
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0020231Research ArticleCell BiologyHomo (Human)Emerin Caps the Pointed End of Actin Filaments: Evidence for an Actin Cortical Network at the Nuclear Inner Membrane Emerin Stabilizes the Pointed End of F-ActinHolaska James M 
1
Kowalski Amy K 
1
Wilson Katherine L klwilson@jhmi.edu
1
1Department of Cell Biology, The Johns Hopkins University School of MedicineBaltimore, MarylandUnited States of America9 2004 24 8 2004 24 8 2004 2 9 e2311 12 2003 24 5 2004 Copyright: © 2004 Holaska et al.2004This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
A Case for a Functional Actin Network in the Nucleus 
X-linked Emery-Dreifuss muscular dystrophy is caused by loss of emerin, a LEM-domain protein of the nuclear inner membrane. To better understand emerin function, we used affinity chromatography to purify emerin-binding proteins from nuclear extracts of HeLa cells. Complexes that included actin, αII-spectrin and additional proteins, bound specifically to emerin. Actin polymerization assays in the presence or absence of gelsolin or capping protein showed that emerin binds and stabilizes the pointed end of actin filaments, increasing the actin polymerization rate 4- to 12-fold. We propose that emerin contributes to the formation of an actin-based cortical network at the nuclear inner membrane, conceptually analogous to the actin cortical network at the plasma membrane. Thus, in addition to disrupting transcription factors that bind emerin, loss of emerin may destabilize nuclear envelope architecture by weakening a nuclear actin network.

Loss of emerin leads to Emery-Dreifuss muscular dystrophy (EDMD). Biochemical studies presented here suggest that emerin drives the formation of an actin-based cortical network at the nuclear membrane, and that network destabilization may contribute to EDMD
==== Body
Introduction
Emery-Dreifuss muscular dystrophy (EDMD) is inherited through mutations in either of two different genes: LMNA, encoding A-type lamins, and STA, which encodes a nuclear membrane protein named emerin (Nagano et al. 1996; Emery 2000; Bengtsson and Wilson 2004). Lamin filaments and emerin interact at the nuclear inner membrane (Burke and Stewart 2002; Holaska et al. 2002). Together, emerin and lamin A form stable tertiary complexes with other binding partners in vitro (Holaska et al. 2003), suggesting that emerin and lamins together provide a structural foundation for oligomeric protein complexes. Mutations in emerin cause the X-linked recessive form of EDMD (Bione et al. 1995; Bonne et al. 2003). Both emerin and lamin A are expressed in most cells, but EDMD disease strikes specific tissues: skeletal muscles, major tendons, and the cardiac conduction system. To explain the tissue specificity of EDMD, it was proposed that emerin might have tissue-specific binding partners such as transcription factors and signaling molecules that regulate gene expression (Wilson 2000; Bonne et al. 2003; Östlund and Worman 2003). There is growing evidence to support “gene expression” models for emerin, as discussed further below. However, a second model, not mutually exclusive, proposes that emerin helps maintain the structural integrity of the nuclear envelope. According to structural models, loss of emerin selectively disrupts tissues under high mechanical stress, such as skeletal muscle and tendons (Bonne et al. 2003; Östlund and Worman 2003). Although this model fails to explain the cardiac conduction phenotype of EDMD, it is consistent with structural defects (aberrant shape and nuclear envelope herniations) seen in nuclei from EDMD patients (Fidzianska and Hausmanowa-Petrusewicz 2003) and in a subset of patients with other diseases linked to mutations in LMNA (“laminopathies”; Holaska et al. 2002; Östlund and Worman 2003). Whereas structural and mechanical roles are expected for lamins, which form nuclear intermediate filaments, mechanical roles for emerin have not been investigated.

Emerin is detected in most human cells tested, except the nonmyocytes of the heart (Manilal et al. 1996). In Caenorhabditis elegans emerin is expressed ubiquitously (Lee et al. 2000; Gruenbaum et al. 2002). Emerin belongs to the LEM-domain family of proteins, which are defined by an approximately 40-residue folded domain known as the LEM domain. In vertebrates, other family members include LAP2β and MAN1 at the nuclear inner membrane (Dechat et al. 2000; Lin et al. 2000; Cohen et al. 2001), LAP2α in the nuclear interior (Dechat et al. 2000), and at least three additional uncharacterized LEM-domain proteins (Lee and Wilson 2004). A major shared function of all characterized LEM-domain proteins is their binding (via the LEM domain) to a small protein named barrier-to-autointegration factor (BAF; reviewed by Segura-Totten and Wilson 2004). BAF is a highly conserved chromatin protein essential for cell viability (Zheng et al. 2000), with direct roles in higher order chromatin structure and nuclear assembly (Haraguchi et al. 2001; Segura-Totten et al. 2002), and gene regulation (Wang et al. 2002; Holaska et al. 2003).

Supporting gene regulation models for EDMD, emerin binds directly to BAF and two other transcription repressors, germ cell-less (GCL; Holaska et al. 2003) and Btf (Haraguchi et al. 2004) as well as an mRNA-splicing factor named YT521-B (Wilkinson et al. 2003). Interestingly, GCL and LAP2β, a LEM-domain protein closely related to emerin, comediate transcription repression in vivo (Nili et al. 2001). On the other hand, emerin also has a growing number of structural or anchoring partners, including a spectrin-repeat (SR) membrane protein named nesprin-1α (Mislow et al. 2002a, 2002b), lamins A and C (Clements et al. 2000; Lee et al. 2001; Sakaki et al. 2001), and lamin B (Dreger et al. 2002). Lamins form type-V intermediate filaments that are critical for the integrity of the nucleus and confer unique elasticity and incompressibility properties to the nuclear envelope (Dahl et al. 2004). Nuclear morphology and lamina architecture are disrupted in a fraction of cells that express disease-causing missense mutations in A-type lamins (Östlund and Worman 2003). However, the line between gene expression and mechanical models for disease has become blurred. For example, the subnuclear localization of Rb, a transcription and cell-fate regulator, depends on both LAP2α and lamin A, since Rb is mislocalized in cells that either lack A-type lamins (Johnson et al. 2004) or express disease-causing lamin-A mutations (Markiewicz et al. 2002). A second putative anchoring partner for emerin is nesprin-1α, an integral nuclear inner-membrane protein with seven SR domains (Zhang et al. 2001; Mislow et al. 2002b). Each SR domain consists of approximately 100 residues and folds into a tightly packed triple α-helical structure (Bennett and Baines 2001). Tandem SR domains, as seen in nesprin-1α, form a rigid, elongated tertiary structure (Djinovic-Carugo et al. 2002). More important, SR domains provide binding sites for other proteins, the specificity of which is determined by exposed residues (Bennett and Baines 2001). SR domains 1–7 (and particularly domains 1–5) of nesprin-1α mediate high-affinity binding to emerin (Mislow et al. 2002a). Interestingly, SR domains 5–7 of nesprin-1α bind directly to lamin A in vitro, suggesting that nesprin-1α, lamin A, and emerin might form stable tertiary complexes. Such complexes have the potential to stabilize lamin filaments at the nuclear envelope, in addition to anchoring and spacing emerin.

