Intercellular Bridges in Vertebrate Gastrulation

The developing zebrafish embryo has been the subject of many studies of regional patterning, stereotypical cell movements and changes in cell shape. To better study the morphological features of cells during gastrulation, we generated mosaic embryos expressing membrane attached Dendra2 to highlight cellular boundaries. We find that intercellular bridges join a significant fraction of epiblast cells in the zebrafish embryo, reaching several cell diameters in length and spanning across different regions of the developing embryos. These intercellular bridges are distinct from the cellular protrusions previously reported as extending from hypoblast cells (1–2 cellular diameters in length) or epiblast cells (which were shorter). Most of the intercellular bridges were formed at pre-gastrula stages by the daughters of a dividing cell maintaining a membrane tether as they move apart after mitosis. These intercellular bridges persist during gastrulation and can mediate the transfer of proteins between distant cells. These findings reveal a surprising feature of the cellular landscape in zebrafish embryos and open new possibilities for cell-cell communication during gastrulation, with implications for modeling, cellular mechanics, and morphogenetic signaling.


Introduction
At gastrula stages, zebrafish embryos have a simple organization, comprised of a few cellular layers and a relatively simple cell morphology. The enveloping layer (EVL) is the outermost part of the blastoderm, surrounding the epiblast, the hypoblast and the yolk syncytial layer (YSL), which lay on top of the yolk mass ( Figure 1B) [1]. Despite having a good understanding of the overall cellular rearrangements required to define the germ layers during zebrafish gastrulation, few in vivo studies have focused on individual cell shapes [2,3]. Past efforts have been mainly concentrated on the cells in the hypoblast, where blebs, filopodia and pseudopodia are notable features [1,4,5]. These cellular protrusions seem to lead the progressing hypoblast cells down the correct path of migration through the embryo [5]. The epiblast has instead been regarded as a single-cell thick epithelium, more homogeneous and simple [1,2]. Compared to the hypoblast, epiblast cells are less motile with simpler morphology and with less dynamic protrusions [5,6].

