Sun2 Is a Novel Mammalian Inner Nuclear Membrane Protein*

Sun protein (Sun1 and Sun2) cDNAs were previously cloned based on the homology of their C-terminal regions (SUN (Sad1 and UNC) domain) with the Caenorh- abditis elegans protein UNC-84 whose mutation disrupts nuclear migration/positioning. In this study, we raised an anti-Sun2 serum and identified Sun2 in mammalian cells. In HeLa cells, Sun2 displays a nuclear rim-like pattern typical for a nuclear envelope protein. The Sun2 antibody signal co-localizes with nuclear pore and INM markers signals. The rim-like pattern was also observed with the recombinant full-length Sun2 protein fused to either EGFP or V5 epitopes. In addition, we found that a recombinant truncated form of Sun2, extending from amino acids 26 to 339, is sufficient to specify the nuclear envelope localization. Biochemical analyses show that Sun2 is an 85-kDa protein that is partially insoluble in detergent with high salt concentration and in chaotropic agents. Furthermore, Sun2 is enriched in purified HeLa cell nuclei. Electron microscopy analysis shows that Sun2 localizes in the nuclear envelope with a sub-population present in small clusters. Additionally, we show that the SUN domain of Sun2 is localized to the periplasmic space between the inner and the outer nuclear membranes. From our data, we conclude that Sun2 is a new mammalian inner

In eukaryotic cells, the nuclear envelope, which separates the nucleoplasm from the cytoplasm, is composed of the inner and outer nuclear membranes (INM 1 and ONM, respectively), the latter membrane being continuous with the endoplasmic reticulum. The two membranes are separated by a thin lumen and are joined at nuclear pores (1)(2)(3). Underlying the INM is a meshwork of various lamin isoforms (4) that are in close contact with the INM and its resident proteins (5). Characterized INM proteins in mammalian cells include the lamin B receptor (6), lamin-associated polypeptides 1 and 2 (Lap1 and Lap2) (7), emerin (8), and Man1 (9). These proteins possess a hydrophilic N-terminal region that protrudes into the nucleoplasm as well as one or more hydrophobic regions leading to predicted single or multispanning transmembrane domains. INM proteins are immobilized in the nuclear envelope through their interaction with lamins and/or heterochromatin (10,11).
Beside their structural role, lamina components also exert additional functions through their ability to interact with effector proteins involved in various regulatory processes. For example, the lamin B receptor directly binds to HP1 (12), a heterochromatin protein involved in transcription repression. Lap2 (13,14), emerin (15), and Man1 (16) interact with the barrier-to-autointegration factor, a 10-kDa DNA-binding protein (17). The range of regulatory functions performed by INM proteins is widened further by the recent discovery that Lap2␤ and emerin interact directly with a transcription repressor, Germ Cell-less (18,19), and that emerin binds in vitro to a splicing-associated factor, YT521-B (20). The discovery that mutations in lamin A/C and emerin are implicated in Emery-Dreyfuss muscular dystrophy (21)(22)(23), and a growing spectrum of human diseases has sparked a resurgence of interest in the lamina/nuclear envelope interface. Collectively, these diseases are now termed "laminopathies" (24 -26). The Hutchinson-Guilford progeria syndrome, associated with a de novo mutation of lamin A (27,28), is the latest laminopathy described to date.
Additional proteins have been reported to associate with the nuclear envelope. These include nurim, which is identified in a visual screen of a GFP fusion library (29), and LUMA and the murine homolog of Sun1 (KIAA0810) discovered in a proteomic screen of the nuclear envelope (30). A recent proteomic study (31) reveals yet another 67 potentially new INM proteins, among them Sun2.
