To condense or not to condense: Wnt regulation by centrosome-nucleated biomolecular condensates

Abstract

Colorectal cancer is the third leading cause of cancer deaths worldwide, and, thus, understanding genetic changes driving cancer development and progression is critical. A key breakthrough occurred 31 y ago when scientists cloned the tumor suppressor gene adenomatous polyposis coli (APC), mutated in families with a predisposition to colon cancer. Subsequent work revealed that APC is part of the destruction complex (DC), a multiprotein machine that negatively regulates a key developmental signaling pathway, the Wnt pathway (1). Wnt signaling regulates cell fate choice and adult stem cell maintenance in all animals. However, key questions remain about the mechanisms by which the DC works. In PNAS, Lach et al. (2) provide exciting insights. Because of the power of Wnt signaling, evolution crafted mechanisms to keep signaling firmly off in the ligand’s absence. Combined work in Drosophila, mice, and cultured colorectal cancer cells provided important insights (3). The key event is regulating stability of the Wnt effector β-catenin (βcat; Fig. 1A ). When present, it works with TCF family transcription factors to activate Wnt target genes. To keep the pathway off, a multiprotein complex including APC, another large scaffolding protein—Axin—and two protein kinases—glycogen synthase kinase-3 (GSK3) and casein kinase 1-alpha (CK1)—captures and phosphorylates βcat, and then delivers it to an E3 ubiquitin ligase, targeting it for proteolytic destruction. In tissues where signaling is active, Wnt ligands bind a multiprotein receptor that turns down DC activity, thus stabilizing cytoplasmic and nuclear βcat and activating gene expression. In APC mutant colorectal cancers, βcat is stabilized even when signaling is off, activating signaling constitutively. Fig. 1. Centrosomes nucleate formation of Wnt regulatory DCs. (A) Canonical Wnt regulation. (B) When Axin is overexpressed in cells, it forms multiple puncta. (C) Hypothetical multiprotein DC biomolecular condensate. (D) Axin expressed at endogenous levels assembles into one or two puncta colocalizing with centrosomes. A–C were produced by laboratory alumna Kristi Schaefer; D is from Max Wilson. This textbook description sounds tidy, leaving the impression that the Wnt regulatory mechanism is solved. However, key questions remain. For example, Axin alone can act as a scaffold to bring the two kinases to βcat (4), leaving APC’s function unclear. More puzzling, a series of studies suggested that a four-protein DC is not the whole story. When Axin is expressed in cultured cells, it assembles into multiple “puncta,” into which are recruited APC, the kinases, and βcat (e.g., refs. 5 and 6; Fig. 1B ). Axin’s DIX domain mediates protein polymerization (7), potentially driving assembly of puncta containing many copies of each DC protein (Fig. 1C ). This suggested that the DC might be a biomolecular condensate (reviewed in ref. 8). These large, non-membrane-bound structures assemble by multivalent protein interactions, some mediated by intrinsically disordered regions, thus generating large protein complexes by “liquid–liquid phase separation” (9). DC proteins can phase separate in vitro (e.g., ref. 10). Further, elevated resolution microscopy revealed that Axin puncta seen in cultured mammalian cells when Axin levels are elevated are structured multiprotein complexes (11, 12). Similar complexes are seen in vivo in Drosophila embryos, with Axin expressed only a few-fold above endogenous levels, and their subcellular localization and assembly are regulated by upstream Wnt signals (13), suggesting that they are functional. More than 100 Axins assemble in each punctum. However, other groups took alternate approaches, building DCs from the bottom up from purified proteins, and defining molecular interactions involved in assembly (14, 15). This was quite successful, reconstituting, in one case, not only βcat phosphorylation but also ubiquitination by the E3 ligase. The in vitro protein complexes were substantially smaller than those seen in vivo, with ≤10 Axins in each. These data raised a key question: What is the true assembly state of the DC in vivo? Lach et al. (2) take this on via a highly multidisciplinary approach. They note that phase-separating systems can exhibit switch-like responses to changes in concentration, potentially explaining differences in DC assembly in vitro and in vivo. This occurs, in part, because newly synthesized Axin or APC proteins must choose between joining an existing complex or nucleating a new one. Their concentrations and posttranslational modifications determine their location in a high-dimensional phase diagram, governing whether these proteins remain diffuse, spontaneously demix into existing droplets, or nucleate new condensates. Thus, precisely controlling protein expression levels is key to defining the true assembly state. To visualize DCs at endogenous levels, Lach et al. tagged the endogenous βcat gene with a fluorescent protein, and visualized cells live. βcat accumulated at the plasma membrane in cadherin-based cell adhesion complexes, while