-1 was present at the NE, where it colocalized with green fluorescent protein (GFP)-tagged nuclear pore protein Nup43 and to some extent also in the cytoplasm (Figure 1B,B′ and B″ and Figure S1) using monoclonal antibody (mAb) K55-460-1 generated against a polypeptide encompassing the two central coiled-coil domains (aa 344–653; Figure 1A). Upon fractionation of total cell lysates on discontinuous sucrose gradients, Sun-1 was exclusively present in the fractions of highest density, which correspond to the nuclear fractions and ER membranes (15) as we find the ER marker PDI in these fractions as well (Figure 1C). The Dictyostelium KASH domain protein interaptin (11) showed a similar distribution but was also present in lighter fractions resembling the distribution of PDI (Figure 1C). Coimmunofluorescence studies using calreticulin–GFP and calnexin–GFP expressing cells as well as PDI antibodies supported this localization (Figure S1). Sun-1 was partially extracted from nuclei by addition of Triton-X-100, urea or a combination of both reagents, indicating that Sun-1 is hydrophobic and strongly associated with the NE (Figure 1D). Similarly, mammalian Sun-2 is resistant to Triton-X-100 and urea extraction, characteristics that are thought to be because of its immobilization on lamins (16,17). Distinct from higher eukaryotes, D. discoideum lacks lamins; thus, Sun-1 may bind to other nuclear components.
To investigate the targeting and retention of Sun-1, we first addressed its membrane topology in proteinase K protection assays using intact nuclei. A 70 kDa Sun-1 fragment representing the complete C-terminus including the transmembrane domain (aa 291–905) was protected from proteolysis (Figure 1E). As the low molecular mass of proteinase K enables it to diffuse into the nucleoplasm and degrade epitopes on the INM, we were unable to determine the precise localization of Sun-1 at the INM and/or ONM from these assays. In conclusion, Sun-1 is a type II transmembrane protein, which exposes its N-terminus to either the nucleoplasm and/or the cytoplasm and its C-terminus to the PNS. A location of the N-terminus to the nucleoplasm is, however, favored based on the findings that all SUN domain proteins studied so far interact directly or indirectly with lamins, although their INM targeting is lamin independent (18–21).
As Dictyostelium lacks lamins, we hypothesized that Sun-1 may be retained in the INM through interaction with chromatin. We did not expect DNA sequence specificity for Sun-1, as the chromatin binding is supposed to anchor Sun-1. Therefore, we analyzed the protein–DNA interaction by polymerase chain reaction (PCR) amplification of the housekeeping gene actin-8. In chromatin immunoprecipitation (ChIP) experiments, chromatin was coprecipitated with the endogenous Sun-1 (Figure 2A), indicating that chromatin binding may represent a lamin-independent INM retention mechanism and that the N-terminus of Sun-1 is most likely involved in chromatin binding. Next, we subjected purified recombinant Sun-1 N-terminus [glutathione S-transferase (GST)-Sun1N400] to electromobility shift assays (EMSA) to validate whether the Sun-1 N-terminus interacts directly with DNA. GST-Sun1N400 interacted directly with radiolabeled DNA, whereas GST did not (Figure 2B). Furthermore, we analyzed the DNA binding using non-induced and induced bacterial total lysates containing GST-Sun1N400 or GST-Sun1N800, a polypeptide encompassing the Sun1N400 sequences, using the Southwestern technique. Both GST-Sun1N400 and GST-Sun1N800 interact directly with radiolabeled DNA, whereas non-induced bacterial lysates or a GST-fusion protein containing the coiled-coil domains of Sun-1 (Sun1CT) did not show DNA interaction (Figure 2C and Figure S2). In addition, both GST-Sun1N400 and GST-Sun1N800 interact not only with the DNA sequence of D. discoideum hp1 but also with that of D. discoideumactin-8 and mammalian DNA, suggesting that the Sun-1 N-terminus does not recognize specific DNA sequences (Figure S2). Collectively, these data obtained from ChIP and EMSA demonstrate that the Sun-1 N-terminus binds to chromatin and DNA. Distinct from mammalian Sun-1 that possesses a zinc-finger domain at the N-terminus, the N-terminus of D. discoideum Sun-1 does not contain known DNA-binding motifs, but it is likely that Sun-1 has a cryptic DNA-binding domain with strong DNA-binding affinity that allows the INM retention of Sun-1. By extension, this interaction may be responsible for the INM targeting and retention of SUN domain proteins in other organisms.