UNC-84 is another nuclear envelope protein that is involved in nuclear anchoring and migration during Caenorhabditis elegans development. In Schizosaccharomyces pombe, the Sad1 protein localizes at the spindle pole body and also at the nuclear envelope when overexpressed (32). Interestingly, UNC-84 and Sad1 share a common C-terminal motif of ϳ200 amino called the SUN (Sad1 and UNC) domain (33). In mammals, the SUN domain is present in two proteins whose cDNAs where isolated and termed Sun1 and Sun2 (34). In an initial effort to understand better the biological function of the highly conserved SUN domain, we raised an anti-Sun2 serum and analyzed the biochemical and localization properties as well as the topology of the endogenous Sun2 protein in mammalian cells. * This work was supported in part by National Institutes of Health Grants GM42259, AI35884, and AI20015 (to P. D. S.), the "Fondation Leon Fredericq," and the Muscular Dystrophy Association (to D. M. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ¶

EXPERIMENTAL PROCEDURES
Sun2 Serum-Rabbit were immunized with a recombinant protein spanning amino acids 262-492 of the KIAA0668 Sun2 cDNA (35) fused to GST (Fig. 1). The cDNA encoding that fragment was obtained from a PCR reaction using KIAA0668 cDNA as a template with an EcoRI forward primer and a NotI reverse primer. The PCR product was cloned in-frame with the GST moiety of pGEX-4T1 and the resulting construct transformed in DH5␣ bacterial cells. Bacterial cultures were grown until mid-log phase and passed twice on a French press, and the recombinant protein GST/Sun-(262-492) was purified on a glutathione-Sepharose column (Pierce). One milligram of fusion protein was used per rabbit immunization (Biosolutions). The anti-Sun2 serum IgG fraction (referred as the Sun2 serum) was purified on a protein A-Sepharose column according to the recommendations of the manufacturer (Sigma).
Cell Extraction and Immunoblotting-HeLa cells, routinely maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and grown at 80% confluence, were rinsed twice in PBS, lifted with PBS-0.02% EDTA, washed twice in PBS, and swollen on ice for 10 min in a hypotonic buffer (10 mM MgCl 2 , 1 mM KCl, 5 mM HEPES, pH 7.4) containing protease inhibitors. Cells were disrupted by 30 strokes in a Dounce homogenizer and centrifuged 5 min at 15,000 ϫ g in a tabletop centrifuge. The pellets were resuspended in PBS-1% Triton X-100, PBS-8 M urea, or PBS-1% Triton X-100/8 M urea and stirred at 4°C for 15 min. Each extraction was centrifuged at 15,000 ϫ g for 5 min. Pellet and supernatant fractions were then analyzed by immunoblotting. Extraction of whole HeLa cells with increasing salt concentrations was carried out under the same conditions. Nitrocellulose membranes were blocked with 10 mM Tris, pH 7.5, 150 mM NaCl, and 0.1% Tween 20 (TBST) containing 5% nonfat dry milk for 1 h at room temperature and incubated with the appropriate primary antibody overnight at 4°C. The secondary antibody was added for 1 h at room temperature, and the blots were developed by ECL following the recommendations of the manufacturer (PicoSignal, Pierce). Antibodies used in immunoblotting experiments included the IgG fraction of the Sun2 serum (1:1000), an anti-lamin B1 antibody (1:500, Santa Cruz Biotechnology), an anti-emerin antibody (1:200, Santa Cruz Biotechnology), and an anti-EGFP antibody (1:5000, Abcam). For peptide competition, a master mixture of a 1:1000 dilution of the Sun2 serum was divided in 200-l aliquots containing increasing amounts of the competitor protein. Each aliquot was added to total HeLa cell lysates blotted on a membrane clamped in a Miniprotean II multiscreen device (Bio-Rad).
Nuclei Purification-Nuclei purification was carried out as described previously (30). HeLa cells were disrupted in a Dounce homogenizer after resuspension in a 0.25 M sucrose buffer containing 0.05% Nonidet P-40 and a protease inhibitor mixture. The homogenate was mixed with a 2.1 M sucrose buffer to 1.5 M of final sucrose concentration. 0.25 M of sucrose buffer, the homogenate, and 2.1 M of sucrose solutions were successively underlaid and centrifuged at 100,000 ϫ g for 90 min in a Ti-50.2 rotor (Beckman). The pelleted nuclei were resuspended in PBS, flash-frozen in liquid nitrogen, and stored at Ϫ80°C.
