Actomyosin stiffens the vertebrate embryo during crucial stages of elongation and neural tube closure

Dorsal elongation and neural tube closure during early development are inherently mechanical events, yet little is known about the mechanical principles that drive these movements. Morphogenetic movements are guided both by the application of forces that are generated in the embryo and the mechanical resistance that embryonic tissues exert in response to those forces. Embryonic tissues can resist forces through deformation of anatomical structures or by the molecular deformation of constituent materials(Wainwright et al., 1976). Additionally, anatomical structures within the embryo are themselves composed of physical materials that exhibit molecular deformation, but their contribution to the anatomical structure depends on how they are assembled into the structure. To identify the relative contribution of these two aspects of mechanical resistance requires the ability to measure tissue elasticity experimentally after structures in the embryo or the molecular interactions in materials the make up the embryo have been altered(Koehl, 1990).

The notochord, a defining anatomical feature of chordate embryos, dominates early morphogenesis as both a source of molecular signals and later as a mechanical structure. The notochord first acts as an organizing center-releasing growth factors to induce midline structures such as the neural notoplate (later the floorplate) and hypochord(De Robertis and Kuroda, 2004; Stemple, 2005). The second function of the notochord later in development is as a mechanical structure,providing a stiff `backbone' for muscle attachment in the swimming tadpole(Hoff and Wassersug, 2000; Wassersug, 1989). The earliest mechanical role of the notochord during morphogenesis occurs as notochord cells change shape and undergo mediolateral intercalation during convergent extension (Shih and Keller,1992), and then extends faster than somitic mesoderm as the dorsal axis elongates (Keller et al.,1989). Experimental embryology and computer simulations suggest that elongation of the notochord and the attached notoplate shape the vertebrate neural plate (Jacobson and Gordon, 1976). Other dorsal tissues, including the paraxial somitic mesoderm (Wilson et al.,1989) and the neural plate(Elul and Keller, 2000; Ezin et al., 2006), also contribute to dorsal extension.

In contrast to the clear-cut role for the notochord in patterning, there is qualitative support for and against a mechanical role of the notochord in dorsal elongation and neural tube closure. Malacinski and co-workers(Malacinski and Youn, 1981; Youn and Malacinski, 1981)demonstrated that Xenopus embryos without notochords could converge,extend and elongate nearly as well as embryos with notochords. Evidence supporting a mechanical role for the notochord in dorsal extension is found in both amphibian embryos (Kitchin,1949; Lehman and Ris,1938) and zebrafish mutants(Talbot et al., 1995), in which notochord tissues have been ablated, which produces embryos with shortened axes; however, these developmental defects may be due to losses of other dorsal tissues and failure to establish dorsal tissue architecture. The generation of force by the notochordal region is required to produce the elongation in the amphibian neural plate(Jacobson and Gordon, 1976). Failure of dorsal tissue elongation is thought to be one of the principle causes of neural tube defects in frog(Wallingford and Harland,2002), mouse (Copp et al.,2003; Ybot-Gonzalez et al.,2007) and humans (Kibar et al., 2007). These studies suggest an important mechanical role for dorsal mesodermal tissues, as they apply forces to the overlying neural plate,extend the dorsal axis and the close the neural tube without failing or buckling.

The extracellular matrix and the cytoskeleton are the most likely contributors to the molecular basis of tissue mechanics relevant to embryogenesis. ECM has been identified as a major mechanical contributor to tissue stiffness in echinoderm embryos(Davidson et al., 1999), adult tissues (Levental et al.,2007; Vincent,1990) and tumors (Paszek et al., 2005). Collagen type II fibers form some of the stiffest fibers identified in animals (Levental et al., 2007) and are a strong candidate for stiffening the notochord in later dorsal tissues (Adams et al.,1990). Likewise, fibrillin could also contribute to dorsal tissue stiffness analogous to its contribute to tissue stiffness in the lung(Waters et al., 2002), where fibrillin and elastin undergo cyclic extension and contraction and prevent viscoelastic creep (Findley et al.,1989) that would otherwise lead to rapidly failing `stretched-out'hearts and lungs. Laminin in its many heteromeric forms may also contribute to dorsal tissue stiffness (Georges et al.,2006). Each of these candidate ECM components, fibronectin(Lee et al., 1984), fibrillin(Skoglund et al., 2006) and laminin (Wedlich et al., 1989)are assembled into fibrils in frog dorsal tissues through these stages,suggesting that ECM contributes to tissue stiffness from mid- to late-gastrulation and beyond during neurulation.

Tissue mechanics crucially depend on factors such as Rho-GTPase that regulate the actomyosin cytoskeleton to direct cell migration, cell mechanics and assembly of the extracellular matrix(Paszek et al., 2005). Numerous studies have implicated the actin cytoskeleton as a major contributor to the mechanical properties of single cells (for a review, see Luby-Phelps, 2000). Additional crosstalk may couple cell-cell adhesion to actomyosin contractility in frog embryos (Kofron et al., 2002; Tao et al., 2007), where activated cadherin in cell-cell adhesions has been shown to control cortical F-actin assembly. In contrast to a growing consensus on the molecular factors that modulate dynamic F-actin assembly(Pollard and Borisy, 2003) and the mechanics of crosslinked F-actin gels in purified systems(Gardel et al., 2006; Janmey and Weitz, 2004),little is understood about the contribution of the actin cytoskeleton to mechanical properties of tissues.