Sun-1 forms homodimers and higher oligomers
Many SUN domain proteins contain predicted coiled-coil domains through which homodimerization might occur (17). Analysis of the Dictyostelium Sun-1 sequence revealed that the central pair of coiled-coil domains harbors hot spots of conserved amino acid residues such as R, Y, F, H and M, which have the potential to promote strong protein–protein interactions, as well as L and I that contribute moderate protein–protein interfaces (Figure 2D) (22,23). We tested the capability of Sun-1 to oligomerize using recombinant SunCT1 (residues 344–653) that encompasses the pair of coiled-coil domains. First, we used circular dichroism (CD) spectrum analysis to confirm that the bacterially expressed polypeptide had folded. The CD spectrum showed two minima near 210 and 220 nm typical of an α-helical protein and indicated that the protein had folded (Figure S3). In the presence of the cross-linking reagent glutaraldehyde, the polypeptide formed dimers and higher oligomers (Figure 2E). Notably, the native non-cross-linked SunCT1 sample contained some amount of dimers (approximately 80 kDa) and trimers (approximately 100 kDa) that were not disassembled into monomers (40 kDa) under denaturing conditions (Figure 2E). To further strengthen a self-interaction of SunCT1, the native and cross-linked protein was analyzed by gel filtration chromatography using a Sephadex G-75 column. In both cases, Sun1CT eluted primarily in fraction 10 corresponding to the dimer and trimer. The monomer eluted in fraction 11 (Figure 2F).
N-terminal truncation of Sun-1 abolishes INM targeting
To investigate whether the INM targeting and retention of Sun-1 requires chromatin binding, the construct GFP-ΔNSun-1 in which the Sun-1 N-terminus was replaced by a GFP tag was stably expressed in the wild-type AX2 genetic background that did not affect the viability of the cells. GFP-ΔNSun-1 is localized to the NE like the endogenous Sun-1, suggesting that the N-terminus is dispensable for the NE targeting of Sun-1 (Figure 3A,E). As GFP-ΔNSun-1 contains the complete C-terminus, it is likely that that the C-terminus of Sun-1 determines its NE localization.
Within the mutant strain, we compared the distribution of GFP-ΔNSun-1 in the nuclear membranes with that of the endogenous Sun-1 using sequential digitonin and Triton-X-100 permeabilization. First, the plasma membrane was selectively permeabilized with digitonin to address cytoplasmic epitopes on the ONM. The polyclonal rabbit GFP antibodies (pAb GFP) detected the N-terminus of GFP-ΔNSun-1 at the NE facing the cytoplasm (Figure 3B). Under this condition, the mAb K55-460-1 failed to stain GFP-ΔNSun-1 and the endogenous Sun-1, confirming that both proteins locate their C-terminus in the PNS (Figure 3F). The inaccessibility of the coiled-coil domains to the mAb K55-460-1 in the PNS also proved the NE integrity upon digitonin treatment. In the second step, all cellular membranes were permeabilized with Triton-X-100 application that enables the mAb K55-460-1 to access the coiled-coil domains of both GFP-ΔNSun-1 and the endogenous Sun-1. In the analysis, we frequently observed separation of the INM from the ONM at several sites of the GFP-ΔNSun-1-positive nuclei (Figure 3C,D).
Intriguingly, at sites of the separated nuclear membranes, the endogenous Sun-1 was targeted to the INM and was also observed in the nucleoplasm (Figure 3C,D, arrow), whereas GFP-ΔNSun-1 accumulated in the ONM (Figure 3D, arrowhead), indicating that the chromatin binding of Sun-1 is essential for INM retention presumably through its N-terminus. As a control, pAb GFP staining after Triton-X-100 permeabilization did not reveal separation between the INM and the ONM, and we assume that GFP-ΔNSun-1 was restricted to the ONM (Figure 3G,H). In immunoelectron microcopy studies using GFP antibodies, the ONM accumulation of GFP-ΔNSun-1 was confirmed (Figure 3I,I′, arrowheads) as well as the separation of the INM and ONM (Figure 3I,I′, asterisks) implying that Sun-1 may regulate the spacing of the NE lumen, probably by interaction with an ONM protein. According to this hypothesis, GFP-ΔNSun-1 in the ONM may disturb this interaction by binding with its C-terminus to the yet unknown ONM partner leading to the separation of the INM and the ONM. To some extent, GFP-ΔNSun-1 was also detected in the INM that might be because of its heterodimerisation with the endogenous Sun-1 allowing the translocation of the truncated protein to the INM (Figure 3I,I′, arrows).
