Mechanical properties of the nuclear envelope have implications for cell and nuclear architecture as well as gene regulation. Using isolated Xenopus oocyte nuclei, we have established swelling conditions that separate the intact nuclear envelope (membranes, pore complexes and underlying lamin filament network) from nucleoplasm and the majority of chromatin. Swelling proves reversible with addition of high molecular mass dextrans. Micropipette aspiration of swollen and unswollen nuclear envelopes is also reversible and yields a network elastic modulus, unaffected by nucleoplasm, that averages 25 mN/m. Compared to plasma membranes of cells, the nuclear envelope is much stiffer and more resilient. Our results suggest that the nuclear lamina forms a compressed network shell of interconnected rods that is extensible but limited in compressibility from the native state, thus acting as a `molecular shock absorber'. In light of the conservation of B-type lamins in metazoan evolution, the mechanical properties determined in this investigation suggest physical mechanisms by which mutated lamins can either destabilize nuclear architecture or influence nuclear responses to mechanical signals in Emery-Dreifuss muscular dystrophy, cardiomyopathy, progeria syndromes (premature `aging') and other laminopathies.
Cells respond to mechanical cues from their environment in part through changes in gene expression. Compression-induced changes in the shapes of chondrocyte nuclei, for example, correlate with changes in cartilage composition and density (Guilak, 1995). Endothelial cells likewise modify gene expression in response to shear stress (Papadaki and Eskin, 1997). In endothelial cells, physical force is transmitted via a pathway from integrins to the cytoskeleton to the nuclear envelope and nuclear interior (Maniotis et al., 1997). The nature of this proposed mechanical coupling, particularly across the nuclear envelope, is not yet understood. Models suggest `bridging' interactions by conserved spectrin-repeator Sun-domain-containing proteins situated in both membranes of the nuclear envelope (Lee et al., 2002; Zhang et al., 2002; Zhen et al., 2002). In somatic cells, most nuclear inner membrane proteins are anchored to lamins, which form stable filament networks near the inner nuclear membrane and in the nuclear interior (Gruenbaum et al., 2003). The nuclear envelope also has physical connections to chromatin via lamins, which bind directly to chromatin (Moir et al., 1995) and to chromatin-binding nuclear membrane proteins (Burke and Stewart, 2002; Gant and Wilson, 1997). Although the exact mechanisms governing force transduction are unknown, nuclear envelope stiffness is a hypothesized regulator of the effects of cellular forces on chromatin or gene expression.
Although the typical cell nucleus is spheroidal, some differentiated cells undergo dramatic but regulated changes in nuclear morphology. Examples include multi-lobed nuclei in neutrophils (Yabuki et al., 1999) and elongated nuclei in spermatids (Dadoune, 2003). Aberrations in nuclear morphology, more than cell morphology, also identify cancerous tissue (Bissell et al., 1999). At least eight diseases have been linked to mutations in A-type lamins or lamin-binding proteins of the nuclear inner membrane. These diseases include Emery-Dreifuss muscular dystrophy, dilated cardiomyopathy with conduction system disease (Burke and Stewart, 2002), and Hutchinson-Gilford progeria syndrome, a devastating `premature aging' syndrome (De Sandre-Giovannoli et al., 2003; Eriksson et al., 2003). Heterozygous mutations in laminin B receptor (LBR), a membrane protein that binds lamin B, cause the dominant Helger-Puet anomaly of white blood cell nuclear shape; homozygous defects in LBR are linked to bone and cartilage disorders and developmental delay (Hoffmann et al., 2002). Proposed disease mechanisms include defects in tissue-specific gene expression coupled to mechanical weakness of the nuclear envelope (Morris, 2001; Wilson, 2000; Zastrow et al., 2004).
The Xenopus oocyte (XO) nucleus has a diameter about twenty times larger than typical somatic nuclei, but its chromatin content is similar. Thus the mechanical properties of the XO nuclear envelope are less influenced by chromatin. Closely associated with the XO nuclear envelope is a meshwork of primarily B-type lamin filaments (Lourim and Krohne, 1993), which are conserved in metazoan evolution and essential for mammalian cell viability (Cohen et al., 2001; Harborth et al., 2001). Lamins have central roles in nuclear architecture and nuclear integrity (Gruenbaum et al., 2003), and nuclei assembled in lamin-depleted Xenopus egg extracts are extremely fragile (Newport et al., 1990). Structurally, lamin filaments in the XO nucleus form a square-mesh lattice (Aebi et al., 1986) that anchors nuclear pore complexes (NPCs) (Davis, 1995; Stoffler et al., 2003). NPCs allow free diffusion of solutes and small molecules, but regulate the passage of molecules larger than 40 kDa (Suntharalingam and Wente, 2003).
