Abstract
Physical cues, experienced during early embryonic development, can influence species-specific vertebral numbers. Here we show that mechanical stretching of live chicken embryos can induce the formation of additional somites and thereby modify early segmental patterning. Stretching deforms the somites, and results in a cellular reorganization that forms stable daughter somites. Cells from the somite core thereby undergo mesenchymal-to-epithelial transitions (MET), thus meeting the geometrical demand for more border cells. Using a Cellular Potts Model, we suggest that this MET occurs through lateral induction by the existing epithelial cells. Our results indicate that self-organizing properties of the somitic mesoderm generate phenotypic plasticity that allows it to cope with variations in the mechanical environment. This plasticity may provide a novel mechanism for explaining how vertebral numbers in species may have increased during evolution. Additionally, by preventing the formation of transitional vertebrae, these self-organization qualities of somites may be selectively advantageous.
Introduction
A segmented spine is the characteristic feature of the vertebrate body plan, which provides mechanical support, flexibility, and protection of the spinal cord. Vertebral numbers vary considerably among species, ranging from six in frogs to several hundred in snakes (Richardson, Allen, Wright, Raynaud, & Hanken, 1998). The evolvability of the vertebral column allows vertebrates to adapt to diverse habitats and acquire matching locomotor styles by tuning the number of body segments (Galis et al., 2014a).
Patterning of the vertebrate body originates early in embryogenesis when the paraxial mesoderm on both sides of the midline segments into somites, which contain the predecessor cells of vertebrae, ribs, muscles and skin. Somites are cell blocks in which a core of mesenchymal cells, the somitocoel, is surrounded by an epithelial layer (Kulesa, Schnell, Rudloff, Baker, & Maini, 2007; Martins et al., 2009); they also impose a segmented organization on the peripheral nervous system. The sequential partitioning of the paraxial mesoderm appears to be imposed by a molecular clock and a travelling wave of maturation created by a system of signalling gradients (Hubaud & Pourquié, 2014). This complex signalling network appears well conserved throughout the vertebrate phylum. The variation in somite numbers (and vertebral numbers) between species is presumably caused by mutations that lead to changes in the speed of the clock period or the elongation rate of the mesoderm, which both affect segmentation rate and somite size (Gomez et al., 2008; Gomez & Pourquie, 2009; Herrgen et al., 2010).
Despite the intricately controlled network of the clock-and-wavefront mechanism, vertebral numbers in fish (Beacham & Murray, 1986; Hubbs, 1922; Tibblin, Berggren, Nordahl, Larsson, & Forsman, 2016), amphibians (Jockusch, 1997; Peabody & Brodie Jr., 1975), reptiles (Osgood, 1978), mammals (Lecyk, 1966) and birds (Lindsey & Moodie, 1967) may be influenced during early embryonic development by environmental cues such as temperature, salinity or light conditions. Mechanics may be another physical cue that induces different phenotypes of segmental patterning. In the framework of the Extended Evolutionary Synthesis, it has been argued according to the ‘side-effect hypothesis’, that morphological novelties in the body plan can result not only from genetic rearrangements, but also from mechanical cues exerted on self-organizing developing tissues (Laland et al., 2015; G. Müller, 1990; G. B. Müller, 2003). Somitic mesoderm possesses a certain capacity for self-organization, as somite-like structures can form ectopically in the absence of a clock or a wavefront (Dias, de Almeida, Belmonte, Glazier, & Stern, 2014). The size of epithelializing somites strongly correlates with embryonic growth (Tam, 1981), and hampering axial elongation of the embryo leads to disorganized somites (Stern & Bellairs, 1984). Somite formation also requires the contraction of cells in the PSM (Duess, Fujiwara, Corcionivoschi, Puri, & Thompson, 2013) and mechanical adhesion to the surrounding fibronectin matrix (Hubaud, Regev, Mahadevan, & Pourquié, 2017; Martins et al., 2009).