Moreover, we analyzed the nuclear and cell size as well as the integrity of the NE in GFP-ΔNSun-1 as well as in Sun-1-depleted cells. In comparison with AX2 cells that exhibit a diameter of 10–15 μm in 98% of cells, 30% of the GFP-ΔNSun-1 cells and 10% of the Sun-1 RNAi cells displayed an increase in nuclear and cellular diameter (Figure 3J). To examine the integrity of the NE, the 4′-6-diamidino-2-phenylindole (DAPI) staining was superimposed with the GFP-ΔNSun-1 image (Figure 3K), whereas the NE of Sun-1 RNAi cells was visualized by immunofluorescence of interaptin that localized properly to the NE (Figure 3L). The nuclear and cellular expansion upon GFP-ΔNSun-1 expression and Sun-1 RNAi was accompanied by severe nuclear deformations, such as formation of NE blebs that did not contain DNA (Figure 3K,L, arrowheads). We observed NE deformations in 95% of the huge GFP-ΔNSun-1 cells nuclei and in 65% of the huge Sun-1 RNAi cells (Figure 3J). The extent of abnormal nuclear morphology correlated with nuclear and cellular volume expansion, which suggests that GFP-ΔNSun-1 and Sun-1 RNAi cells suffer from inefficient co-ordination of nuclear and cell division. The relatively milder effect on nuclear and cellular expansion and NE deformation in Sun-1 RNAi cells compared with GFP-ΔNSun-1 cells may be because of the limited efficiency of the Sun-1 depletion that was achieved (approximately 40% of AX2 levels) (Figure 3J). However, we were not able either to improve the efficiency of Sun-1 downregulation or to generate a sun-1 knockout strain, as both events are most likely lethal for D. discoideum, indicating that Sun-1 may play an essential role.
GFP-ΔNSun-1 disconnects the centrosome from the nucleus
The nuclear and cellular enlargement suggests defects in nuclear and cell division; thus, we investigated the fidelity of chromosome segregation in GFP-ΔNSun-1 cells by evaluation of the karyotype. In AX2 cells, the genome is distributed on six chromosomes (Figure 4A). In marked contrast, GFP-ΔNSun-1 cells displayed a significant increase of aneuploid nuclei containing different numbers of chromosomes ranging from three to five (Figure 4A) confirming a loss of chromosomes during segregation. In concordance with the numerical chromosome aberrations, we observed reduced proliferation capacity and a fivefold increase in cell death upon GFP-ΔNSun-1 expression as well as Sun-1 RNAi (Figure 4B).
As faithful chromosome segregation depends on the centrosome stability, we studied the centrosome-nucleus connection in GFP-ΔNSun-1 and Sun-1 RNAi cells. In AX2 cells, each nucleus is intimately connected to one centrosome during interphase (Figure 4C). In comparison, in 90% of the GFP-ΔNSun-1 cells the centrosomes are located far away from the nuclei (Figure 4D,E). In regular-sized cells, the NE formed protrusions extending toward the centrosome, which bridged a distance up to one third of the cell diameter (Figure 4D). Notably, GFP-ΔNSun-1 decorated NE protrusions that were attached to the periphery of the centrosomes, indicating that GFP-ΔNSun-1 may participate in a putative protein complex, which forms the centrosome-nucleus connection (Figure 4D, boxes; Figure 4G). Although the centrosome-nucleus vicinity was abolished in the regular-sized GFP-ΔNSun-1 cells, the centrosome number correlated strictly with that of the nuclei as in AX2 cells (Figure 4C,D). On the contrary, NE protrusions were absent in the huge GFP-ΔNSun-1 cells. Instead, the nuclei in almost all huge cells were completely disconnected from the centrosome. Moreover, we observed a dramatic centrosome hyperamplification and cells with two nuclei had up to five centrosomes (Figure 4E). Although centrosome hyperamplification was also found in Sun-1 RNAi cells, fewer nuclei have lost the vicinity to the centrosomes (Figure 4F). Probably, the remaining amount of Sun-1 can maintain the centrosome-nucleus connection to some extent.
The quantification of centrosome and nuclear number revealed that the centrosome-nucleus number strictly correlated within AX2 cells, whereas GFP-ΔNSun-1 cells showed diverse combinations in centrosome-nucleus number. Although the nuclear number was not dramatically increased in GFP-ΔNSun-1 and Sun-1 siRNA cells, GFP-ΔNSun-1 exhibited a severe centrosome hyperamplification. Sun-1 siRNA cells were remarkable in that they mainly had a single nucleus that in the majority of cases was associated with one centrosome. Few cells had supernumerary centrosomes (Figure 4H). As the downregulation of Sun-1 did not enhance centrosome hyperamplification to the severe extent that was observed in GFP-ΔNSun-1 expressing cells, it appears likely that the remaining amount of Sun-1 in Sun-1 RNAi cells is sufficient to tether the chromatin to the centrosome and thereby to maintain the centrosome-nucleus connection and mitotic spindle stability.