In this study XO nuclear envelopes were manipulated by osmotic swelling with high molecular mass dextrans and micropipette aspiration to determine the elastic properties of the nuclear envelope, independent of the nucleoplasm. We show that given sufficient input force, the lamina network can expand uniformly to nearly twice its original surface area. Our findings provide the first quantitative measures of nuclear envelope mechanics and suggest that the native state of the XO nuclear lamina is a relatively compressed network of nearly rigid rods.
Materials and Methods
Xenopus oocyte collection and nuclear isolation
Fresh Xenopus laevis oocytes were obtained from sectioned ovaries (kind gifts from J. Foskett) obtained using an approved animal protocol (Jiang et al., 1998). Oocytes were separated and manually defolliculated using forceps. To obtain nuclei we dissected oocytes at the equator between the animal and vegetal hemispheres and manually teased the nucleus away from the cytoplasm into buffered isolation medium [BIM: 140 mM KCl (Fisher Scientific, Pittsburgh, PA, USA), 10 mM Hepes (Sigma-Aldrich, St Louis, MO, USA) pH 7.3 by KOH]. Various concentrations of 518 kDa dextran or 500 kDa dextran sulfate (Sigma-Aldrich) were added to BIM to control the degree of nuclear swelling during isolation. Other reagents such as sucrose, MgCl2, CaCl2, NiCl2 and other heavy metals or divalent salts (Fisher Scientific) were included in BIM for specific experiments. Swollen nuclei were also exposed to various concentrations of Triton X-100 (Sigma-Aldrich) in BIM. Nuclei were visualized at room temperature with either a Nikon TE300 inverted microscope or Nikon SMZ1500 stereoscopic zoom microscope with a beam splitter and attached video-rate camera, JE8242 (Javelin Systems, Torrance, CA, USA).
Isolated nuclei were aspirated into 197 or 278 μm inner diameter glass pipettes (Drummond Scientific, Broomall, PA, USA) attached to a water-filled manometer with reservoirs of adjustable height. Pressures ranging from 1 to 40 mm water were applied with a combination of syringe suction and reservoir height adjustment via micrometer, and pressures were measured with a pressure transducer (Validyne, Northridge, CA, USA).
Fluorescent labeling of nucleoplasm, lamina and membrane
Isolated nuclei were incubated with YO-PRO, an intercalating monomeric cyanine reagent that stains double and single stranded DNA and RNA (Molecular Probes, Eugene, OR, USA). The nuclear membrane was visualized using the lipophilic dye FM4-64 (Molecular Probes). The nuclear lamina was labeled by first exposing isolated nuclei to a 0.1% solution of Triton X-100 in BIM for 30 minutes and then incubating with murine anti-lamin B2 antibody X223 (Lourim et al., 1996) followed by TRITC-conjugated anti-mouse antibody (Sigma-Aldrich). Fluorescence was visualized at room temperature on a Nikon TE300 inverted microscope with a 4× (0.13 NA) objective using a liquid nitrogen cooled CCD camera (Roper Scientific, Trenton, NJ, USA).
Video images were captured from standard videocassette recorder tape using Scion Image (Scion Corporation, Frederick, MD). Lengths and areas of captured digital images were measured using Scion Image.
