Wide and high resolution tension measurement using FRET in embryo

During embryonic development, physical force plays an important role in morphogenesis and differentiation. Stretch sensitive fluorescence resonance energy transfer (FRET) has the potential to provide non-invasive tension measurements inside living tissue. In this study, we introduced a FRET-based actinin tension sensor into Xenopus laevis embryos and demonstrated that this sensor captures variation of tension across differentiating ectoderm. The actinin tension sensor, containing mCherry and EGFP connected by spider silk protein, was validated in human embryonic kidney (HEK) cells and embryos. It co-localized with actin filaments and changed FRET efficiencies in response to actin filament destruction, myosin deactivation, and osmotic perturbation. Time-lapse FRET analysis showed that the prospective neural ectoderm bears higher tension than the epidermal ectoderm during gastrulation and neurulation, and cells morphogenetic behavior correlated with the tension difference. These data confirmed that the sensor enables us to measure tension across tissues concurrently and with high resolution.

In this study, we introduced a FRET-based tension sensor into Xenopus embryos and measured tissue tension. The sensor was first validated in cultured cells and embryos, and then applied to the gastrula/neurula ectoderm. The tension sensor demonstrated a clear distinction between prospective neural ectoderm and epidermal ectoderm. We also found that the neural and lateral epidermal ectoderms underwent similar morphogenesis, whereas their individual cells behaved differently.

Results
Validation of tension sensor in cultured cells and embryos. The first reported actinin tension sensor 16 contained a FRET domain composed of Venus and Cerulean connected by an elastic linker derived from a spectrin repeat. This FRET domain was inserted between spectrin repeat domains 1 and 2 in α-actinin. However, this FRET domain had unexpectedly low FRET efficiency when the FRET domain alone was expressed in a cultured cell 16 . Therefore, we replaced the FRET domain with one containing mCherry and EGFP linked by an elastic linker derived from a spider silk protein 15 , and named it ActTS-GR (Fig. 1a). For a negative control that always shows high FRET efficiency, the FRET domain was attached to the C-terminal of an intact actinin (hiActTS-GR) (Fig. 1b). The tension sensor was validated in HEK cells and the dorsal ectoderm of Xenopus gastrula. HEK cells were transfected with plasmids encoding the sensor. Embryos were injected with 500 pg of the sensor mRNA into the dorsal animal marginal zone at the four-cell stage, and incubated until the gastrula stage. The tension sensor showed partial colocalization with actin filaments ( Supplementary Fig. 1a,b). In the dorsal ectoderm, actin filaments were abundantly distributed around the apical-lateral adherens junction. The tension sensor also colocalized with the actin filaments, marking the cell-cell interfaces. To test their dynamics, we measured fluorescence recovery after photobleaching (FRAP). Recovery curves were identical between ActTS-GR and mCherry-tagged actinin in HEK cells and embryos ( Supplementary Fig. 1c,d), indicating that ActTS-GR has similar dynamics to actinin.
Next, we calculated the FRET index of the tension sensors in living HEK cells and embryos. In addition to ActTS-GR and hiActTS-GR, we co-transfected a pair of mutant non-fluorescent ActTS-GR (Fig. 1c) into HEK cells (Fig. 1f) and co-injected them into embryos to measure inter-molecular FRET (Fig. 1i). As expected, hiActTS-GR had a high FRET index, the non-fluorescent mutants pair had a low FRET index, and ActTS-GR had an intermediate value both in HEK cells and embryos ( Fig. 1d-k). In embryos, the FRET index of the mutants pair was exceptionally high, but we concluded that it was caused by autofluorescence of embryos since un-injected tissue also returned a FRET index similar to that of the mutants pair (data not show). To further confirm these results, we conducted acceptor photo-bleaching with fixed samples as an alternative method for estimating FRET efficiency without background noise 21 . After mCherry was bleached locally, the EGFP fluorescence intensity of ActTS-GR and hiActTS-GR increased in the mCherry bleached region and calculated FRET efficiencies were consistent with the FRET indices ( Supplementary Fig. 2a,b,d-f,h). Importantly, FRET efficiency of the mutants pair was significantly lower in both HEK cells and embryos ( Supplementary Fig. 2c,d,g,h), indicating that intermolecular FRET is low. These results suggest that ActTS-GR, when creating dimers and bridging actin filaments, was under tension, and the FRET domain was stretched (Fig. 1l).