cytoplasmic levels were low in the absence of Wnt signaling. Wnt signaling stabilized cytoplasmic βcat, as expected. What was less expected was that, in the absence of Wnt signals, βcat also accumulated in one or two bright, spherical puncta near the nuclei (2). These resembled liquid-like biomolecular condensates, fusing and splitting. Importantly, puncta stability was regulated by Wnt signaling: Adding Wnt ligands led to loss of puncta. To confirm this, Lach et al. used an optogenetically controlled Wnt receptor, which was activated by light. Activating the receptor also led to puncta dissolution. Similar puncta were observed with Axin tagged at its endogenous locus, eliminating worries about overexpression (Fig. 1D ). The assembled puncta contain all DC proteins, which could come in and out of puncta readily and rapidly, consistent with known biomolecular condensate properties. Together, these data suggested the puncta are the active DC. Most intriguing, the one or two puncta per cell colocalized with centrosomes, the cell’s major microtubule organizing center (2). Lach et al. confirmed that puncta and centrosomes exclusively colocalized when Axin was expressed at endogenous levels, but, when they used inducible expression to elevate Axin levels, multiple extracentrosomal DC puncta formed. These data suggested centrosomes regulate DC assembly by serving as nucleation sites. At the lower endogenous levels of Axin, centrosomes might locally concentrate Axin or other DC proteins enough to trigger Axin multimerization and thus phase transition. Lach et al. first tested this by mathematically modeling DC assembly and function, contrasting predicted rates of βcat phosphorylation with or without a “nucleator.” Strikingly, their models predicted that nucleation strongly accelerates βcat phosphorylation over a wide range of parameters, improving DC efficiency. Lach et al. (2) then created a tool to test whether condensation elevates Wnt regulatory activity in vivo: a version of the DC kinase GSK3 that oligomerizes in response to light. Strikingly, inducing GSK3 oligomerization rapidly increased GSK3 concentration specifically in centrosomal puncta, without triggering assembly of extracentrosomal puncta. Further, GSK3 oligomerization prevented Wnt signaling from turning down the DC, and inhibited Wnt-induced human embryonic stem cell differentiation, putting the work into the context of normal Wnt signaling. Together, this work provides key insights into Wnt signaling, helping reconcile earlier in vitro, cell culture and in vivo experiments concerning the DC assembly state (2). These data reveal that, at least in cultured mammalian cells, DCs form phase-separated biomolecular condensates even when proteins are all expressed at endogenous levels. Lach et al.'s modeling suggests this increases efficiency of targeting βcat for phosphorylation and thus destruction. This work also suggests that earlier experiments suggesting the presence of many such puncta likely resulted from Axin overexpression, confirming the very tight concentration threshold for condensate formation. In their system, centrosomes nucleate condensates, triggering condensation just below the normal concentration threshold. This aspect will have broader relevance, as similar concentration-dependent behaviors and nucleation mechanisms may regulate other important biomolecular condensates. This work (2) also raises many exciting questions. Which protein interactions mediate centrosomal recruitment? Which molecular interactions then trigger condensation? Candidates include Axin polymerization, APC’s Armadillo repeats, and APC’s enigmatic CID/R2-B motif (11, 16, 17). We also must define how APC protein truncations seen in colorectal tumors affect condensate assembly or function. The data open questions about mechanisms that regulate condensate. It was striking that GSK3 oligomerization rendered DCs immune to disassembly by Wnt signaling—in addition to phosphorylating βcat, GSK3 also phosphorylates Axin and APC, and GSK3 phosphorylation of APC’s CID/R2-B motif may regulate assembly (11). Handoff to the E3 ubiquitin ligase is the next step in targeting βcat for destruction. We must determine whether and how the E3 ligase is recruited to centrosomal condensates. The E3 F-box protein TrCP can localize to centrosomes (18), and in vitro work supports physical interaction of E3 ligase and DC proteins (12, 19). While localization of individual DC proteins to centrosomes was known, the centrosome’s role in nucleation was a surprise. However, it opens new questions. Both mouse embryos (20) and Drosophila (21) lacking centrioles can survive without noticeable defects in Wnt signaling, revealing that centrosomes are not essential for this process. What other cellular structures might serve as nucleation sites? Finally, as the authors note (2), it will be important to use their system in partnership with in vitro approaches to define how Wnt signaling acts through the Wnt receptor complex to activate Dishevelled and turn down DC activity. It will be exciting to see how complementary approaches help tackle these key issues in cell, developmental, and cancer biology.