Model of osmotic swelling as Poiseuille flow
The movement of buffer through NPCs, which caused nuclear envelope swelling, was modeled as Poiseuille flow of BIM (or water) of viscosity η through rigid pores in a continuous membrane. We assumed that diffusion of water across the two lipid bilayers was insignificant compared to flow through the pores. The number of pores per area p, the inner pore radius a (Akey, 1991) and the pore length l (Mazzanti et al., 2001) were obtained from the Xenopus oocyte literature. From these numbers, a filtration coefficient ϕ was determined as:
The filtration coefficient, which represents the resistance of the membrane to flow, was related to the volume flux of water JV and the driving force. The driving force was defined as the balance between an effective osmotic pressure Π and the resisting pressure ΔP due to the membrane:
Initially, the trans-membrane ΔP was assumed to be negligible and the flux was proportional only to ϕ and Π. The osmotic pressure, which represents a solute gradient across the membrane, stayed relatively constant throughout swelling since the nucleoplasm did not expand or disperse in the entrained isolation media. When swelling reached equilibrium there was no influx (JV is zero and ΔP is equal to Π), and the final effective membrane pressure ΔP was calculated as a function of the filtration coefficient ϕ and the initial volume influx JV, init or (dV/dt)init/SAinit, where (dV/dt)init is the change in volume with time SAinit is the initial surface area. The volume influx was determined as a geometric function of the kinetic data obtained from the initial phase of swelling (see Results):
The Law of Laplace relates membrane tension to trans-membrane pressure and radius of the nucleus r.
The effective strain on the nuclear envelope was determined from the final area expansion αfinal, which was defined as the area of the swollen nuclear envelope SAfinal less the area prior to swelling SAinit, divided by the initial area.
Swelling separates the nuclear envelope and lamins away from the nucleoplasm
When isolated into buffered isolation medium (BIM), Xenopus oocyte (XO) nuclei remained spherical but swelled two-fold in diameter (Fig. 1A,B; see Materials and Methods). The nuclear lamina co-expanded with the membranes, as shown by indirect immunofluorescence staining of both unswollen and swollen nuclei (Fig. 1A,B). The apparently equal dilation at all points on the envelope was consistent with a relatively homogeneous, two-dimensional network of peripheral lamin filaments. In contrast, the nucleoplasm and chromatin did not swell, but remained the same size in both swollen and unswollen nuclei (Fig. 1A,B), probably due in part to polymerization of nuclear actin upon isolation of nuclei in BIM buffer (Fig. 1C). Polymerization of actin within XO nuclei after extraction has been reported previously (Pederson and Aebi, 2002). When visualized as either an off-center phase-dense body in bright-field microscopy, or using the fluorescent DNA/RNA dye YO-PRO, the bulk of nucleoplasmic actin and chromatin typically remained attached to one side of the swollen envelope. There was little or no evidence of envelope distortion at the attachment region, and it would appear that the site of attachment was random. A small fraction of YO-PRO-stained DNA/RNA remained associated with the swollen envelope.
Swelling was due to the entry of phase-light buffer. We concluded that the nuclear envelope remained largely intact, and buffer entered through NPCs because, as shown below, swelling was reversible by changing the extracellular solutes. Furthermore, small particles in the buffer-filled space between nuclear envelope and nucleoplasm exhibited Brownian motion very similar to particles outside the nucleus (data not shown). These buffer-filled, swollen nuclei were well suited to evaluating the mechanical properties of the nuclear envelope and its associated lamina network, largely independent of nuclear contents.
Micropipette aspiration of single nuclei
We labeled nuclear membranes with a fluorescent dye, and then aspirated the nuclear envelope, stepwise, into a micropipette (of radius Rp). This experiment demonstrated that the XO nuclear envelope was extremely deformable (Fig. 2). Small lipid vesicles and granules bound to the envelope were also labeled. When aspirated, the swollen nuclear envelope deformed in response to the applied stress. The enclosed material, mostly buffer, offered little elastic resistance. The nuclear membranes wrinkled near the entrance of the micropipette (Fig. 2, white arrow). For large extension lengths (L) of membrane into the micropipette (L>Rp, as in Fig. 2 panels 4-9), these folds indicate that the nuclear envelope and lamina behave as a solid-like network that resists membrane extension, since a fluid membrane under such pressures would smooth to relax any imposed bending strain (Discher et al., 2002). By tracking the distance between intense fluorescence spots along the projection length, the membrane is seen to stretch locally as it was being pulled into the micropipette. Some of these fluorescent markers, including the pair bracketed in Fig. 2, reveal membrane stretching up to 300% from the point at which the surface markers became visibly distinct (Fig. 2, panel 5) up to complete aspiration of the nucleus (Fig. 2, panel 9).