We also examined the sensor function by experimentally changing the tension. HEK cells expressing either ActTS-GR or hiActTS-GR were examined with FRET before and after incubation in either the actin polymerization inhibitor, cytochalasin b, the Rho-associated protein kinase (ROCK) inhibitor, Y-27632, or a hypotonic solution. When the transfected cells were incubated in 2.05 μM of cytochalasin b, the filament-like distribution disappeared and the fluorescence aggregated in clusters, indicating disruption of the actin filaments (Fig. 2a,b). The FRET index of ActTS-GR but not of hiActTS-GR was increased after the cells were treated with cytochalasin b (Fig. 2c). Likewise, treating HEK cells with 10 μM of Y-27632 increased the FRET index of ActTS-GR, indicating a decrease in tension (Fig. 2d). In contrast, incubation in 90% distilled water decreased the FRET index of ActTS-GR but not of hiActTS-GR, indicating that the cells were swollen, and tension on the actin filament was increased (Fig. 2e). Together, these data indicated that ActTS-GR could respond to tension changes in HEK cells.
In embryos, FRET images were obtained under a hypertonic solution. Switching the embryos from 1x Steinberg's solution (SS) to 2x SS increased the FRET index of ActTS-GR but not of hiActTS-GR (Fig. 3), suggesting that the embryos were shrunk by the hypertonic challenge, releasing tension on the actin filament network across the embryonic surface. In case the constructs were toxic, injected embryos were incubated until the larval stage, and no apparent phenotype was observed (data not shown). These results confirmed that the construct functioned as a tension sensor inside cultured cells as in previously reported constructs, and also inside Xenopus embryos without evident toxicity.
Tension measurement in Xenopus early embryo ectoderm. Having confirmed the functionality of the sensor, we then measured the variation in tension between differentiating tissues. In Xenopus gastrula, the dorsal ectoderm extends along the anterior-posterior axis and converges along the medial-lateral axis. The dorsal ectoderm becomes a neural plate in the neurula stage and undergoes neural tube closure, disconnecting from the epidermal ectoderm. At the same time, the vegetal ectoderm covers the whole embryo and forms the epidermis. The tensile properties of the neural plate and the epidermal ectoderm are different in an axolotl embryo 22 . Embryos were injected with 1 ng of ActTS-GR or hiActTS-GR mRNA into the animal marginal zone at the four-cell stage and observed from the gastrula stage to the neurula stage. Confocal images were taken from the vegetal view, and the tension was measured around a blastopore and in the dorsal marginal zone. We calculated the FRET ratio by dividing the acceptor fluorescence intensity by the donor fluorescence intensity 16 . This method cannot be used to evaluate FRET between the co-transfected/co-injected fluorescence mutants pair, but the acquisition of microscopic images is easy, and thus it is useful to analyze a large living tissue with high temporal resolution. We confirmed that FRET ratios calculated in the experiments described above were consistent with the FRET indices calculated in the same experiments (data not shown). The FRET ratio of ActTS-GR showed a clear separation between the prospective neural ectoderm and the epidermal ectoderm. The prospective neural ectoderm had a lower FRET ratio, and this difference was observed from the late gastrula stage (Fig. 4a-c).
In the neurula stage, the cell shapes of the epidermis and the neural plate were different. The cells in the neural plate remained isotropic in appearance, whereas the cells in the lateral epidermis were elongated along the antero-posterior axis (Fig. 4d). To investigate when and how this difference emerged, we injected 400 pg of membrane-tethered-GFP mRNA into an animal marginal zone at the four-cell stage and measured the deformation of cells and tissues (Fig. 5). We first traced cells manually and calculated tissue strain rates 23 . During the late gastrula stage, groups of cells in both prospective neural ectoderm and epidermal ectoderm underwent a similar rotation and deformation (Fig. 5c,d). They elongated roughly along the antero-posterior axis. Next, we measured the width of the cells and the groups of cells along the elongating axis (Fig. 5e). Both of them increased width in the epidermal ectoderm. In contrast, the width of the cells in prospective neural ectoderm did not change during tissue morphogenesis. Although the two differentiating tissues deformed similarly, their cells behaved differently. At gastrulation, the dorsal ectoderm is pulled antero-posteriorly by the underlying mesoderm 24,25 . A previous study using axolotl embryo measured the tensile property of the ectoderm by stretching the specimen of the ectoderm, and found that the epidermal ectoderm was easier to deform than the neural ectoderm 22 . Our data suggest that the cells in prospective neural ectoderm generated higher tension on actin filaments than the epidermal ectoderm cells, and the higher tension appeared as cell stiffness against outer forces (Fig. 5f). Both the prospective neural ectoderm and the lateral epidermal ectoderm were pulled antero-posteriorly by the mesoderm. The prospective neural ectoderm deformed by rearranging its stiff cells, whereas, in the epidermal ectoderm, the soft cells themselves deformed.