On the basis of these observations, we hypothesized that if somitic mesoderm is self-organizing under the influence of biomechanical cues, mechanical stretching may then suffice to induce morphological changes in the segmented vertebrate body plan.
To test this hypothesis, we developed a novel experimental setup to apply controlled strains to live chicken embryos. Time-lapse imaging confirmed that, given sufficient strain, mesodermal patterning can indeed be modulated (Fig S1). This suggests that mechanical cues can affect morphogenesis and may be an explanatory factor for variation in vertebral numbers in the scope of the ‘side-effect hypothesis’.
Results
To test the effect of mechanical stretching on somitogenesis, stage HH8-9 chick embryos (Hamburger & Hamilton, 1951) were cultured ex ovo in modified submerged filter paper sandwiches (Schmitz, Nelemans, & Smit, 2016) and stretched along their body axis twice for 51 to 54 minutes at 1.2 μm/s (Fig S1). As a result, the embryos experienced strain values of 23 ± 3% (average ± SD) after the first pull and 19 ± 3% after the second pull. Approximately 12 hours after manipulation, the mesodermal segmentation of one third (19/57) of all stretched embryos was clearly disturbed, unlike that of control embryos (Fig 1A and B). Time-lapse microscopic imaging (Schmitz et al., 2016) revealed that somites budded off from the PSM at regular intervals of 79 ± 8 minutes in controls and 80 ± 6 minutes in stretched samples. Stretched somites were more elongated than those in control embryos (Fig 1C-F, Movie S1 and Movie S2). These deformed somites then regularly divided into what we call “daughter somites”, and their morphology was consistent with somite divisions observed in N-cadherin and Cadherin-11 knockout mice (Horikawa, Radice, Takeichi, & Chisaka, 1999; Kimura et al., 1995). During daughter-somite formation, an invagination along the mediolateral plane of the deformed somites appeared simultaneously to their separation from the PSM (Fig 1F, Movie S2). It took several hours from the first appearance of this mediolateral invagination for the somite to divide completely. Somite division in stretched embryos appeared unilaterally or bilaterally, and often resulted in daughter somites of different sizes (Movie S2, Fig S4, Fig S5 and Fig S8).
Daughter somites appeared as stable, rounded and clearly separated cellular spheres (Fig 1B). They were morphologically similar to control somites, with epithelial cells organized radially around a somitocoel of mesenchymal cells (compare Fig 2C and G). The daughter somites were enclosed by a fibrous extracellular matrix (ECM) staining positively for fibronectin (Fig S4). While somites in control embryos were rounded (Fig 2A), those in stretched embryos were elongated, while any daughter somites were variable in size (Fig 2B, Fig S4 and Fig S5). Small daughter somites consisting of a few epithelial cells only were also observed (Fig S4). After fixation in transitional stages, the apical actin cortices of somites showed discontinuities along their mediolateral plane, indicating openings of the epithelial sheet under influence of the mechanical deformation (Fig 2D, E and Fig S6). At these locations, mesenchymal somitocoel cells appeared elongated, presumably having undergone mesenchymal-to-epithelial transitions (MET), and were integrated into the existing epithelium (Fig 2E and Fig S6). Upon stretching, there was also strong ectopic expression of EphA4 in the somitocoels (Fig 2I); this was not present in control embryos (Fig 2H) and indicates stretch-induced MET. During normal somitogenesis, cell-cell signalling between receptor EphA4 in the rostral part of somite S-I and its ligand ephrin B2 in the caudal half of somite S0 has two effects: it induces formation of the somite gap (Watanabe, Sato, Saito, Tadokoro, & Takahashi, 2009), but also establishes epithelialization at somite boundaries by initiating a columnar morphology and cell polarity via apical redistribution of β-catenin (Barrios et al., 2003).