Notably, both emerin and lamin A also bind G-actin in vitro (Sasseville and Langelier 1998; Fairley et al. 1999). Actin binds two regions in the lamin-A tail (Zastrow et al. 2004). Both α- and β-actin bind emerin in vitro, and emerin coimmunoprecipitates with actin from cell lysates (Fairley et al. 1999; Lattanzi et al. 2003). The significance of these findings was unclear, in part because nuclear actin has been both documented and debated for over 35 y (Pederson and Aebi 2002). However there is a growing consensus that nuclear actin is no artifact (Pederson and Aebi 2002; Bettinger et al. 2004). Both α- and β-actin have been shown, definitively, to reside in the nucleus (Scheer et al. 1984; Gonsior et al. 1999; Olave et al. 2002; Lattanzi et al. 2003) and to form short filaments in the nucleus (Clark and Rosenbaum 1979). Actin and actin-related proteins (Arps) are required for chromatin remodeling and transcription (Olave et al. 2002; Percipalle et al. 2003). Also interesting is that polymerase II–dependent mRNA transcription requires a nuclear-specific myosin I motor (nuclear myosin I; Nowak et al. 1997; Pestic-Dragovich et al. 2000). Thus, actin probably has a variety of roles in the nucleus.

To test the hypothesis that emerin forms multiprotein complexes in vivo, we affinity-purified emerin-binding proteins from nuclear extracts of HeLa cells. We identified actin itself plus several actin-binding proteins as bona fide emerin-associated proteins, and we further discovered that emerin stimulates actin polymerization in vitro by binding and stabilizing the pointed end of growing filaments. These results suggest that emerin contributes to the formation of an actin cortical network at the nuclear inner membrane.

Results
We used affinity chromatography to purify emerin-binding complexes from HeLa nuclear extract; as the negative control, beads were conjugated to bovine serum albumin (BSA). Mass spectrometry (data not shown) and Western blotting identified β-actin as a major emerin-binding protein (Figure 1A). Six other proteins, including nuclear-enriched αII-spectrin, were also identified and will be reported in full elsewhere (J. M. H. and K. L. W., unpublished data). Consistent with previous reports (Fairley et al. 1999; Lattanzi et al. 2003), antibodies specific for emerin coprecipitated actin from HeLa cell nuclear lysates, as shown by Western blotting with immune serum (Figure 1B, Im). Only background levels of actin were precipitated by preimmune sera (Figure 1B, PI). These findings led us to hypothesize that emerin might bind filamentous actin (F-actin).

Figure 1 Affinity Purification of Emerin-Associated Proteins
(A) Immunoblot of HeLa nuclear lysate proteins (L), or proteins affinity-purified using either BSA beads or emerin beads (see Materials and Methods), probed with antibody against actin.

(B) HeLa nuclear lysate proteins (L) were immunoprecipitated using either immune (Im) or preimmune (PI) serum 2999 against emerin, resolved by SDS-PAGE, and Western blotted using antibodies specific for actin (upper panel) or emerin (lower panel), in succession.

(C) Cosedimentation assays using F-actin and purified, recombinant wild-type emerin (residues 1–222). G-actin (2 μM) was polymerized and then incubated in the absence or presence of 2 μM emerin. Emerin was incubated alone in polymerization buffer as a negative control. After 30 min samples were centrifuged 1 h at 100,000g, resolved by SDS-PAGE, and stained with Coomassie blue. L, load (100%); S, supernatant (100%); P, pellet (100%).

(D) F-actin column was used to determine the affinity of F-actin for emerin. The Kd was 480 nM for the experiment shown; range was 300–500 nM, n = 8.

(E) Binding of wild-type (WT) or mutant emerin protein to F-actin beads. Recombinant emerin proteins were incubated with F-actin beads, and bound emerins were eluted with SDS-PAGE sample buffer, resolved by SDS-PAGE, blotted, and probed with antibodies against emerin (“bound”; all emerin mutants are recognized by this antibody; Lee et al. 2001; Holaska et al. 2003). The input amounts (10%) of each emerin mutant (“load”) were visualized either by immunoblotting (top row, top panel) or Coomassie staining (top row, bottom panel).

(F) Domains in emerin required for binding to BAF, lamin A, transcription repressor GCL, or actin (Lee et al. 2001; Holaska et al. 2003; present study). Asterisks indicate EDMD disease-causing mutations.

Emerin Binds F-Actin with High Affinity
Emerin was first tested for binding to F-actin in a cosedimentation assay. Actin filaments were incubated 30 min in the presence or absence of recombinant emerin (4 μM) and then pelleted at 100,000g. Approximately 75% of input emerin pelleted in the presence of F-actin, compared to 15% in the absence of F-actin (Figure 1C), demonstrating that emerin binds polymerized actin in vitro. The stoichiometry of this interaction was one emerin molecule per approximately 300 actin monomers, demonstrating that emerin binds actin filaments with an average length of 0.9 μm (data not shown). Emerin binds F-actin with high affinity, as determined by binding to an F-actin column (Kd = 480 nM, range = 300–500 nM, n = 8; Figure 1D). The F-actin columns were also used to screen selected emerin mutants for binding to F-actin. Wild-type emerin and EDMD disease-causing mutants S54F and P183H, and alanine substitution mutant m196 (Ser196Ser197 to Ala196Ala197; Holaska et al. 2003), bound efficiently to an F-actin column, whereas 12 other tested mutants, including EDMD disease-causing mutant Q133H, showed no significant binding to F-actin (Figure 1E). Similar results were seen in coimmunoprecipitation assays using antibodies against actin to immunoprecipitate F-actin in the presence of recombinant wild-type or mutant emerin proteins (data not shown). Fifteen additional emerin missense mutants, including LEM-domain mutants, were tested in blot overlay assays; mutants m11, m24, m30, m40, m207, m214, and m217 bound detectably to actin, whereas no significant binding was detected for mutants m45A, m45E, m61, m141, m145, m161, and m206 (data not shown; see Holaska et al. 2003 for details of mutations). Thus, three independent assays all showed that emerin binds F-actin. Furthermore, based on the positions of missense mutations that blocked binding to F-actin, this recognition involved almost the entire nucleoplasmic domain of emerin, with the notable exception of the LEM-domain (Figure 1F). The putative actin-binding domain in emerin overlaps with both the lamin- and repressor-binding (GCL) domains described previously (Holaska et al. 2003).

