Bridging the divide

Abstract

Ring canals are made from arrested cleavage furrows, and provide direct cytoplasmic connections among sibling cells. They are well documented for their participation in Drosophila oogenesis, but little is known about their role in several somatic tissues in which they are also found. Using a variety of genetic tools in live and fixed tissue, we recently demonstrated that rapid intercellular exchange occurs through somatic ring canals by diffusion, and presented evidence that ring canals permit equilibration of protein among transcriptionally mosaic cells. We also used a novel combination of markers to evaluate the extent of protein movement within and across mitotic clones in follicle cells and imaginal discs, providing evidence of robust movement of GFP between the 2 sides of mitotic clones and frequently into non-recombined cells. These data suggest that, depending on the experimental setup and proteins of interest, inter-clonal diffusion of protein may alter the interpretation of clonal data in follicle cells. Here, we discuss these results and provide additional insight into the impact of ring canals in Drosophila somatic tissues.

Introduction

Across all kingdoms of life, cell division is the foundational mechanism by which organisms multiply, grow, and develop. Many of the specific mechanisms of cell division are conserved across the animal kingdom, making alterations to this process especially interesting topics of study. One example of an alternative outcome of cell division is the formation of a stable intercellular bridge, formed when a cleavage furrow arrests prior to abscission of the plasma membrane that connects the dividing cells. Here referred to as ring canals (RCs), these structures are best known for their essential role during Drosophila oogenesis where they facilitate both diffusion and directed transport of cellular contents between the supporting nurse cells and rapidly developing oocyte.¹ Similar structures are also present during gametogenesis in many other species throughout the animal kingdom, including humans. However, it is not clear how RCs contribute to gametogenesis in other animals. Directed transport through RCs has not been observed outside of Drosophila oogenesis, and evidence indicates that although they are essential for spermatogenesis in mice, loss of germline RCs in female mice does not significantly alter fertility from 0–6 mo of age.² Other studies have offered hints at the ability of syncytial cells to share proteins, metabolites, and organelles, but these hypotheses have remained largely unexplored.^3-7 A more comprehensive picture of how ring canals affect cellular communication and other processes will shed light onto the tremendous conservation of this interesting deviation from normal cytokinesis.

Electron microscopy (EM) studies demonstrated the existence of RCs in Drosophila somatic tissues including the ovarian follicle cells, larval imaginal discs, and pupal legs.^8-10 Somatic RCs in all these tissues were consistently found to have an electron dense backbone 50 nm thick with an average lumenal diameter of approximately 250 nm. The variety of tissues containing RCs make Drosophila an interesting and technically accessible system for studying the formation and function of RCs. The first of these studies to examine intercellular movement through somatic RCs used the dye Lucifer Yellow and demonstrated the potential for direct intercellular exchange.¹¹ However, perhaps due to the lack of evidence for movement of protein or informative RC mutants, intercellular communication through RCs was not widely appreciated as a significant feature of follicle cell biology.

In 2 recent papers, we further explored somatic RC function by determining whether somatic RCs permit intercellular exchange of protein, at what rate might this exchange occur, and how far these proteins go.^12,13 The results suggest somatic RCs can provide cells with a mechanism to equilibrate protein levels and compensate for disparities in transcription between connected cells.

Intercellular Exchange

Photomanipulation of GFP and photoactivatable GFP (PAGFP) in live tissue provided convincing evidence of direct intercellular exchange of cytoplasmic protein. Photoactivation of PAGFP in single cells was followed by the rapid appearance of activated PAGFP in cells connected by RCs. Quantitation and computational modeling of PAGFP movement between follicle cells indicated that intercellular exchange occurs by diffusion. Imaginal disc cells, which also form RCs, similarly exhibited exchange of activated PAGFP between cells. In a complementary approach, repeated photobleaching of single follicle cells expressing GFP-tagged proteins resulted in a loss of GFP fluorescence in neighboring cells connected by RCs (Fluorescence Loss in Photobleaching, or FLIP). Thus, both GFP and GFP-tagged endogenous proteins move freely through somatic RCs of multiple tissues.¹³