However, the expression pattern for cMeso1, the key initiator of somite rostro-caudal polarity in the unsegmented PSM in chicken (Morimoto et al., 2007), was not influenced by the stretching (Fig 2K), and the expression pattern of caudal somite marker Uncx4.1 (Schrägle, Huang, Christ, & Pröls, 2004) showed that daughter somites did not gain a new rostro-caudal genetic identity (Fig 2M). Altogether, our data suggest that mechanical stretching did not affect the unsegmented PSM or the rostro-caudal polarization of the somites. However, mechanical stretching did lead to continuation of EphA4-mediated epithelialization of mesenchymal cells from the somitocoel in somites undergoing daughter-somite formation.
To identify the conditions under which mechanically deformed somites would successfully reorganize into daughter somites, we used the Cellular Potts model (CPM) and implemented a cell-based computer simulation with the open-source package CompuCell3D (Glazier & Graner, 1993; Swat et al., 2012). The somite consisted of a core of non-polarized mesenchymal cells surrounded by a layer of polarized, epithelial cells (Dias et al., 2014). We simulated a somite embedded within an elastic extracellular matrix (ECM; Fig 3A) and mimicked stretching by applying axial tension to the ECM (Fig 3B, Movie S3). To narrow down the possible mechanisms for somite reorganization under mechanical deformation, we tested different rules for cellular behaviour in the deformed somite (Fig S10).
We found that daughter-somite formation in silico could successfully be induced only if MET was initiated when mesenchymal core cells came into contact with the basal or lateral membranes of epithelial cells (Fig 3C, C’ and D). To validate this model prediction in vivo, we calculated the relative fraction of mesenchymal and epithelial cells in the equatorial cross-section of control somites and divided somites. This showed that the epithelial cell fraction was indeed significantly higher in daughter somites than in controls, and also matched the in silico prediction independent of the size of the initial somite (Fig 3E, Fig S11). As the stretching in vivo caused no significant changes in the apoptosis and proliferation rates (Fig 3F), we conclude that the increase in the epithelial cell fraction was due to MET.
To determine the relationship between mechanical deformation and somite division, we compared the aspect ratios of stretched dividing somites (prior to division) and stretched non-dividing somites with those of control somites (Fig 3G, Fig S7). These measurements show that somite division is only possible beyond an aspect ratio threshold of about 2 in vivo or 2.5 in silico. The corresponding receiver operating characteristics (ROC) curves (Goodenough, Rossmann, & Lusted, 1974; Hanley & McNeil, 1982; Lusted, 1971; Metz, 1978) show that the somite aspect ratio is indeed an excellent predictor of somite division in vivo and in silico, indicating that daughter-somite formation is a highly mechanically-determined process (Fig 3H).
Discussion
We found that mechanical stretching of live chick embryos can induce a slow reorganization of somites to form two or more well-shaped and stable daughter somites. The complete division of a stretched somite into daughter somites took several hours, indicating that daughter-somite formation is an active process of tissue reorganization, rather than acute mechanical disruption. The resulting formation of new epithelial borders took place in the somitic mesoderm, i.e. outside the functional range of the molecular cues described in the clock-and-wavefront mechanism (Hubaud & Pourquié, 2014). Remarkably, the segmentation clock speed (estimated by the average somite formation time) and the genetic segmentation (A-P-polarity of the somites) stay robust during severe deformations imposed on the embryo. As the FGF/WNT gradients are intracellular by mRNA or protein inheritance (Dubrulle & Pourquié, 2004), the cells may retain their local information, despite their different spacing after stretching. Interestingly, our data show that the organization of forming somites is not final after their separation from the anterior tip of the PSM, but that somites can still adapt to their environment.