Transfection and Immunofluorescence Microscopy-HeLa cells were grown overnight on coverslips in 24-well plates. Transfection was carried out with FuGENE 6 following the recommendations of the manufacturer (Roche Applied Science). Routinely, 1 g of plasmid DNA and 3 l of FuGENE 6 were added to 100 l of Opti-MEM (Invitrogen). The mixture was incubated at room temperature for 30 min, and 20 l/well was added to the plate. After a 20-h incubation, cells were washed and fixed in PBS containing 2.5% paraformaldehyde for 30 min at room temperature. After three washes with PBS, permeabilization was carried out for 10 min at room temperature in PBS containing 10% goat serum and 0.2% saponin (Sigma). Detection of emerin in immunofluorescence microscopy was carried out on HeLa cells and fixed for 5 min in methanol at Ϫ20°C followed by a rinse in cold acetone. Digitonin permeabilization experiments (0.004%, Calbiochem) were carried out on paraformaldehyde-fixed cells for 15 min on ice. Primary antibodies were added for 1 h at room temperature. After several washes with PBS, the secondary antibody coupled to Alexa 564 or Alexa 488 (Molecular Probes) was added for 1 h at room temperature. Cells were washed, and the coverslips were mounted on glass slides with Mowiol (Calbiochem). For direct fluorescence microscopy with the EGFP constructs, cells were fixed, washed, and directly mounted on glass slides. Antibodies used in fluorescence microscopy included a monoclonal anti-Nup 153 (1:2, clone XB-10), a monoclonal anti-V5 conjugated to fluo-rescein isothiocyanate (1:2000, Invitrogen), an anti-␣-tubulin (1:1000, clone DM 1A, Sigma), an anti-giantin polyclonal antibody (1:750, Covance), an anti-emerin antibody (1:20, Novo Castra), and an anti-protein-disulfide isomerase monoclonal antibody (1:250, Stressgen). The Sun2 serum was used at a 1:1000 dilution in all of the immunofluorescence experiments. For peptide competition in immunofluorescence microscopy, GST/Sun-(262-492) and GST proteins were used at final concentrations of 7 and 100 g/ml, respectively. All of the slides were examined with an Axiovert 100 M (Zeiss) fluorescence microscope.
Immunolabeling of Frozen Thin Sections-HeLa cells were lifted in PBS-10 mM EDTA, pelleted, and immediately fixed with 2% paraformaldehyde and 0.2% glutaraldehyde in PBS, pH 7.2, at room temperature for 2 h. After two rinses in PBS, cells were embedded in 10% gelatin and processed for ultramicrotomy as described elsewhere (36). Ultrathin sections were incubated with blocking buffer containing 10% goat serum. Immunolabeling was carried out with the Sun2 serum or the preimmunization serum at different dilutions for 2 h followed by 12-nm gold-labeled goat anti-rabbit IgG (Jackson Immunoresearch Laboratories) for 1 h. After washing, the sections were stained with uranyl acetate and embedded in methylcellulose. The sections were viewed on a Zeiss-902 electron microscope.
Proteinase K Protection Assays-Following the recommendations of the manufacturer, 1 g of the pcDNA 3.1 Topo/V5-His plasmid containing the Sun2 cDNA ( Fig. 1) was included in a 25 l of TNT coupled reticulocyte lysate reaction (Promega) containing [ 35 S]methionine (Amersham Biosciences) either without or with 2 l of canine pancreatic microsomal membranes (Promega). Aliquots of the synthesis reactions were further incubated on ice for 30 min with a final concentration of 100 g/ml proteinase K (Promega) either alone or with a final concentration of 1% Triton X-100. Proteolytic digestions were terminated by the addition of PMSF and boiling Laemmli buffer. Samples were transferred in duplicate to a nitrocellulose membrane and visualized on a PhosphorImager (Amersham Biosciences). An immunoblot was then performed with the anti-Sun2 serum and the anti-V5 antibody.

RESULTS
An analysis of the Sun2 amino acid sequence ( Fig. 1, KIAA0668) using SMART (37) predicts a signal sequence (amino acids 1-18) as well as two coiled-coil domains (amino acids 429 -460 and 500 -527). The program TMHMM2 (38) predicts one (amino acids 238 -260) or two transmembrane domains (amino acids 204 -226 and 238 -260) depending on the omission or the presence, respectively, of the putative signal sequence in the input file. With the exception of the conserved SUN domain (amino acids 545-742), Sun2 does not show any further homology with other mammalian proteins.
