Recent developments in the field of liquid-liquid phase separation (LLPS) have led to a transformation in our understanding of the biogenesis of subcellular membrane-less compartments. As more and more phase-separated condensates are being discovered, there is considerable interest in exploring key factors (proteins) involved in the organizations and physiological functions of the compartments. However, we are still at an early stage of understanding the precise regulation and the biochemical processes inside the condensates, lacking a global view of the interactions within/among the compartments (and the environment).

For many years, the field of compartmentalization was limited to the study of membrane-bound organelles. The presence of these functionally and structurally distinct compartments is the essential feature of eukaryotic cells. In the 1980s, small granules that behaved as fluid droplets were identified in the cytosol, and these droplets were observed to fuse together into larger assemblies known as non-membranous organelles (Strome and Wood, 1982). High-resolution imaging studies (and other methods of determining molecular composition) have revealed that membrane-less compartments generally exhibit similar dynamics and similar assembly pathways, although their position, composition, and function may differ (Mitrea et al., 2018). P granules, a type of membrane-less compartment found in Caenorhabditis elegans, were the first biomolecular condensate observed to form via LLPS (Brangwynne et al., 2009). These early observations concerning P bodies greatly promoted the development of this field, furthering our understanding of the physical processes driving the formation of organelles. Subsequently, evidence was provided demonstrating the involvement of LLPS in the formation of additional membrane-less organelles, including nuclear Cajal bodies (in plant and animal cells, Frey et al., 1999; Sleeman et al., 2011; Riback et al., 2020) or the homologous nucleolar body (in budding yeast, Verheggen et al., 2001), nuclear speckles (Cotto et al., 1997; Chiodi et al., 2000; Brangwynne et al., 2011; Spector and Lamond, 2011; Tripathi et al., 2012), stress granules (Buchan and Parker, 2009; Kato et al., 2012; Youn et al., 2019; Yang P. et al., 2020), and the carboxysome (a well-studied subcellular compartment in cyanobacteria responsible for sequestering and concentrating Rubisco enzymes for CO₂ fixation) (Wang et al., 2019). Moreover, evidence was presented that LLPS may be involved in forming bacterial inclusion bodies (IBs) (Baneyx and Mujacic, 2004; Singh and Panda, 2005; Sabate et al., 2010; Chebotareva et al., 2013; Azaldegui et al., 2021; Su et al., 2021). Although membrane-less organelles have no enclosing membrane, the condensates have been demonstrated to maintain (for hours to days) stable, coherent structures capable of compartmentalizing and concentrating specific sets of molecules and exchanging material with surrounding components (Shin and Brangwynne, 2017).

An understanding of the principles underlying the formation and function of biomolecular condensates is vital for any in-depth investigation of the physiology and pathophysiology of biological processes and systems. Using up-to-date imaging, structural, and computational methods, scientists have been able to study the features of LLPS in vitro and in vivo (Alberti et al., 2018; Bracha et al., 2018; Mitrea et al., 2018). However, an in-depth study of LLPS in the comparatively small prokaryotic cell remains technically challenging, limiting our understanding of the molecular basis and biological function of compartmentalization in microorganisms. Nevertheless, recent developments in this field have yielded an extraordinary leap in understanding. In this review, we highlight representative examples of phase-separated condensates observed in microbial cells. Using these examples, we summarize the underlying mechanisms accounting for the composition, function, and the assembly/disassembly of microbial membrane-less compartments. We have also highlighted a series of challenges and future perspectives in this exciting area.

In Tiebackx (1911) firstly reported that coacervation was achieved through liquid-liquid phase separation (LLPS). Then, the concept of LLPS was applied in organic chemistry, especially in polymer chemistry (Jong and Kruyt, 1929). In this context, coacervation can be attained using either a mixture of oppositely charged polyelectrolytes (complex coacervation) or a polymer capable of self-association (self-coacervation) (Jong and Kruyt, 1929; Gabryelczyk et al., 2019). When a homogeneous polymer solution of macromolecules undergoes LLPS, two different phases are formed, a phase of concentrated molecules (dense phase) and a dilute molecule-depleted phase (dilute phase). The dense phase resembles liquid droplets (Alberti et al., 2019), and molecules in this phase can move quickly and are free to exchange interactions with multiple other molecules. Because molecules in the dense phase are highly likely to experience random molecule-molecule collisions, the potential for these molecules to complete biochemical reactions is high.