High-resolution confocal imaging indicates that, during daughter-somite formation, the demand for additional border cells is satisfied by the recruitment of mesenchymal cells from the somitocoel into the existing epithelium. The mechanical deformation creates discontinuities in the apical actin cortices of stretched somites. At the resulting interfaces, mesenchymal cells from the somite core undergo mesenchymal-epithelial transitions (MET) and get integrated into the somitic epithelium. Similarly, it has been shown previously that the development of normal somitic epithelia involves a continuous addition of cells from the somitocoel by accretion and egression (Martins et al., 2009). Our observations indicate that, by adding mesenchymal cells to the epithelium via a similar cellular behaviour during stretching, the epithelium adapts to the changing environment. Under sufficient deformation, this can lead to daughter-somite formation.
We were able to model daughter-somite formation in silico only if MET was initiated when mesenchymal core cells came into contact with the basal or lateral membranes of epithelial cells. This suggests a contact-induced mechanism triggering MET of somitocoel cells during daughter-somite formation. In vivo, we observed ectopic expression of EphA4 without cMeso1 expression in strained somites, although EphA4 is thought to be downstream of cMeso1 (Watanabe et al., 2009). This indicates that EphA4 expression in stretched somites is maintained or reinitiated, suggesting an additional mechanosensitive pathway that leads to EphA4 upregulation during somite division that is independent from, or redundant to, cMeso1. A contact-induced MET mechanism, as suggested here for daughter-somite formation, could underlie general epithelial self-organization during development and homeostasis of epithelia under mechanical stress (Jackson, Kim, Balakrishnan, Stuckenholz, & Davidson, 2017; Martins et al., 2009). This cell behaviour could be mediated via epithelial membrane-based signalling (Baum & Georgiou, 2011; Campbell, Casanova, & Skaer, n.d.), for example on the level of Eph and ephrin binding.
We show that chick somite formation is phenotypically plastic under changing biomechanical conditions. As the geometry of a somite in a stretched embryo predicts with remarkable reliability if it will reorganize into daughter somites or not, daughter-somite formation is highly mechanically determined. This supports the idea that, like temperature, light regime or salinity (Tibblin et al., 2016), mechanical forces can be an additional cue to the induction of different phenotypes of segmental patterning during embryonic development. To further explore the role of mechanical cues in inducing phenotypic plasticity of vertebral numbers, it will be necessary to study the effect of modified mesodermal patterning in stretched embryos on later embryonic development. Though in principle possible (Nagai, Sezaki, Nakamura, & Sheng, 2014), transplanting stretched embryos back into a host egg could prove technically challenging due to the deformation of the embryos. It may therefore be more promising to transplant daughter somites into host chick embryos in the egg and see how skeletal patterning might be influenced.
It has been suggested that the wide range of vertebral numbers between vertebrate species could have evolved for two reasons: (1) The spatial dissociation between axis regionalisation via Hox gene expression and segmentation patterning, and (2) the evolvability of the segmentation clock’s period and its relation to the axial growth of the developing embryo (Gomez et al., 2008; Gomez & Pourquie, 2009; Herrgen et al., 2010). However, if the phenotypic plasticity of somite formation shown in our experiments can indeed translate into different vertebral numbers in the later embryo, our findings could indicate an alternative route for vertebrate body plan evolution. Following the ‘side-effect hypothesis’ (G. B. Müller, 2003), natural selection may act on body proportions, leading to changes in the geometry and mechanical loading of the somitic mesoderm. As an initial by-product, the phenotypic plasticity of somites that we show in our experiments then may lead to increasing somite numbers. Later, maintenance of a consistent alteration of the somite number over generations may then become consolidated at the genetic level by natural selection, i.e. via genetic assimilation (Braendle & Flatt, 2006; Fusco & Minelli, 2010), thus providing robustness to the development of the newly acquired body plan. A correlation between axial growth of the embryo and somite size has been shown in mice (Tam, 1981). While daughter-somite formation may be useful as a source of skeletal variation, it remains to be clarified whether and how the developing embryo may cope with the missing A-P polarity of the stretch-induced daughter somites.