Emerin Regulates Actin Polymerization
The above results suggested that in vivo emerin might (a) use actin filaments as anchors, (b) stabilize F-actin networks, or (c) actively influence actin dynamics. To test these models, we first used reactions containing 5% pyrene-labeled actin (final actin concentration, 2 μM) to determine if emerin influenced actin polymerization in vitro. Results were graphed as the rate of emerin-induced polymerization (R) divided by the control rate (cR; rate of actin polymerization in the absence of emerin). At concentrations ranging from 0.1 to 4.4 μM, emerin increased the rate of pyrene–actin polymerization 4-fold (mean = 6.2 ± 2.2, n = 32; one experiment is shown in Figure 2A). These experiments also yielded an equilibrium affinity of 420 nM (Figure 2A), consistent with our previous results (480 nM; see Figure 1D).

Figure 2 Emerin Stimulates Actin Polymerization
(A) Graph of a representative experiment (n = 32) showing that emerin increases the rate of actin polymerization 4-fold. R, rate of polymerization in the presence of emerin; cR, control rate of polymerization (actin alone). These data also yielded an equilibrium binding affinity of emerin for actin of 420 nM.

(B) Representative graph (n = 17) in which each recombinant emerin mutant protein (1.0 μM) was added to 2.0 μM G-actin, and polymerization rates were calculated. R, rate of polymerization in the presence of emerin; cR, control rate of polymerization (actin alone). Stars indicate EDMD disease-causing mutations.

(C) Critical concentration assays were performed in the presence or absence of 625 nM emerin. Actin was polymerized in the absence (barbed-end growth) or presence of 5 nM gelsolin–actin seeds (pointed-end growth) for 16 h at room temperature. Barbed-end growth with (□) or without (▪) emerin. Pointed-end growth with (○) or without (•) emerin. Actin monomers (Δ).

We next tested eight emerin mutants (1 μM) in the pyrene–actin polymerization assay (Figure 2B). Five mutants (Q133H, m151, m164, m192, and m198) that failed to bind actin in vitro did not stimulate actin polymerization; instead, they reduced the rate of actin polymerization slightly, by 5%–40% (Figure 2B). Mutant 196, which had wild-type binding to F- and G-actin in coimmunoprecipitation assays, stimulated actin polymerization approximately 50% as well as wild-type emerin (Figure 2B). The two disease-causing mutants with apparently normal binding to F-actin, S54F and P183H, enhanced the rate of actin polymerization at least as well as wild-type emerin (Figure 2B). Critical concentration assays were done to determine if emerin acted on the barbed or pointed end of growing actin filaments. Pointed-end growth was examined by capping filaments with gelsolin. Emerin had no significant effect on the critical concentration of barbed-end growth (Figure 2C, + emr/no emr), but it increased the critical concentration for pointed-end growth by 2.3- to 2.7-fold (Figure 2C, + gelsolin no emr/+ gelsolin + emr). We therefore hypothesized that emerin, like tropomodulin (Fowler et al. 2003), might stabilize growing filaments by capping the pointed end.

Emerin Binds F-Actin at the Pointed End
The vast majority of actin-binding proteins that influence subunit addition do so by binding the barbed end (dos Remedios et al. 2003). However, because emerin failed to influence the critical concentration for barbed-end growth (Figure 2C), we used gelsolin–actin seeds to test the hypothesis that emerin binds the pointed end. Gelsolin binds and caps the barbed end of actin filaments (Burtnick et al. 2001), thereby restricting subunit addition to the pointed end only. We measured the extension of gelsolin–actin dimers (10 nM) in the presence of 2 μM actin plus 0 to 2 μM wild-type emerin (Figure 3A). Emerin blocked actin polymerization in a concentration-dependent manner, supporting our model that emerin binds and caps the pointed end of actin filaments. Based on this assay, the affinity (Kd) of emerin for F-actin was 430 nM (range, 300–500 nM, n = 12; Figure 3A). Interestingly, the affinity determined by either direct binding (see Figure 1D) or activity measurements (see Figures 2A and 3A) differed by a maximum of 2-fold. We conclude that emerin binds F-actin with an affinity of 300–500 nM.

Figure 3 Emerin Binds the Pointed End of Actin Filaments
(A) Gelsolin–actin seeds were incubated with increasing concentrations of wild-type emerin residues 1–222. Emerin significantly reduced the rate of subunit addition at the pointed end, with an apparent Kd of 430 nM (range, 300–500 nM, n = 12). R, rate of polymerization in the presence of emerin; cR, control rate of polymerization (actin alone).

(B) Emerin inhibits depolymerization of actin filaments (2 μM) preformed from gelsolin–actin seeds, with an apparent Kd of 380 nM (range, 350–450 nM, n = 6). R, rate of depolymerization in the presence of emerin; cR, control rate of depolymerization (actin alone).

(C) Rhodamine–phalloidin-stabilized actin filaments were formed from 2 μM actin, then capped at the barbed end by the addition of 100 nM capping protein, and finally diluted 2-fold in the presence of buffer or 1 μM emerin, GST, or tropomodulin (Tmod). Samples were then incubated with actin (3.2 μM) and Alexa-488 phalloidin (3.2 μM) for 2 min, diluted 1:500, placed on polylysine-coated coverslips, and viewed by fluorescence microscopy. Bar is 1 μm and applies to all panels.