The N-terminal Region of Sun2 Confers Its Nuclear Envelope Localization-We initially examined the localization of Sun2 recombinant constructs in direct and indirect immunofluorescence microscopy. HeLa cells, transfected with the recombinant human Sun2 cDNA (KIAA0668) fused to either EGFP or a V5 epitope, displayed a nuclear rim-like pattern ( Fig. 2A). Accordingly, recombinant SUN domain-containing proteins from different organisms (Sad1 from S. pombe, UNC-84 from C. elegans, and human Sun1) have been shown to localize in the nuclear envelope (30,32,34). To delineate the region of Sun2 that is essential for its nuclear envelope localization, Sun-(26 -339) and Sun-(262-742) covering either the N-or C-terminal region of the protein were fused to EGFP (Fig. 1) and transfected in HeLa cells. Sun-(26 -339) by itself displayed a rim-like pattern that co-localized with the nuclear pore marker Nup153 (Fig. 2B). On the contrary, Sun-(262-742), corresponding to the C-terminal region of Sun2 downstream of the transmembrane region ( Fig. 1), accumulated in the Golgi as indicated by its co-localization with giantin, a Golgi marker (Fig. 2C). This result indicates that amino acids 26 -339, which include the putative transmembrane span(s), are sufficient to specify the nuclear envelope localization of the whole protein. This result excludes any involvement of the SUN domain in the protein localization. Furthermore, the absence of the putative signal sequence (amino acids 1-18) in the Sun-(26 -339) construct indicates that this stretch of hydrophobic amino acids is not Sun2 Is an INM Protein required for the proper nuclear envelope localization of Sun2.
The Endogenous Sun2 Protein Displays a Nuclear Envelope Rim-like Pattern-A GST protein fused to a fragment of Sun2 corresponding to amino acids 262-492 (Fig. 1, GST/Sun 262-492) was used to raise an anti-Sun2 serum. We chose that region both for its lack of homology with other proteins and for the good purification yield from bacteria. The crude serum was purified on a protein A-Sepharose column to obtain the IgG fraction, hereafter called the Sun2 serum.
In immunofluorescence microscopy, the Sun2 serum detected a rim-like pattern around the nuclei as well as one or several bright spots in the cytoplasm (Fig. 3A). To correlate the rimlike pattern with the anti-Sun2 immunoreactivity, immunofluorescence was carried in the presence of either GST or GST/Sun-(262-492) proteins. The addition of GST/Sun-(262-492) effi-ciently prevented the detection of the rim-like pattern (Fig. 3A, Sun2ϩGST/Sun 262-492), whereas the same experiment carried out with the GST protein alone did not (Fig. 3A,  Sun2ϩGST). Therefore, these results showed that the rim-like pattern originates from the anti-Sun2 immunoreactivity.
HeLa cells were co-labeled with the Sun2 serum and an anti-Nup 153 antibody, a nuclear pore marker, and emerin, an INM marker. The results showed that the Sun2 rim-like pattern co-localized with both markers in indirect immunofluorescence (Fig. 3, B and C). Interestingly, naturally occurring mitotic HeLa cells displayed a Sun2 rim that reformed around the chromatin of early telophase cells and co-localized with emerin (Fig. 3C). In electron microscopy, the nuclear envelope localization of Sun2 was confirmed by the detection of gold particles above the nuclear lamina. Interestingly, some area of the nu- clear envelope displayed clusters of gold particles (Fig. 3D). Overall, these results indicate that the rim-like pattern corresponds to the localization of the endogenous Sun2 protein in vivo and further confirm the efficiency of the Sun2 serum raised for this study.