We also carried out rescue experiments with Sun-1 RNAi cells by expressing a full-length GFP-tagged Sun-1 in order to prove that the observed effects were specific. Such cells had normal nuclei, and centrosome amplification was no longer observed, whereas centrosomes were often detached from the nuclei (Figure S4A). The latter phenotype appears to be caused by GFP-Sun-1 overexpression as detachment of centrosomes was also observed in AX2 cells overexpressing GFP-Sun-1. In fact, this strain exhibited the highest number of cells with detached centrosomes (Figure S4A,B). Furthermore, in sequential permeabilization experiments, GFP-Sun1 was also detected at the ONM in AX2 cells (Figure S4C).
The observed aneuploidy in GFP-ΔNSun-1 cells prompted us to analyze the distribution of DNA during mitosis. In contrast to AX2 cells in which the chromosomes are partitioned and connected to the centrosomes during mitosis (Figure 4I,I″), we found that chromosomes were detached from the centrosomes and lost along the spindle in GFP-ΔNSun-1 cells (Figure 4J,J″), indicating that Sun-1 may facilitate chromosome attachment to the centrosome during mitosis. In addition, centrosome-nucleus detachment and centrosome hyperamplification in GFP-ΔNSun-1 cells yielded the formation of monopolar and multipolar mitotic spindles that represent various forms of defective spindles, whereas some bipolar spindles performed asymmetric nuclear division (Figure S5).
Sun-1 competes with interaptin for NE localization
SUN and KASH domain proteins interact through their C-termini in the PNS to provide a physical connection of the nucleoskeleton with the cytoskeleton (16–19). In D. discoideum, interaptin exhibits the conserved domain architecture of the KASH domain proteins that includes an N-terminal actin-binding domain, a central stretch of coiled-coil repeats and a C-terminal transmembrane domain with a short tail harboring a KASH motif (11). In AX2 cells, both interaptin and Sun-1 were localized in the NE (Figure 5A,A′ and B,B′). Surprisingly, while in an interaptin-deficient strain Sun-1 was targeted to the NE (Figure 5C,C′ and D,D′), Sun-1 was displaced from the NE when interaptin was overexpressed (Figure 5E,E′ and F,F′), which implies a competitive localization of Sun-1 and interaptin in the NE of D. discoideum, that is in contrast to the vertebrate model (3).
Comparison of the interaptin C-terminal tail with those from C. elegans Anc-1, Drosophila melanogaster Msp-300/Nesprin and mammalian Nesprin-1 and -2 pointed to the terminal amino acids PT as the relevant KASH motif for NE targeting. Previously, we have shown that the truncated interaptin GFP-IntCT, which encompasses the C-terminal transmembrane domain and the KASH motif, displaces the endogenous interaptin from the NE (11) (Figure 5G,G′). Now, we show that the terminal amino acids PT are essential for NE targeting of interaptin, as proteins carrying a mutation of proline to alanine (GFP-IntP/A, Figure 5I,J′) or a truncation of PT (GFP-IntΔPT, Figure 5K,L′) failed to localize to the NE and did not compete with endogenous interaptin for NE localization (Figure 5I,K). As interaptin was absent from the NE protrusions and interaptin mutants did not suffer from centrosome-nucleus disconnection or centrosome hyperamplification (data not shown), it is apparently not involved in the centrosome-nucleus complex but may have other functions particularly during development.
NE localization of Sun-1 was abolished upon overexpression of GFP-IntCT (Figure 5H,H′) but was not affected by GFP-IntCT-P/A or GFP-IntCT-ΔPT (Figure 5J,J′ and L,L′), which demonstrates that the interaptin C-terminus exerts a dominant-negative effect on the NE localization of Sun-1 and that the KASH motif is involved in a competition with Sun-1. Conversely, overexpression of GFP-ΔNSun-1 reduced the NE localization of interaptin (Figure 5M,M′), indicating that Sun-1 and interaptin may compete for a common binding partner in the NE. Similarly, C. elegans UNC-84 may interact indirectly with the KASH domain protein Anc-1 (24), implying that their interaction may be mediated by an additional partner. In D. discoideum, Sun-1 may interact with a centrosome-attached KASH domain protein to establish centrosome-nucleus connection, whereas nuclear positioning may require KASH domain proteins other than interaptin.
Furthermore, we observed in live cell imaging analysis that nuclei of GFP-ΔNSun-1 cells experienced severe deformations but were able to move, although they were disconnected from the centrosomes. In comparison, wild-type nuclei in which the NE was labeled by GFP-IntCT maintained a nearly perfectly round shape (Figure 6).