Determination of the network elastic modulus by micropipette aspiration
The nucleoplasm was highly deformable in unswollen nuclei (Fig. 3A). Nevertheless to test whether nucleoplasm influenced the mechanical resistance of the envelope, we aspirated swollen nuclei starting at a point distant from the nucleoplasmic body (Fig. 3A). The resistance of the envelope and its attached lamina was the same whether we aspirated at positions distal to, or near, the nucleoplasmic body. Resistance was specifically determined from the imposed pressure and resulting deformation. The pressure applied by the micropipette imposes a membrane tension on the nuclear envelope. This tension is calculated from the Law of Laplace and plotted against the dimensionless extension length (L/Rp), yielding a straight line (Fig. 3B).
The slope of the tension versus (L/Rp) line gives an extensional stiffness, E (Evans and Skalak, 1980), which can also be referred to as a network elastic modulus as discussed below. Swollen and unswollen nuclei had very similar average Es of 24 mN/m and 28 mN/m, respectively (Table 1). The statistically identical Es for swollen and unswollen nuclei indeed demonstrate very little resistance from the nucleoplasm and suggest that XO nucleoplasm has very low intrinsic stiffness. Lastly, consistent with elastic deformation, rather than plastic or permanent deformation, aspiration of nuclear envelopes was reversible upon graded release of the pressure (triangles in Fig. 3B).
Reversible swelling of nuclei
Nuclear swelling was also reversible and, with later analysis (see the Discussion) indicative of a dilational elasticity. After swelling in BIM, nuclei were transferred into BIM containing 50 mg/ml high molecular mass (500 kDa) dextran. The dextrans returned swollen nuclei to their original unswollen size (Fig. 4A). Given that swelling was spherically symmetric (Fig. 1), relative changes in nuclear size can be reported in terms of a dimensionless increase (α) in envelope surface area (SA). Micropipette aspiration of `de-swollen' nuclei showed that the E of the nuclear envelope lamina network was unaffected by swelling, consistent again with a highly elastic structure.
As a further control, swollen nuclei were exposed to various concentrations of the non-ionic detergent Triton X-100. With increasing Triton concentrations, nuclear surface area decreased and there was obvious weakening of the membrane. Detergent treatment perforates nuclear membranes, allowing free entry of larger solutes and buffer. In 5% Triton the nuclear envelopes became ∼50% smaller as visualized by bright-field microscopy (Fig. 4A). Nuclei nonetheless maintained their spherical shape (data not shown), suggesting that the underlying lamina network remained intact. Interestingly, dextrans were more effective than Triton at restoring swollen nuclei to their original size. Dextran-induced changes in osmolarity may provide a physical driving force that restores the native state of the nuclear envelope lamina.
Rates of nuclear swelling were highly reproducible, indicative of rate-limiting fluid movement through NPCs. When swollen in BIM, nuclear surface area (α) increased at a constant rate for 300-400 seconds and reached equilibrium by 400-450 seconds (Fig. 4B). The rate and extent of nuclear swelling could be precisely controlled by varying the concentration of 500 kDa dextrans (extent shown in Fig. 4C). Visualization of fluorescently conjugated dextrans confirmed their exclusion from nuclei (data not shown), as expected from their large mass. These results confirmed that swelling was due to buffer entering the nucleus in osmotically unbalanced samples. Charge was not a factor, since swelling was blocked equally well by either neutral dextran or negatively charged dextran sulfate (Fig. 4C). In contrast, swelling was unaffected by small, neutral molecules such as sucrose, at concentrations equi-osmolar to dextrans or higher. Sucrose and similar molecules are commonly used to osmotically swell lipid vesicles (Rawicz et al., 2000). Their ineffectiveness here confirmed that swelling was due to fluid and solute transport through NPCs, not direct permeation through nuclear membranes. Divalent salts such as MgCl2, CaCl2 and NiCl2 also had no effect on nuclear swelling (data not shown). We concluded that dextrans, excluded from the nucleus by their large mass, suppress swelling by retaining co-solutes that help to osmotically balance large osmolytes (DNA and proteins) inside the nucleus. Interestingly, nuclei were not significantly compressed even at extremely high levels of dextran (Fig. 4C).
One obstacle to quantifying mechanical properties of the somatic nuclear envelope is its association with chromatin, which is mediated by envelope lamins themselves as well as other membrane proteins (Foisner, 2003; Gant and Wilson, 1997; Gruenbaum et al., 2003; Moir et al., 1995). Isolated chromosomes have a Young's modulus of approximately 5 kPa (Houchmandzadeh et al., 1997), which is similar to a stiff collagen gel and within the range of elasticities measured for whole nuclei in somatic cells [1-8 kPa (Caille et al., 2002; Guilak et al., 2000)]. To minimize chromatin contributions, it is possible to make nuclear envelope `ghosts' stripped of chromatin, but their preparation involves harsh chemical and physical manipulation (Graham and Rickwood, 1997) that may compromise nuclear envelope integrity. In contrast, the XO nuclear envelope has proved, in this work, to be readily accessible to manipulation and characterization.