Detection of the Endogenous Sun2 Protein-Three main proteins migrating at 65, 75, and 85 kDa (Fig. 4A, lanes Ϫ) were detected with the anti-Sun2 serum, whereas the preimmunization serum did not show any comparable signals (Fig. 4A, lane  P). The addition of the GST protein alone to the anti-Sun2 serum did not modify the immunoblot pattern, whereas the competition with increasing amounts of the GST/Sun-(262-492) fusion protein efficiently prevented the detection of the 85and 65-kDa proteins (Fig 4A, arrows). The detection of the 75-kDa protein was also prevented with higher amounts of the GST/Sun-(262-492) protein. Immunoprecipitation studies were carried out with the anti-Sun2 serum. HeLa cells were extracted with PBS-1% Triton X-100 and centrifuged at 15,000 ϫ g. The supernatant (SN1) was collected, and the pellet was resuspended in PBS containing 1% Triton X-100 and 8 M urea, centrifuged at 15,000 ϫ g, and dialyzed against PBS to obtain SN2. Immunoprecipitation carried out on SN1 and SN2 mostly pulled down the 85-kDa protein (Fig. 4B). The 85-kDa protein was also specifically enriched in HeLa cell nuclei purified on sucrose cushion in comparison with the 75-and 65-kDa immunoreactive proteins (Fig. 4C). The possibility that the 75and 65-kDa immunoreactive bands arose from alternative splicing of the Sun2 mRNA was examined. The Sun2 mRNA of HeLa cells was amplified with different pairs of primers, but the resulting fragments displayed a size and a restriction enzyme profile that was predicted by the Sun2 cDNA (not shown).
Biochemical Properties of Sun2-The biochemical properties of these proteins were examined. HeLa cells, swollen in a hypotonic buffer, were disrupted by 30 strokes in a Dounce homogenizer and centrifuged at 15,000 ϫ g. Under these conditions (Fig. 5A, Total), the 65-kDa protein whose detection was specifically prevented by the GST/Sun-(262-492) peptide (Fig.  4A) was completely recovered from the supernatant of this HeLa cells were labeled with the Sun2 serum and an anti-emerin antibody and counterstained with DAPI. Goat anti-rabbit IgG coupled to Alexa 488 and goat anti-mouse IgG coupled to Alexa 594 antibodies were used for immunodetection. D, frozen sections of HeLa cells were processed for immunogold electron microscopy with the anti-Sun2 serum as described under "Experimental Procedures." A goat anti-rabbit secondary antibody coupled to 12-nm gold particles was used. n, nucleus; c, cytoplasm. Bar ϭ 100 nm.

Sun2 Is an INM Protein
centrifugation. Indeed, gels allowing a better resolution of the 65-kDa region (not shown) indicated that only the upper band of the immunoreactive doublet at 65 kDa (see Fig. 4A) partitioned specifically in the supernatant, whereas the lower band, which was not competed by the GST/Sun-(262-492) peptide, was retained in the pellet. All of the other immunoreactive species were pelleted in these conditions (Fig. 5A, Total). The crude membrane fraction of the hypotonic shock (Lane P, Total) was further extracted in PBS-1% Triton X-100 or PBS-8 M urea and centrifuged at 15,000 ϫ g. The 75-kDa protein was efficiently solubilized in both conditions, but the 85-kDa protein showed a distinct extraction pattern. It was partially resistant to both PBS-1% Triton X-100 and PBS-8 M urea extraction (Fig.  5A, 1% Tx and 8M Urea). All of the immunoreactive proteins were efficiently solubilized in PBS containing both 1% Triton X-100 and 8 M urea (Fig. 5A, 8M Ureaϩ1% Tx). A duplicate filter was immunoblotted with an anti-lamin B1 antibody to control for the total amount of sample loaded in each well as well as that for the extraction conditions. The detergent insolubility of the endogenous 85-kDa protein was confirmed in immunofluorescence microscopy. HeLa cells were extracted with PBS-1% Triton X-100 on ice for 5 min prior to the paraformaldehyde fixation step. The results (Fig. 5B) show that, in these conditions, the Sun2 immunofluorescence signal persisted around the nuclei while most of the proteindisulfide isomerase immunoreactivity, an endoplasmic reticulum marker, was removed by the detergent pretreatment. The extraction of the 85-kDa protein in PBS-1% Triton X-100 supplemented with increasing salt concentration was also examined in parallel with emerin, a well characterized INM protein.
The 85-kDa protein was retained but gradually solubilized from the pellet of whole HeLa cells extracted with increasing salt concentration (Fig. 5C) similar to emerin, a well characterized INM protein whose extraction properties have been described previously (39,40). The same extraction profile also was observed for the Sun-(26 -339) recombinant construct fused to EGFP (Fig. 5C).
