Historically, membrane-less compartments have been described as granules, foci, cellular bodies, cellular aggregates, and bodies. Recently, the general term ‘biomolecular condensates’ (BMCs) has been proposed to describe their ability to spatially organize biomolecules [1]. BMC formation is most often driven by the physical process of phase separation, whereby the macromolecular mixture separates into two phases: a dense phase of concentrated molecules and a dilute phase (Figure 1) 1, 2. Phase separation is facilitated by ‘driver’ proteins/nucleic acids with multivalent weak interactions occurring in a homotypic fashion or in combination with other molecules in heterotypic fashion [3]. Phase separation is often promoted by proteins with large intrinsically disordered regions (IDRs), but IDRs are not present on all drivers of phase separation [4]. BMCs can have various material properties, ranging from liquid-like to gel-like or solid-like, and can achieve various levels of macromolecular complexity [5]. Drivers typically act as scaffolds with other protein/nucleic acid ‘client’-binding sites (Figure 1) 6, 7. Clients cannot phase separate on their own, and recruitment occurs through direct interactions with the driver/scaffold or indirect associations with another client 1, 6. Some BMCs have been found to enrich hundreds of different clients 8•, 9•, 10•, 11, and some form complex shell–core–like architecture [12]. Furthermore, BMCs are dynamic structures and provide semipermeable barriers that enable constituent molecules to rapidly exchange between BMCs and the surrounding dilute phase on a time scale of seconds to minutes [1]. By concentrating on specific biomolecules, BMCs help to promote biochemical pathways and prevent unfavorable reactions inside the cell 1, 7, 13. Additionally, BMCs sequester molecules from their enzymes to inhibit biochemical activity, differentially partition diverse substrates to increase substrate specificity, or can buffer fluctuations in cellular concentration of molecules 1, 14, 15, 16•. In eukaryotes, BMCs have been found in most cellular compartments and extracellular milieu, and bacterial BMCs have been found in the cytoplasm, periplasm, and external structures such as biofilms 17•, 18, 19. Despite the protein-centric view of BMCs, RNA or polyphosphate also appears to be able to form BMCs [20], suggesting that a diverse array of biomolecules may be capable for driving condensation. The largest class of BMCs are ribonucleoprotein (RNP) BMCs, which will be the focus of this review 1, 21, 22, 23.

Most of what we know about RNP BMCs has been revealed in eukaryotic systems. RNP BMCs have been reported in the cytoplasm, nucleus, mitochondria, and chloroplasts of eukaryotic cells and are involved in a broad range of biological functions, including gene expression, RNA metabolism, and signaling 1, 23, 24. The nucleolus, cajal bodies, and paraspeckles are located inside the nucleus of eukaryotic cells and function in RNA metabolism and various aspects of gene expression [1]. Cytoplasmic BMCs such as P-bodies (PBs) and stress granules (SGs) are thought to regulate RNA degradation, mRNA translation, signaling pathways, and stress responses 1, 25. Mitochondrial degradosome-foci (D-foci) have been found to play a role in mitochondrial RNA degradation [26]. In bacteria, RNP BMCs have been observed in the cytoplasm and in the nucleoid 18, 27. RNA polymerase and Rho phase separate, suggesting that transcription can be compartmentalized in BMCs 28, 29. In addition, Hfq and RNase E, which are involved in mRNA regulation and decay, have also been identified to form BMCs 17•, 30. Finally, DEAD-box RNA helicases (DDXs) from eukaryotes and bacteria form BMCs, suggesting their role in BMC organization is conserved [31]. These observations indicate that RNP BMC organization is broadly utilized in both eukaryotes and prokaryotes, in both RNA biosynthesis as well as RNA processing/decay. Furthermore, condensates have been shown to improve the catalytic activity of ribozymes [32], suggesting condensates were present in the RNA world before modern cells.

The RNA decay machinery is compartmentalized in BMCs in eukaryotes, bacteria, and mitochondria and chloroplasts (Table 1), suggesting biomolecular condensation of RNA decay machinery is an ancient organizational strategy. Since the early 1990s, both SGs and PBs were found to contain mRNA decay machinery and have been topics of in-depth studies due to their importance in human disease 33, 34. SGs and PBs were subsequently observed from yeast to mammalian and plant cells 35, 36, suggesting they are conserved across eukaryotes. PBs are found constitutively in unstressed cells and increase in size and number when subjected to stress conditions and recognized as the sites of mRNA degradation/storage in the cell [37]. Moreover, the major cytoplasmic 5′-3′ exonuclease Xrn1 accumulates in PBs and SGs. SGs have been observed to occur across various stresses but, in contrast to PBs, are not constitutively present in unstressed cells. Due to the broad distribution of PBs and SGs, these structures have been proposed to be utilized across all eukaryotic cells. Organellar mRNA decay BMCs originated from bacterial endosymbiosis and have been found in mitochondria and in chloroplasts 26•, 38. D-foci have been suggested to be constitutive and the site for the mitochondrial RNA degradation containing the mitochondrial degradosome components SUV3 and PNPase [26], which are partially conserved with bacterial RNA degradosomes [39]. Chloroplast SGs are mainly associated with mRNAs, poly(A)-binding proteins, and small ribosomal units and are induced upon stress 38, 40.