Discussion
In summary, our experiments demonstrated that our tension sensor was sufficiently functional in embryos. ActTS-GR performed the best among FRET-based tension sensors we tested (data not shown). They were not invasive in that the injected cells engaged in normal tissue development and the embryos underwent the complete developmental process. The tension sensors distinguished between the prospective neural ectoderm and the epidermal ectoderm, when they were just beginning differentiation. The difference in measured tensions between the two tissues were consistent with the previous report 22 . Neural plate development has been reported to require a change in the cadherin complex components 26,27 , which might be responsible for the tension change.
In an epithelial tissue where actin filaments localized mainly around the apico-lateral adherens junction, ActTS-GR lined the cell-cell interfaces. With resolution high enough to distinguish each cell-cell interface, tension on the actin filament around the interface can be measured separately. Since the actomyosin network is a major generator of cortical contractile force 28-31 , the measured tension should correspond to the tension on the cell-cell interface. Similarly to laser ablation, correlation analysis between the direction of the cell-cell interface and its tensions will provide the directional components of the tension on a cell sheet. A recent study showed that medial myosin pulses also generate a force that is transmitted to the cell-cell interface 4 . The tension measurement now requires a subcellular and high temporal resolution. ActTS-GR measures inner tension. Measurement of the outer force should also be possible with sensors derived from cadherin, talin, or other cell-cell adhesion proteins. These FRET-based tension sensors have limitations. First, the tension range that can be measured by the sensor is narrow, and a variety of elastic linkers are needed to visualize tension distribution across other tissues. Second, it is difficult to express the tension sensor evenly across a large tissue by injecting mRNA, and establishment of a constitutively expressing transgenic strain is needed for accurate analysis.
This FRET-based sensor measures tension across tissue concurrently and with cellular resolution. The measurement can be performed in a living tissue. And the most importantly, this method does not require a special instrument, making it instantly available to all researchers. In future applications, this sensor will provide an easy but high spacial and time resolution sequential measurement of tension during morphogenesis, cell rearrangement, and differentiation.

Constructs.
The DNA fragment of flagelliform sequence was synthesized, and cloned into the vector carrying CMV promoter. To generate a tension sensor module, mCherry was fused to N-terminus of flagelliform linker (GPGGA) 8 and EGFP was fused to C-terminus of the linker 15 . The constructed module was fused between spectrin repeat domains of human alpha actinin 1 16 and named ActTS-GR. For high FRET control, the module was fused with C-terminus and named hiActTS-GR. Non-fluorescent control (intermolecular FRET control) was generated by replacing an amino acid residue in chromophore of fluorescence domain. 66 th tyrosine of EGFP domain was replaced leucine and 72 nd tyrosine of mCherry domain was replaced to leucine, respectively. For mCherry-tagged actinin, mCherry was fused to C-terminus of human alpha actinin 1 with two linking amino acids between them. For mRNA in vitro synthesis, the constructs were cloned into the vector carrying SP6 promoter.
Cell culture and transfection. HEK 293 cells were plated in 4-well coverglass chamber (Iwaki) in a total volume of 1 ml high-glucose DMEM culture medium, containing 10% heat-inactivated FBS, 100 U/ml penicilin and 100 μg/ml streptomycin, and incubated at 37 °C and 5% CO 2 . For transfection, 1:1 mixture of 0.323 mg/ml polyethylenimine (Wako) and 100 ng/μl DNA with 75 mM NaCl was incubated for 30 min at room temperature, and 25 μl of the mixture was added into 0.5 ml cell culture. Cells were incubated for 1-2 days after transfection.
Xenopus embryos and micro-injection. mRNA were injected into 4 sites of dorsal animal marginal zone or 8 sites of animal marginal zone. Injections were performed in a 5% ficoll solution. Embryos were cultured in 1x Steinberg's solution. For osmotic perturbation, embryos were transferred between dishes with different concentration of SS in them.