All together, these results indicate that the 85-kDa protein is a new mammalian INM protein that corresponds to the endogenous Sun2 protein. This is supported further by the fact that its molecular weight corresponds to the molecular weight predicted for the protein encoded by the KIAA0668 cDNA.
Sun2 Is a Type II Transmembrane Protein-To determine the orientation of Sun2 in biological membranes, we carried out proteinase K protection assays. Sun-(1-742)/V5 (Fig. 1) was synthesized in vitro in the presence of [ 35 S]methionine with or without canine pancreatic microsomes. Proteinase K digestion was then carried out with or without 1% Triton X-100. Samples were transferred to a nitrocellulose membrane and visualized on a PhosphorImager (Fig. 6A, top panel). The in vitro synthesized protein displayed the expected molecular weight and was detected by both the anti-Sun2 serum and the V5 monoclonal antibody (Fig. 6A, arrows). Upon proteinase K digestion, a radiolabeled 65-kDa fragment (asterisk in Fig. 6A, top panel) was detected specifically in samples containing microsomes but lacking 1% Triton X-100 as is expected from a proteinase Kprotected fragment of a translocated protein. The same protected fragment was recognized both by the anti-Sun2 serum and the V5 antibody (asterisk in Fig. 6A, bottom panel). The size of this fragment was in agreement with the location of the predicted transmembrane region within Sun2 and indicates that it is the C-terminal region of Sun2 that is protected from proteinase K digestion by its translocation inside the microsome lumen. It is notable that a 70-kDa fragment also was protected but most probably represented a proteinase K digestion product intermediate, because it was not observed reproducibly and its abundance was higher and inversely related to the abundance of the 65-kDa protected fragment when lower proteinase K concentrations were used (data not shown).
To confirm that the Sun domain localizes between the INM and the ONM, paraformaldehyde-fixed HeLa cells were treated with digitonin that selectively permeabilizes the cytoplasmic membrane but not the nuclear envelope. The permeabilization of HeLa cells with 0.004% digitonin prevented the anti-Sun2 serum to detect its target while tubulin, used as a control of the cytoplasmic membrane permeabilization, was efficiently detected. Higher digitonin concentrations gradually allowed the detection of the Sun2 rim-like pattern (not shown). Collectively, these results show that the Sun2 epitope located downstream of the transmembrane domain ( Fig. 1) points to the periplasmic space and that Sun2 is a type II transmembrane protein with the SUN domain protruding in the periplasmic space located between the INM and the ONM. DISCUSSION Sun1 (KIAA0810) and Sun2 (KIAA0668) cDNAs were initially isolated from human brain (35,41). The two proteins encoded by these cDNAs show strong homology in their Cterminal regions, which comprises ϳ200 amino acids, with the C-terminal region of the C. elegans protein UNC-84 (34) and with the Sad1 protein from S. pombe (32). This region of homology defines the SUN domain. In C. elegans, UNC-84 localizes to the nuclear envelope and its mutation prevents nuclear migration/positioning. In S. pombe, the endogenous Sad1 protein, whose deletion is lethal, localizes to the spindle pole body and to the nuclear envelope when overexpressed. Therefore, the conserved SUN domain could define a functionally new class of essential nuclear envelope proteins. We show that the recombinant Sun2 protein also localizes in the nuclear envelope. The Sun2 rim-like immunofluorescence pattern is comparable with the previously described nuclear envelope localization of recombinant forms of Sun1 (KIAA0810) transfected in  Fig. 1) was used to synthesize the full-length protein in vitro with [ 35 S]methionine in the absence (Ϫ) or the presence (ϩ) of canine-pancreatic microsomal membranes. Aliquots of the synthesis reaction were digested for 30 min on ice with either proteinase K alone or with proteinase K supplemented with 1% Triton X-100. Proteinase K was inhibited with PMSF, and the samples were added directly to boiling Laemmli buffer. Duplicate samples were transferred to a nitrocellulose membrane, and the radiolabeled proteins were visualized with a Phospho-rImager (top panel). The same membranes then were immunoblotted with either the anti-Sun2 serum or the anti-V5 antibody (bottom panel). B, paraformaldehyde-fixed HeLa cells were permeabilized with 0.04% digitonin for 15 min on ice and probed with the anti-Sun2 serum and anti-␣-tubulin antibody used as a cytoplasmic membrane permeabilization control.