Our model of diffusion-mediated exchange is supported by analysis of RC structure by electron microscopy (EM), which reveals no structural elements that could regulate movement between cells. Immediately inside the RC backbone is a layer of circumferential filamentous actin and, in most cases, a layer of rough ER (rER).^11,13 Previous measurements of follicle cell RCs reported the limiting diameter of the filamentous layer to be 230–260 nm, but did not account for additional occlusion by the rER.¹¹ We derived the effective inside diameter of follicle cell RCs by using a computational model of diffusion that recapitulates the observed exchange of PAGFP in vivo, and arrived at 80–220 nm (Fig. 1). For comparison, the inside diameter of the Drosophila gap junction is predicted to be slightly larger than the 1.4 nm diameter of mammalian gap junctions.^14,15 Aside from the peripheral actin filaments and rER, the remaining RC lumen lacks any cytoskeletal or other physical structures that might serve selective or regulatory functions. Such a large size is likely beyond the reach of interaction-mediated regulatory mechanisms, such as those known to control transport across the nuclear pore, into the cilia, and through plasmodesmata.^16-18 The presence of rER in the RC lumen is interesting, but it is not clear how it might contribute to intercellular communication or RC function. Exploration of the cellular autonomy of proteins that transit the ER would be interesting to pursue.

Figure 1. Somatic ring canal structure. The thickness and distribution of putative rER in the RC varies significantly and is represented here with a minimum and maximum observed thickness. (PM) Plasma membrane, (Backbone) electron-dense region of the RC, (F-Actin) filamentous actin. All units are nanometers.^11,13

Notwithstanding the observed diffusion of cytoplasmic proteins and absence of physical regulatory structures, several GFP-labeled protein-traps did not move through RCs within the timeframe of the FLIP assay; among them are several ribosomal, ribonucleoprotein (RNP), and mRNA processing-related proteins.¹² Based on the wide range of sizes of these proteins and the work of others in characterizing diffusion rates in cytoplasm,¹⁹ size does not seem to be a significant discriminating factor for movement through RCs. However, the FLIP analysis was restricted to less than 60 min, so movement of proteins or protein complexes that diffuse over the course of hours—potentially including the ribosomes—could be missed. Nonetheless, we speculate that proteins anchored in specific subcellular localizations or involved in tightly regulated complexes have little opportunity for diffusion through RCs. In this view, the ability of any protein to diffuse through RCs, and the rate thereof, is determined primarily by the subcellular localization and binding properties of that protein. Consistent with this hypothesis, FLIP analysis of nls.GFP demonstrated a slower rate of exchange when compared with non-localized, cytoplasmic GFP (McLean P and Cooley L; unpublished).

The number and nature of proteins that do not readily cross RCs supports a model of a highly structured cytoplasm in which many cytoplasmic processes can remain cell-autonomous in spite of direct connections with other cells. For example, generation of twin-spot MARCM clones in follicle cells, which relies on the segregation of a non-diffusing membrane-tethered fluorescent reporter away from its cognate siRNA,²⁰ suggests that little or no spreading of siRNA can occur between cells. Additional evidence for this idea is the lack of significant cell-cycle coordination among cells connected by RCs.¹²