In order to investigate the mechanical properties of embryonic tissues and isolate the relative contribution of anatomical structures and molecular composition to those properties, we use the frog embryo to microsurgically isolate dorsal tissues, manipulating both tissue architecture and molecular composition, and then measuring the time-varying stiffness (i.e. Young's modulus or mechanical resistance) (Janmey et al., 2007) along the anterior-posterior direction that results from these manipulations. We use standard engineering measures of material properties so that our results can be compared directly with theoretical and empirical studies conducted on a variety of biomaterials, structures and organisms (Koehl, 1990). In agreement with a previous study of Xenopus embryonic tissues(Moore et al., 1995), we find that dorsal isolates also behave like a conventional viscoelastic material capable of elastically supporting applied forces. We find dramatic increases in tissue stiffness at later stages and equally large differences between the stiffness of germ layers that form dorsal tissues. Using antisense morpholinos to block fibronectin fibril assembly and acute-acting actomyosin disrupting drugs, we find that embryonic tissue stiffness at these stages originates in the actin cortex rather than in the fibronectin extracellular matrix. We propose that dorsal tissues form an integrated composite structure that is dependent on actomyosin contraction that generates force and resists catastrophic failure to drive dorsal axis elongation and neurulation.

Xenopus laevis embryos are obtained by standard methods(Kay and Peng, 1991) and cultured to stage 16 (Nieuwkoop and Faber,1967), at which time dorsal axial and paraxial tissues (dorsal isolate) are microsurgically removed from the embryo(Fig. 1A). Dorsal isolates can be harvested from embryos from mid-gastrulation (stage 11.5) to neural tube stages (stage 22) and beyond. The dorsal isolate can be reproducibly subdivided even further along the mediolateral or dorsal ventral axis.

To visualize ECM and pMLC, tissue explants and whole embryos were fixed in either 3.7% formaldehyde or 3% TCA in PBS (see Davidson et al., 2004). Fibronectin fibrils were localized with mAb 4H2 directed against Xenopus fibronectin (1:500), fibrillin-2 fibrils with mAb JB3 (1:500;Developmental Studies Hybridoma Bank), laminin 1 with a rabbit pAb (1:100;cat# L-9393, Sigma) and activated myosin light chain (pMLC) with a rabbit pAb(1:200; ser19 phosphorylated; Cell Signaling Technology) and visualized with a rhodamine-conjugated goat anti-mouse or anti-rabbit IgG (Jackson ImmunoResearch Laboratory). To visualize F-actin in embryos, we fixed embryos using two media: either a modified Dent's fixative [replacing methanol with isopropanol (Dent et al.,1989; Munro and Odell,2002)] or 4% paraformaldehyde with 0.25% glutaraldehyde(Tao et al., 2007) in PBS. F-actin was visualized by incubating fixed tissues for 3 hours with bodipy-FL phallacidin (5U/ml PBST; Invitrogen). Following immunofluorescence or phallacidin staining whole embryos were bisected or cut into en face fragments, dehydrated in methanol or isopropanol (to preserve phallacidin binding to F-actin) and cleared in Murray's clear(Davidson and Wallingford,2005). Single optical sections and z-series of explants and whole embryo fragments were collected with a confocal laser scan head(SP5, Leica Microsystems) mounted on an inverted compound microscope (DMI6000,Leica Microsystems) using image acquisition software (LASAF, Leica Microsystems). Maximum projection and reslicing of z-series stacks,and collection of intensity profiles were carried out with ImageJ (v. 1.38,Wayne Rasband, NIH).

The nanoNewton force measurement device (nNFMD) has been described previously (Davidson and Keller,2007). Briefly, the device for measuring nanoNewton to microNewton scale forces was based on the earlier `Histowiggler' design(Moore et al., 1995). Prior to the start of an experimental run, the anterior or posterior face of a tissue explant is brought into contact with a sensitive force probe. An experimental run begins as the explant is moved into contact with the force probe and compressed by 15 to 20% of its original length. The explant is held in place for 180 seconds as a record of resistive force is collected. The experimental run concludes as the explant is moved away from the force probe and placed in fixative [MEMFA (Sive et al.,2000)]. Stiffness (i.e. Young's modulus) is calculated from the resistive force measured during the stress-relaxation test, the cross-sectional area of the explant measured from the fixed sample and the compressive strain observed from a time-lapse sequence of the stress-relaxation test.

Dorsal tissues isolated from the late gastrula stage embryo exhibit viscoelastic behavior like most `soft' biological tissues(Fig. 1D)(Wainwright et al., 1976). This tissue, referred to as `dorsal isolate', contains all three germ layers in their proper laminar context from the spinal cord/hindbrain boundary to approximately 200 μm from the blastopore. The transverse cross-section of the dorsal isolate is consistent along its anterior-posterior length. Germ layers are separated from each other by tissue interfaces that delineate axial notochord from surrounding tissues, and paraxial pre-somitic mesoderm from neural ectoderm and endoderm (Fig. 1B).