Sun-1 is a chromatin-binding protein of the INM
In this study, we have shown that Sun-1 is an INM protein in D. discoideum, which extends its N-terminus into the nucleoplasm and the C-terminus into the PNS. The localization and the topology of Sun-1 are conserved with that of SUN domain proteins in other species (3,17). We have provided evidence that the central coiled-coil domains promote dimerization and oligomerization of Sun-1 in vitro. Based on this, it is possible that Sun-1 forms dimeric and/or oligomeric complexes in the INM that serve as a platform for ONM proteins.
In general, SUN domain proteins do not contain nuclear localization signal sequences; therefore, their specific targeting mechanism to the INM was unexplained. Some SUN domain proteins from higher eukaryotes, such as C. elegans matefin/Sun-1 and mammalian Sun-1 and Sun-2 interact with lamins but can localize to the INM after depletion of lamins, indicating that targeting of the SUN domain proteins may involve other nuclear components (18,20). Here, we provide evidence that Sun-1 is targeted and retained in the INM through binding to chromatin through its N-terminus in D. discoideum that is a lamin-free model system. Notably, the N-terminus of Sun-1 is dispensable for NE localization per se but determines its targeting to the INM as GFP-ΔNSun-1 mostly failed to localize to the INM and accumulated in the ONM, demonstrating that chromatin binding of the Sun-1 N-terminus represents an INM retention mechanism. Mislocalization of GFP-ΔNSun-1 causes separation of the INM and ONM, indicating that Sun-1 regulates the spacing of the NE through interaction with an ONM protein that may be bypassed by GFP-ΔNSun-1 in the ONM. Dilation of the PNS, nuclear deformation and the increased nuclear volume both in GFP-ΔNSun-1 and in Sun-1 RNAi cells suggest that Sun-1 may define the spacing of the NE lumen and the integrity of NE morphology.
Sun-1 establishes centrosome-nucleus connection
We observed loss of centrosome-nucleus juxtaposition because of NE protrusions toward the centrosome in regular-sized cells, whereas the nuclei in huge GFP-ΔNSun-1 cells were completely disconnected from the centrosomes and dramatic centrosome hyperamplification occurred that caused chromosome missegregation, aneuploidy and a fivefold increase in cell death. Our findings agreed well with the neoplastic characteristics of malignant tumors in which centrosome hyperamplification was always associated with aneuploidy (25–29).
To date, the mechanisms for centrosome amplification are not completely understood and so far most efforts were focused on regulation of centrosome duplication. Interestingly, a centrosome-nucleus cross talk seems to be mediated by shuttling centrosomal and nuclear proteins between these organelles. In particular, during mitosis, the nuclear proteins p53, Orc2, Orc6 and RAD51 are required at the centrosome to control its duplication only once during the cell cycle (30–33). Conversely, the Ran/importin-dependent shuttle of the centrosomal proteins centrin-1 and pericentrin/kendrin to the nucleus is proposed to control MT dynamics and nucleation at the centrosome (34), whereas centrin-2 functions in a complex with Xeroderma pigmentosum group C protein during nucleotide excision repair (35). According to this shuttling model, the centrosome-nucleus juxtaposition may facilitate the rapid signaling of regulatory proteins between these organelles to co-ordinate their accurate duplication and division. Our data suggest that the physical attachment of centrosome and chromatin co-ordinate their duplication in a concerted fashion and that Sun-1 participates in a complex with further centrosome-specific proteins of the ONM to establish centrosome-nucleus juxtaposition in D. discoideum. Thus, loss of centrosome-nucleus vicinity (coupling) disables the cross talk of these organelles, thereby causing centrosome and chromosome instability.
Recent studies revealed that in S. cerevisiae, the SUN domain protein Mps3 is required as a positive regulator for the duplication of the spindle pole body (14). In contrast, C. elegans matefin/Sun-1 has been reported to act as a suppressor for the centrosome duplication promoting kinase Zyg-1 (36). However, our findings agree rather with the negative role of C. elegans matefin/Sun-1 as overexpression of GFP-ΔNSun-1 or Sun-1 RNAi induces centrosome hyperamplification. Furthermore, D. discoideum Sun-1 may also involve cell cycle regulators to couple the centrosome and nuclear duplication.
In addition to their mechanical role in nuclear positioning, SUN domain proteins are also involved in more diverse cellular processes, such as organization of the telomeres, influence on the fat metabolism and germ line maturation (7,20,37,38). In meiotic cells, however, S. cerevisiae Mps3, S. pombe Sad1, C. elegans matefin/Sun-1 as well as mouse Sun-1 may facilitate chromosome pairing (8,10). This aspect cannot be investigated in D. discoideum as the laboratory strain AX2 occurs as a haploid organism and under laboratory conditions does not undergo meiosis.