Micropipette aspiration of swollen and unswollen nuclei gave similar values (∼25 mN/m) for the network elastic modulus, E. Thus the nucleoplasm of the XO nucleus is not stiff enough to contribute any resistance to aspiration. An easily deformable nucleoplasm is most likely unique to XO nuclei because of their extremely large size and low chromatin density. Swollen and unswollen XO nuclei appear to provide a powerful experimental system for studying the mechanics of intact nuclear envelope lamin networks.
Dilational elasticity from swelling XO nuclei
The initial kinetics of buffer flow into the nucleus during swelling (Fig. 4B) was used to estimate the elastic modulus for nuclear envelope dilation. From equations developed as a model for osmotic swelling (see Materials and Methods) we estimate that a fully swollen nucleus is under a significant tension (Fig. 5). The slope of the best fit line of tension versus α indicates an approximate dilation modulus, K, for the XO nuclear envelope of ∼390 mN/m (Fig. 5). Somewhat lower area dilation moduli are seen for single bilayers of pure lipid (150-250 mN/m) (Rawicz et al., 2000). For two closely apposed bilayers in a multilamellar vesicle, which better approximates the two bilayers of the nucleus, the apparent membrane modulus increased relative to a unilamellar vesicle by 1.6-fold (Kwok and Evans, 1981). This gives a range of lipid moduli close to our finding here for intact nuclear envelopes.
The similar values for the membrane dilation moduli of nuclear envelopes and double bilayer lipid vesicles might suggest that membrane properties dominate nuclear envelope swelling. However, large differences in apparent lysis strain revealed properties unique to the nuclear envelope. Red cell membranes and lipid vesicles are limited in area expansion by lipid-lipid or lipid-protein interactions and lyse when the surface area increases by no more than 5% (Evans et al., 1976). In striking contrast, our results show that the nuclear membrane surface area can expand by 60%. This difference might in part result from either the ability of NPCs to dilate (Danker and Oberleithner, 2000), changes in membrane curvature at nuclear pores, smoothing of superficial membrane folds or combinations thereof.
Nuclear envelope elastic moduli
Stretching, dilating and bending of a cell membrane are generally considered independent modes of membrane deformation. With the extensional modulus E representing membrane stretching in micropipette aspiration, and with K obtained from nuclear swelling assays, it should come as no surprise that these quantities are numerically different. The structural origins of E seem relatively clear. Not only was E the same for swollen and unswollen nuclei, which thus eliminates the contributions of nucleoplasm, membrane-associated protein networks are generally known to resist deformation during micropipette aspiration (Evans and Skalak, 1980). Thus, in our experiments E measures the resistance of the lamin network to both shear and dilation, thus making E an effective two-dimensional Young's modulus for the lamin network.
Comparisons with E in other membrane systems are insightful. The red blood cell membrane is an especially well studied lipid-plus-protein-network, with a widely reported E of ∼0.01 mN/m based largely on micropipette aspiration measurements (Mohandas and Evans, 1994). The nuclear lamina's E of ∼25 mN/m is about three orders of magnitude stiffer. The difference is undoubtedly due to the larger stiffness of proteins in the lamin network, since protein stiffness is linearly related to the network E (Boal, 2002). Protein filament networks in both systems have a similar density, but spectrin filaments, which constitute the red cell network, are significantly softer than lamins. For filamentous proteins, persistence length is a measure of stiffness, and spectrin has a small persistence length of just 2.5-10 nm (Discher et al., 1998; Lenormand et al., 2003). Electron microscopy images of lamins suggest that they are far stiffer (Aebi et al., 1986) with persistence lengths much longer than that of spectrin. Moreover lamins are type V intermediate filament proteins, and other intermediate filament proteins have persistence lengths measured to be in the range of 100-1000 nm (Hohenadl et al., 1999).