The first step in quantifying the viscous and elastic properties of these tissues is to determine the time (t)-dependent Young's modulus, E(t). Throughout this study, we chose to compare the Young's modulus at the 180 second time-point in our compression test as E(180) represents the static mechanical properties of the embryo in vivo. We refer to E(180) throughout this study as tissue `stiffness', as this value has been obtained from an unconfined uniaxial compression test. The Young's modulus at earlier time-points reflects the changing viscous response of the tissue to a suddenly applied strain, as well as variable loading conditions such as stiction and compliance of our force transducer. The stiffness immediately after strain is applied, E(0), after 90 seconds, E(90), and after 180 seconds, E(180), along with standard deviations are provided in supplementary tables for each tissue measured [see Tables S1-S11 in the supplementary material; in addition we provide the parameters k<sub>1</sub> (Einf), k<sub>2</sub> (Esp), andη(viscosity) representing the Standard Linear Solid (SLS) Model(Findley et al., 1989) fitted using nonlinear regression techniques (NLREG ver. 3.2; Brentwood TN) for each tissue measured]. Statistical tests of significance of E(180) values were carried out pair-wise for sets of explants from each clutch with the non-parametric Mann-Whitney U-test(Sokal and Rohlf, 1994) using commercial software (SPSS v. 16; Chicago, IL). In the course of measuring the E(t) of embryonic tissues, we found several complications to fitting the data with the SLS model. First, the initial elastic response [E(t) for t<10 seconds] is significantly higher than the parameters describing the initial elastic response (k<sub>1</sub>+k<sub>2</sub>) from a best-fit SLS model (see Fig. S2B in the supplementary material). This high value may reflect static friction (i.e. stiction) of the tissue explant with the holder, or may indicate high frequency mechanical compliance of our force transducer. Second,we occasionally observe spontaneous mechanical contractions during compression tests (data not shown). These contractions appear as `bumps' 30 seconds to 1 minute in length in which the E(t) rises transiently and then falls back. These contractions are similar to spontaneous contractions observed with other mechanical test devices in our laboratory(von Dassow and Davidson,2009). Applying nonlinear regression to fit these traces results in low-quality SLS parameters that consistently underestimate the observed Young's modulus. In order to enable comparison of our data with those collected by other laboratories and for comparison with other materials, we include the nonlinear regression `best-fit' SLS model parameters of k₁, k₂, and η for all tests in Table S1-S11 (see supplementary material). We chose to use the measured E(180) rather than the SLS model-calculated k₁, because, as discussed above, fitted SLS model parameters introduce errors not found in measured data.

The dorsal tissues of the Xenopus embryo are made up of elements from the three germ layers (see Fig. S3A in the supplementary material):endoderm, neural and mesoderm, which is further subdivided into axial notochord, paraxial-medial somitic mesoderm and paraxial-lateral somitic mesoderm. Although we cannot mechanically test any single one of these tissues, we can construct tissue explants with duplicated structures, e.g. the two-notochord explants (Fig. 2A,2A′),or explants lacking particular structures, e.g. the notochord-less explant(Fig. 2A). We can use the composite nature of the explant, the measured stiffness (E), and the measured cross-sectional areas (A) to write a set of algebraic equations for the general form of the superposition principle(Christensen, 1991):E_total × A_total=Σ E_i ×A_i. The special case for the dorsal isolate (DI) is written:E_DI × A_DI=E_noto ×A_noto + E_{medial mesoderm} × A_{medial mesoderm} + E_{lateral mesoderm} × A_{lateral mesoderm} + E_neural × A_neural +E_endo × A_endo. This equation relates the dependence of the stiffness of the dorsal isolate (DI) (see Fig. S3A in the supplementary material) on its component parts: notochord (noto),paraxial-medial mesoderm (medial mesoderm), paraxial-lateral mesoderm (lateral mesoderm), neural plate (neural) and endoderm (endo) (see Fig. S3B in the supplementary material). Measured stiffness and cross-sectional areas from sets of explants can be written as sets of algebraic equations where we can`solve' for the mechanical contribution of the tissue that has been excised.

After early gastrulation movements, endoderm, mesoderm and neural ectoderm are tightly coupled into a dorsal block of tissue; this dorsal tissue then narrows mediolaterally and lengthens anteroposteriorly. As it is not possible to measure the mechanical properties of tissue in vivo, we instead measured the properties of the dorsal isolate (Fig. 1A) (see Materials and methods), an explant that contains all three germ layers in their proper laminar architecture and undergoes the morphogenetic movements of convergence and extension.