Daughter-somite formation could also help to explain an interesting phenomenon. Transitional vertebrae, i.e. vertebrae with morphological characteristic of two adjacent spinal regions (for example lumbar area and sacrum), result from incomplete homeotic shifts of axial identity defined by Hox gene expression. Generally, as more than one mutation is needed for complete transformations, incomplete homeotic transformations (and therefore transitional vertebrae) are frequent (Alkema, van der Lugt, Bobeldijk, Berns, & van Lohuizen, 1995; Charité, de Graaff, & Deschamps, 1995; Horan, Wu, Wolgemuth, & Behringer, 1994; Kostic & Capecchi, 1994; Li, Kawasumi, Zhao, Moisyadi, & Yang, 2010; Rancourt, Tsuzuki, & Capecchi, 1995; van den Akker et al., 2001; Varela-Lasheras et al., 2011). However, these transitional vertebrae are statistically underrepresented in several amniote species with variable trunk vertebrae numbers, including Triturus newts and lizards ((Kaliontzopoulou, Llorente, & Carretero, 2008; Slijepčević, Galis, Arntzen, & Ivanović, 2015), F. Galis, personal communication, October 2017). This suggests a developmental mechanism which favours complete numbers of trunk vertebrae over transitional vertebrae that might hamper mechanical function (Galis et al., 2014b; Slijepčević et al., 2015). Assuming that additional trunk somites form bilaterally via daughter-somite formation, both newly created somites would have the same axial identity as their mother somite, given that the Hox identity is determined mainly before the somite is formed, during the earliest stages of somite formation (Carapuço, Nóvoa, Bobola, & Mallo, 2005; Dubrulle, McGrew, & Pourquié, 2001; Mallo, Wellik, & Deschamps, 2010). This would generate vertebrae of similar regional identity with better mechanical performance than that of potentially disadvantageous transitional vertebrae with a different morphology (Galis et al., 2014a).
Somites’ self-organizational properties provide a promising basis for further exploration into the physical component of somite formation and the causal role of mechanics in body-plan evolution.
Materials and Methods
Embryo preparation and culture medium
HH8-9 chicken embryos were explanted using filter paper carriers (Chapman, Collignon, Schoenwolf, & Lumsden, 2001) and cultured ex ovo as modified submerged filter paper sandwiches (Chapman et al., 2001; Schmitz et al., 2016), in Pannett-Compton (PC) saline (Pannett & Compton, 1924; Schmitz et al., 2016; Voiculescu, Papanayotou, & Stern, 2008), mixed with freshly harvested thin albumen in a 3:2 ratio. Silicone sheets protected embryos in culture from convection of the medium, thereby avoiding damage. Filter paper carriers were prepared as described recently (Schmitz et al., 2016). Additionally, four holes were cut out from corners of the carriers (Fig S1C) to hook the filter paper sandwiches onto the pins of the motorized arms of the stretching setup (Fig S1A).
Stretching protocol, axial deformation and somite deformation
Embryos were exposed to a standardized stretching protocol in a custom-made embryo stretcher (Fig S1C). In two consecutive stretching intervals, a slow displacement of the computer-controlled metal arm (see red arrow in Fig S1) extended the filter paper sandwiches by 3.7 to 3.95 mm at a speed of 1.2 μm/s. At this speed, each stretch took 53 to 55 min. We calculated the mechanical strain for the first and the second stretching as relative length change compared to the axial length before stretching. Results are presented in S1 Table. The first stretch lead to 23 ± 3 % strain (average and standard deviation over all 21 embryos presented in S1 Table). The second pull caused 19 ± 3 % strain. For details on the embryo stretching, somite formation time and determination of somite deformation by aspect ratio, see SI Materials and Methods.