(D) Actin (2 μM) was incubated with gelsolin–actin seeds (500 nM) in the presence of rhodamine–phalloidin (2 μM). These red filaments were diluted 10-fold and incubated with buffer alone or with 1 μM emerin or tropomodulin (Tmod) for 10 min, followed by incubation with actin (2 μM) and Alexa-488-labeled (green) phalloidin (2 μM) for 2 min. Samples were diluted 1:500, placed on polylysine-coated coverslips, and viewed by fluorescence microscopy. Bar is 1 μm and applies to all panels.

(E–H) Alexa-488-labeled emerin (green) was incubated 30 min with actin filaments stabilized by Alexa-546 phalloidin (red) and centrifuged at 100,000g to recover filaments, which were diluted 1:500 for viewing.

To independently confirm that emerin binds the pointed end, we tested the effect of emerin on actin depolymerization in two separate assays. First, gelsolin–actin seeds were incubated with 2 μM actin and grown in the absence of emerin. The resulting filaments were then diluted to 0.2 μM actin in the presence of increasing concentrations of emerin (0–2 μM), and assayed immediately. Preformed actin filaments depolymerized rapidly in the absence of emerin (Figure 3B, arrow), as expected. Supporting our model, depolymerization was slowed up to 8-fold by emerin (Figure 3B). To independently confirm that emerin blocked subunit addition at the pointed end, preformed actin filaments were capped at the barbed end with capping protein (CapZ), a high-affinity barbed-end binding protein (Cooper and Schafer 2000), then incubated in the presence or absence of emerin, and diluted into 0.2 μM actin; emerin slowed the rate of depolymerization by 10-fold (data not shown). Based on these three assays, we conclude that emerin binds and protects the pointed end of actin filaments in vitro, thereby stabilizing actin filaments.

Fluorescent actin polymerization assays were done to demonstrate visually that emerin blocks pointed-end growth of single actin filaments. Rhodamine–phalloidin-stabilized actin filaments (red) were preformed from 2 μM actin, then capped on the barbed end with capping protein, and diluted 2-fold into a final concentration of 1 μM emerin. Pointed-end growth was then initiated by increasing actin to 3.2 μM in the presence of 3.2 μM Alexa-488 phalloidin (green) for 2 min. In the absence of emerin (Figure 3C, buffer or GST), actin filaments containing both red and green segments are seen (Figure 3C), demonstrating pointed-end growth. The average lengths of the growing filament segments (green) as measured for buffer and GST were 1.45 ± 0.3 μm and 1.47 ± 0.3 μm, respectively (n = 30). However, in the presence of either emerin or tropomodulin, a pointed-end binding protein, most red filaments lacked green segments (Figure 3C), consistent with capped pointed ends. In the presence of emerin or tropomodulin, the lengths of the green segments were 0.05 ± 0.09 μm (n = 60) and 0.09 ± 0.13 μm (n = 50), respectively. Single filament assays were also done using small red filaments formed from gelsolin–actin seeds (Figure 3D). Here, actin (2 μM) was incubated with gelsolin–actin seeds (500 nM) and rhodamine–phalloidin (2 μM). The resulting red filaments were diluted 10-fold and incubated with 1 μM emerin for 10 min, then incubated 2 min with actin (2 μM) and Alexa-488-labeled (green) phalloidin (2 μM). In the absence of emerin, single filaments contained both short red (gelsolin–actin seeds) and longer green (pointed-end growth) segments (Figure 3D). The average length of these growing filament segments was 2.1 ± 0.6 μm (n = 40). However, when emerin was present, the preformed filaments remained predominantly short and red, demonstrating that emerin blocks pointed-end growth (Figure 3D). Similar results were obtained in control reactions containing tropomodulin, the pointed-end binding protein (Figure 3D, Tmod). The average lengths of growing filaments incubated with emerin or tropomodulin were 0.1 ± 0.1 μm (n = 40) and 0.1 ± 0.11 μm (n = 40), respectively. These experiments also show that emerin does not stimulate branching (Figure 3D). To directly visualize emerin bound to actin filaments, green (Alexa-488-labeled) emerin was incubated with red (Alexa-546 phalloidin) actin filaments (Figure 3E–3H). Under these conditions 85% of labeled emerin molecules were bound to actin filaments; of these, 92% were localized at a filament end (n = 306). Interestingly, 10% of actin-associated emerin proteins localized to “aster-like” structures (Figure 3G and 3H), presumably due to the aggregation of emerin proteins on different actin filaments.

Discussion
This work shows for the first time that a nuclear membrane protein, emerin, is a pointed-end F-actin-binding protein. Similar to the activity of tropomodulin (Fowler 1997; Cooper and Schafer 2000; Fowler et al. 2003), emerin caps the pointed end, thereby stabilizing the growing filament. Only three other pointed-end binding proteins have been reported: the Arp2/3 complex (Mullins et al. 1998), tropomodulin, and mSWI/SNF, a component of a nuclear complex that remodels chromatin structure (Rando et al. 2002). The Arp2/3 complex initiates filament branching at the cell surface (Mullins et al. 1998; Mullins and Pollard 1999). We have no evidence that emerin initiates branching. Instead, emerin behaves most like tropomodulin, which binds the pointed end of F-actin with high affinity (Kd = 110 nM) and stimulates actin polymerization by stabilizing the actin filament (Fowler et al. 2003).

Our analysis of 15 emerin missense mutants suggested that the actin-binding region in emerin overlaps with regions required for binding to lamin A (Lee et al. 2001), transcription factors GCL and YT521-B (Holaska et al. 2003; Wilkinson et al. 2003), and nesprin-1α (J. M. H. and K. L. W., unpublished data). However this overlap does not necessarily imply that actin competes with these other proteins. Indeed, despite similar overlap, GCL and lamin A can form stable ternary complexes with emerin in vitro (Holaska et al. 2003). Further work is needed to determine if F-actin cobinds or competes with lamin A, nesprin-1α, or other emerin-binding proteins.

A Proposed Actin Network at the Nuclear Envelope
We propose that emerin stabilizes and promotes the formation of a nuclear actin cortical network, analogous to the actin cortical network at the plasma membrane (Figure 4). Another LEM-domain protein, LAP2β, also an integral nuclear inner-membrane protein, was 20-fold less active than emerin in actin polymerization assays (data not shown), suggesting that LAP2β binds actin with an affinity 20-fold lower than that of emerin. Other LEM-domain proteins have not yet been tested for binding to actin. Whether emerin has specialized roles involving actin, or shares this function with other nuclear membrane proteins, are both interesting possibilities. An actin-based cortical network could help anchor emerin and possibly other nuclear membrane proteins and lamin filaments, contributing significantly to the structural integrity of the nuclear envelope and potentially reinforcing sites of chromatin attachment (Figure 4).