As biological tissues can exhibit a wide range of mechanical behaviors from rigid elastic to fluid-like and can actively change shape unpredictably in response to applied forces, we carried out a series of mechanical tests to determine whether standard techniques could be used to study dorsal isolates. First, we assessed the amount of strain present in the embryo as large amounts of pre-strain can significantly alter the interpretation of mechanical tests of stiffness (Zamir and Taber,2004b). To measure the degree of pre-strain in dorsal isolates, we collected images of the epithelial surface of whole embryos before microsurgery and compared those images with ones collected immediately after microsurgery. We measured strain in five embryo-explant pairs and have found the surface layer in dorsal isolates contract around 10% isometrically (see Fig. S1A-D in the supplementary material). Thus, some small degree of pre-strain is present in the embryo but we do not see large amounts of elastic recoil as observed when biological tissues within hydrostat-like structures,such as the embryonic heart, are removed from their micro-environment(Zamir and Taber, 2004a). We then determined the response of dorsal isolates to applied force, the linearity of their elastic response to that force, whether the response changes with repeated application of force and whether tissue became stiffer in response to higher applied forces. Single dorsal isolates at stage 16 were placed into the nanoNewton force measurement device (nNFMD)(Fig. 1C) and subjected to a nine-minute uniaxial compression test (see Fig. S1E in the supplementary material) demonstrate that the elastic resistance of dorsal isolates reaches steady-state by 180 seconds. Repeated series of five, 3-minute compression/1-minute relaxation tests (see Fig. S1F in the supplementary material) or five 1-minute steps of incrementally increasing strain (see Fig. S1G in the supplementary material) on the same dorsal isolate reveal that isolates do stiffen slightly with repeated cycles or increasing strain but remain viscoelastic.

Taken together, these mechanical tests demonstrate that dorsal tissues behave like conventional viscoelastic materials with an immediate viscous response to applied strain. Dorsal isolates are able to maintain a long term residual stiffness and their residual Young's modulus (hereafter referred to as `stiffness') can be reliably and reproducibly measured at the conclusion of a single 180-second unconfined uniaxial compression test. The goal of the remainder of this paper is to characterize stage- and tissue-specific variation in these tissues, and to test the role of the ECM and actomyosin cytoskeleton as contributing factors.

Since a previous study (Moore et al.,1995) found increasing stiffness in dorsal involuting marginal zone tissues early in gastrulation (between stages 10+ and 11.5) coincident with the onset of convergent extension, it is important to know whether tissues continue to stiffen during the subsequent rapid phase of narrowing and elongation that precedes neurulation. Dorsal isolates excised from whole embryos converge and extend at the same rate as the identical tissues in whole embryos and preserve the relationship between endoderm, mesoderm and neural ectoderm, as seen in whole embryos (Fig. 1B). To ensure that the same tissues were tested at both early and late stages, we prepared dorsal isolates at the end of gastrulation (stage 13)and aged them until co-cultured control embryos reached the defined stage(with the exception of the mid-gastrula stage isolate). Single explants were placed into the nanoNewton force measurement device(Fig. 1C) and subjected to a 180-second stress-relaxation protocol applied along the anteroposterior axis of the isolate (Fig. 1D). In preliminary studies, we found clutch-to-clutch variation in the mechanical properties of embryonic frog tissues (von Dassow and Davidson, 2009) (data not shown). To compensate for this variance, we typically measured and compared tissue stiffness changes with developmental stage within the same clutch. Previous measurements found a threefold increase in axial mesodermal tissue stiffness from 3 to 10 Pa from early gastrula to mid-gastrula stages(Moore et al., 1995). With the updated force measurement device, we found dorsal tissues continued to increase stiffness along the anteroposterior axis from 13 Pa to 85 Pa(Fig. 1E; representative data shown), an increase of more than sixfold from late gastrula (stage 13) to early neural tube stage (stage 22).

As the notochord is a central structure during gastrulation and neurulation, and undergoes substantial changes in architecture during this period, we decided to test the contribution of the notochord to the stiffness at early neurula stages (stage 16). To accomplish this, we microsurgically made dorsal isolates with two-notochords and zero-notochord(Fig. 2A). To make a dorsal isolate with two notochords, we recombined a left-half explant with notochord with a right-half explant with notochord. The medial edges of the left and right explants were then held in apposition with a glass coverslip fragment until they healed together, typically 30 to 60 minutes. Dorsal isolates without notochord were assembled from the two remaining lateral fragments. To control for the possibility that microsurgery altered the stiffness of dorsal isolates we also made `sham-operated' control dorsal isolates that were split axially and then re-combined. After healing, sham-operated controls,two-notochord and zero-notochord isolates retain the capacity for elongation and achieve rates similar to intact dorsal isolates(Fig. 2B). To confirm that the architecture of these explants contained properly positioned dorsal tissues,we carried out immunofluorescence to reveal fibrillar fibronectin and collected stacks of confocal sections of stained explants(Fig. 2C).

