Extreme nuclear branching in healthy epidermal cells of the Xenopus tail fin

Nuclear shape is highly conserved across cell types and species, with most healthy cells having round or ellipsoid nuclei. A few healthy cell types exhibit non-ellipsoid morphologies, including neutrophils, which have a distinct lobular structure that allows them to extravasate to areas with damaged tissue (Pillay et al., 2013; Rowat et al., 2013). Frequently, perturbations in nuclear morphology are associated with disease. Well-studied examples include progeria (Chen et al., 2014; Dahl et al., 2006; Goldman et al., 2004; Schirmer et al., 2001; Verstraeten et al., 2008), muscular dystrophy (Bonne et al., 1999), neurodegeneration (Frost et al., 2016) and cancers (Denais and Lammerding, 2014; Fu et al., 2012; Shah et al., 2013). HeLa cells in particular demonstrate nuclear morphological variation, which includes blebbing and ruffling of the nuclear membrane and dysregulation of multiple nucleoskeletal components (Wiggan et al., 2017). However, it is largely unclear what general mechanisms allow cells to acquire non-ellipsoid nuclear morphologies or how these morphologies might influence tissue and cellular function. One barrier to understanding extreme morphological variation of the nucleus is the dearth of models where nuclear morphology varies in the absence of disease. Here, we characterize Xenopus tropicalis tadpole fin margin epidermal cells, which have a non-ellipsoid, branched nuclear architecture. These striking nuclear morphologies arise during tail development and persist late into metamorphosis.

The nucleus derives its shape from interactions between the nucleoskeleton and the actin cytoskeleton. The nucleoskeleton is a complex network of lamin filaments, associated proteins, and families of linker of nucleoskeleton and cytoskeleton (LINC) complexes (Chang et al., 2015; Chen et al., 2014; Davidson and Lammerding, 2014; Denais and Lammerding, 2014; Fu et al., 2012; Goldman et al., 2004; Schirmer et al., 2001; Vergnes et al., 2004; Zwerger et al., 2013). Alterations in nuclear lamina composition, particularly the relative levels of A-type and B-type lamins, enable changes in not only nuclear shape but also nuclear deformability (Swift et al., 2013). Changes in the ratio of these protein types allow the formation of nuclear lobes and a highly deformable nuclear envelope in neutrophils, which in turn enables passage through small capillaries. Perturbation of B-type lamins or of Lamin B receptor (LBR) has deleterious effect on neutrophil migration (Dreesen et al., 2013; Rowat et al., 2013). More recent studies of interactions between perinuclear actin and the nuclear envelope have also clarified that the rigidity of the actin cap and the degree of actin polymerization directly affect nuclear shape and tissue stiffness (Swift et al., 2013; Wiggan et al., 2017). Variation in nuclear morphology is therefore predicted to have consequences for the biophysical function of the associated tissue, although relatively little is known about the mechanism by which other nuclear functions are modulated or constrained by extreme shape change (Dahl et al., 2006; Pajerowski et al., 2007; Rowat et al., 2013; Zwerger et al., 2013).

The structural organization of the nuclear lamina scaffolds functional domains within chromatin and serves to protect the genome (Peric-Hupkes et al., 2010; Shah et al., 2013; Solovei et al., 2013). Chromatin–lamina interactions are important for appropriate gene regulation. Canonically, heterochromatin or repressed regions of the genome are associated with the nuclear lamina (Fraser et al., 2015; Mattout et al., 2015a; Peric-Hupkes et al., 2010). Alterations in heterochromatin propagation are linked to changes in nuclear morphology caused by laminopathies (Davidson and Lammerding, 2014; Dreesen et al., 2013; Perovanovic et al., 2016; Shah et al., 2013). Hutchinson–Gilford progeria syndrome (HGPS) is a laminopathy that causes premature aging and is associated with mutations in the gene encoding lamin A, LMNA, that disrupt prelamin A cleavage, leading to gross changes in nuclear morphology (Goldman et al., 2004; Dahl et al., 2006; Verstraeten et al., 2008; Chen et al., 2014). As cultured cells with lmna HGPS mutations undergo more passages, they acquire progressively more nuclear ruffling and alterations of heterochromatin, resembling senescent cells rather than proliferative cells. Similar alterations in heterochromatic regions are seen in Lamin B1-depleted cells and cancer cells (Perovanovic et al., 2016; Shah et al., 2013). This suggests that alteration of the nucleoskeleton can contribute to large-scale changes in chromatin reorganization and gene expression that contribute to aging or other pathologies.

In this study, we have characterized nuclear branching in the fin epithelium of X. tropicalis tadpoles. The thin epithelium of the tadpole is made up of flattened epidermal cells that overlie a mesenchymal core (Tucker and Slack, 2004). The specialized cell biological and biophysical properties of the epithelial cells allow rapid regeneration and sinusoidal swimming movements. We show that branched morphologies of the nuclear lumen, chromatin and nuclear lamina arise during development in a heterogenous population of epidermal cells that make up the fin periphery. Cells with branched nuclei contain epigenetic marks of active enhancers and inactive chromatin throughout the nucleoplasm, and have active cell cycles. We found that actin filaments, but not polymerized microtubules, are necessary to maintain branched nuclear morphology. We also found that functional epidermal Lamin B1 is required for both nuclear branching and for proper development of the fin and tail.