Figure 4 Model in Which Emerin Binding to the Pointed End of F-Actin Stabilizes an Actin Cortical Network at the Nuclear Inner Membrane
Our model is based on the actin cortical network at the cell surface of erythrocytes, except that lamin filaments also anchor to emerin-based junctional complexes. Spectrin heterodimers bind short actin filaments at the erythrocyte membrane; we therefore speculate that nuclear isoforms of αII-spectrin (J. M. H. and K. L. W., unpublished data) act similarly. Direct binding of emerin and αII-spectrin has not yet been tested. Nuclear isoforms of protein 4.1, which are essential for nuclear assembly (Krauss et al. 2002), have the potential to cross-link short actin filaments and spectrin filaments at the inner membrane (IM). Further work is necessary to test our model and identify other components of this proposed nuclear actin cortical network. OM, nuclear outer membrane.

Since lamin A also binds G-actin in vitro (Sasseville and Langelier 1998), we are currently testing the actin-binding properties of lamin A. Because emerin forms stable complexes with lamin A in vitro (Clements et al. 2000; Lee et al. 2001), and because the nuclear envelope localization of emerin depends on lamins, we speculate that emerin might interlink multiple filament networks (actin, spectrin, and lamins) at the nuclear envelope. This model will be tested in future experiments by determining whether lamin A and actin compete for binding to emerin, or form trimeric complexes. Such complexes could significantly reinforce the mechanical properties of the nuclear envelope. Our nuclear actin cortical network model is further supported by the properties of the nesprin family of nuclear membrane proteins, which includes nesprin-1α (see Introduction) and NUANCE. Nesprin-1α binds directly to both emerin (Kd = 4 nM) and lamin A (affinity undetermined; Mislow et al. 2002a). NUANCE is a large (796 kd), alternatively spliced isoform of nesprin that localizes to the nuclear envelope and nucleoplasm and binds F-actin (Zhen et al. 2002).

The organization of the membrane skeleton in erythrocytes (red blood cells) includes integral membrane proteins (e.g., Band 3), anchoring proteins (ankyrin), spectrin filaments, and “junctional complexes” (short actin filaments, protein 4.1, adducin, tropomodulin, and tropomyosin; Delaunay 2002). Tropomodulin and tropomyosin stabilize the junctional complex. Spectrin filaments (α/β-spectrin heterodimers) attach to junctional complexes through direct binding to protein 4.1, adducin, and actin. At the inner nuclear membrane, our working model is that emerin stabilizes junctional complexes (Figure 4), consisting of short actin filaments, nuclear-specific αII-spectrin (McMahon et al. 1999), and nuclear isoforms of protein 4.1 (Krauss et al. 2002). This model is the first step toward understanding the structural function of nuclear actin.

Materials and Methods

Antibodies and proteins
A pan-actin antibody (Sigma, St. Louis, Missouri, United States; catalog #A-5060) was used at 1:1,000 for immunoblotting. An antibody specific for β-actin (Sigma; catalog #A-5316) was used at 1:10,000 for immunoblotting and 1:1,000 for immunoprecipitation. Our rabbit polyclonal emerin antibody (serum 2999), described previously (Lee et al. 2001), was used at 1:20,000 for immunoblotting and 1:2,000 for immunoprecipitation. Purified chicken actin was a kind gift of Doug Robinson (Johns Hopkins Medical School). Purified rabbit actin was purchased from Cytoskeleton. (Denver, Colorado, United States; catalog #AKL95 and #AKL99). Alexa-488 actin (#A12373), Alexa-594 actin (#A34050), Alexa-488 phalloidin (#A-12379), rhodamine–phalloidin (#R-415), and Alexa-546 phalloidin (#A-22283) were purchased from Molecular Probes (Eugene, Oregon, United States). Purified CapZ (capping protein) was a kind gift from John Cooper (Washington University, St. Louis). The full nucleoplasmic domain of wild-type emerin (residues 1–222) and corresponding mutants (detailed in Holaska et al. 2003) were expressed in bacteria and purified as described (Lee et al. 2001; Holaska et al. 2003). Emerin protein was labeled with Alexa-488 (Molecular Probes, catalog #A20000) per manufacturer's instructions.

Affinity purification using emerin-conjugated beads
Wild-type emerin residues 1–222 (comprising the entire nucleoplasmic domain of emerin and lacking the transmembrane domain) or BSA (as a negative control) were coupled to Affigel-15 beads (Bio-Rad, Hercules, California, United States) per manufacturer's instructions. Nuclear extracts were prepared by hypotonic lysis (Offterdinger et al. 2002) from 1010 HeLa-S3 cells, obtained as frozen cell pellets from the National Cell Culture Center. For each affinity purification, we incubated 50 mg of nuclear lysate proteins with 2 ml of either emerin beads (0.5 mg/ml) or BSA beads in binding buffer (50 mM HEPES, 250 mM NaCl, 0.1% Triton X-100) for 4 h at 4 °C. Beads were collected by centrifugation at 500g, washed five times with binding buffer, and eluted with SDS-PAGE sample buffer. Dr. Robert Cole at the Johns Hopkins Mass Spectrometry Facility performed MALDI-TOF. Actin and αII-spectrin were two of seven emerin-associated proteins identified unambiguously in this work, which will be reported separately (J. M. H. and K. L. W., unpublished data).

F-actin-binding assays
F-actin columns were assembled as described (Forero and Wasserman 2000). Equal amounts of purified recombinant wild-type and mutant emerin proteins (residues 1–222) were incubated with each column in PBS containing 0.1% Triton X-100 (PBST) for 1 h at 22 °C. After washing beads five times with PBST, bound proteins were eluted and resolved by SDS-PAGE, and detected either by Coomassie blue staining, or by immunoblotting with rabbit serum 2999 against emerin.

Coimmunoprecipitation assays were performed as described (Lee et al. 2001). Briefly, equal masses (5 μg) of actin and either wild-type or mutant emerin were incubated 2 h, then incubated 4 h with protein-A Sepharose-coupled antibodies against emerin or actin. The beads were washed five times with buffer, and bound proteins were eluted with SDS-sample buffer, resolved by SDS-PAGE, then blotted and probed with antibodies specific for either actin or emerin.