The fivefold increased cell death in both GFP-ΔNSun-1 and Sun-1 RNAi cells as well as the unavailability of a sun-1 knockout strain demonstrate that Sun-1 may play an essential role for cell viability. The mild extent of centrosome hyperamplification in Sun-1 RNAi cells may be because of the residual amount of Sun-1 that may be sufficient to propagate the centrosome-nucleus connection and proper spindle formation (data not shown) but failed to maintain the nuclear and cellular morphology. Our data demonstrate that Sun-1 may have two functions: during interphase, it defines the integrity of the NE and establishes centrosome-nucleus connection by interaction with chromatin, whereas during mitosis, Sun-1 ensures the accuracy of chromosome segregation and genome stability by tethering chromosomes to the centrosome. Similarly, depletion of C. elegans matefin/Sun-1 was observed with abnormally condensed chromatin and enhanced apoptosis (20,39). In contrast to the vertebrate model, the KASH domain protein interaptin does not bind to Sun-1 to position the nucleus on F-actin or the centrosome, although it is likely that interaptin may have other functions in the NE during the development of D. discoideum. Yet, the identification of the centrosome linker of that complex is a future challenge that will provide more insight into the function of the SUN domain proteins.
Taken together, we propose that Sun-1 forms dimers or higher oligomers in vivo that are retained in the INM by binding to chromatin with their N-termini. The Sun-1 C-terminus interacts with an ONM protein and a centrosome linker in the PNS to define the spacing of the NE lumen and to ensure the centrosome-nucleus vicinity. Alternatively, Sun-1 may bind indirectly to interaptin or other KASH domain proteins to position the nucleus on the actin cytoskeleton (Figure 7). Truncation of the Sun-1 N-terminus abrogates chromatin binding as well as the INM localization of GFP-ΔNSun-1. Low amounts of GFP-ΔNSun-1 can be escorted in a complex with the endogenous Sun-1 to the INM, whereas the majority of GFP-ΔNSun-1 accumulate in the ONM and counteract the connection of the centrosome linker with Sun-1, resulting in the formation of NE protrusions (Figure 7). These NE protrusions were likely formed because of the mechanical force emerging between the cytoskeleton and the nucleus leading to disconnection of the centrosomes from the nucleus. Subsequently, increase in centrosome-nucleus distance leads to loss of the centrosome connection promoting the centrosome hyperamplification and genome instability such as aneuploidy. Based on this, we propose that SUN domain proteins will also play an important role during carcinogenesis in mammalian cells.
Material and Methods
Generation of Sun-1 constructs and D. discoideum strains used
For GST-SunCT1, a PCR fragment extending from cDNA position 1032–1959 (encompassing the central two coiled-coil domains) was generated as a 5′-XmaI and a 3′-XhoI fragment that was cloned into the expression vector pGEX4T2. The GST-SunCT1 polypeptide was expressed in the Escherichia coli strain XL1-Blue. GFP-ΔNSun-1 containing the nucleotide region 853–2718 as a ClaI fragment encoding the aa 283–905 was generated using the vector pDEX-79 (40). Full-length Sun-1 cDNA was also cloned into pDEX-79 to generate a GFP fusion for rescue experiments. To generate a NE marker, a full-length cDNA encoding D. discoideum nucleoporin Nup43 was cloned into pDEX-79. GFP-IntΔPT was generated by excluding the final six base pairs during PCR amplification using GFP-IntCT as template that contains the C-terminal transmembrane domain and part of the tail sequence of interaptin (11). GFP-IntP/A was generated by site-directed mutagenesis using GFP-IntCT. Selection of D. discoideum strain AX2 transformants was with G418 (4 μg/mL).
The Sun1 RNAi construct contains the sequence 5′-AAGAGCTTAAACTAGTTAAACTT-3′ that targets the Sun-1 cDNA at the position 1187–1209 bp. The target sequence is flanked by complementary sequences that form a hairpin. Selection of transformants was by blasticidin (3.5 μg/mL).
AX2 was used as host strain for all transformations. Cells expressing GFP-tagged calreticulin and calnexin as ER markers were obtained from Dr G. Gerisch (41). The interaptin-deficient strain abpD− was described previously (11).
