All of us in the business talk about the blood-brain barrier, but. . .no, I’m not going to end this sentence with “. . .none of us do anything about it”, because how it should end is “very few of us really stop to think about what it is”. What makes this (and similar structures) more of a barrier than other layers of cells? Some in the crowd will at this point say “tight junctions”, which is true and does indeed place you over hoi polloi, but what, then, is a tight junction?
There’s been a lot of work on that question over the years. It comes down to tighter cell adhesion, and you see changes on both sides of the cell membrane, with claudin and occludin adhesion proteins (among others) gathered on the outer surfaces and forming a dense net of tiny strands between adjacent cells. These generally are anchored in the cell membrane, with an extracellular tail that goes out and associates with those from the adjacent cell surface. There’s also a corresponding bunch of proteins on the intracellular side, and these associate with both the membrane proteins just mentioned, on their intracellular loops and tails, and with other cytosolic proteins. Two of the important ones are ZO1 and ZO2, with that “ZO” standing for zonula occludens, an older term for the tight junctions as noted by microscopy.
OK, how do those proteins work? There we are überfragt, as they say in German, “over-asked” about a question that we don’t have good answers to. But this new paper (from Dresden, the Max Planck Institute and the TU-Dresden) has some really interesting developments. They find that the ZO1 and ZO2 proteins in fact form an intracellular condensate at the point of a tight junction – yep, those darn condensate droplet thingies again (here’s a nice recent overview of this fast-evolving field). It had already been found that these proteins assemble via the interactions of several domains in their structure, into what appeared to be homodimers and homo-oligomers as well as complexes with those other tight junction proteins. But photobleaching experiments also indicated that this was a very dynamic process, with bleached-out proteins being replaced within seconds from the rest of the cytoplasm. How does that work?
Apparently though formation of a droplet of concentrated protein, which makes sense, when you think about it. That gives all those interaction domains a role in forming that phase-separated material, while also allowing a dynamic exchange with the surroundings as other species partition in and out of the droplets. This has already been noted in the structure of the neuronal synapse, and indeed, the ZO proteins are homologs of some of the synaptic ones, so it’s good to see this principle extended. (And we can look for even more of this sort of thing; cells presumably have a lot of uses for fluid-but-localized, structured-but-not-too-structured assemblies of this sort).
Fluorescent tagging of the proteins (and some high-quality imaging work) shows them forming small droplets on the inside of the cell membrane which merger into larger ones on a time scale of seconds. When they overexpressed the ZO proteins in cells (HEK293) that don’t form tight junctions, it appeared that liquid condensates formed spontaneously once the protein concentration got up to around 8 or 9 micromolar. As shown at right in a confluent layer of MDCK cells, the condensate formed a pretty continuous layer around the inside of the cell membranes, which seems to be partially mediated by interaction with actin. That was demonstrated by adding Latrunculin A, which breaks up actin polymerization, and similarly broke up the condensate layer into discrete beads and droplets. Immunostaining showed that other known tight-junction proteins were also enriched in those same spots, which acted very much like a viscoelastic fluid in time-resolved studies.
Now, that 8 micromolar in the HEK293 cells is much higher than the real concentration in tight-junction-forming cells, as you would imagine, so while that tells you that these proteins can form condensates, it also tells you that the real situation is more complex. That’s no surprise – when these phase-separated things have been analyzed, they generally contain a whole list of different proteins (many in small amounts), and often as not, at least so far, plenty of RNA species as well, although that may be an effect of the sorts of condensates studied so far (such as stress granules and transcriptional complexes). This is one of the big challenges in this area, actually – figuring out what’s going on in vivo and what all these constituents are doing.
The way stations along that path include the sorts of cellular overexpression experiments just described, as well as even more simplified ones in vitro. This paper has some of those as well, with the three ZO protein subtypes expressed in pure full-length form with fluorescent tags. In high-salt buffer, all three were freely soluble to high concentrations, but under more physiological conditions they spontaneously separated out at low micromolar levels into droplets which merged over time. It had already been demonstrated that the phosphorylation state of these proteins was involved in their tight-junction behavior, and once these were artificially phosphorylated in vitro (using casein kinase) they stayed in solution and did not form condensates at all. Indeed, phosphorylation state is thought to be a key regulatory switch in formation and dissipation of other condensates, which one can easily imagine while thinking about polar interactions.
And the proteins that you’d expect to interact with these condensates in the cell, the cytoplasmic tails of the claudin proteins and so on, do indeed sequester into the condensate droplets in vitro, which is a nice result. Cytoskeletal proteins such as afadin and cingulin do as well (as does monomeric actin), which could provide the bridge between the tight junction regions themselves and the rest of the cytoskeleton (a connection observed by microscopy in earlier studies). Some of these get up to 40-fold enhanced in the condensate phase as compared to outside it, and these high concentrations may be essential to things like concentrating claudin receptors in the local cell membrane. There’s a lot to learn about this sort of behavior – what are these liquid phases like, considered as solvents? What partitions into them, and what’s excluded? It’s a question that applies to proteins, to RNAs, to small biomolecules, and surely to small-molecule drugs as well.
The paper also has a good deal of work on mutations and truncations of the ZO proteins in various domains and the effect this has on condensate formation. There are a number of mutants whose cellular behavior has already been described, and matching these up with condensate experiments strongly suggests that condensate formation is essential for tight junction formation and function. There are quite a few twists and turns in the story, though – anyone really into this subject should read the paper in detail.
Overall, the authors hypothesize that the ZO proteins are recruited to their binding partners at the membrane until they reach a concentration threshold that tips things over into phase separation, a process that’s been seen in other condensate systems, but there are plenty of details (such as phosphorylation state and presentation of various protein domains) that still need to be worked out. There are plenty of details that need to be worked out in this whole field! And we’re seeing that happen right now. Future generations of students will take all this stuff for granted and wonder how anyone could not have known it – we’re the people who get to experience the bridging. Enjoy the sensation!
Update: I also have to mention this other recent paper from the IST (Austria) and University College, London on the phase separation of ZO proteins. They identify a role in tight-junction mechanosensing as well, as demonstrated in zebrafish embryos, which sheds light on the role of such tension-sensing mechanisms in development and how it manages to work at all. These papers make very interesting complements to each other!