Using the nNFMD, we found two-notochord isolates had a twofold greater stiffness over zero-notochord dorsal isolates(Fig. 2D). We can compare the stiffness of these differently sized explants as stiffness is independent of the shape and size of the tissue (Koehl,1990). Close inspection of the tissue architecture of explants with two notochords revealed that the two notochords were separated from each other by a piece of notochord-adjacent paraxial mesoderm(Fig. 2C, asterisk). To rule out the effect of this small piece of notochord-adjacent paraxial mesoderm, we again prepared two-notochord and zero-notochord dorsal isolates cutting closer to the boundary between the notochord and somite(Fig. 2A′). Staining of these explants for fibronectin fibrils confirmed that we had completely removed the medial notochord-adjacent paraxial mesoderm from the two notochord isolates (Fig. 2C′). However, after removing the small amount of midline tissue, we found no significant differences in the stiffness between isolates with only two notochords and those without notochords(Fig. 2D′).

In order to rule out the effects of microsurgical recombination, we compared the stiffness of dorsal isolates cut from dorsalized embryos. Our strategy was to alter the early patterning of embryos to generate a larger notochordal field and then use these embryos to microsurgically excise dorsal isolates. We chose LiCl to dorsalize early embryos and measured the consistency and degree of dorsalization with the Dorso-Anterior Index [DAI(Stewart and Gerhart, 1990)]. By regulating the timing and limiting the dosage of LiCl to 0.3 M(Fig. 2E), we could consistently generate embryos with a DAI between 7 and 8(Fig. 2F). DAI 7/8 embryos have a single deep archenteron containing more than double the normal amount of notochord, as judged both by chordin expression in dorsal isolates(Fig. 2G), and fibronectin fibril localization in transverse confocal sections(Fig. 2H). Neither dorsalized embryos nor dorsal isolates made from them undergo much elongation (data not shown). Assured that dorsal isolates made from dorsalized embryos contained large quantities of notochord, we measured their axial stiffness and found that they did not differ significantly from control dorsal isolates made from the same clutch (Fig. 2I)(note: a single clutch from the first three clutches showed dorsalized tissue was stiffer and prompted a more extensive test of two additional clutches,none of which showed significant differences in stiffness). Thus, two independent methods of producing excess notochord in dorsal isolates failed to show any increase in stiffness. This surprising result prompted us to evaluate paraxial somitic mesoderm as a potential source of the stiffness of the dorsal isolate.

Paraxial mesoderm consists of medial and lateral tissues that have distinct fates in the tadpole (Keller,2000). As our first microsurgical attempts to make explants with double notochords instead revealed a stiff material lodged between the two notochords, we suspected the medial fragments from paraxial mesoderm made a larger contribution to overall stiffness. To test this hypothesis, we again prepared zero-notochord explants from stage 16 embryos (here referred to as LMML for Lateral-Medial-Medial-Lateral isolates) consisting of two complete arrays of paraxial mesoderm and compared them with isolates that contain only lateral mesoderm (LL, Lateral-Lateral isolates)(Fig. 3A). Both of these tissues are similar in appearance when stained for fibronectin fibrils(Fig. 3B); however, isolates containing medial somitic mesoderm are clearly identified by their expression of the prospective muscle marker XmyoD(Hopwood et al., 1989) (left and middle columns in Fig. 3C),which is absent in LL explants (right-most column in Fig. 3C). Furthermore, LL explants retain the elongation capacity of intact dorsal isolates(Fig. 3D). Confident of our ability to microsurgically separate lateral from medial paraxial tissues we made LL and medial-medial paraxial (MM) explants combining both left- and right-hand tissues from the same embryo(Fig. 3A). Measuring the axial stiffness of these tissues we found LL explants were significantly less stiff than MM explants (Fig. 3E). The higher stiffness of isolates containing medial notochord-adjacent paraxial mesoderm supports the hypothesis that it is not the notochord but rather notochord-adjacent paraxial tissues that contribute most to the axial stiffness of dorsal isolates at mid neural plate stages.

The MM explant consists of neural and endodermal tissues in addition to medial paraxial mesoderm; each of these tissues could contribute to the increased stiffness of the MM explant. To determine the contribution of these other tissues to axial stiffness, we made explants from intact dorsal isolates from which either endoderm or neural ectoderm had been removed(Fig. 4A). Each of these explants had been used in previous studies of segmentation(Wilson et al., 1989) where they were reported not to extend as well as intact dorsal isolates. We are able to prepare mesoderm-only explants; however, these explants are too thin to test in the nNFMD. Comparing the stiffness of no-neural dorsal isolates with control explants, we found they were not significantly different(Fig. 4B). By contrast, we found no-endoderm isolates were significantly stiffer than control dorsal isolates (Fig. 4C). This last result indicates that the endoderm is the weakest tissue in the dorsal isolate.

Stiffness measurements of reconstructed dorsal isolates reveal significant differences between tissues but cannot directly quantify the stiffness of any single tissue. By definition, the stress relaxation protocol reports a representative bulk time-dependent stiffness that is independent of the geometry of the dorsal isolate (Koehl,1990; Vincent,1990). Another limitation is due to sensitivity of the nNFMD;smaller pieces of tissue do not produce sufficient compression-resistant forces to determine their stiffness.