Although Xenopus has long served as a model for epidermal cell biology and nuclear composition, there has been little examination of nuclear morphology in the differentiated fin. We conducted whole-embryo DAPI stains of the X. tropicalis tadpole fin, which revealed an unexpected elaborately branched distribution of DNA in the fin marginal cells (Fig. 1A). Although examples of potentially branched nuclear morphologies can be observed in earlier literature, these have not been described in detail (Davis and Kirschner, 2000). Our first goal was to establish whether branching was confined to chromatin or was shared by the nucleoplasm and envelope (Fig. 1B). To this end, cleavage-stage embryos were injected with either a cocktail of mRNA encoding H2B–RFP and Lamin-B3–GFP (LmnB3–GFP) to label histones (chromatin) and the nuclear lamina respectively, or with GFP bearing a nuclear localization signal (Nuc–GFP) (Fig. 1C). Tadpoles were reared to Nieuwkoop and Faber (NF) stage 41, and then live images were taken in the anterior-most third of the fin margin. Visualization of Nuc–GFP confirmed that the nucleoplasm in cells of the fin margin is also a highly branched structure (Fig. 1D). Consistent with our observations for DAPI, H2B–RFP labeling of chromatin also revealed a branched distribution in the nuclei of fin margin cells. LmnB3–GFP localization showed that the nuclear compartment is also branched (Fig. 1E). Thus, the entire nucleus of fin marginal cells is branched.

We next asked what the spatiotemporal distribution of nuclear branching is during Xenopus development. Because injected mRNA has a limited lifetime, we utilized immunohistochemistry to explore endogenous nuclear structure in the epidermis and other tissues through development. Tadpoles were fixed at various stages and stained with Lamin B1 antibody to show the nuclear periphery and, chromatin labeled with DAPI. We found that by late neurula stages (NF stage 18), the nuclear envelope is ruffled and irregular, though the chromatin distribution is still largely ellipsoid (Fig. 1F). We note that ruffling of the nuclear envelope is found throughout the fin epidermis, spanning multiple cell types (Fig. S1). At NF stage 18, ruffling of the nuclear envelope is more pronounced. As the embryo enters tailbud stages, the distribution of both chromatin and the nuclear lamina becomes gradually more branched, with defined branches appearing by NF stage 26, multiple branches evident per nucleus by NF stage 36, and the most elaborate degree of branching reached by NF stage 41 (Fig. 1E). The absolute number of branches per nucleus is variable, ranging from 2 to 13 (Fig. S1). Branching persists in the distal tail but is reduced through late tadpole stage and limb formation (Fig. 1F). All nuclear branching is lost shortly before the onset of tail reabsorption, and epidermal cells of the adult frog are not branched (data not shown). Although non-ellipsoid nuclear structures are also visible in some other cell types, notably granulocytes and/or neutrophils, nuclear branching was only observed in epidermal cells. Epidermal cells of the head at stage 41 showed some minor lobulation, whereas nuclei of other tissues such as the heart and somites were ellipsoid (Fig. S1). The only structure in which we identified branched nuclei outside the tail was the surface epithelial cells covering the prospective cornea (Fig. S1).

Because perturbations of nuclear morphology are associated with pathology in many cell types (Li et al., 2016; Wang et al., 2008), we asked whether epidermal cells with branched nuclei showed hallmarks of cellular damage or senescence. These could include nuclear envelope rupture, mitochondrial damage or cell cycle exit. To assess subcellular signs of cell damage, we utilized transmission electron microscopy (TEM) to assess nuclear envelope integrity and mitochondrial abundance. Micrographs reveal diverse nuclear structures (Fig. 2A), including clearly demarcated branched nuclei enclosed by bilayer nuclear envelopes. Upon close examination of the nuclear envelope, we see that it is a continuous bilayer in cells with branched nuclei, with an average of 18 nm between the inner and outer leaflets and containing nuclear pores with a mean diameter of 60 nm (Fig. 2B). The integrity of the nuclear envelope is also supported by the even distribution of Nuc–GFP within the nuclear lumen (Fig. 1D) and by the continuous distribution of Lamin B1 in branched nuclei, with no evidence of leaks, partitions or ruptures (Fig. 1E). Cells with branched nuclei also have numerous mitochondria with abundant cristae (Fig. 2A). These observations suggest that cells with branched nuclei are not undergoing apoptotic or senescent processes that would be reflected in nuclear envelope breakdown, low mitochondrial numbers or loss of mitochondrial cristae.

TEM did reveal some atypical features in branched nuclei, including a lack of well-defined regions of perinuclear increased electron density in micrographs that would be indicative of heterochromatic regions, or of clear identifiable nucleoli (Fig. 2A). To determine whether there were nucleoli present in nuclear branches we analyzed localization of the nucleolar marker fibrillarin (Brangwynne et al., 2011). We found that cells with branched nuclei did contain foci of fibrillarin (Fig. 2D), suggesting the presence of nucleoli, and that the average number of foci per nucleus did not change significantly between branched (1.60) and unbranched nuclei (1.69) (Fig. 2C). We did note that foci of fibrillarin did not correspond to apparent foci of H2B (arrowhead in Fig. 2D, right panel).