Like eukaryotes, numerous bacterial RNP BMCs have now been identified. In 2018, the first bacterial RNP BMC was discovered, termed bacterial RNP-bodies (BR-bodies) containing the α-proteobacterial RNase E RNA degradosomes [17]. Initial studies of Escherichia coli RNase E suggested it formed helical filament structures in the cytoplasm [41]; however, TIRF microscopy revealed events of foci-fusion [42], suggesting they are RNP BMCs. Additional evidence suggests Cyanobacteria and Mycobacterium smegmatis RNase Es form RNP BMCs 43, 44, 45. Furthermore, Muthunayake et al. also demonstrated that when comparing bacteria from each major clade that possess RNase E, distinct patterns of charge blocks are present within the IDRs of RNase E, which promotes phase separation [46]. While RNase E is the most common mRNA decay enzyme, some bacteria use other RNases for controlling mRNA decay. In Bacillus subtilis, RNase Y initiates mRNA decay and form RNase Y degradome BMCs [47], and Helicobacter pylori RNase J degradosomes form BMCs [48]. Overall, these observations suggest that BR-bodies containing the mRNA decay machinery are broadly distributed across bacteria. Other bacterial RNP BMCs outside mRNA decay also been identified, including RNAP/NusA 29, 49, Rho [28], Hfq [20], and DDXs [31]; however, ‘BR-bodies’ will be used subsequently to describe ‘RNA degradosome’ condensates. While many of these RNP BMCs have been only observed in a single bacterial species, BR-bodies appear to be widely distributed across bacteria [46].

Despite the diverse RNA decay machinery in eukaryotes, their organelles, and bacteria, the proteomes of PBs, SGs, and BR-bodies appear to contain similar types of nonhomologous enriched enzymes, suggesting evolutionary convergence. Hundreds of proteins in PBs and SGs have been identified and characterized through purification by particle sorting or differential centrifugation followed by mass spectrometry or by enzyme-catalyzed proximity labeling followed by mass spectrometry 8•, 9•, 10•, 11. Over 100 proteins were also identified in Caulobacter crescentus BR-bodies isolated via differential centrifugation followed by mass spectrometry [50]. Analysis of their proteomes revealed several categories of enriched proteins involved in many post-transcriptional processes, including DDX, small interfering RNA (siRNA)/micro RNAs (miRNA)/small regulatory RNAs (sRNA) factors, major mRNA decay nucleases (RNase E and Xrn1), accessory ribonucleases, translation repressors, and components of the deadenylation complex, are present in PB, SG, and BR-body BMCs (Table 2). Surprisingly, 5′-decapping enzymes are present in PBs and SGs, but bacterial 5′-decapping enzymes are not enriched in BR-bodies (Table 2). The enrichment of decapping enzymes in PB and SG has been suggested to the presence of decapped translationally repressed mRNAs in these structures. While bacterial mRNAs do not contain a canonical 7-methyl guanylate (m7G) cap, decapping enzymes RppH and NudC also are crucial for the stability of mRNAs where they can remove 5′-end triphosphate (5′-PPP) or 5′-NAD(H) from RNA 51, 52, 53, and resulting 5′-P ends have enhanced cleavage by RNase E [52], but these enzymes are not enriched in BR-bodies [50]. While it is known that RNase E cleavage is accelerated with a 5′-P, the coordination of decapping with BR-bodies is not yet understood.

While most of the proteomics experiments are performed in bulk, this approximates each BMC as a homogenous mixture of the enriched proteins. However, an interesting finding is that, while a subset of BR-body proteins appears to uniformly be enriched in BR-bodies, certain BR-body–enriched proteins are present in a subset of BR-bodies, leading to heterogeneity in BR-body clients (Figure 2). RhlB, Aconitase, and PNPase are enriched uniformly in BR-bodies and known to stoichiometrically assemble into the RNA degradosome [54]. Additionally, they are colocalized with RNase E in vivo and strongly recruited into RNase E droplets over other RNP BMCs in vitro, suggesting that the RNA degradosome likely composes the core clients within each BR-body. In contrast to the core BR-body clients, accessory proteins, RhlE, Hfq, and RNase D, are contained in a subset of cellular BR-bodies (Figure 2) [50]. This heterogeneous client composition in BR-bodies may promote distinct biochemical activities. In that way, perhaps BR-body RNA decay activity may be regulated upon the recruitment of these specialized factors.

In eukaryotic PB and SG, evidence also exists, which suggests specialization. The PB proteome revealed that PB proteins form a dense protein network, which is three times denser than the SG protein network [9]. This network is mostly composed of RNA-binding protein (up to two-thirds of PB-enriched proteins), including many DDXs (four times more compared to SG-enriched helicases), suggesting stronger PB importance [9]. Importantly, the protein composition of SGs varies according to stress, cell type, and disease state. For example, 20% of proteins were specific to arsenite-induced SGs versus heat-induced SGs 8•, 11. Overall, these observations suggest that SGs assemble with different proteins in response to different stress conditions. The specific stress-inducing SGs can also influence the material properties of SGs [55]. In yeast PB, core protein components were found to be highly concentrated and uniformly present in PB [56] and were able to reconstitute in vitro with similar dynamics and stoichiometries [57]. When scaffold molecules of PBs were deleted, the partitioning of many PB components was altered, whereas when client molecules were deleted, only the immediate interacting components of the clients were affected, suggesting that deleting scaffold molecules is likely to have heterogeneous effects on the PB composition [56]. While PBs contain a more complex core interaction network than BR-bodies, the reconstitution of both structures in vitro should allow the functional dissection of how specialized clients might impact biochemical functions.