Size and Organization of Follicle Cell Syncytia

The 1000+ follicle cells that encapsulate each germline cyst are produced in approximately 2 lineages, each derived from a stem cell that resides in the germarium. One daughter cell from each stem cell division differentiates into a follicle cell and undergoes a total of ~9–10 additional divisions over approximately 3 d. Our live imaging of dividing follicle cells confirmed that somatic RCs are a product of incomplete cytokinesis, thereby constraining all cells in a syncytium to their respective lineage. Most of the mitotic divisions occur in the context of an established epithelium, which restricts mixing of cells from different lineages.²¹ Whole-egg chamber counts of follicle cell RCs vs. nuclei demonstrate that once egg chambers emerge from the germarium, ~90% of divisions result in the formation of a visible RC. If the remaining 10% of divisions completed cytokinesis without producing a RC and were randomly distributed, we would expect to find much smaller groups of syncytial cells within each 500+ cell lineage. In fact, the observed connectivity among follicle cells, or size of the RC-based syncytia, averaged 8 cells in size with a large variance of 1 to 38 cells. Thus, small groups of cells with syncytial cytoplasm exist within much larger lineages of follicle cells. Interestingly, expression of transgenes in follicle cells with the Gal4/UAS system often produces a patch-like distribution of protein. Using FLIP and photoactivation experiments, we confirmed that these patches of expression correspond to the underlying syncytia.¹³

Follicle cell syncytia, evaluated directly by PAGFP activation experiments and indirectly by mosaic GFP expression, have no consistent configuration, orientation, or polarity with respect to egg chamber axes or other morphological features. For example, mosaic GFP patches frequently cross the domains of rhomboid²² and Broad²³ expression in stage 10a and 10b egg chambers, emphasizing the cellular autonomy of the pathways responsible for dorsal appendage formation (McLean P, Du J and Cooley L; unpublished). A closer inspection of follicle cell divisions revealed a random mitotic spindle orientation with respect to existing RCs.¹² This, in combination with a stochastic closure of 10% of cleavage furrows, is consistent with the observed random configuration, orientation, and distribution of syncytium sizes within the epithelium.

Equilibration of Protein Among Transcriptionally Disparate Cells

The expression patterns of GFP-protein traps provided the first suggestion for a role of follicle cell RCs in equilibrating expression between cells. Using FLIP we identified several non-diffusing proteins, such as RpL30, that had notable cell-to-cell variations in protein level. However, proteins that can diffuse (e.g., Cam, Oda, and Men-b) have uniform protein levels across the epithelium.¹² A possible explanation for this difference is that intercellular diffusion equilibrates the levels of proteins that are able to move through RCs; evidence of transcriptional variation between cells would provide compelling evidence of this model. Detecting subtle differences of transcript abundance in individual cells of an epithelium, however, represents a significant technical challenge. As an alternative, we used the exaggerated mosaicism of the Gal4/UAS system to examine the relationship between the distributions of transcript and protein expressed from a UAS-GFP transgene in follicle cells. The GFP distribution appears as patches of cells with dramatically different amounts of protein, and in some groups of cells, with no protein at all. Since the patches correspond to syncytial clusters of cells, the absence of GFP suggests that none of the cells in such a patch express Gal4-dependent transcripts. Indeed, a co-stain for GFP transcript revealed that only individual or small clusters of cells contained significant amounts of transcript, meaning that the much broader distribution of protein resulted from protein diffusing out of the cells of origin.¹³ Though these experiments used the exaggerated mosaicism afforded by the Gal4/UAS system, they demonstrate that RCs do allow the equilibration of protein among cells with different levels of transcription. It is therefore likely that other proteins capable of intercellular diffusion would similarly equilibrate within syncytia regardless of transcriptional disparities, potentially contributing to the overall health of these epithelial cells.

Whether the exaggerated mosaicism of Gal4/UAS transgenes is a result of the exogenous nature of the expression system or simply an amplification of endogenous inconsistencies in expression is not yet clear. In favor of the latter, relocation and truncation of the RpL23 promoter is sufficient to generate similarly exaggerated mosaicism independent of Gal4/UAS amplification (McLean P, MacDonald P, Cooley L; unpublished). As GFP-protein traps do not typically affect the promoter region, this could explain why no GFP trap lines were observed to have exaggerated or patchy mosaic expression similar to Gal4-driven GFP or the RpL23::GFP transgene. Additionally, recent work suggests that mosaicism in follicle cells is a manifestation of epigenetic instability that gradually resolves as the follicle cells complete differentiation,²⁴ which could be exaggerated by Gal4/UAS amplification.