To better understand whether cells with branched nuclei contain both active and inactive chromatin domains, we used immunofluorescence and live imaging to examine the distribution of histone modifications associated with active enhancers (H3K27ac and H3K4me1), and heterochromatin (H3K9me3 and HP1β). Through immunofluorescence imaging we found that H3K27ac, H3K4me1 and H3K9me3 are all distributed broadly in branched nuclei at NF stage 41 (Fig. 2E); the distribution seems uniform for H3K27ac, but H3K9me3 seemed more concentrated in foci, and H3K4me1 may be excluded from some regions of the nuclear periphery. To better characterize the distribution of heterochromatin, we used a GFP fusion of the heterochromatin-binding protein HP1β (Mattout et al., 2015b; Cheutin et al., 2003). We found that HP1β–GFP broadly colocalizes with H2B (Fig. 2F). Although heterochromatin is typically enriched at the nuclear envelope, we did not observe a clear enrichment of HP1β at the nuclear periphery; however, we observed foci of HP1β–GFP fluorescence in the nucleus corresponding to foci in H2B. Taken together, this suggests that active enhancers and heterochromatin are widely distributed in nuclear branches.

In many cell types, breakdown of ellipsoid nuclear morphology is a hallmark of senescence, cell cycle dysregulation or genomic instability (Dahl et al., 2006; Goldman et al., 2004; Schirmer et al., 2001; Wang et al., 2008). In particular, keratinocytes are known to acquire aberrant nuclear morphologies following terminal differentiation and cell cycle exit, and in premature aging syndromes (Gdula et al., 2013; McKenna et al., 2014). We therefore wanted to determine whether cells with branched nuclei in the keratin-rich tadpole epidermis were undergoing an active cell cycle. We utilized immunofluorescence of phosphorylated histone H3 (pH3) to mark mitotic nuclei, and Lamin B1 to mark the nuclear periphery. We found numerous examples of pH3-positive cells that retain branched nuclei (Fig. 3A). Examination of chromatin morphology in pH3-positive cells suggests that nuclei remain branched and are still enclosed by a branched nuclear envelope through prophase but form a condensed metaphase plate while the nuclear envelope breaks down. Chromatin distribution remains unbranched through anaphase. Daughter cells establish independent branching patterns and re-form the nuclear envelope at late telophase.

To better characterize nuclear envelope and chromatin dynamics through mitosis, we conducted live imaging through visualization of H2B–RFP and GFP-labeling of the plasma membrane (Mem–GFP) to track individual nuclei throughout mitosis (Movie 1, single frames in Fig. 3B). These confirmed our initial observations that nuclei are initially branched, form morphologically normal metaphase plates that segregate into two well-defined populations at anaphase, and are re-enclosed by the nuclear envelope following telophase, with the nucleus beginning to re-form branches ∼21 min after cytokinesis (Movie 1). Nuclear branching patterns in daughter cells do not typically recapitulate those of the mother cell, nor do both daughters show the same branching patterns. Additionally, branches seemed relatively stable once formed after the completion of mitosis, and did not appear to be reabsorbed but did exhibit some ‘wiggling’ within the branches. In nuclei not undergoing mitosis, the number and relative positions of branches can remain stable for 2 h or more. Both fixed and live imaging therefore demonstrate that cells with branched nuclei are able to undergo mitosis.

We next sought to determine what molecular mechanisms enable nuclear branching in the fin margin. In mammalian cells, perturbations of nucleoskeleton components lead to nuclear shape deformation. These include mutations in LMNA, which lead to nuclear blebbing in progeroid syndromes (Chen et al., 2014; Dahl et al., 2006; Goldman et al., 2004; Perovanovic et al., 2016; Verstraeten et al., 2008), mutations or duplications of LMNB1 or LBR, which disrupt nuclear flexibility and extravasation in neutrophils (Dreesen et al., 2013), perturbations of the sun and nesprin components of the LINC complex (Chang et al., 2015; Hatch and Hetzer, 2016; Kim and Wirtz, 2015), or alterations in the abundance, orientation or phosphorylation of actin, which contribute to nuclear morphological disruption in HeLa cells (Ho et al., 2013; Kim and Wirtz, 2015; King and Lusk, 2016; Ramdas and Shivashankar, 2015; Webster et al., 2009; Wiggan et al., 2017; Zwerger et al., 2013). Therefore, we decided to pursue whether similar components were required for nuclear branching in the fin margin.

First, we observed actin localization in cells with branched nuclei. We found no apparent bias of actin localization to tips or bases of branches (Fig. 4A). To determine whether actin filaments were necessary for nuclear branches, we incubated stage 41 tadpoles with latrunculin B (Lat B), which disrupts actin filament formation and other actin-dependent processes in Xenopus at non-lethal doses (Lee and Harland, 2007). To monitor the effect of this inhibitor on actin filaments and nuclear morphology, we injected embryos at cleavage stages with mRNA encoding H2B–RFP and the actin-binding protein Utrophin–GFP. We found that treatment with Lat B results in breakdown of the actin cytoskeleton beginning at 25 min post treatment. At this time, foci of Utrophin-GFP were visible (Movie 2; Fig. 4B). Nuclear branches were gradually lost after actin destabilization and were lost more slowly in nuclei that initially had more numerous or complex branches. Nuclear branches reformed after washout of Lat B. Actin filaments visualized by labeling with LifeAct began to be visible 25 min after Lat B removal along with some nuclear deformation, with similar kinetics to what was observed for the loss of actin filaments. By 125 min after Lat B removal, new branches were fully formed although the branching patterns were not conserved relative to their initial pre-treatment distribution (Fig. 4C).