Preparation of total cell lysates and intact nuclei
Cells were washed twice with Soerensen phosphate buffer (2 mm Na2HPO4, 14.6 mm KH2PO4, pH 6.0) before resuspending in TMS buffer [50 mm Tris/HCl, pH 7.4, 100 mm NaCl, 5 mm MgCl2, 250 mm sucrose, 1 mm ethylenediaminetetraacetic acid (EDTA), 1 mm EGTA, 1 mm DTT, 1 mm benzamidine, 1 mm phenylmethylsulphonyl fluoride (PMSF)]. Total cell lysates were obtained by passage through Nuclepore membrane (5 μm diameter, Whatman) and used in further experiments. Intact nuclei were collected after lysis through Nuclepore membrane by spinning for 5 min at 4000 g.
Proteinase K protection assay
Intact nuclei were isolated from AX2 cells and incubated in proteinase K protection assay buffer (10 mm Tris/HCl, pH 7.4, 250 mm sucrose, 1 μg/mL proteinase K) with or without addition of 0.5% Triton-X-100 (v/v) on ice. Samples were collected from both reactions after 5, 10, 30, 45 min, and the proteinase K was immediately inactivated by adding PMSF to a final concentration of 1 mm and heating (95°C, 5 min) in sodium dodecyl sulfate (SDS) sample buffer. The samples were further analyzed by SDS–PAGE and Western blot.
About 5 × 107 cells per ChIP reaction were harvested and lysed in TMS buffer (see above). To reduce the viscosity of genomic DNA and disrupt the NE, total cell lysates were sonified twice with 10 pulses. The appropriate antibodies were coupled to protein A–Sepharose beads and then incubated with sonified total cell lysate (2 h on a vertical rotator, 4°C). Unspecific protein and DNA binding were removed by five times washing with PBS and once with TE buffer (10 mm Tris/HCl, pH 8.0, 1 mm EDTA) and divided into two aliquots. One aliquot of each ChIP reaction was used for elution of DNA by adding 100 μL of TE buffer containing 1% SDS. After phenol/chloroform extraction and ethanol precipitation, DNA was resuspended in 50 μL water; 5 μL of the DNA was subsequently used for PCR analysis using actin-8 gene-specific primers. The second aliquot was subjected to SDS–PAGE and Western blot analysis.
Electromobility shift assay and Southwestern blotting
GST-Sun1N400 (aa 1–135) was cloned using EcoRI/NsiI fragment and GST-Sun1N800 (aa 1–255) using EcoRI/NcoI fragment into pGEX-4T3 (Amersham Pharmacia). Protein purification from E. coli XL1-Blue using GST-Sepharose beads and elution from the beads was carried out following user’s manual (Amersham Pharmacia). About 20 or 40 ng of each protein and GST for control were mixed with 2000 cpm radiolabeled DNA probe (720 bp D. discoideum hp1-fragment produced by EcoRI digestion of pGEM-Teasy-Ddhp1) in the EMSA buffer containing 50 mm Tris/HCl, pH 8.0, 0.5 mm EDTA, 50 mm NaCl, 0.1% Triton-X-100. The protein–DNA binding was carried out for 3 h at room temperature. The samples were separated on 5% polyacrylamide gels containing 12.5 mm Tris/Cl, pH 8.4, 95 mm glycine and 0.5 mm EDTA. DNA was visualized by autoradiography of dried gels.
Southwestern analysis of GST-Sun1N400 and GST-Sun1N800 were performed as described (42). Radiolabeled DNA probes (D. discoideum hp1 and actin-8 or human CAP2 gene sequences) were incubated overnight with the protein membrane at room temperature. After three times washing with the renaturation buffer, the Southwestern blot was exposed to an X-ray film.
Metaphase arrest of cell division (karyotyping)
Cells were allowed to attach to coverslips for 2 h before adding nocodazole to a final concentration of 33 μm to the medium. The incubation with nocodazole was maintained for 4 h to increase the number of cells arrested in metaphase. The solution was replaced by cold water for 10 min at 4°C. Chromosomes were then fixed on the coverslips in ethanol/acetic acid 3/1 (v/v) for 1 h on ice, followed by a change of fresh fixative for an additional 10 min on ice, before proceeding with DAPI staining.
In vitro cross-linking and gel filtration chromatography
To address the multimerisation behavior, GST was removed from SunCT1 by thrombin cleavage. Ten micrograms of SunCT1 was incubated at room temperature in the phosphate potassium buffer (pH 7.4) containing 0.001% (v/v) glutaraldehyde. The reaction was stopped by addition of glycine to a final concentration of 0.1 m. After 5, 10 and 20 min, samples were analyzed by SDS–PAGE (10% acrylamide) and Western blotting. Native and chemically cross-linked SunCT1 were also analyzed by gel filtration chromatography using a Sephadex G-75 column. About 100-μl fractions were collected and analyzed by SDS–PAGE followed by Western blotting (43). Analytical gel filtration employed the SMART system (GE Healthcare). For calibration, BSA (66 kDa), ovalbumin (43 kDa) and chymotrypsin (25 kDa) were used.