Immunostaining. For actin filaments staining, HEK cells expressing ActTS-GR were incubated with 1 μM
Sir-actin (SPIROCHROME) for 30 minutes at room temperature. For embryos, vitelline membrane were removed surgically and the embryos were immediately fixed in MEMFA for 30 minutes. The samples were permeabilized in 1% PBT for an hour, followed by an incubation in blocking buffer for an hour. The samples then stained with Alexa Fluor 488 phalloidin at 4 °C overnight, washed with PBS.
Confocal imaging. An Olympus FV1200-D confocal microscope was used for imaging all samples in this study. An UPLSAPO40XS silicon immersion objective was used for imaging HEK cells. An UPLSAPO10X2 objective was used for imaging embryos. Embryos were imaged on 35 mm/glass base dish (Iwaki), in Steingerg's solution for FRET imaging and immuno-stained samples, or embedded in 1.2% UltraPure L.M.P Agarose (invitrogen)/Steingerg's solution for membrane-tethered-GFP injected embryos. Immuno-stained images of HEK cells show maximum intensity projections of Z-stack covering lower part of the cells. Images of HEK cells for FRET analysis were single slice. Images of embryos show maximum intensity projections of Z-stacks. FRAP analysis. The initial 4 images were acquired and then a spot in a cell (HEK cells) or a spot on a cell-cell junction (embryos) was photo-bleached for mCherry. The samples were then imaged every 0.128898 seconds for 12.11641 seconds (HEK cells) or every 0.064904 seconds for 6.10097 seconds (embryos). Average intensity in the spot was measured by FV10-ASW, the microscope controller software.
FRET analysis. FRET index and FRET ratio are the quantitative indicator of the FRET efficiency. FRET index was calculated with a custom script of ImageJ. The calculation algorithm follows the intensity-based FRET index calculation 32 with an assumption that unquenched donor (D) is proportional to acceptor (A) signal intensity. Acceptor and donor spectral bleed-through (ASBT and DSBT) were obtained from A and quenched donor qD as ASBT = c a A and DSBT = c d qD, where coefficients c a and c d were obtained from images of cells and embryos expressing non-fluorescent mutant constructs. Then pure FRET (PFRET) is PFRET = uFRET−ASBT−DSBT, where uFRET is uncorrected FRET (an image of the acceptor fluorescence channel excited by laser for the donor). Then D is D = Qd + PFRET × γ, where γ is a factor of the fluorescent pair, and the FRET efficiency E is proportional to FRET index: For an embryo, FRET calculation was performed on every single slice and the maximum FRET index of Z-stack was projected.
FRET ratio was calculated on ImageJ, dividing A by qD. Prior to dividing the acceptor image by the donor image, an area that did not express tension sensors was masked, where the mask was derived from the donor image to cover an area under the lower threshold. The lower threshold was determined based on an intensity of marginal zone. The calculation was done with images of the maximum intensity projection of Z-stacks. When the tension is low and the two fluorescent domains are enough close to each other, excited EGFP lose an energy by FRET so that the fluorescence intensity of EGFP was lowered. When the tension is high, FRET does not occur and EGFP emits full fluorescence. Therefore, ratio between EGFP and mCherry reflects the FRET efficiency. Statistical significance was assessed by student t-test on Numbers (Apple).
Acceptor photo-bleaching. For acceptor photo-bleaching, HEK cells were fixed in 4% paraformaldehyde/ PBS for 20 minutes and then washed with PBS. Embryos were fixed in MEMFA for 1 hr at room temperature and then washed in PBS. FRET efficiency was calculated based on increase of donor fluorescence that was quenched before acceptor was bleached 21 .
Calculation of tissue and cell deformation. For each cell, its center location in plane x,y was taken manually, and for its velocity (v x , v y ), change of the location from 40 minutes before was taken as an average velocity and used for calculation below. A group of 20-30 cells were chosen so that the cells were adjacent to each other and the group did not split in two groups along the cell rearrangement. A tissue velocity gradient tensor, . An elongation rate and its direction were given by an eigenvalue and eigenvector of  E T . The eigenvalue +1 was taken as the symmetric strain rate value, and the (1,2) element of Ω T was taken as the spin rate value. An antero-posterior, tissue, and cell elongation rate represented changes of bounding box width. The bounding boxes were tilted along the antero-posterior axis or the elongating direction. The antero-posterior axis were determined according to a closing neural tube. Tissue and cells elongating direction was corrected by Ω T . Since an embryo was spherical, the projection on a plane caused a distortion, but it could be ignored as long as relative elongation rates between cells and its group was issued.