Because actin has known roles in compressing nuclei (Versaevel et al., 2012; Vishavkarma et al., 2014; Wiggan et al., 2017) and epithelial cells are extremely flat, we measured nuclear height and surface-area-to-volume ratios of nuclei with intact and Lat B-perturbed actin networks. Both untreated and Lat B-treated tadpoles had comparable nuclear height (5.6 and 4.1 μm, respectively, Fig. 4D). However, the nuclear surface-area-to-volume ratio decreased from 1.405 in untreated to 1.134 in Lat B-treated animals (Fig. 4E). Nuclear volume also decreased by ∼7.5% and the surface area decreased by ∼25.4% in Lat B-treated animals (data not shown). Taken together, this suggests that loss of branches decreases the amount of nuclear membrane (surface area) relative to the volume. However, the lack of change in nuclear height suggests that actin is not acting compressively, but instead forms branches orthogonally to the plane of the tissue.

To confirm that loss of nuclear branches was not specific to Lat B treatment, we utilized cytochalasin D (Cyto D), which inhibits actin polymerization and also disrupts other actin-dependent processes in Xenopus (Lee and Harland, 2007). Treatment with either Cyto D or Lat B results in a rapid loss of nuclear branching in the fin margin, as revealed through visualization of LmnB3–GFP and H2B–RFP (Fig. 4F).

We next asked whether microtubules contributed to nuclear branches, as they have been shown to play a role in maintaining nuclear morphology (Tariq et al., 2017). We utilized the microtubule polymerization inhibitor nocodazole at non-lethal doses (Dutta and Kumar Sinha, 2015). Although nocodazole treatment noticeably disrupts spindle formation in tadpoles, it does not affect nuclear morphology relative to DMSO-treated controls (Fig. 4G).

Through quantification of changes in nuclear morphology for all cytoskeleton perturbations using a circularity measurement (Fig. S2), we found that there was a statistically significant increase in epidermal nuclear circularity when tadpoles were treated with either Cyto D or Lat B, but not nocodazole (Fig. 4H). These results indicate that nuclear branches require intact F-actin, but not microtubules.

We next asked whether nuclear branching relies on specific components of the nuclear lamina. Modulation of nuclear lamina components has been shown to regulate tissue elasticity in mammals; greater amounts of lamin A contribute to stiffer tissue, whereas Lamin B is crucial for nuclear envelope flexibility in neutrophils (Mattout et al., 2015a,b; Peric-Hupkes et al., 2010; Perovanovic et al., 2016; Solovei et al., 2013; Towbin et al., 2012). Xenopus tropicalis have four lamin genes; lmna, lmnb1, lmnb2 and the germline-specific lmnb3 (Session et al., 2016). We first sought to determine whether any of these components were preferentially enriched or depleted in fin marginal cells containing branched nuclei. To this end, we isolated fin margin tissue or whole embryo tissue, and quantified expression of lmnb1, lmnb2 and lmna using quantitative real-time PCR (qRT-PCR) (Fig. S2). We found that expression of lmnb1 is significantly upregulated in the fin margin relative to the whole embryo (2.6-fold increase), whereas lmnb2 is unchanged and lmna mRNA levels is reduced in the fin margin. To determine whether this upregulation reflected a functional role for lmnb1 in the fin margin or in nuclear branching, we used CRISPR/Cas9 to create mutations in lmnb1 by co-injecting gene-specific single guide RNA (sgRNA) together with humanized Cas9 protein in F0 tadpoles (Bhattacharya et al., 2015; Nakayama et al., 2013). To track nuclear and cell morphology, we again co-injected these embryos with mRNA encoding H2B–RFP and Mem–GFP. We used high-resolution melt analysis to confirm gene-specific mutations (Fig. S2). Upon analyzing nuclear morphology in F0 tadpoles at stage 41, we found that only lmnB1 mutant embryos (Fig. 5A, Whole embryo lmnb1 CRISPR) had markedly reduced branching, and instead displayed nuclei with crescent or elongated obloid shapes. This effect was confined to lmnb1 mutants and was not induced by injection of a scrambled version of the lmnb1 sgRNA (Fig. 5A, Scrmbl CRISPR).

To confirm that these effects were intrinsic to epidermal cells, and not secondary to any effect of whole-embryo perturbation, we next targeted our injections specifically to the ventral animal blastomeres at the 8-cell stage, which give rise to the epidermal lineage (Bauer et al., 1994; Moody, 1987). These epidermal-only lmnb1 mutant embryos (Fig. 5A, Epidermal lmnb1 CRISPR) also exhibited reduced nuclear branching in the fin epidermis. As a further confirmation, we generated a Xenopus form of dominant-negative Lamin B1, based on the mammalian protein domain structure (Goldman et al., 2004). This dominant-negative Lamin B1 (Lmnb1-rod) contains only the rod domain, which is thought to disrupt the lamin network and LINC complex interactions when overexpressed (Goldman et al., 2004). We co-injected mRNA encoding this mutant protein into epidermal blastomeres at the 8-cell stage, together with mRNA encoding H2B–RFP and Mem–GFP. Epidermal cells expressing Lmnb1-rod also exhibited reduced nuclear branching at stage 41 (Fig. 5A). We quantified nuclear circularity in lmnb1-perturbed tadpoles and found that epidermal cells in Whole embryo lmnb1 CRISPR, Epidermal lmnb1 CRISPR and Lmnb1-rod tadpoles exhibited a statistically significant increase in nuclear circularity compared with wild type, whereas there was no significant difference between wild-type and Scrmbl CRISPR tadpoles (Fig. 5B). Taken together, these results argue that nuclear branching depends on functional Lamin B1 in the epidermis.