To determine the stiffness of each of the germ layers, we formulated a simple architectural model of the dorsal isolate. Based on superposition principles of composite structures(Christensen, 1991), we constructed a simple mechanical model of the dorsal isolate that allows an estimate of the stiffness of the component tissues of the dorsal isolate (see Materials and methods). In essence, each tissue contributes to the stiffness of the whole tissue explant according to its individual stiffness and to its contribution to the full transverse cross-sectional area of the tissue explant. Each component tissue is analogous to a spring and the full explant to a series of springs in parallel. Force applied to the anterior or posterior ends of the explant is distributed to all the springs to bring about the same degree of compression in each. In this way, we account for tissue mechanical properties that cannot themselves be isolated from their surroundings or are not compatible with the nNFMD. Qualitatively, we can consider the contribution of the various tissues to the stiffness of recombinant explants: (1) when we remove a relatively stiff tissue the resulting recombinant is less stiff; (2)when we remove an easily deformed tissue the resulting recombinant is stiffer;and (3) when we remove a tissue that is equal to the stiffness of the whole isolate we see no difference in stiffness. Using the composite model and comparing the stiffness of endoderm-free, neural-free, notochord-free,two-notochord, LiCl-notochord and LL versus MM explants, we conclude that there are three levels of tissue stiffness in the dorsal isolate. From this analysis, we rank the relative stiffness of the components of the dorsal isolate: (1) the endoderm is less stiff by an order of magnitude than the intact dorsal isolate (with stiffness between 2 and 11 Pa; 6 to 22% stiffness of the intact dorsal isolate); (2) the neural and notochord are equivalent to the stiffness of the intact dorsal isolate (between 40 and 60 Pa), and the paraxial mesoderm is stiffest (between 70 and 100 Pa; 140 to 170% of the stiffness of the intact dorsal isolate) and nearly twofold stiffer than the notochord or neural tissues. A more quantitative evaluation of specific tissue stiffness will require development of more spatially sensitive methods either by excising and testing smaller tissue explants, or by measuring spatial variation within intact explants or embryos. With our measurements of anatomical contributions to tissue stiffness, we turned to assessing the molecular contribution to tissue stiffness.

We suspected that fibronectin or other ECM components such as fibrillin and laminin were responsible for the stiffness of the paraxial mesoderm during dorsal elongation of the vertebrate embryo from mid-gastrulation through neurulation based on the pronounced assembly of fibronectin fibrils(Davidson et al., 2004) and deposition of fibrillin (Skoglund et al.,2006) around the paraxial mesoderm(Fig. 5A). To reduce the assembly of fibronectin ECM, we injected fertilized eggs with a mix of two antisense morpholinos directed against the 5′ start of the two fibronectin pseudoalleles expressed at these stages (6 μM each; FNMO). At these doses, FNMO has been previously shown to knock down fibronectin synthesis and its subsequent assembly into fibrils in Xenopus embryos(Davidson et al., 2006). Dorsal isolates from injected embryos at stage 13 showed no difference in stiffness from control uninjected embryos(Fig. 5B, left-most bars);furthermore, additional measurements of stiffness of stage 16 dorsal isolates showed FNMO-injected explants increased in tissue stiffness, just as control dorsal isolates (Fig. 5B,right-most bars). Each tissue explant tested was stained for fibronectin,demonstrating reduced assembly of fibronectin fibrils in FNMO-injected explants (Fig. 5C, top panel). Thus, dorsal isolate stiffness was unchanged after fibronectin fibrillogenesis was severely blocked.

As it has been suggested that assembly of other ECM proteins might depend on fibronectin fibril assembly (Sivakumar et al., 2006), we investigated whether we could also assess the contribution of fibrillin and laminin extracellular matrix to the stiffness of the dorsal isolate by mid-gastrula stages. Staining FNMO injected dorsal isolates revealed reduced fibrillin and laminin fibril assembly. Fibrillin shows strong reduction in the assembly in FNMO injected dorsal isolates(Fig. 5C, middle panel),whereas laminin assembly is both reduced and disorganized(Fig. 5C, bottom panel). Quantitative analysis of staining across the somite-notochord boundary demonstrates that, in addition to the 60% reduction of fibrillar fibronectin,fibrillin and laminin are reduced by 22% and 29%, respectively, in FNMO-injected dorsal isolates (Fig. 5C′). The collateral inhibition of fibrillin and laminin assembly suggests that fibrillar ECM may not contribute to tissue stiffness during mid-gastrulation and neurulation.

To test the contribution of the actin-based cytoskeleton to tissue stiffness, we prepared dorsal isolates and treated them with a panel of acute-acting drugs. Previous studies have shown that latrunculin B (latB)effectively de-polymerizes F-actin within whole embryos(Fig. 6A)(Benink and Bement, 2005; Lee and Harland, 2007) and Y27632, a Rho-kinase inhibitor, blocks myosin II activation(Maekawa et al., 1999; Narumiya et al., 2000). Both latB (incubated 20 minutes) and Y27632 (incubated for 60 minutes) reduced tissue stiffness by at least 50% in dose-dependent manners(Fig. 6B,C, respectively). Treatment with high doses of latB over long times causes tissue explants to irreversibly dissociate. LatB, like another F-actin depolymerizing drug(cytochalasin D), causes disassembly of the fibronectin matrix(Davidson et al., 2008) (data not shown). However, as reduced fibronectin knockdown does not alter tissue stiffness, we propose that the reduction in stiffness after treatment of dorsal isolates with 0.6 μM latB is due to reduced F-actin rather than to disruption of fibronectin fibrils. Combinations of both latB and Y27632 show that the effect of latB dominates and that reduced myosin II contractility makes little additional contribution to tissue stiffness(Fig. 6D). Two additional compounds have been reported to stabilize F-actin [jasplakinolide(Cramer, 1999)] or to increase myosin II contractility [calyculin A (Yam et al., 2007)]; however, neither jasplakinolide (up to 10 μM for 60 minutes) nor calyculin A (40 nM for 20 minutes) produced significant changes in tissue stiffness (see Table S8 in the supplementary material).