FIG. 5. Extraction properties of Sun2.
A, HeLa cells were submitted to a hypotonic shock and disrupted in a Dounce homogenizer. The supernatant (SN, Total) and pellet (P, Total) of this homogenate were obtained by centrifuging the total homogenate at 15,000 ϫ g for 5 min. The latter pellet was resuspended in 1% Triton X-100 or 8 M urea or 8 M urea, 1% Triton X-100, incubated on ice for 15 min, and centrifuged. Volume to volume aliquots of pellet and supernatant of each extraction condition were analyzed with the Sun2 serum. A duplicate nitrocellulose membrane was immunoblotted with an anti-lamin B1 polyclonal antibody. B, immunofluorescence microscopy detection of Sun2 in HeLa cells extracted with PBS-1% Triton X-100 prior to the fixation step. Cells were probed with the anti-Sun2 serum and an antiprotein-disulfide isomerase antibody. C, whole HeLa cells were extracted in PBS-1% Triton X-100 supplemented with the indicated NaCl concentrations. Volume to volume aliquots of pellet and supernatant of each extraction condition were analyzed with the Sun2 serum, an anti-lamin B1 antibody, and an antiemerin antibody. , UNC-84 in C. elegans embryos (34,42), and Sad1 in S. pombe (32). Our results further demonstrate that a deletion mutant extending from amino acids 26 to 339 of Sun2 is sufficient to specify the nuclear envelope localization of the whole protein. Indeed, this fragment when fused to either EGFP (EGFP/Sun-(26 -339)) or a V5 epitope (data not shown) displayed a rim-like pattern that also co-localizes with the endogenous nuclear pore protein Nup153. In contrast, no rimlike pattern was observed when the EGFP/Sun-(262-742) was expressed in HeLa cells. In this latter case, the signal was both uniformly spread and accumulated in a bright structure identified as the Golgi. Therefore, this result excludes any involvement of either the SUN domain (amino acids 545-742) or the putative signal sequence (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18) in the nuclear envelope localization of Sun2. The possibility that a stretch of basic amino acids ( 126 RRRR 130 ) could be the determinant for the nuclear envelope localization of Sun2 was examined; however, the mutant forms bearing a 126 RRRR 130 to 126 AAAA 130 mutation displayed the same nuclear envelope rim-like pattern as the wildtype forms (data not shown).

Sun2 Is an INM Protein
We raised an anti-Sun2 serum to identify the endogenous Sun2 protein both in microscopy and immunoblot. Peptide competition studies with the IgG fraction of the anti-Sun2 serum showed that three main bands of 65, 75, and 85 kDa were competed specifically by the GST/Sun-(262-492) protein. The immunoaffinity purification of the Sun2 anti-serum against the GST/Sun-(262-492) protein did not modify the immunoblot pattern (data not shown). Nevertheless, our data indicate that the 75-and 65-kDa proteins are cross-reactive proteins. First, in reverse transcription-PCR experiments, we did not obtain any result that would suggest the existence of alternative splicing of the Sun2 mRNA that could account for the detection of the 75-and 65-kDa proteins (data not shown). Second, the affinity of the anti-Sun2 serum was stronger for the 85-kDa protein as indicated by the persistence of the 85-kDa signal and the disappearance of the 75-and 65-kDa signals when the amount of HeLa cell lysate was serially diluted (not shown). Third, in the immunoprecipitation experiments, the anti-Sun2 serum specifically pulled down the 85-kDa protein whose molecular mass was predicted by the Sun2 cDNA (KIAA0668).