Results and Discussion
To study cell shape in the zebrafish epiblast, we generated high cellular contrast by targeting photoconvertible protein Dendra2 [7] to the cell membrane, Tol2-mediated integration [8] of a fusion of Dendra2 to a lipid anchor recognition sequence [9,10] (Lyn-Dendra2 referred to here as mDendra2) resulted in mosaic labeling ( Figure 1A), without the need for cell transplantation or single blastomere injection, which might disrupt cellular and intercellular morphologies. This high-contrast labeling of a subset of the cells permitted confocal microscopy to obtain highresolution images of the cell boundaries. Surprisingly, we detected the presence of cell protrusions extending up to ten cell diameters in length ( Figure 1) in these mosaically labeled mDendra2 embryos during the early phases of gastrulation. Further imaging revealed that these protrusions were not filopodia; instead, they seemed to be intercellular bridges linking distant pairs of cells (see below). To rule out the possibility that these protrusions were an artifact of the expression of mDendra2, we imaged un-injected embryos at the same stage of development using differential interference contrast (DIC) light microscopy, and confirmed the presence of the previously undescribed intercellular bridges ( Figure 1E). The protrusions are thin and contain little free cytoplasm, as labeling cells with rhodamine dextran or cytoplasmic Dendra2 did not highlight the protrusions (see Figure S1A and data not shown).
Three-dimensional imaging revealed that the cells linked by intercellular bridges resided within the epiblast layer; bridges were not observed linking cells in either the EVL or the hypoblast. A quantitative analysis of 51 intercellular bridges from 38 independent embryos at late gastrula stages showed that the protrusions averaged 215 mm in length (see Figure 2E). The imaging confirmed the presence of pseudopodia and other cellular extensions in the epiblast, but the length of such projections was a small fraction of the length of the intercellular bridges. Interestingly, previous reports of the cell morphologies during chick neurulation [11] and of deep cells in the medaka fish suggested the presence of intercellular bridges; in medaka a fraction of fixed and dissociated deep cells were found to be linked by short (up to 30 mm) intercellular extensions [12].
Time lapse studies of embryos show that the intercellular bridges form when a mitotic pair of daughter cells maintain a membrane tether, which is the lengthened as the cells move further apart and as neighboring cells subsequently intercalate between them (Figure 2A-A'''; see Movie S1). To document this association with mitoses, we captured confocal time-lapse movies of embryos ubiquitously expressing the chromatin marker Histone 2B fused to the red fluorescent protein mCherry (H2B-mCherry) [13] and watched as bridges were formed by the mDendra2labelled cells. The intercellular bridges first appeared between dividing cells at the blastula stage; dividing cells only rarely gave rise to intercellular bridges at the gastrula stage. This formation mechanism, as a persistent mid-body from mitosis, differs from the active extension of the membrane observed for cytonemes in in the wing imaginal disc of Drosophila melanogaster [14] or for the protrusion of thin filopodia during sea urchin gastrulation [15].
The intercellular bridges persisted throughout gastrulation ( Figure 2C-C'; see Movie S2), even as the cells executed morphogenetic motions such as radial intercalation. They were still visible as late as at the 10-13 somites stage (Figure S1B,C and Movie S3). While somewhat variable, about one of five cells were endowed with intercellular bridges during gastrulation. The bridges were not disrupted by subsequent mitotic activity of the cells (Movie S4). Cells equipped with intercellular bridges mitosed normally, suggesting that the bridges do not alter the oriented cell divisions present throughout gastrulation stage [16,17]. Occasionally, we observed the fragmentation of an intercellular bridge, apparently when a migratory cell breaks through it (Movie S5). We have not captured the disruption of enough bridges to know if there are other mechanisms that act to reduce the number of intercellular bridges at later stages of development.
Quantitative analysis of the intercellular bridges during gastrulation showed that they lack any obvious orientation or preferential position. We superimposed bridges of 30 cell pairs from 20 different embryos at the mid-gastrula stage ( Figure 2D), using the embryonic shield and the anterior neural boarder as morphological landmarks. The bridges do not respect boundaries in the fate map, bridging between neural and non-neural territories as well as within the same embryonic regions. We measured at mid-gastrula the angle between the putative anteroposterior axis in the neural plate and the shortest path between cells equipped with intercellular bridges (44 pairs from 23 independent embryos) and no obvious orientation was found (data not shown). This lack of alignment with the body plan and the intercellular bridges is expected because the axes of division in the blastula are not related to the later body plan or organ fate maps [1,2].
Further characterization of intercellular bridges shows that there are minor variations in diameter (up to 1 mm) and significant variations in length, reaching lengths up to 350 mm (average span of 215 mm; Figure 2). The bridges are extraordinary in length when put in context with the individual cell diameters of 20-25 mm. The total surface area of a typical epiblast cell soma is around 1200 mm 2 , almost identical to the surface area of the longest intercellular bridge of 1 mm diameter (1100 mm 2 ). Thus, the surface area of these membrane bridges increases the amount of cell surface of the linked cells, and the total present in the embryo.
The lumen of each intercellular bridge appears to be filled with cytoskeletal components and devoid of bulk cytoplasm. To sample the cytoskeleton present inside intercellular bridges we mosaically labeled cells within the embryos with membrane-targeted mCherry together with EGFP-b-Actin or EGFP-a-Tubulin. The images revealed that b-Actin was present at a low level for most if not all of the length of the intercellular bridges; a-Tubulin was detected only in the proximal portion of the intercellular bridges ( Figures S1 and data not shown). In extremely long intercellular bridges, the contents seemed to thin in the middle, with the central part of the structure almost devoid of cytoskeletal components. The presence of cytoskeletal proteins within the intercellular bridges suggests they could be used to exert tension, and potentially acts as a cellular mechanical mediator. To verify that the intercellular bridges are continuous between pairs of cells we photoconverted mDendra2 in one cell body and followed the motions of the photoconverted pool towards the other cell ( Figure 3 and Movie S6). In all cases, the mDendra2 moved along the length of the intercellular bridge, eventually reaching the plasma membrane of the sister cell ( Figure 3B-C'). The photoconversion of mDendra permitted us to measure the time required for a tagged protein to move along the length of the intercellular bridge. The movement seems faster than expected by diffusion alone, reaching distances of 100 mm in about 30 minutes (3.4 mm/min,). This fast rate is not particular to mDendra2; Dendra2 anchored via CD8 (CD8-Dendra2) [9,10] was found to move at rate of 3.3 mm/min. The movement of the membranelinked Dendra2 probes seems linear with respect to time, sug- gesting that it is due to some active process rather than by diffusion alone ( Figure 3D). This rate of transfer along the intercellular bridges is sufficiently fast to permit a role in cell-cell communication during gastrulation. In cultured mammalian cells, tunneling nanotubes have been shown to be capable of organelle transfer [18], retroviral infection [19] and immune surface receptor transfer [20]. The motion of membrane-linked Dendra2 along the intercellular bridges seen here offers up the intriguing possibility that the intercellular bridges perform similar biological functions in zebrafish embryogenesis.

