PerspectiveGene Expression

Dynamic condensates activate transcription

See allHide authors and affiliations

Science  27 Jul 2018:
Vol. 361, Issue 6400, pp. 329-330
DOI: 10.1126/science.aau4795

Every aspect of human function, from proper cell differentiation and development to normal cellular maintenance, requires properly timed activation of the necessary genes. This requires transcription of genomic DNA into messenger RNA (mRNA), accomplished by RNA polymerase II (RNA Pol II), which initiates transcription at gene promoters. This highly regulated process requires hundreds of proteins that must go to the promoter in a coordinated manner. Although many of these proteins are already organized into large and stable protein complexes, and so travel as a group, the process still requires coordination of many individual proteins and preformed complexes so that they are all in the same place on genomic DNA at the same time. This problem has been appreciated for years and has led to models such as “transcription factories,” where components are organized and ready to act on a gene that goes to the cellular location of the factory (1). On pages 378, 379, and 412 of this issue, Chong et al. (2), Sabari et al. (3), and Cho et al. (4), respectively, argue that special protein domains, which interact with each other to form fleeting or more persistent interactions, form biomolecular condensates that concentrate the transcription machinery. Some of these condensates might even form droplets, generating a liquid phase separated from the rest of the nucleus. Phase separation is a phenomenon familiar to anyone who has made a salad dressing: The oil and vinegar exist as two separated liquids. Phase separation in cells creates membraneless organelles that, in this case, provide the organization necessary for productive transcription (5).

Dynamic transcription machinery clustering during gene activation

Transcription factors (TFs) and coactivators condense into high-concentration clusters in the nucleus. Condensation is mediated by low-complexity disordered regions (LCDRs) in these proteins. These clusters can incorporate RNA Pol II through transient interactions to efficiently activate gene transcription.


Several distinct types of proteins are needed for transcriptional activation. Gene-specific transcription factors (TFs) bind to specific regions of the genome. They then interact with large complexes needed for the transcription process, including a key complex called Mediator, which, in turn, interacts with RNA Pol II to increase transcription from a promoter (6). Enhancer sequences, which are physically removed from the promoter in the genome, also increase transcription and can be bound by other specialized proteins (7). Regions of proteins involved in these processes were analyzed in the three studies, with the notable finding that several different transcriptional regulatory components each contain protein domains that form condensates in cells. These condensates increase the effective concentration of components needed for transcriptional activation and allow organization of those components via numerous cooperative interactions within the condensates, thus providing an attractive mechanism for combining factors in a timely fashion to generate transcription.

The visualization of these condensates required the use of imaging technologies to characterize the behavior of individual protein domains, which was compared to the behavior of entire complexes. The domains were fused to fluorescent proteins to allow visualization with lattice light-sheet imaging (8). This allowed sufficient spatial and temporal resolution to see condensates, which display as puncta of fluorescence, formed by these domains in living cells and to characterize the dynamics of these condensates (see the figure). Domains of the TFs, FET [composed of FUS, EWS, and TAF15 (TATA-box binding protein–associated factor 15)] and SP1 (specificity protein 1) form puncta. Thus, these protein domains cluster with each other instead of freely diffusing separately from each other. Similarly, the enhancer binding factors BRD4 (bromodomain-containing protein 4) and MED1 (mediator of RNA Pol II transcription subunit 1) were seen in discrete puncta. MED1 is one of more than 20 proteins that comprise the Mediator complex (9), and the entire Mediator complex can also be seen as puncta. RNA Pol II, itself composed of 12 subunits, also forms clearly delineated puncta. There is evidence for multiple molecules important for transcription being incorporated into the same phase-separated condensates (10). Chong et al. and Cho et al. found that RNA Pol II colocalized with TFs or Mediator puncta in live cells, respectively.

These puncta are formed via interactions between domains that are called either low-complexity domains (LCDs) or intrinsically disordered regions (IDRs). These domains have limited types of amino acids and are characterized as disordered according to their predicted secondary structure (11). To unify the terminology, we refer to these domains as low-complexity disordered regions (LCDRs). The current hypothesis is that proteins with these domains form networks, based, in part, on hydrophobic interactions, that are individually short lived and that allow for dynamic interplay that can create liquid-like properties (12). All three studies offer support for this hypothesis by showing that 1,6-hexanediol, which impairs hydrophobic interactions, can disrupt the structures. They also all used fluorescence recovery after photobleaching (FRAP) to show that molecules move in and out of these puncta rapidly, indicating that the components that make up the puncta are dynamic and not solid aggregates. Finally, Sabari et al. and Cho et al. show, with live-cell movies, that the puncta can merge together, just as water droplets will form a bigger droplet when they interact on a glass surface. Thus, the puncta have liquid-like characteristics.

The studies explore the role for specific protein sequences in self-association. In Chong et al., examination of the LCDRs of TFs shows that there is specificity in the interactions. Factors can self-associate (for example, FET LCDRs), but certain interactions between separate factors (for example, FET LCDRs and the SP1 LCDR) do not occur. This might be due to differences in the sequences of the LCDRs. Mutational analysis of the EWS LCDR demonstrated that 29 tyrosine residues were required for LCDR interactions. Similarly, Sabari et al. found that the MED1 LCDR sequence was dominated by serine residues, which were required for self-association and liquid droplet formation in vitro.

The extent to which phase separation is a necessary element of transcriptional activation is called into question by Chong et al., who found that TF LCDR self-assembly is transient at physiological concentration, in the range of seconds, and thus not consistent with phase separation into isolated droplets. This raises the question of when phase separation is an important part of the mechanism as opposed to a side effect of more transient interactions. Transient interactions between LCDRs might play a critical role in organizing components without a need for a stably phase-separated state. One possibility is that some interactions need only be transient, whereas others require greater stability. Perhaps the “potency” of domains to phase-separate (that is, the concentration needed to achieve phase separation) varies depending on need. For example, in Cho et al., Mediator and RNA Pol II both have properties consistent with being phase-separated at normal physiological concentration. This is presumed to help the two complexes interact, but the interaction between the two complexes is transient. Activation of transcription by necessity requires that some interactions be transient: RNA Pol II must initiate transcription and elongate the transcript by interacting with different complexes (13). The ability of the activation components to move onto the next round of transcription, and thus switch contacts to a new RNA Pol II, is likely to be important for genes that are being rapidly transcribed. Thus, in transcriptional activation, there are sound theoretical reasons to have interactions that not only increase effective local concentration (as transient interactions between LCDRs would) but that also allow those components to be integrated with, instead of separated from, other nuclear components.

By contrast, in other regulatory settings, phase separation might be helpful or even necessary for appropriate regulatory functions of transcription, for example, interactions involved in stable repression (14, 15) or, perhaps, long-range enhancer interactions. The continued development of technologies and experimental strategies to determine the importance of phase separation will be an exciting area to follow. Are these condensates central to most nuclear functions and thus a general regulatory mechanism with multiple distinct specificities and temporal characteristics? How much of nuclear function occurs in phase-separated domains? Do long noncoding RNAs, which are prevalent in the nucleus, contribute to the potency of phase separation and/or the organization of phase-separated domains? When is phase separation essential to the processes that generate the regulatory organization needed for life?


View Abstract

Stay Connected to Science

Navigate This Article