As disrupting actomyosin reduced tissue stiffness by nearly half, we wondered whether F-actin or active myosin II could account for the much larger differences in stiffness seen between stages or between germ layers. By confocal sectioning whole embryos stained with phallacidin to localize F-actin and whole embryos stained with an antibody that recognizes the phosphorylated serine of the myosin regulatory light chain (pMLC) to localize activated myosin II (Lee et al., 2006),we were able to resolve stage specific and germ layer-specific changes in actomyosin. Surprisingly, neither F-actin(Fig. 7A,B) nor pMLC(Fig. 7C) shows gross changes in intensity between stages 13 and 16 (embryos were fixed and stained together, and confocal sectioned with the identical settings). However, there are considerable changes in the organization of pre-somitic mesoderm into a dorsal and ventral leaflets. The interface between these two layers, known as the pre-myocoel (Keller,2000), shows increased levels of F-actin localization(Fig. 7B,C, arrowheads) that may indicate the assembly of a new load-bearing mechanical structure within the pre-somitic mesoderm. In contrast to the lack of F-actin differences between stages 13 and 16, the endoderm, the most compliant tissue in the embryo, shows dramatically lower levels of F-actin. We suspect that this does not simply reflect the larger size of endoderm cells (as large mesendoderm or bottle cells show intense F-actin, data not shown) but instead reflects repression of F-actin assembly within these cells. Thus, we may conclude that some, but by no means all, of the spatial and temporal variation in stiffness is due to absolute levels of F-actin or activation of myosin II.

In this paper, we have investigated the architectural or structural mechanical properties of dorsal tissues in the frog embryo. Using biophysical and bioengineering tools, we described the stage-dependent stiffening of axial tissues in the frog embryo from gastrula to neurula stages. Dorsal tissues in the frog embryo stiffen over 8 hours, during which time the embryo elongates more than 100% and closes the neural tube. We analyzed dorsal isolates at the mid-neurula stage in the middle of this rapid phase of elongation(Wilson and Keller, 1991) to determine the contribution of various regions of the dorsal isolate to its stiffness. Through a process of microsurgical elimination, we discovered that it is not the notochord but a notochord-adjacent domain of paraxial mesoderm that is responsible for more than 60% of the mechanical stiffness of the dorsal axis.

Molecularly unrelated processes may alter stiffness in the same way. In order to determine the molecular source of tissue stiffness, we tested whether stiffness of the paraxial mesoderm is controlled by fibronectin matrix assembly or by the cytoskeleton in dorsal cells. Surprisingly, the fibronectin ECM does not contribute to stiffness at either late gastrula or mid neurula stages, when a complex fibrillar matrix consisting of laminin, fibrillin and fibronectin assembles at interfaces between germ layers. Instead, tissue stiffness appears to be controlled by the contractile state of the actomyosin cytoskeleton.

The classification of the dorsal isolate, or any material for that matter,as viscoelastic, does not imply any particular mechanism is responsible for its physical response to applied force. We have chosen to represent the viscoelasticity by a spring and dashpot network for several reasons: (1) the parameters from the fitted spring and dashpot network allow comparison of embryonic tissues with synthetic materials and with biological tissues whose resistive force increases with increasing applied strain; and (2) these measured values can be used by theorists interested in simulating gastrulation and neurulation. In contrast to the generalized springs and dashpots that make up a viscoelastic model, it is not correct to think of the viscoelastic response of a biological tissue as a `passive' response but instead as a property of a living tissue. One could imagine a more complex response to applied force in which the tissue mimics a viscoelastic response; however, we think this interpretation is both unlikely and unnecessary. Viscoelastic mimicry would require a complex mechano-sensory feedback network capable of actively generating resistive forces to match applied forces. By contrast,theoretical and experimental analyses of biological polymers predict viscoelastic behaviors like those seen in frog embryonic explants(Flory, 1953; Vincent, 1990). Further studies will be needed to test the predictions of polymer mechanics, to search for mechano-sensory feedback, and to resolve the molecular, cellular and architectural sources of tissue stiffness.