One potential source of nuclear morphological variation derives from the physical forces exerted against the nucleus either by intracellular compression, such as through perinuclear actin, or by extracellular compression, such as that exerted by endothelial cells during neutrophil extravasation. Tadpole fin nuclei encounter a unique extracellular force profile during swimming. We therefore tested whether the mechanical forces sustained by the fin during swimming were required for nuclear branching in the fin. To this end, we utilized dorsal-posterior explants at the neurula stage. These explants give rise to tails with fins that lack muscle (Tucker and Slack, 2004). We then compared the circularity of nuclei in the fin margin of explants and stage-matched wild-type tadpoles and found no significant change in circularity (Fig. 6A). This suggests that the mechanical forces exerted on the fin during swimming are not necessary to induce nuclear branching.

We concluded by investigating the relationship between nuclear branching and development. We observed that tadpoles with lmnb1 mutations had tail defects, sloughing of the fin epidermis, were inefficient swimmers and developed edema, likely due to decreased locomotion (Fig. 6B,C). We found that nuclear morphology and fin morphology were correlated; in lmnb1 sgRNA-injected tadpoles that had no tail defect, we did not observe changes in nuclear morphology (Fig. S2). Tails were statistically significantly shorter and narrower in tadpoles with lmnb1 mutations compared to wild type (Fig. 6D,E). This suggests that lmnb1-dependent nuclear branching may be necessary for proper tail formation and function, but does not rule out the possibility that tail formation requires Lamin B1-dependent gene regulation.

Across tissues and species, nuclear morphology is generally ellipsoid. There are very few cases of healthy epithelial cell types with highly irregular nuclear morphologies; the mandibular gland epithelium of the wax moth Ephestia kuehniella is a notable example, which exhibits a branched nuclear morphology similar to the morphology of the tadpole fin (Buntrock et al., 2012). Here, we describe an epithelial epidermal tissue, the Xenopus tadpole fin margin, which exhibits a highly branched nuclear morphology. In this epithelial tissue, a heterogeneous population of cell types, including secretory cells, goblet cells and multiciliated cells, all have highly irregular branched nuclear morphology. The degree of nuclear branching we describe is more extreme than in most other instances of nuclear morphological changes seen among both healthy and diseased vertebrate cells. Like neutrophils, these fin margin cells develop branched nuclear morphology over the course of development, but unlike neutrophils, nuclear branching is lost in the tail at the end of metamorphosis.

Xenopus has long served as a model organism for nuclear morphology, including molecular and cell biological characterization of nuclear envelope components, and the cell biological consequences of their perturbation. Overexpression of specific lamin components has been observed to alter both nuclear shape and size in oocyte nuclei from X. laevis, and nuclear size scaling in early Xenopus embryos is dependent on cytoplasmic volume as well as the nuclear transport factors importin α and NTF2 (Good et al., 2013; Jevtić and Levy, 2015; Jevtić et al., 2015; Levy and Heald, 2010). Recently, Xenopus has served as a source model for proteomic studies of nuclear composition (Wühr et al., 2015). Morphological variation in nuclei later in embryogenesis has not been examined in depth. Here, we found that nuclear branching begins late in neurulation, well after the initial specification of epidermal fate, at a developmental stage similar to when multiciliated cells begin to undergo apical emergence (Sedzinski et al., 2016). Nuclear branching is dramatically elaborated as the tail elongates, and by late tailbud stages highly branched nuclei are found both in multiciliated cells and their goblet cell neighbors. Our data suggest that nuclear branching is a general property of the fin margin epithelium in Xenopus. The extremity of morphological variation observed highlights that these nuclei are a valuable model for nuclear diversity; they are easily imaged in a whole-organism system that is readily modulated both genetically and through treatment with small molecules.

Xenopus epidermal fin cells exhibit branched nuclear morphologies while maintaining an active cell cycle. This is in contrast to the non-ellipsoid nuclear morphologies that occur in post-mitotic cells. In the epidermis, keratinocyte nuclei undergo flattening and become irregular after cell cycle exit; these phenotypes become more extreme with aging or in specific disease scenarios (Yang et al., 2011). Among actively cycling cells, nuclear morphological perturbations such as blebbing or nuclear ruffling are common in cancer cells but very infrequent in healthy cell types (Denais and Lammerding, 2014; Fu et al., 2012; Pillay et al., 2013; Shah et al., 2013). We found that nuclear branching is common in mitotic epidermal cells in the Xenopus tail fin, with rapid collapse of nuclear branching ∼7 min before metaphase and re-formation of nuclear branches becoming apparent 21 min after cytokinesis (Movie 1). Branched nuclei do appear to undergo complete nuclear envelope breakdown, and we have not found evidence of karyomeres or chromosome-specific nuclear envelopes as are seen in the early mitoses of zebrafish and Xenopus (Lemaitre et al., 1998; Schoft et al., 2003). Following mitosis, the branched structures formed by the two daughter cells are distinct and do not recapitulate the nuclear morphology of the mother cell.