As early as 120 years ago, Wilson raised that protoplasm might be constructed by condensed liquid-droplet-like granules (Wilson, 1899). However, the commonly held view considered the cytoplasm as a fluid-like homogeneous mix of soluble proteins and compounds. It is only in the past decade, studies revealed that the cytosol does not act simply as a continuous medium but demonstrates complex rheological characteristics (Brangwynne et al., 2009; Guo et al., 2014; Shin and Brangwynne, 2017). The bacterial cytoplasm displays properties characteristic of glass-forming liquids and can solidify to resemble soft glass, depending on the metabolism, component sizes, and non-steric interactions (Parry Bradley et al., 2014; Xiang et al., 2021). Modern microscopy techniques reveal that the many proteins in bacteria tend to form large complexes targeted to specific regions within the cytosol (Abbondanzieri and Meyer, 2019). Some of the regions exhibit remarkable liquid droplet-like behaviors and undergo rapid assembly and disassembly in response to stress or cell signaling events (Sehgal et al., 2020). Evidence for this phenomenon includes the gathering of RNA degradosomes into bacterial ribonucleoprotein bodies (BR-bodies) displaying liquid-like behavior in Escherichia coli, Bacillus subtilis, and Caulobacter crescentus (Al-Husini et al., 2018, 2020; Hamouche et al., 2020). According to Hyman’s hypothesis (proposed by Hyman et al., 2014), the formation of phase-separated condensates in eukaryotes occurs through three main steps (Hyman et al., 2014; Strom et al., 2017; Peng and Weber, 2019): nucleation; rearrangement; and supersaturation (Figure 1). A saturation concentration, C_sat, was defined such that: for C < C_sat, the molecules are diffuse in solution; and for C > C_sat, dense droplets form. If the concentration consistently increases, the liquid-like condensate may change into its gel-like or solid states (Figure 1; Nandana and Schrader, 2021). Notably, C_sat values are not fixed, but vary with the concentration of condensate components (Dzuricky et al., 2020; Riback et al., 2020; Zhang et al., 2021). In yeast P bodies, seven proteins are present at high concentrations (5–15 mM), forming the “core” of the condensate, and 24 additional P-body proteins are present at lower concentrations (<2.6 mM) (Xing et al., 2020). It is important to note that Hyman’s hypothesis is not limited to eukaryotic cells (Azaldegui et al., 2021). It can also be applied to bacteria, which were once considered amorphous “bags of enzymes” lacking membrane-bound organelles (Al-Husini et al., 2018). For example, E. coli FtsZ, a well-studied tubulin homolog that is essential for cytokinesis, is capable of forming crowding-induced condensates (Monterroso et al., 2019). In vitro experiments indicate that FtsZ-rich droplets are formed only when FtsZ is in a complex with nucleoid-associated inhibitor SlmA (which antagonizes FtsZ polymerization while binding to specific sites on the E. coli chromosome), and that concentrations of SlmA greater than 40 μM (far above the physiological concentration) are required for FtsZ condensates to form (Han et al., 2012).

Although the fundamental role played by phase separation in the spatiotemporal organization of essential microbial processes has recently been revealed (Table 1 and Supplementary Table 1), many of these processes have been scarcely explored in bacteria, largely owing to their small sizes and to resolution limits. Nevertheless, ten bacterial LLPS systems have already been identified (Azaldegui et al., 2021), and the number is increasing. These observations support the proposal that LLPS in microbes may be more the rule than the exception. In E. coli, aggregated proteins can be disaggregated during environmental stresses by chaperones, and their spatio-temporal localization is changed in the process (Winkler et al., 2010). In rod-shaped bacterial cells, cell poles are special regions for the localization of signaling and sensing proteins, and here proteins like MreB exhibit random movement (Lopian et al., 2010; Van Teeffelen et al., 2011; Govindarajan et al., 2013; Shi et al., 2018). Furthermore, a high-throughput tagging pipeline of C. crescentus proteins revealed 153 proteins with patchy or spotty subcellular localization patterns (Werner et al., 2009). Together, the above observations provide evidence that LLPS may be widely involved in subcellular organization across different microorganisms, although compartments remain to be discovered.

Although many key questions regarding the organizing principle and physicochemical driving forces of phase separation remain unanswered, in many cases, weak and reversible multivalent interactions between proteins and/or nucleic acids have been demonstrated to be important drivers of biomolecular condensates (Banani et al., 2017). Several different theories explaining phase separation in condensates have been proposed, positing a role for electrostatic interactions, cation-π interactions, aromatic interactions, volume exclusion/crowding, surface tension, or the permeability rate of molecules. With these theoretical frameworks, it may now be possible to explain how the assembly, composition, dynamics, physical properties, and biochemical functions of these biomolecular condensates are regulated. Here, we focus on known mechanisms involved in driving LLPS in microorganisms.