To measure emerin binding to single actin filaments, actin was polymerized by the addition of KMEI buffer (2 mM MgCl, 50 mM KCl, 10 mM imidazole [pH 7. 0], and 2 mM EGTA). After 30 min, the indicated form of emerin (4 μM, recombinant wild-type or mutant emerin residues 1–222, with or without conjugation to Alexa-488) was added to the filaments and incubated 30 min; we lastly added Alexa-546 phalloidin (final concentration, 0. 33 μM). These red F-actin polymers were then pelleted at 100,000g for 60 min. For experiments with unlabeled emerin, corresponding load, supernatant, and pellet fractions were resolved by SDS-PAGE and stained with Coomassie blue. For experiments with phalloidin-546-labeled F-actin, filaments were viewed using a Zeiss Axiovert 200 fluorescent microscope (Zeiss, Oberkochen, Germany) and images captured using a Quantix CCD camera (Photometrics, Huntington Beach, California, United States) attached to an Apple G4 computer using IPLab (version 3.6; Scanalytics, Fairfax, Virginia, United States) software.

Actin polymerization and depolymerization assays
Actin polymerization assays were performed per manufacturer's instructions (Cytoskeleton). Rabbit actin (2 μM; Cytoskeleton, catalog #AKL95) plus pyrene–actin (0.1 μM; Cytoskeleton, catalog #AP05) were used in all assays, unless otherwise stated. Actin polymerization was measured in a fluorimeter (Fluoromax 2; SPEX, Edison, New Jersey, United States), with excitation wavelength 365 nm and emission wavelength 407 nm, and plotted using DataMax-Std (version 2.2; SPEX, Edison, New Jersey, United States). Graphs were refined using Cricketgraph III (version 1.0, Computer Associates, Smithfield, Rhode Island, United States) and Kaleidagraph (version 3.5.1, Synergy Software, Reading, Pennsylvania, United States). Pyrene–actin was always present at 5% of total actin. Increasing concentrations of recombinant emerin were added just prior to initiating actin polymerization. The actin depolymerization assays were performed exactly as described (Mullins et al. 1998) with increasing amounts of recombinant emerin protein. Briefly, F-actin was polymerized by adding 2 μM G-actin to gelsolin–actin seeds (100 nM), then diluted 10-fold in the absence or presence of increasing concentrations of recombinant emerin (residues 1–222). Both gelsolin and actin were obtained from Cytoskeleton (catalog #HPG5 and #AKL95, respectively). Gelsolin–actin seeds were made exactly as described (Blanchoin et al. 2000). Alternatively, 2 μM G-actin was polymerized in the absence of gelsolin for 2 h. Polymerized filaments were then incubated with or without 100 nM CapZ. Subsequent polymerization and depolymerization assays were assayed as described above.

Actin polymerization in the presence or absence of emerin was monitored by fluorescent microscopy, as described (Blanchoin et al. 2000; Amann and Pollard 2001). Samples were diluted 1:500–1:1000, viewed on a Nikon Eclipse E600W microscope (Nikon, Tokyo, Japan), and images were captured with a Q-imaging Retiga Exi CCD camera (Q Imaging, Burnaby, British Columbia, Canada) using IPLab (version 3.9.2) attached to an Apple G5 computer. Images were converted to TIFF images and lengths of filaments were measured in Photoshop version 7.0 (Adobe Systems, San Jose, California, United States).

These studies were funded by the National Institutes of Health (F32 GM067397 and T32 HL07227 to JMH and R01 GM48646 to KLW) and the American Heart Association Scott B. Deutschman Memorial Research Award (KLW). We are grateful to Dyche Mullins, Tom Pollard, and Doug Robinson for advice. We thank the National Cell Culture Center for providing the HeLa cell pellets. Protein identification was done by the AB Mass Spectrometry/Proteomics Facility at Johns Hopkins School of Medicine (http://www.hopkinsmedicine.org/msf) with support from National Center for Research Resources shared instrumentation grant 1S10-RR14702, the Johns Hopkins Fund for Medical Discovery, and the Institute for Cell Engineering.


Conflicts of interest. The authors have declared that no conflicts of interest exist.


Author contributions. JMH conceived and designed the experiments. JMH performed the experiments and analyzed the data. AKK and KLW contributed reagents/materials/analysis tools. JMH and KLW wrote the paper. KLW was the principal investigator.

Academic Editor: Marc W. Kirschner, Harvard University

Citation: Holaska JM, Kowalski AK, Wilson KL (2004) Emerin caps the pointed end of actin filaments: Evidence for an actin cortical network at the nuclear inner membrane. PLoS Biol 2(9): e231.