In cells with branched nuclei we found that both active enhancer chromatin marks and heterochromatin are distributed throughout the nucleus. Normally, regions of chromatin that are transcriptionally repressed are associated with the nuclear lamina and the periphery of the nucleolus (Mattout et al., 2015a,b; Peric-Hupkes et al., 2010; Perovanovic et al., 2016; Solovei et al., 2013; Towbin et al., 2012). Through TEM examination of branched epidermal nuclei, we saw no increase in electron density around the nuclear periphery, nor did we find evidence of increased electron density representing a nucleolus in most nuclei. However, we found that cells with branched nuclei did contain foci of fibrillarin, a component of the nucleolus, in regions distinct from the densest chromatin as represented by H2B fluorescence. Interestingly, we do see puncta of increased intensity of HP1β and H3K9me3 but have not yet seen corresponding regions of increased density through use of TEM. Taken together, this suggests that cells with branched nuclei may partition heterochromatin without anchoring these regions to the nuclear lamina. Future research will examine the organization of heterochromatin in branched nuclei, and the organization of specific chromatin domains within these nuclei.

Previous work has shown a role for the nuclear lamina, LINC complexes and cytoplasmic actin in the shaping of the nucleus (Chen et al., 2014; Hatch and Hetzer, 2016; Hatch et al., 2013; Ho et al., 2013; Kim and Wirtz, 2015; King and Lusk, 2016; Lammerding et al., 2006; Ramdas and Shivashankar, 2015; Webster et al., 2009; Wiggan et al., 2017). Here, we have shown that both an intact actin network and Lamin B1 are necessary to maintain nuclear branches. Our working model therefore suggests that actin and Lamin B1 filaments contribute to the formation and/or maintenance of nuclear branches (Fig. 6F). It remains unclear how F-actin and Lamin B1 interact with other nuclear envelope components, and particularly LINC complexes, in order to carry out these roles. Additional studies are needed to understand how Lamin B1 and F-actin each contribute to nuclear branching.

Actin is known to play a role in compressing the nucleus through manipulation of stress fibers during migration or passage through narrow openings (Versaevel et al., 2012; Vishavkarma et al., 2014; Wiggan et al., 2017). In laminopathies, nuclei lose rounded morphologies and adopt more irregular architectures. Previous studies have also shown that gaps in the nuclear lamina allow blebbing (Hatch et al., 2013). Conversely, the fin margin nuclei appear to have a fully functional lamina network with no apparent gaps. There was no obvious localization of actin to the base or tips of nuclear branches indicative of actin pushing or pulling on the nucleus, but loss of actin caused nuclei to increase in circularity, suggesting that actin does contribute to nuclear branching by some other mechanism. We also found that loss of actin did not increase nuclear height, suggesting that the extracellular matrix or other force-transmitting molecules cause the flattening of this tissue. We did find that there was a decrease in nuclear surface-area-to-volume ratio when F-actin was lost, as well as a modest decrease in total nuclear volume. These observations both suggest that nuclear envelope distribution, and possibly quantity, are closely linked to F-actin in these nuclei. While it is clear F-actin is necessary to maintain nuclear branches, it is unclear how nuclear actin or actin­-binding proteins contribute to maintenance and establishment of nuclear branches. Previous studies have shown a relationship between cell spreading and nuclear actin polymerization, raising the possibility that, in this flattened epithelium, nuclear actin may contribute to nuclear branch formation (Keeling et al., 2017; Plessner et al., 2015). Another possibility that we have not yet been able to test explicitly is the role for intra-nuclear actin filaments (Baarlink et al., 2017; Kalendová et al., 2014; Oda et al., 2017), which may also contribute to nuclear branching. Our time-lapse movies show that nuclear morphological change can be tracked closely in time with cytoplasmic F-actin disruption. We therefore favor the hypothesis that cytoplasmic F-actin is critical to nuclear morphology, but intranuclear actin may also contribute to the formation or stabilization of branches.

Perturbations of nuclear branching have deleterious effects on the formation of the fin and consequently on its downstream function. Although we have been able to show that specific nucleoskeletal components are required for nuclear branching, the ultimate role of branched nuclear morphologies in tail fin cell and tissue function remains open. Nuclear branching may play a role in genomic organization or gene regulation, as discussed above, or in fin biomechanics.

The thin epithelium of the tadpole fin is made up of flattened epidermal cells that overlie a mesenchymal core. Its specialized cell biological and biophysical properties allow rapid regeneration and sinusoidal swimming movements (Tucker and Slack, 2004). To accommodate this structure, a flattened nuclear structure would be advantageous, and nuclear branching could impart biophysical properties necessary for tissue function. The elastic modulus of the nucleus has been shown to be different from that of the cytoskeleton. The irregular nuclear structure could aid in creating a more uniform elastic modulus of the tissue, as opposed to localized regions of differential stiffness (Guilak et al., 2000; Kha et al., 2004; Pajerowski et al., 2007). The requirement of Lamin B1 to maintain nuclear branches suggests that nuclear branching could be modulating tissue stiffness (Kha et al., 2004; King and Lusk, 2016; Pajerowski et al., 2007; Swift et al., 2013; Verstraeten et al., 2008; Zwerger et al., 2013).

In conclusion, we have shown that the fin epithelium of the X. tropicalis tadpole tail contains a heterogenous population of cells that have branched nuclear structures. These cells with branched nuclei are healthy and have active cell cycles. Additionally, we have shown that nuclear branching depends on an intact actin network and Lamin B1. We determined that forces incurred from swimming are not necessary to induce nuclear branches; however, loss of nuclear branching resulting from lmnb1 mutations decreases swimming efficiency and impedes tail and fin development. These cells offer a novel system to study extreme nuclear morphological variation in a healthy tissue.

