The molecular and physical mechanisms underlying the anisotropy of lumen formation are an emerging area of research. In this study, by searching for a mechanism that could explain the anisotropy of hepatocyte apical lumina, we discovered the existence of specific extensions of the apical membrane sealed by TJs in the lumen between two adjacent hepatocytes. The best analogy we could find for these structures are the bulkheads of boats, ships, and planes. Bulkheads provide structural stability and rigidity, strengthening the structure of elongated vessels. From the physics of thin shells, formation of a tubular lumen with inner pressure and no outlets, such as the forming BC, requires anisotropy of surface tension and/or rigidity of the wall (Landau and Lifshitz, 1986; Berthoumieux et al., 2014). The apical bulkheads are structural elements that can provide such anisotropy and mechanical stability to the elongating cylindrical lumen under inner pressure. Interestingly, they follow a quasi-periodic pattern, whose distance is in the range of the diameter of the lumen, as in human-made constructions, where bulkheads are load-bearing structures. Here, they provide forces required for maintaining a nonspherical lumen. One can consider the cylinder with bulkheads as a “chain of spheres,” which is mechanically stable. The bulkheads in ships can also act as (semi)watertight compartments to prevent seeping of water to other parts of the ship. Similarly in the BC, they may act as valves ensuring directionality of bile flux in a nonperistaltic contractility. Additionally, the bulkheads may serve as hot-spots of contractility to facilitate bile flux, as shown in vivo (Watanabe et al., 1991; Meyer et al., 2017). Mechanistically, the position of the bulkheads could be determined by mechano-sensing mechanisms coupled to the tension and local curvature through the actin cortical mesh (Meyer et al., 2020). The elongation of the apical lumen also entails the movement and rearrangement of cell–cell contacts, which are accompanied by the formation of new bulkheads (Fig. 1 d and Fig. 2). Upon loss of the bulkheads caused by Rab35 down-regulation, the apical surfaces of hepatocytes lose their anisotropic growth, and the elongated lumina convert into spherical. Remarkably, we succeeded in reengineering liver tissue structure by down-regulation of Rab35 in vivo. This resulted in the modification of the cell polarity of hepatocytes, which, instead of forming BC, self-organized into tubular epithelial structures resembling bile ducts. It will be interesting to assess whether such morphological changes have consequences on hepatocyte cell fate and function.

We showed that the apical bulkheads are present in embryonic and adult liver, suggesting that they are not a cell culture artifact but have physiological relevance. In addition, the elongation assisted by apical bulkheads does not rely on cell division and therefore can explain the BC extension in quiescent differentiated hepatocytes in later stages of liver development (Yang et al., 2017). Their dynamic and adaptable nature fit the requirements of a growing, branching, and fusing BC network in vivo. To our knowledge, these structures were never described before despite several ultrastructural studies of liver from different species. They were probably not observed by EM before or mistaken for folds and ramifications due to the complexity of the BC network in the liver, or interpreted as septa in 2D EM sections of adult hepatocytes (Kawahara and French, 1990). Their visualization requires a 3D EM reconstruction.

We obtained several cues to the mechanisms underlying the apical bulkheads formation from the morphological analysis and functional screen by RNAi. First, the bulkheads are characterized by a T-shaped arrangement of TJs, which seals the two halves of the bulkheads (Fig. 1 b and Fig. 3 f). To our knowledge, this organization is unprecedented in polarized cells. Second, given that the TJs are connected to actin filaments, it is no surprise that the bulkheads contain F-actin transversally to the lumen elongation, thus introducing anisotropy in apical surface tension. Third, by a focused RNAi screen for established regulators of cell polarity, we found that the small GTPase Rab35 is required for the formation of the apical bulkheads and hepatocyte lumen shape. Based on previous work (Kouranti et al., 2006; Klinkert and Echard, 2016; Dambournet et al., 2011; Bhat et al., 2020; Zhang et al., 2009; Chevallier et al., 2009; Salvatore et al., 2018; Allaire et al., 2013; Marat et al., 2012; Jewett et al., 2017; Egami et al., 2011; Klinkert et al., 2016), Rab35 may contribute to the formation of these structures directly or indirectly, and we envision the following nonexclusive possibilities. Rab35 is a regulator of endosomal recycling (Kouranti et al., 2006; Klinkert et al., 2016; Mrozowska and Fukuda, 2016) and may control the intracellular distribution and function of apical recycling endosomes to deliver transmembrane proteins, e.g., junction components, at the site of bulkheads initiation and/or growth. The T arrangement of the TJs could originate from the junctions longitudinal along the tubule (horizontal bar in the T), and Rab35 may support the zip-up along the ridgeline (vertical bar of the T), either from the bottom or from the top of the tube. However, in addition to protein transport, generation of the apical bulkheads may require the formation of a mechanical support, either by delivering molecules to specific areas of the apical surface or by the apical vesicles anchored to the cytoskeleton (actin and microtubules) to project force into the apical bulkheads. The presence of clusters of vesicles at the base of the bulkheads as visualized by EM supports this view. Preliminary results suggest that Rab35 indeed localizes to sub-apical vesicles (Bebelman and Zerial, unpublished data). In addition, Rab35 is also known to coordinate membrane trafficking with the organization of the actin cytoskeleton (Klinkert and Echard, 2016; Chua et al., 2010). It may regulate actin remodelling to form the F-actin of the bulkheads, similar to its function in promoting the formation of F-actin–rich tunneling nanotubes in neuronal cells (Bhat et al., 2020). In the context of the apical bulkheads, it would orient the filaments between the TJs and the vesicles at the base, providing the aforementioned mechanical function. Rab35 could regulate the local phosphoinositide content via, e.g., inositol polyphosphate 5-phosphatase OCRL, nucleation and/or dynamics of the F-actin at the bulkheads, e.g., via MICAL1 or unknown hepatocyte-specific effectors (Chaineau et al., 2013; Dambournet et al., 2011; Frémont et al., 2017). Alternatively, Rab35 could play an indirect role by modulating signaling pathways, e.g., integrin-based cell adhesion (Allaire et al., 2013) and/or gene expression. Also, the function of genes implicated in hepatocyte polarity, e.g., Par1b, Pard3, Cldn2, Cldn3, and Lkb1 (cAMP-Epac-MEK-AMPK pathway regulating BC network formation) should be revisited specifically in the context of the bulkheads and anisotropy of lumen elongation (Wang et al., 2014; Fu et al., 2010; Son et al., 2009; Grosse et al., 2013; Slim et al., 2013; Homolya et al., 2014; Fu et al., 2011; Woods et al., 2011).

Our data thus suggest that transversal mechanical coupling between hepatocyte apical surfaces underlies the formation of BC and provide new insights into the longstanding problem of lumen morphogenesis in embryonic liver.