Abbreviations
Arpactin-related protein

BAFbarrier-to-autointegration factor

BSAbovine serum albumin

EDMDEmery-Dreifuss muscular dystrophy

F-actinfilamentous actin

GCLgerm cell-less

SRspectrin-repeat
==== Refs
References
Amann KJ  Pollard TD   The Arp2/3 complex nucleates actin filament branches from the sides of preexisting filaments Nat Cell Biol 2001 3 306 310 11231582 
Bengtsson L  Wilson KL   Multiple and surprising new functions for emerin, a nuclear membrane protein Curr Opin Cell Biol 2004 16 73 79 15037308 
Bennett V  Baines AJ   Spectrin and ankyrin-based pathways: Metazoan inventions for integrating cells into tissues Physiol Rev 2001 81 1353 1392 11427698 
Bettinger BT  Gilbert DM  Amberg DC   Opinion: Actin up in the nucleus Nat Cell Biol 2004 5 410 415 
Bione S  Small K  Aksmanovic VM  D'Urso M  Ciccodicola A    Identification of new mutations in the Emery-Dreifuss muscular dystrophy gene and evidence for genetic heterogeneity of the disease Hum Mol Genet 1995 4 1859 1863 8595407 
Blanchoin L  Pollard TD  Mullins RD   Interactions of ADF/cofilin, Arp2/3 complex, capping protein and profilin in remodeling of branched actin filament networks Curr Biol 2000 10 1273 1282 11069108 
Bonne GYR  Beroud C  Boriani G  Brown S  de Visser M    108th ENMC International Workshop, 3rd Workshop of the MYO-CLUSTER project: EUROMEN, 7th International Emery-Dreifuss Muscular Dystrophy (EDMD) Workshop, 13–15 September 2002, Naarden, The Netherlands Neuromuscul Disord 2003 13 508 515 12899879 
Burke B  Stewart CL   Life at the edge: The nuclear envelope and human disease Nat Rev Mol Cell Biol 2002 3 575 585 12154369 
Burtnick LD  Robinson RC  Choe S   Structure and function of gelsolin Results Probl Cell Differ 2001 32 201 211 11131832 
Clark TG  Rosenbaum JL   An actin filament matrix in hand-isolated nuclei of X. laevis  oocytes Cell 1979 18 1101 1108 574804 
Clements L  Manilal S  Love DR  Morris GE   Direct interaction between emerin and lamin A Biochem Biophys Res Commun 2000 267 709 714 10673356 
Cohen M  Lee KK  Wilson KL  Gruenbaum Y   Transcriptional repression, apoptosis, human disease and the functional evolution of the nuclear lamina Trends Biochem Sci 2001 26 41 47 11165516 
Cooper JA  Schafer DA   Control of actin assembly and disassembly at filament ends Curr Opin Cell Biol 2000 12 97 103 10679358 
Dahl KN  Kahn SM  Wilson KL  Discher DE   The nuclear envelope lamina network has elasticity and a compressibility limit suggestive of a molecular shock absorber J Cell Sci 2004 In press 
Dechat T  Vlcek S  Foisner R   Review: Lamina-associated polypeptide 2 isoforms and related proteins in cell cycle-dependent nuclear structure dynamics J Struct Biol 2000 129 335 345 10806084 
Delaunay J   Molecular basis of red cell membrane disorders Acta Haematol 2002 108 210 218 12432217 
Djinovic-Carugo K  Gautel M  Ylanne J  Young P   The spectrin repeat: A structural platform for cytoskeletal protein assemblies FEBS Lett 2002 513 119 123 11911890 
dos Remedios CG  Chhabra D  Kekic M  Dedova IV  Tsubakihara M    Actin binding proteins: Regulation of cytoskeletal microfilaments Physiol Rev 2003 83 433 473 12663865 
Dreger CK  Konig AR  Spring H  Lichter P  Herrmann H   Investigation of nuclear architecture with a domain-presenting expression system J Struct Biol 2002 140 100 115 12490158 
Emery AE   Emery-Dreifuss muscular dystrophy—A 40 year retrospective Neuromuscul Disord 2000 10 228 232 10838246 
Fairley E  Kendrick-Jones J  Ellis J   The Emery-Dreifuss muscular dystrophy phenotype arises from aberrant targeting and binding of emerin at the inner nuclear membrane J Cell Sci 1999 112 2571 2582 10393813 
Fidzianska A  Hausmanowa-Petrusewicz I   Architectural abnormalities in muscle nuclei: Ultrastructural differences between X-linked and autosomal dominant forms of EDMD J Neurol Sci 2003 210 47 51 12736087 
Forero C  Wasserman M   Isolation and identification of actin-binding proteins in Plasmodium falciparum  by affinity chromatography Mem Inst Oswaldo Cruz 2000 95 329 337 10800190 
Fowler VM   Capping actin filament growth: Tropomodulin in muscle and nonmuscle cells Soc Gen Physiol Ser 1997 52 79 89 9210222 
Fowler VM  Greenfield NJ  Moyer J   Tropomodulin contains two actin filament pointed end-capping domains J Biol Chem 2003 278 40000 40009 12860976 
Gonsior SM  Platz S  Buchmeier S  Scheer U  Jockusch BM    Conformational difference between nuclear and cytoplasmic actin as detected by a monoclonal antibody J Cell Sci 1999 112 797 809 10036230 
Gruenbaum Y  Lee KK  Liu J  Cohen M  Wilson KL   The expression, lamin-dependent localization and RNAi depletion phenotype for emerin in C. elegans 
 J Cell Sci 2002 115 923 929 11870211 
Haraguchi T  Koujin T  Segura-Totten M  Lee KK  Matsuoka Y    BAF is required for emerin assembly into the reforming nuclear envelope J Cell Sci 2001 114 4575 4585 11792822 
Haraguchi T  Holaska JM  Yamane M  Wilson KL  Hiraoka Y   Emerin binding to Btf, a death-promoting transcriptional repressor, is disrupted by a missense mutation that causes Emery-Dreifuss muscular dystrophy Eur J Biochem 2004 271 1035 1045 15009215 
Holaska JM  Wilson KL  Mansharamani M   The nuclear envelope, lamins and nuclear assembly Curr Opin Cell Biol 2002 14 357 364 12067659 
Holaska J  Lee K  Kowalski A  Wilson K   Transcriptional repressor germ cell-less (GCL) and barrier-to-autointegration factor (BAF) compete for binding to emerin in vitro J Biol Chem 2003 278 6969 6975 12493765 
Johnson BR  Nitta R  Frock RL  Mounkes L  Barbie DA    A-type nuclear lamins regulate retinoblastoma protein function by promoting correct subnuclear localization and providing protection from proteasomal degradation Proc Natl Acad Sci U S A 2004 101 9677 9682 15210943 
Krauss SW  Heald R  Lee G  Nunomura W  Gimm JA    Two distinct domains of protein 4.1 critical for assembly of functional nuclei in vitro J Biol Chem 2002 277 44339 44346 12171917 