Adult organisms simply cannot be supported by the extremely low stiffness found in early Xenopus embryos [as low as 3 Pa in some cultures (our results) (Moore et al.,1995)]. This paper begins to resolve the striking disparity between the highly deformable tissues of early embryos and the stiff tissues found in adults [more than 1000-fold stiffer(Levental et al., 2007)]. Early Xenopus embryos are some of the most deformable cellular tissues yet measured, but increase their stiffness by 10- to 50-fold in as little as 8 hours. Tissue stiffening occurs as the embryo establishes the basic vertebrate body plan, shapes the neural plate and folds the neural tube(Davidson and Keller,1999).

In order to estimate the stiffness of individual tissues that make up the dorsal axis, we devised a simple analytical model based on the principles of superposition of composite materials(Christensen, 1991) (see Materials and methods). We find that endoderm lining the archenteron roof plate has the lowest stiffness, nearly matching the stiffness of dorsal involuting marginal zone explants at early gastrula stages(Moore et al., 1995), and indicates that increasing stiffness is not a necessary consequence of development. The stiffness of microsurgically enhanced explants with two notochords does not differ from the stiffness of explants where the notochord had been removed (Fig. 2D′). Likewise, the stiffness of explants with expanded notochords was no different from control explants(Fig. 2I) and explants without a neural plate did not differ in stiffness from intact dorsal isolates(Fig. 4B). These results reveal a complex architecture within the early embryo that sets the stage for the initiation of organogenesis that follows neural tube closure.

Our finding that stiffness of dorsal axial tissues vary in time and position suggests these mechanical properties may serve multiple roles during embryogenesis: (1) allowing tissues to serve as a mechanical scaffold for the action of cell-generated forces (Stern,2004; Trinkaus,1984); (2) provide positional cues to pattern cell identity; or(3) may simply reflect ongoing steps in cell specification and differentiation. At the present, we can speculate only: that deformable tissues in the early gastrula are more compatible with large movements, such as involution and epiboly; or that later convergence and extension or folding movements of the neural epithelium may require stiffer tissues. It is also intriguing that germ layers at this early stage exhibit different stiffnesses. In principle, these spatial variations in the micro-environment may provide positional information to embryonic cells by triggering alternative differentiation pathways. Studies of cultured precursor cells grown on stiff or deformable substrates differentiate according to the mechanical properties of the substrate, e.g. soft substrates generate adipose cells and stiff substrates osteocytes (McBeath et al.,2004). Mechanical cues such as these can equal the strength of growth factors in contributing to cell fate choices(Engler et al., 2006). In the case of the frog embryo, the outer layer of pre-somitic mesoderm faces stiff neural tissues and may use that information to direct cells to a myotome fate,while the inner layer faces much more deformable endoderm and could use that information to initiate a sclerotome fate. The different mechanical environments seen by the inner versus the outer layer of the presomitic mesoderm may modulate the cell fate choices of these two adjacent tissues. Future studies combining biomechanics, signal transduction and classical embryology will be needed to address these prospective roles for spatial and temporally regulated tissue mechanics.

Fibronectin-fibrils do not contribute to the stiffness of dorsal tissues,and whereas other ECM components may be present at these stages, we suspect that they also do not make substantial contributions. Four lines of evidence argue against a contribution of these three ECM components to dorsal tissue stiffness: (1) dorsal tissues with 60% less fibronectin showed no difference in stiffness from control embryos; (2) these same dorsal isolates exhibited collateral reduction of fibrillin and laminin 1 fibrils; (3) the assembly of fibronectin fibrils are a pre-requisite to formation of fibrillin fibers(Kinsey et al., 2008) and may be required for laminin polymerization(Colognato and Yurchenco,2000); and (4) laminin-1 and fibrillin are strongly localized to surface of midline axial tissues such as the notochord(Fig. 5) and removal of the notochord or addition of a second notochord does not change the stiffness of isolate from control isolates (Fig. 2). Of course, there may be other fibrillar ECM components that we cannot yet visualize, but, if they follow similar patterns of assembly and require a pre-existing fibronectin ECM, they too are unlikely to contribute to the stiffness of dorsal tissues in the embryo.

The pathways that regulate actin dynamics during cell movement also control aspects of tissue stiffness. Contractility within the actomyosin cytoskeleton can alter its apparent stiffness as it generates pre-stress (see Wainwright et al., 1976). Pre-stress can increase the apparent stiffness of a material (known as`strain-hardening') and is often used by civil engineers in construction of bridges and skyscrapers. Strain-hardening in response to pre-stress occurs in both cells (Stamenovic, 2005)and reconstituted actomyosin gels (Gardel et al., 2006). In addition, complex regulatory networks can control the organization and assembly of actin and myosin II within motile cells. Xenopus embryos, like all other embryos yet studied, exhibit a wide variety of different cell behaviors that are regulated by these control networks. Our study, by depolymerizing F-actin or inhibiting myosin II contractility, has found that these same networks may also control spatial and temporal patterns of stiffness within the embryo. Surprisingly, although these acute drug treatments reduce tissue stiffness by half, these changes are much smaller than the six- to ten-fold spatiotemporal changes measured in the first half of our study. Further work will be needed to identify both anatomical and molecular factors responsible for patterning the large-scale changes in stiffness as the embryo develops and the role that spatial and temporal changes in stiffness plays in morphogenesis.