Conclusion
During zebrafish gastrulation, a significant fraction of epiblast cells are endowed with intercellular bridges that extend to connect epiblast cells as far away as 350 mm. These intercellular bridges are formed by the maintenance of a cytoskeleton-filled membrane tether between pairs of dividing cells. The bridges persist for several hours. The rate of transfer of integral (CD8-Dendra2) and peripheral (mDendra2) membrane proteins along the intercellular bridges is fast enough to mediate cell-cell communication during gastrulation and neurulation. The formation, persistence, and material transfer along the intercellular bridges provide an unexpected pathway for the transfer of materials between cells in the zebrafish embryo. Their presence has important implications for not only membrane homeostasis, but also on the mechanical and signaling processes in embryonic patterning and development.

Zebrafish
Raising, maintaining, injection and spawning of wild-type AB zebrafish in house colony were performed as previously described [21].

DNA Constructs
To generate embryos with high expression levels distributed in a mosaic subset of cells we used Tol2-mediated transient expression [8]. Briefly, the synthetic transposase mRNA and the transposon donor plasmid pMTB, which contains a b-Actin promoter flanked by minimal Tol2 cis-sequences necessary for transposition [22,23], were co-injected into the zebrafish fertilized eggs. The Tol2 construct is excised from the donor vector and eventually integrated into the genome of somatic cells. To target the photoconvertible protein Dendra2 [7], the red fluorescent protein mCherry [13] to the plasma membrane, we used as a lipid anchor 2 repeats of a myristoylated and palmitoylated NH2-terminal MGCIKSKRKDNLNDDE signal sequence from Lyn kinase as previously described [9,24]. The fusion constructs were generated by polymerase chain reaction primer extension, and further subcloned with ClaI and BamHI as the cloning sites into the transposon donor vector pMTB. Mouse CD8-Dendra2, a fusion protein between transmembrane mouse lymphocyte marker CD8 and Dendra2, was generated by subcloning a PCR product encoding the mouse CD8 (mCD8) gene [10] into the pCS2+Den-dra2 vector with BamHI and XmaI as the cloning sites, generating a new ORF with Dendra2 fused in frame to the 39 of mCD8. Finally, mCD8-Dendra2 was sub-cloned into the transposon donor vector pMTB with BamHI and SnaBI as the cloning sites. EGFP-a-Tubulin and EGFP-b-Actin were generated by subcloning both PCR products from the pEGFP-a-Tubulin vector and the pEGFP-b-Actin vector (Clontech), respectively, into the transposon donor vector pMTB vector with ClaI and SnaBI as the cloning sites. The H2B protein was tagged with mCherry [13] at its carboxyl terminus as previously described [25]. The H2B-mCherry gene were then sub-cloned into the transposon donor vector pMTB.

Imaging, Dendra2 Photoconversion and Quantification
Injected Embryos were raised at 28uC in 30% Danieau solution [26] and developed indistinguishably from un-injected counterparts. Prior to imaging, specimens were oriented and placed in 1% low melting point agarose. Live images were obtained with a LD C-APO 40x/1.1 water objective on a Zeiss LSM 5 Exciter microscope setup. Photoconversion of Dendra2 was achieved by a 40 second exposure to intense 405 nm laser light. Images were processed using Adobe Photoshop CS3 (Adobe Systems), and analyzed using Imaris (Bitplane AG). The MATLAB (The MathWorks, Inc.) curve fitting toolbox was used to calculate the behavior of photoconverted Dendra moving along the intercellular bridges. Movie S1 Intercellular Bridge Formation (related to Figure 2). Time-lapse imaging capturing the formation of an intercellular bridge after cell division at late blastula stage. Nuclei are labelled with H2B-mCherry (red) and cellular boundaries are labelled with mDendra2 (green). Scale bar 25: mm.

(AVI)
Movie S2 Intercellular Bridge Persistence During Gastrulation (related to Figure 2). Time-lapse imaging showing the stability of an intercellular bridge (red arrowheads) labelled with mDendra2 (green) from mid-gastrula to the end of gastrulation. Scale of the grid: 25 mm.