Immunohistochemistry
After the pulling experiments, the embryos and age-matched controls were fixed in 4% paraformaldehyde overnight in PBS at 4°C. Permeabilization in PBST + 0.15% Triton-X-100 lasted for 1.5 hours. Blocking was performed for 2 hours in PBST + 2% BSA + 5% normal goat serum. The following antibody was used: fibronectin mouse-anti-chicken (B3/D6-s, Hybridoma bank). The antibody was diluted in PBST with 1% BSA. Embryos were incubated in primary antibody solution for 24h at 4°C, followed by extensive washing in PBS and incubation with appropriate Alexa Fluor-conjugated secondary antibody (1:500, Molecular Probes). Embryos were stained for F-actin using Alexa Fluor 546 Phalloidin (1:200, Molecular Probes) and for nucleic DNA using DAPI (1 μg/ml). Cell proliferation and apoptosis staining was performed using following antibodies: rabbit polyclonal anti-cleaved caspase-3 (1:200, Cell Signaling) and rabbit polyclonal anti-phosphohistone-H3 (1:400, Cell Signaling) with the appropriate Alexa Fluor-conjugated secondary antibodies (1:500, Invitrogen) and DAPI for nucleic DNA (1 μg/ml). For details on mesenchymal and epithelial cell counts, see SI Materials and Methods.
In situ hybridizations
In situ hybridizations were performed by standard procedures. Embryos were fixated in freshly prepared 4 % PFA in PBS. The embryos were pre-treated with proteinase-K in PBST at 37°C with agitation for 3 minutes. During staining, embryos were incubated in NTMT containing 4.5 μl NBT (75mg/ml in 70% DMF) and 3.5 μl BCIP (50mg/ml in 100% DMF) per 1.5 ml. Pulled embryos and age-matched controls were stained in the same wells for the same time, as much as possible. After the staining had been stopped, the embryos were photographed in glycerol 80% in H2O with a Leica DFC320 camera on a Leica MZ75 microscope.
Cellular Potts model of somite division
To understand the influence of mechanical stretching on somites and the somite division observed in vivo, we constructed a two-dimensional mathematical model based on the Cellular Potts Model (Glazier & Graner, 1993; Graner & Glazier, 1992), representing a cross-section through a three-dimensional somitic tissue. For the modelling details, see SI Materials and Methods.
Competing interests
The authors declare that there are no conflicts of interest.
Author Contributions
B.N. and M.S. performed the experiments and analysed the data. H.T. and R.M. performed the simulations. T.S. and R.M. supervised the project. All authors wrote the manuscript.
Movie S1: Control embryo
Time-lapse movie of the development of a control chick embryo (stage HH9+), cultured ex ovo, mounted in submerged filter paper sandwiches in the stretch setup for 19 hours, without stretching. Timer is in hours.
Movie S2: Stretched embryo showing somite divisions
Time-lapse movie of the development of an experimental chick embryo (stage HH10), cultured ex ovo, mounted in submerged filter paper sandwiches in the stretch setup for 17,5 hours, and stretched at a speed of 1.2 μm/s along the anterior-posterior (AP) axis, in two pulls. First the overview is shown, afterwards a zoom in at the mesoderm. The stretching deforms the embryos slowly but substantially, while development progresses without damage. During the deformation, somites divide into daughter somites of different sizes, as marked by the white arrowheads in the zoom. For example: the first arrowhead shows an asymmetric somite division (lower left), while the second arrowhead shows a symmetric division (upper left).
Movie S3: Somite epithelialization and division in silico
Simulation video of dividing in silico somites, consisting out of frames of every 500th Monte Carlo Step. We tested different rules in our Cellular Potts model concerning cellular behavior in the deformed somites, to find out the most likely mechanism for the somite division to take place. Daughter somite formation in stretched somites could only successfully be induced when a MET of mesenchymal core cells upon contact to the basal or lateral membranes of epithelial cells was allowed.
Acknowledgements
We are grateful to Julio Belmonte for his advice on the simulations at the early stages of model development. We thank Stuart A. Newman and Frietson Galis for valuable comments during the final preparation of the manuscript.