Use of X. tropicalis was carried out under the approval and oversight of the Institutional Animal Care and Use Committee at the University of Washington, an Assessment and Accreditation of Laboratory Animal Care-accredited institution. Ovulation of adult X. tropicalis and generation of embryos by in vitro fertilization were according to published methods (Khokha et al., 2002; Sive et al., 2010). Fertilized eggs were de-jellied in 3% cysteine in 0.11 × modified frog Ringer's solution (1/9th MR) for 10–15 min. Embryos were reared as described previously (Khokha et al., 2002). Staging was assessed by the Nieuwkoop and Faber (1994) method.

DNA plasmids were linearized at appropriate restriction sites and mRNA was transcribed with Sp6 mMessage mMachine kits (Ambion). mRNA was injected into embryos at the 1–8-cell stage at a dose of 100 pg/embryo, depending on experiment. The following constructs were generated in pCS2+ and linearized with NotI: Nuc–GFP, H2B–RFP and Mem–GFP (generous gifts from Richard Harland, University of California Berkeley, USA); Utrophin–GFP and LifeAct–GFP (generous gifts from John Wallingford, University of Texas Austin, USA); LmnB3–GFP (generous gift from Daniel Levy, University of Wyoming, USA). Lmnb1-rod-only and fibrillarin–GFP (generous gift from Clifford Brangwynne, Princeton University, USA) were generated from pCS107 that was linearized with KpnI. Finally, GFP–HP1β was linearized from a pBCHGN construct (Addgene #17651; Cheutin et al., 2003) with KpnI.

Terminally anesthetized Xenopus tropicalis embryos were fixed for 20 min in MEMFA at room temperature. Embryos were permeabilized by washing for 3×20 min in PBS+0.01% Triton X-100 (PBT). Embryos were blocked for 1 h at room temperature in 10% CAS-block (Invitrogen, 00-8120) in PBT. Then embryos were incubated in primary antibodies in 100% CAS-block overnight at 4°C. Embryos were then washed 3×10 min at room temperature in PBT and re-blocked for 30 min in 10% CAS-block in PBT. Secondary antibodies were diluted in 100% CAS-block and incubated for 2 h. Embryos were then washed 3×20 min in PBT. Whole embryos or isolated tails were mounted on slides in Vectashield containing DAPI (Vector Laboratories, H-1500). Images were acquired with a Leica DM 5500 B and ORCA-flash 4.0LT camera. Anti-Lamin B1 (Abcam, 16048), and anti-phospho-H3 (Abcam, 14955) were utilized at a 1:1000 dilution. Anti-H3K27ac (Abcam, 4729), anti-β-tubulin (Sigma, T8535), anti-H3K27Me3 (Abcam, 6002-100), anti-H3K4Me (Abcam, 8895), anti-H3K9me3 (Active Motif, 39162), anti-mouse (Life Technologies, A21422), and anti-rabbit (Life Technologies, A11008) were utilized at a 1:500 dilution. α-tubulin (Invitrogen, 62204) was utilized at a 1:250 dilution.

Branches were counted as the number of termini of the nucleus (Fig. S1), from images of tadpoles with nuclear markers of H2B, DAPI, or nuclear-localized GFP.

Tadpoles were imaged sedated in 0.01% tricaine in 0.11 × MR. Tadpoles were mounted for imaging as previously described (Kieserman et al., 2010; Wallingford, 2010) with the following modifications for actin and Lamin B1 perturbation (Figs 4 and 5). A perimeter of vacuum grease was made on a glass slide. A tadpole was placed in the center of the vacuum grease perimeter with several drops of media containing drug (see ‘Pharmacological inhibitors’ section for details). A glass coverslip was gently pressed into the vacuum grease perimeter over the tadpole. Images were acquired with a Leica DM 5500 B camera. Mitosis and actin perturbation movies were acquired with a Zeiss 880 camera. Gross tadpole morphologies were acquired with a Leica M205 FA camera.

Stage 41 tadpoles were fixed in 2.5% glutaraldehyde+0.1 M sodium cacodylate buffer. Samples were washed four times in sodium cacodylate buffer, postfixed in osmium ferrocyanide (2% osmium tetroxide and 3% potassium ferrocyanide in 0.1 M sodium cacodylate buffer) for 1 h on ice, washed, incubated in 1% thiocarbohydrazide for 20 min, and washed again. Samples were washed and en bloc stained with 1% aqueous uranyl acetate overnight at 4°C. Samples were finally washed and en bloc stained with Walton's lead aspartate for 30 min at 60°C, dehydrated in a graded ethanol series, and embedded in Durcupan resin. All wash steps were performed in 0.1 M sodium cacodylate buffer. Serial sections were cut at 60 nm thickness and viewed on a JEOL-1230 microscope with an AMT XR80 camera (Giarmarco et al., 2017).

Total RNA was isolated from embryos (3–5 per experiment) or fin margin (15–20 per experiment) (Sive et al., 2010). RNA was treated with DNase I (Invitrogen, 18068015). cDNA was synthesized using a SuperScript III first strand synthesis kit (Invitrogen, 18080-051). Quantitative PCR analysis was performed using BioRad iCycler PCR machine, iQ Sybr Green mix (BioRad, 1708862) and analysis software. Primer sequences are as follows: lmnb1 forward 5′–AACCAGAACTCATGGGCAAC–3′, reverse 5′–ACTGTTGTGCGCTGTGCTAC–3′; lmnb2 forward 5′–ACAGGCATTGGATGAACTCC–3′, reverse 5′–TCAAGCTTGGCCTGATAGGT–3′; lmna forward 5′–ACTGTACCGATTCCCACAGC–3′, reverse 5′–GAGGAGCTGAGCTGGACAGT–3′; odc (orthithine decarboxylase) forward 5′–TTTGGTGCCACCCTTAAAAC–3′, reverse 5′–CCCATGTCAAAGACACATCG–3′.