Lattanzi G  Cenni V  Marmiroli S  Capanni C  Mattioli E    Association of emerin with nuclear and cytoplasmic actin is regulated in differentiating myoblasts Biochem Biophys Res Commun 2003 303 764 770 12670476 
Lee KK  Wilson KL   All in the family: Evidence for four new LEM-domain proteins Lem2 (NET-25), Lem3, Lem4 and Lem5 in the human genome. In: Hutchinson C, Bryant J, Evans D, editors. Communication and gene regulation at the nuclear envelope 2004 Oxford (United Kingdom) BIOS Scientific In press 
Lee KK  Gruenbaum Y  Spann P  Liu J  Wilson KL   
C. elegans  nuclear envelope proteins emerin, MAN1, lamin, and nucleoporins reveal unique timing of nuclear envelope breakdown during mitosis Mol Biol Cell 2000 11 3089 3099 10982402 
Lee KK  Haraguchi T  Lee RS  Koujin T  Hiraoka Y    Distinct functional domains in emerin bind lamin A and DNA-bridging protein BAF J Cell Sci 2001 114 4567 4573 11792821 
Lin F  Blake DL  Callebaut I  Skerjanc IS  Holmer L    MAN1, an inner nuclear membrane protein that shares the LEM domain with lamina-associated polypeptide 2 and emerin J Biol Chem 2000 275 4840 4847 10671519 
Manilal S  Nguyen TM  Sewry CA  Morris GE   The Emery-Dreifuss muscular dystrophy protein, emerin, is a nuclear membrane protein Hum Mol Genet 1996 5 801 808 8776595 
Markiewicz EDT  Foisner R  Quinlan RA  Hutchison CJ   Lamin A/C binding protein LAP2alpha is required for nuclear anchorage of retinoblastoma protein Mol Biol Cell 2002 13 4401 4413 12475961 
McMahon LW  Walsh CE  Lambert MW   Human alpha spectrin II and the Fanconi anemia proteins FANCA and FANCC interact to form a nuclear complex J Biol Chem 1999 274 32904 32908 10551855 
Mislow JMK  Holaska JM  Kim MS  Lee KK  Segura-Totten M    Nesprin-1α self-associates and binds directly to emerin and lamin A in vitro FEBS Lett 2002a 525 135 140 12163176 
Mislow JMK  Kim MS  Davis DB  McNally EM   Myne-1, a spectrin repeat transmembrane protein of the myocyte inner nuclear membrane, interacts with lamin A/CJ Cell Sci 2002b 115 61 70 
Mullins RD  Heuser JA  Pollard TD   The interaction of Arp2/3 complex with actin: Nucleation, high affinity pointed end capping, and formation of branching networks of filaments Proc Natl Acad Sci U S A 1998 95 6181 6186 9600938 
Mullins RD  Pollard TD   Structure and function of the Arp2/3 complex Curr Opin Struct Biol 1999 9 244 249 10322212 
Nagano A  Koga R  Ogawa M  Kurano Y  Kawada J    Emerin deficiency at the nuclear membrane in patients with Emery-Dreifuss muscular dystrophy Nat Genet 1996 12 254 259 8589715 
Nili E  Cojocaru GS  Kalma Y  Ginsberg D  Copeland NG    Nuclear membrane protein LAP2beta mediates transcriptional repression alone and together with its binding partner GCL (germ-cell-less) J Cell Sci 2001 114 3297 3307 11591818 
Nowak G  Pestic-Dragovich L  Hozak P  Philimonenko A  Simerly C    Evidence for the presence of myosin I in the nucleus J Biol Chem 1997 272 17176 17181 9202039 
Offterdinger M  Schofer C  Weispoltshammer K  Grunt TW   c-erb-3: A nuclear protein in mammary epithelial cells J Cell Biol 2002 157 929 939 12045181 
Olave IA  Reck-Peterson SL  Crabtree GR   Nuclear actin and actin-related proteins in chromatin remodeling Annu Rev Biochem 2002 71 755 781 12045110 
Östlund C  Worman H   Nuclear envelope proteins and neuromuscular diseases Muscle Nerve 2003 27 393 406 12661041 
Pederson T  Aebi U   Actin in the nucleus: What form and what for? J Struct Biol 2002 140 3 9 12490148 
Percipalle P  Fomproix N  Kylberg K  Miralles F  Bjorkroth B    An actin-ribonucleoprotein interaction is involved in transcription by RNA polymerase II Proc Natl Acad Sci U S A 2003 100 6475 6480 12743363 
Pestic-Dragovich L  Stojiljkovic L  Philimonenko AA  Nowak G  Ke Y    A myosin I isoform in the nucleus Science 2000 290 337 341 11030652 
Rando OJ  Zhao K  Janmey P  Crabtree GR   Phosphatidylinositol-dependent actin filament binding by the SWI/SNF-like BAF chromatin remodeling complex Proc Natl Acad Sci U S A 2002 99 2824 2829 11880634 
Sakaki M  Koike H  Takahashi N  Sasagawa N  Tomioka S    Interaction between emerin and nuclear lamins J Biochem (Tokyo) 2001 129 321 327 11173535 
Sasseville AM  Langelier Y   In vitro interaction of the carboxy-terminal domain of lamin A with actin FEBS Lett 1998 425 485 489 9563518 
Scheer U  Hinssen H  Franke WW  Jockusch BM   Microinjection of actin-binding proteins and actin antibodies demonstrates involvement of nuclear actin in transcription of lampbrush chromosomes Cell 1984 39 111 122 6386181 
Segura-Totten M  Wilson KL   BAF: Roles in chromatin, nuclear structure and retrovirus integration Trends Cell Biol 2004 14 261 266 15130582 
Segura-Totten M  Kowalski AM  Craigie R  Wilson KL   Barrier-to-autointegration factor: Major roles in chromatin decondensation and nuclear assembly J Cell Biol 2002 158 475 485 12163470 
Wang X  Xu S  Rivolta C  Li LY  Peng G    Barrier to autointegration factor interacts with the cone-rod homeobox and represses its transactivation function J Biol Chem 2002 277 43288 43300 12215455 
Wilkinson FL  Holaska JM  Sharma A  Malinal SB  Holt I    Emerin interacts in vitro with the splicing-associated factor, YT521-B Eur J Biochem 2003 270 2459 2466 12755701 
Wilson KL   The nuclear envelope, muscular dystrophy and gene expression Trends Cell Biol 2000 10 125 129 10740265 
Zastrow MS  Vlcek S  Wilson KL   Proteins that bind A-type lamins: Integrating isolated clues J Cell Sci 2004 117 979 989 14996929 
Zhang Q  Skepper JN  Yang F  Davies JD  Hegyi L    Nesprins: A novel family of spectrin-repeat-containing proteins that localize to the nuclear membrane in multiple tissues J Cell Sci 2001 114 4485 4498 11792814 
Zhen YY  Libotte T  Munck M  Noegel AA  Korenbaum E   NUANCE, a giant protein connecting the nucleus and actin cytoskeleton J Cell Sci 2002 115 3207 3222 12118075 
Zheng R  Ghirlando R  Lee MS  Mizuuchi K  Krause M    Barrier-to-autointegration factor (BAF) bridges DNA in a discrete, higher-order nucleoprotein complex Proc Natl Acad Sci U S A 2000 97 8997 9002 10908652