Although highly extensible (E) and dilatable (K), the XO nucleus was also found here to have clear limits to compressibility from its native state since high concentrations of dextran in the isolation buffer did not reduce nuclear size below the native state. The chromatin seems too sparse to limit compression and had no notable influence on E. Though the nucleoplasmic actin was polymerized (phalloidin stained) as an artifact of nuclear isolation, aspiration of the nucleus into a micropipette showed that the nucleoplasm contributes little if any resistance to compression. Still, there exists a possibility that the nucleoplasm resists compression from the dextran-induced driving forces. However, we think the nuclear lamina network is more likely to limit compressibility.
Nuclear lamina superstructure
The present understanding of nuclear lamina organization posits a two-dimensional network of orthogonal lamin filaments (Aebi et al., 1986). How might dilation and incompressibility occur in this structure? Square mesh networks are known to collapse locally in response to two-dimensional compressive forces (Tessier et al., 2003). Figure 6 shows Monte Carlo simulations of a square mesh network on two spheres of different diameters. Both networks have the same number of `nodes' or interconnects. Collapsed regions of the mesh are present in the smaller `native' state of the nucleus (Fig. 6, left), but these collapsed regions expand in a fully dilated, swollen nucleus (Fig. 6, right). Interestingly, deconvolution microscopy of interphase nuclei suggests lamins are organized non-homogeneously (Paddy et al., 1990), consistent with the model of the native state proposed from this work.
While locally collapsed network domains might intrinsically oppose further compression, `stabilizer' elements, such as nuclear actin and membrane proteins that bind actin/lamins, might also play active roles. There is growing evidence that actin has multiple roles in the nucleus (Fig. 1C) (Bettinger et al., 2004; Pederson and Aebi, 2002). Indeed, G-actin binds directly to A-type lamins (Sasseville and Langelier, 1998) and also to emerin, an integral membrane protein (Lattanzi et al., 2003). Emerin also binds directly to F-actin, and stabilizes F-actin by capping the pointed end in vitro (Holaska et al., 2004). Other plausible stabilizer elements include the nesprin family of integral nuclear membrane proteins, some isoforms of which bind actin, lamins and emerin (Mislow et al., 2002; Zhang et al., 2002; Zhen et al., 2002). Sun-domain proteins might also act as mechanical stabilizers (Lee et al., 2002). Other structures, including an actin-dependent network of filaments that link NPCs to intranuclear organelles in XO nuclei, do not include lamins (Kiseleva et al., 2004) and are unlikely to contribute to the native compressed state of the lamina. However, we speculate that NPC-linked structures might limit the extent of envelope expansion during swelling.
With or without putative `stabilizers', our findings support the idea of a natively compressed lamina. This collapsed network of stiff lamin rods appears to require significant energy to rearrange. Furthermore lamin filaments must be both stiff and linearly elastic over a wide range of strains in order to explain our linear aspiration results for both swollen and unswollen nuclei. Given that lamins also interact with chromatin (Gruenbaum et al., 2003), our findings have intriguing implications for the lamin-dependence of gene expression (Spann et al., 2002) if, as we propose, the lamina network acts as a `molecular shock-absorber' with elastic extensibility but a clear limit to compressibility from the native state. Given the high conservation of B-type lamins throughout metazoan evolution (Herrmann et al., 2003), the properties of XO nuclei revealed here are likely to be widely applicable to somatic nuclei. A nearly incompressible (though dilatable) envelope would help protect chromatin and nuclear domains from compressive stresses. The proposed `compressed' native state of the nuclear lamina network would also allow for significant strength and flexibility during shear and extension, conferring clear advantages to motile cells crawling through tissues as well as non-motile but actively deforming cells such as muscle cells. Finally, a `compressed' native state might facilitate the transmission of mechanical force signals across the nuclear envelope.
The authors thank J. K. Foskett and G. Dreyfuss at the University of Pennsylvania and members of their laboratories, M. Ferreri-Jacobia, D. Mak, N. Petrenko and J. Yong who supplied oocytes and helpful information. The authors also thank J. Gall at the Carnegie Institute, Baltimore, MD and M. Goulian at the University of Pennsylvania for helpful conversations. This work was funded by grants from the Whitaker Foundation (K.N.D.), the Muscular Dystrophy Association (K.L.W., D.E.D.) and an NIH-BRP grant (D.E.D.).
- Accepted June 14, 2004.
- © The Company of Biologists Limited 2004