Impact of Ring Canals on Clonal Analysis

To more comprehensively analyze the extent of intercellular diffusion, we developed a clonal marking system (STRIT) to permit visual genotyping of individual cells and determination of the extent of diffusion from a defined cellular source. Briefly, in STRIT a mitotic recombination event gives rise to a daughter that produces GFP and inherits 2 copies of the non-diffusible lacO DNA marker, and another daughter that cannot produce GFP and inherits no copies of the lacO DNA marker. Thus, the number of GFP-positive, lacO-negative cells reveals the extent of diffusion of GFP from its source. The results show a functional RC connects over 90% of sister clones, and clones are often connected to non-recombined cells by pre-existing RCs (Fig. 2). Generation of STRIT clones in wing imaginal discs yielded similar results, albeit limited to small clones of up to 2 cells. The frequency and extent of GFP exchange across clone borders make cytoplasmic GFP an unreliable marker for mosaic analysis. Indeed, using GFP to negatively-mark FRT clones homozygous mutant for the SMRTER gene in follicle cells similarly presented evidence of inter-clonal diffusion of both GFP and SMRTER protein.¹³

Figure 2. Syncytia reside within a single lineage and can overlap mitotic clone boundaries. (A) A simplified representation of 5 cellular divisions (left) with successive generations of RCs denoted as colored or gray circles. A more complicated pattern arises from cells in a hexagonal arrangement in which mitotic planes are randomly assigned (right). A single cell is selected as a mitotic clone progenitor (black outline, generation 2) and gives rise to 2 sister clones (black and white fill). (B–D) In this lineage, the patterns of syncytia depend on which RCs remain functional. Possible syncytial arrangements caused by RC loss are shown relative to the clonal cells (solid and dashed black outlines). Examples include (B) loss of the first generation RC (red), (C) the third generation RCs (blue), and (D) a combination of early and late generations. In each case, the mitotic clones remain connected to each other and/or non-recombined cells of a syncytium. RC loss appears biased toward older RCs, so additional divisions will increasingly isolate the clonal cells.

It should be noted, however, that the extent to which diffusion of GFP confounds clonal analysis depends on the diffusion properties of the protein under study. GFP diffusion observed in these experiments likely represents an upper limit of intercellular diffusion of a cytoplasmic protein, owing to the exogenous, monomeric, and non-interacting nature of GFP. Consistent with this idea, FLIP of GFP alone occurred more rapidly than FLIP of any GFP-tagged protein. A large difference in rate of intercellular exchange was also observed in SMRTER clones; GFP was observed at nearly WT (heterozygous) levels while SMRTER protein in these cells was present at much lower levels.¹³ However, a complete explanation for the reduced movement of SMRTER protein into mutant cells should also include its predominately nuclear localization, different rate of protein turnover, and specific physical interactions within the cell. These variables will be different for each protein of interest and should be considered when evaluating the impact of diffusion in clonal experiments. Similarly, the threshold of activity will differ for every protein and experimental setup—how much wild-type protein has to move into mutant cells to complement the mutation? Independent of the ability for a particular protein to diffuse through RCs, the infidelity of GFP to its cell of origin presents a challenge to the interpretation of clonal data. In negatively marked clones, for example, diffusion of GFP will likely result in an apparently smaller clone than anticipated. The phenotype and interpretation of false-positive cells, however, will depend on the net abundance and activity of that protein (Fig. 3).