A Gaussian blur was applied to all images in a data set and a threshold was applied to images. Particles were selected using FIJI (ImageJ) and manually refined. Particles were discarded if the whole nucleus was not in the field of view, if a partial particle was selected based on the original image, if a particle selected comprised two nuclei in the original image, or if a particle selected did not appear on the original image. After manual refinement, the circularity of particles were measured using FIJI (Schöchlin et al., 2014).

Latrunculin B (Sigma, L5288), cytochalasin D (Sigma, C8273) and nocodazole (CalBiochem, 31430-18-9) were resuspended using DMSO as a vehicle. Latrunculin B and cytochalasin D were equilibrated at room temperature for 1 h prior to use. For experiments, inhibitors were diluted to the following final concentrations in 0.11 × MR: 1 μM latrunculin B, 10 μM cytochalasin D (Lee and Harland, 2007), 150 μM nocodazole.

Imaris (Bitplane) was utilized to create 3D renderings and perform the surface area and volume calculations, with a surface area detail level for all treatments of 0.25 μm. Nuclei were excluded if volumes were below 100 μm³ or above 1000 μm³, as these were determined to be incomplete nuclei or fused nuclei, respectively, when images were examined.

CRISPR guides were designed from the V7.1 or V8 gene models on Xenbase and CRISPRscan. Target sites were chosen from UCSC tracks. Guides were chosen using the following criteria: no off-targets predicted, a score greater than 50, and in a region in or as close to exon 1 as possible. We generated site-specific sgRNAs a single oligo for each target site (target sequences listed below) 5’–CTAGCTAATACGACTCACTATAGG–(n18) target sequence–GTTAGGAGCTAGAAATAG–3’. These target sequences are as follows: LmnB1 guide 1 (G1), 5’– GGGAAGAGGTGCGGAGCC–3’; Lmn B1 guide 2 (G2), 5’–GCGGAGCCGGGAAGTGAG–3’; Scrmbl 5’– GGGAAGAGGGCGTGAGCC–3’. PCR was performed as described in Bhattacharya et al. (2015). SgRNA was transcribed using T7 mMachine kit (Ambion). Guides were injected into 1 or 2 cell embryos with 1.5 ng Cas9 (Bhattacharya et al., 2015; Nakayama et al., 2013). All guides were injected at a dose of 400 pg/embryo, all images show animals injected with LmnB1 guide 1.

Lamin B1 dominant-negative constructs were constructed following a similar strategy to that described in Schirmer et al. (2001), beginning with X. laevis Lamin B1 (Xenbase ORFeome clone XICD00712670) with the following primers: rod-only left primer, 5′–GGATCCATGGCCACTGCCACA–3′; right primer, 5′–GAATTCCAGTGGCAGAGG–3′.

To extract genomic DNA, individual tadpoles were lysed by heating at 95°C in 25 mM NaOH and 0.2 mM EDTA. Samples were cooled to room temperature and an equal volume of Tris-HCl 40 mM buffer was added. 2 μl of extracted genomic DNA was utilized in PCRs containing HRM master mix [GoTaq Flexi buffer (Promega, M8901), dNTPs, MgCl₂, DMSO, EvaGreen (Biotium, 31000), Taq polymerase (Quiagen, 201203), nuclease-free water]. The region of interest was amplified (left primer, 5′–GATCTGCAGGAGCTGAATGAC–3′; right primer, 5′–TGTTCCACGGAGATCTTACTGA–3′) for 35 cycles and melted from 60–95°C at 0.1°C increments with EvaGreen fluorescence measured after each temperature change.

Dorsal posterior explants were dissected between stage 15 and 17 as previously described (Tucker and Slack, 2004). Explants were cultured in Danilchik's for Amy (DFA) buffer without antibiotics. Sibling tadpoles were reared in 0.11 × MR as described above.

RStudio was utilized in generating statistics. One-way ANOVA, with Tukey's post hoc was utilized to calculate P-values for circularity and tail morphometric measurements; for qPCR, nucleoli number, and surface area and volume measurements, P-values were calculated with a two-tailed Student's t-test assuming unequal variance.

We thank Ed Parker for assisting with TEM imaging, and the University of Washington vision core facility (NEI P30EY001730). We acknowledge support from the W. M. Keck Center for Advanced Studies in Neural Signaling (NIH S10 OD016240) and the assistance of center manager Dr Nathaniel Peters. We are grateful to Nathaniel Ng of the Enrique Amaya lab (University of Manchester) for testing staging conditions for time-lapse analysis and preliminary movies. Alexander Chitsazan helped with training in R and advised on statistical methods. We thank the Molecular and Cell Biology of Xenopus Course at Cold Spring Harbor for embryology and microscopy training, the Xenopus Quantitative Imaging Course at the Marine Biological Laboratories for training in imaging and statistical analysis. We thank Xenbase for curation of genomic and literature information used to generate materials and conduct analysis. We thank Wills lab members, Emily Hatch of FHRC and John Wallingford of University of Texas Austin for comments on the manuscript. Finally, we thank Daniel Levy, University of Wyoming; Richard Harland, UC Berkley; Cliff Brangwynne, Princeton University and John Wallingford, University of Texas Austin, for materials.