In immunofluorescence microscopy, a nuclear rim-like pattern comparable to other inner nuclear membrane proteins and to the rim-like pattern observed with the recombinant Sun2 protein was detected. The rim-like pattern specifically was attributed to the anti-Sun2 immunoreactivity because the GST/Sun-(262-492) protein efficiently prevented its detection, whereas the GST protein alone did not. Furthermore, Sun2 co-localized with Nup153, a nuclear pore marker, and emerin, an INM marker. Interestingly, as for emerin (43) and the lamin B receptor (11), Sun2 also was recruited in the nuclear envelope of the telophase cells. The fact that the 85-kDa protein corresponds to the protein associated with the nuclei was shown further by its selective retention in purified HeLa cell nuclei. In addition to the rim-like pattern of the endogenous Sun2 protein, some bright spots of signal also were observed inside the nuclei as well as in the cytoplasm of HeLa cells (Fig.  3, A and B). In electron microscopy, gold particles were detected in the nuclear envelope. Interestingly, some regions of the nuclear envelope displayed clusters of gold particles in darkened areas of the nuclear envelope. In addition to their presence in the nuclear envelope, gold particles were also observed but to a much lesser extent in the cytoplasm of HeLa cells.
Fractionation of total HeLa cells with either a detergent (1% Triton X-100) or a chaotropic agent (8 M urea) indicated that the 85-kDa protein was partially insoluble in these conditions. The detergent insolubility also was confirmed in immunofluorescence microscopy. Furthermore, as was observed for emerin, the 85-kDa protein was retained in the pellet of cells extracted with detergent supplemented with increasing salt concentrations. The other immunoreactive bands did not display these properties. Similar biochemical properties have been described for UNC-84 (42), the Sun1 mouse homolog (30), and other INM proteins (7). Therefore, these results suggest that Sun2 is an INM protein that interacts with the nuclear lamina. This conclusion is reinforced by the recent identification of Sun2 in a stringent proteomic analysis of the nuclear envelope aimed at the identification of new INM proteins (31).
We determined the topology of Sun2 in biological membranes. Our results indicate that Sun2 is a type II membrane protein with a C-terminal luminal domain. This result was obtained by proteinase K protection assay of in vitro synthesized Sun2 and further confirmed in digitonin experiments showing that the C-terminal region of Sun2 downstream of the transmembrane domain points to the periplasmic space between the ONM and the INM. A simple model for the topology of Sun2 in the nuclear envelope is presented in Fig. 7. The insolubility of Sun2 both in detergent with increasing salt concentrations and in urea indicates a tight interaction of Sun2 with the nuclear lamina, presumably through the N-terminal region. Indeed, we have shown that the N-terminal region (amino acids 26 -339) of Sun2 was sufficient to specify its nuclear envelope localization and to confer an extraction pattern that was similar to the whole endogenous protein in high salt concentrations buffers. Furthermore, lamin-dependent localization of UNC-84 in the nuclear envelope has been demonstrated in C. elegans (42). Therefore, we speculate that a lamina component immobilizes Sun2 in the nuclear envelope.
A further assessment of Sun2 function will require the identification of its direct binding partners. One candidate is the human ortholog of the C. elegans protein UNC-83, which binds the SUN domain of UNC-84 in vitro (44). A second candidate is the human ortholog of Anc-1, another C. elegans protein that alters nuclear migration when mutated and that is recruited to the nuclear envelope in a UNC-84-dependent manner (45). Interestingly, Anc-1 contains a KASH (Klarsicht/Anc-1/Syne-1 homology) domain, part of which protrudes in the periplasmic space, and a calponin domain. Both domains are found also in several isoforms of mammalian nesprin proteins (46 -48). A model proposing the involvement of these molecules in the actin-dependent anchoring of the nuclei has been proposed recently (49). Therefore, a possible function for Sun2 would be to relay and/or regulate the traction forces necessary to anchor or move the nucleus when required. The validation of this model (Fig. 7) will require the demonstration that nesprins and Sun2 are direct or indirect binding partners as well as the identification of one or several nuclear lamina components that could anchor Sun2 in the nucleus. Our results showing the insolubility of Sun2 as well as the presence of the SUN domain in the periplasmic space allow for the possibility of such a direct or indirect interaction. In addition to this latter structural function of Sun2, a regulatory function can also be hypothesized because it has been demonstrated that the 310 C-terminal amino acids of Sun1 containing the SUN domain directly interact with TRAX (translin-associated factor X), a direct binding partner of the RNA-binding protein TB-RBP (50).
Finally, a recent report shows (51) that Sun2 mRNA expression is up-regulated in human biopsies from patients with ventricular septal defect, a particular type of congenital heart defect in children. This finding could be indicative of the possible involvement of Sun2 in muscular dysfunction as is the case for other nuclear lamina components (26).