Indirect immunofluorescence microscopy
Approximately 1 × 106 cells were transferred onto a coverslip (10 mm diameter) and allowed to adhere for 20 min at room temperature. If not mentioned otherwise, standard immunofluorescence stainings were carried out using ice-cold methanol as fixative (5 min, −20°C). Cells were treated twice for 15 min (room temperature) with blocking solution (1 × PBS containing 0.5% (w/v) BSA and 0.1% (v/v) fish gelatin). The appropriate antibodies were diluted in the blocking solution and applied for 1 h at room temperature; the excess of antibodies was removed by washing with the blocking solution prior to the 1 h of incubation with the according secondary antibodies.
For sequential digitonin/Triton-X-100 permeabilization experiments, cells adhered to the coverslips were fixed by application of 4% (w/v) paraformaldehyde for 20 min. After three times washing with PBS, cells were first permeabilized by short incubation with prechilled digitonin (10 μg/mL in PBS) (5 min, on ice) and washed five times with PBS. Blocking reaction and incubation with the primary antibodies was performed as described above. After removal of excess of the primary antibodies (six washes with PBS), cells were permeabilized by application of 0.2% Triton-X-100 in PBS (10 min, room temperature). The permeabilized sample was blocked before probing with the second epitope-specific antibodies. The two specific primary antibodies were finally visualized by simultaneous incubation with the according secondary antibodies. Confocal images were acquired with an inverted Leica TCS-SP laser scanning microscope using ×40 and ×100 Neofluar oil immersion objectives.
Primary antibodies used in this study were: mouse monoclonal anti-GFP antibody (44), rabbit polyclonal anti-GFP antibody (gift from M. Schleicher), rabbit polyclonal anti-GST antibody (unpublished), mouse monoclonal anti-interaptin antibody (260-60-10) (11), mouse monoclonal anti-PDI-1 antibody (221-135-1) (45), mouse monoclonal anti-Sun-1 antibody (K55-432-2, Western blot analysis, this study), mouse monoclonal anti-Sun-1 antibody (K55-460-1, immunofluorescence, this study), mouse monoclonal antitubulin antibody (K29-359-31) that was generated against a tubulin fraction from the green alga Spermatozopsis similis and which recognizes the D. discoideum centrosome (gift from Dr K. Herkner and Dr M. Melkonian), mouse monoclonal anti-α-tubulin antibody (46), rat monoclonal anti-α-tubulin antibody (YL1/2) (47). The appropriate secondary antibodies were: Cy3-conjugated goat anti-mouse immunoglobulin G (IgG) (Sigma); Cy5-conjugated goat anti-mouse IgG (Sigma), Alexa 568-conjugated goat anti-mouse IgG (Molecular Probes), Alexa 568-conjugated goat anti-rat IgG (Molecular Probes). DNA was stained with DAPI or ToPro-3 (Invitrogen).
Nuclei of GFP-ΔNSun-1 cells were isolated using Nuclepore membranes (5 μm; Corning). The cells were washed twice in cold PBS and then resuspended in Dicty-PHEM (48) containing a protease inhibitor cocktail and PMSF. They were pressed through the filter assembly three to five times. The resulting mixture was layered onto a 30% sucrose cushion and centrifuged at 2200 ×g for 10 min at 4°C. The pellet containing mainly nuclei was resuspended in Dicty-PHEM and centrifuged at 2200 ×g for 5 min onto 12-mm coverslips.
The nuclei were fixed with 2% formaldehyde in Dicty-PHEM for 15 min. They were washed and treated with rabbit anti-GFP antibody (P. A. Silver, Harvard) for 45 min and with 5 nm PAG-Gold (Department of Cell Biology, Utrecht University) overnight. Alternatively, mouse monoclonal anti-GFP antibody (J. Wehland, Braunschweig) was detected using 6-nm goat anti-mouse antibody (Aurion, Biotrend). Dehydration and embedment were carried out according to standard procedures.
The nuclei were sectioned using a Reichert Ultracut E and viewed in a JEOL 1200C equipped with a TEM camera Keenview-10/12 and iTEM imaging system (Soft Imaging System).
We thank Rolf Müller for help throughout the course of this work. We thank Drs M. Melkonian and K. Herkner for providing a centrosome-specific antibody, V. Peche for CAP2 cDNA, Dr G. Gerisch for cells expressing GFP-tagged calreticulin and calnexin and Sabrina Rosenbaum for help with CD spectroscopy. H. X. was a member of the International Graduate School in Genetics and Functional Genomics of the University of Cologne.