Figure 3. Intercellular diffusion across clone boundaries may result in clonal ambiguity. Example of a compromised 8-cell mitotic clone (black outline) with GFP diffusion into 3 cells (light gray cells). Depending on the diffusion properties of protein X (concentration indicated by size), the false-positive GFP cells can have no protein X (None), reduced amounts of protein X (Reduced), or WT levels of protein X (Complete). The phenotype associated with each condition will depend on the activity of that protein. The 3 false-positive GFP cells will, at minimum, affect the interpretation of clone size; in the opposite extreme, they could appear to reveal non-autonomous effects on protein expression or phenotype.

Importantly, there are several ways to evaluate and mitigate the impact of intercellular diffusion when using clonal analysis in tissues with RCs. Protein diffusion can be measured directly by photoactivation or FLIP experiments with appropriately labeled transgenes. Close evaluation of clone size may also be sufficient to determine whether cross-clonal diffusion is occurring; STRIT clone sizes were consistent among egg chambers of similar stages and predictably followed a pattern of geometric growth (Fig. 4), so apparent deviations from this distribution may indicate intercellular diffusion of the clonal reporter. Alternatively, genes may be combined with non-diffusing identifiers such as the LacO reporter elements. Using clonal reporters with transmembrane domains (e.g., CD8, CD2) and other membrane tags (e.g., myristoylation) will also reduce the generation of inappropriately labeled cells, as they appear to have significantly less non-autonomous localization than traditional cytoplasmic and nuclear reporters (McLean P, Cooley L; unpublished). Clone size should also be considered when planning experiments, as small clones will be disproportionately affected by intercellular diffusion compared with large clones. In follicle cells, the 8-cell average syncytium size means that the interior of large clones are unlikely to have a continuous syncytial connection to WT cells. The extent of intercellular diffusion will be further reduced in tissues with a smaller average syncytium sizes, such as the imaginal discs (2–4 cells). Additionally, increasing distance from the WT cells will decrease the likelihood of interference from the diffusion GFP, as evidenced by the observed decreasing concentration gradient away from the cells producing GFP in the STRIT experiments.¹³

Figure 4. Follicle cell divisions follow a pattern of geometric growth. Distribution of all LacO-negative clone sizes from STRIT analysis¹³ indicates that mitosis occurs regularly, making clone sizes predictably dependent on the time from clone formation.

Perspectives

In sum, the work on somatic RCs provides compelling evidence that intercellular movement of protein does occur between follicle cells with disparate levels of transcription, and has the effect of evening protein expression among cells. In follicle cells, this could be an important mechanism for counteracting inconsistent gene expression resulting from epigenetic or other transcriptional variations. It is also possible that the prevalence of RCs in several Drosophila tissues is not indicative of a specific function for RCs, but rather a general inefficiency in the completion of cytokinesis. Answering this question, however, will likely require a more complete understanding of the mechanism driving abscission in Drosophila.

This work is also relevant to understanding cellular and cytoplasmic autonomy. In spite of the robust and rapid diffusion of many cytoplasmic proteins, the independence of various signaling pathways and little cell-cycle synchrony observed in cells connected by RCs reinforces the hypothesis of a highly structured cytoplasm that can operate independently. More broadly, the Drosophila somatic RCs are structurally and topologically similar to the germline RCs of Drosophila males and other animals, which suggests they too permit rapid equilibration of cytoplasmic proteins. Whether equilibration of non-uniformly expressed protein is the primary function or a simple exploitation of RCs is an exciting topic of ongoing and future investigation.

Glossary

Abbreviations:

RC

ring canal

GFP

green fluorescent protein

PAGFP

photoactivatable GFP

FLIP

fluorescence loss in photobleaching

STRIT

syncytial tracing by recombination induced transcription

rER

rough endoplasmic reticulum

EM

electron microscopy

Pav

Pavarotti

RNP

ribonucleoprotein

Flp

flippase

FRT

flippase recognition target

UAS

upstream activating sequence

McLean PF, Cooley L. Protein equilibration through somatic ring canals in Drosophila. Science. 2013;340:1445–7. doi: 10.1126/science.1234887.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.