In cells, phase separation is controlled by the assembly and material state of a variety of chaperone proteins, posttranslational modifications (PTMs), and cellular factors, and these in turn determine the size, assembly rate, and material properties of protein condensates, ensuring that distinct cellular functions can be spatiotemporally coordinated (Wang and Zhang, 2019; Quiroz et al., 2020). The various biological activities coupled with LLPS include the classification of misfolded and unwanted proteins for degradation, chromatin organization, gene expression, the assembly of signaling clusters, actin- and microtubule-based cytoskeletal networks, the asymmetric segregation of cell fate determinants, and the formation of pre- and post-synaptic density signaling assemblies (Marrone et al., 2019; Zhang et al., 2020). The functional mechanisms of biomolecular condensates in microorganism are listed in Table 1. Here, we highlight the main functions and summarize them into two main categories.

Evidence is accumulating that phase transitions may be a general mechanism through which microorganisms regulate cellular functions and rapidly adapt to a changing environment. The functions are presented as a model in Figure 4. However, several issues remain unresolved: What mechanisms regulate the specific recruitment of macromolecules in membrane-less organelles? In particular, why are some molecules allowed entry into these organelles while other molecules are selectively excluded? How (and under what circumstances) are these condensates assembled and disassembled? How can the biochemical reactions inside the condensates be scrutinized? By what mechanisms do some condensates divide further into additional compartments (or structured regions) that perform specialized functions? These questions may be addressed by studying the behavior of microbial cells at length over a relative long time scale, and by studying the structural, dynamic, and thermodynamic aspects of these condensates.

Numerous technical challenges need to be overcome to solve these problems. In microbial condensates (especially condensates from prokaryotic cells), the major hurdle is condensate size (Alberti et al., 2019). While traditional microscopic approaches can be used to detect LLPS in eukaryotic cells, LLPS in prokaryotic cells are typically an order of magnitude smaller (McSwiggen et al., 2019). Thus, prokaryotic condensates in vivo are typically <100–300 nm in diameter (Cho et al., 2018) which is beyond the spatial resolution of light microscopy (∼300 nm, Wenger et al., 2007), resulting in all the condensates appearing spherical. In vitro studies in simulation systems, can be a viable alternative to the in vivo assays. To date, almost all understanding of the protein structures and dynamics involved in bacterial condensates have been garnered from in vitro studies using recombinant proteins. However, these systems are comprised of only one or (at most) two components, and are considerably less complex than in vivo systems, which have properties that are determined by the coexistence of hundreds of thousands of macromolecules and small molecules in a highly confined volume. To mimic the crowded subcellular environment, crowding agents can be added to the in vitro systems. As mentioned above, however, it is not a simple matter to mimic typical condensate viscosity or viscoelasticity. Recently, single-molecule tracking/single-particle tracking (SMT/SPT) super-resolution microscopy and fluorescence correlation spectroscopy (FCS) have proved to be promising tools for investigating the properties of condensates in microorganisms (Gahlmann and Moerner, 2014; Tuson and Biteen, 2015; Sahoo et al., 2016; Jiang et al., 2017; Wei et al., 2017; Maharana et al., 2018; Sieben et al., 2018; Coelho et al., 2020; Gwosch et al., 2020; Peng et al., 2020), while proximity-dependent labeling approaches have been applied to map the protein interactome within the condensates (Govers et al., 2017; Markmiller et al., 2018; Ramanathan et al., 2018). Fluorescence recovery after photobleaching (FRAP), the “gold standard” assay in eukaryotic cells for measuring condensate fluidity and the dynamics of protein exchanges, may also be applied to study bacterial condensates, although model choice and data analysis need to be carefully considered (McSwiggen et al., 2019; Taylor et al., 2019). Together, the application of new fluorescence microscopy techniques to the study of microbial LLPS may prove ground-breaking, creating exciting new perspectives (Cambre and Aertsen, 2020).

To mimic subcellular compartmentalization and control micro-reactions in space and time, artificial membraneless organelles with liquid-like properties have been successfully constructed.

For example, artificial intracellular condensates were de novo designed in E. coli basing on a simple repeat sequence of (Gly-Arg-Gly-Asp-Ser-Pro-Tyr-Ser)XX (where XX is the number of repeats, between 20 and 80). They exhibited controllable dynamics by modulating the molecular weights (number of the repeats, Dzuricky et al., 2020). Protein/RNA coacervates, spider silk protein, and elastic-like protein were also engineered in E. coli with reversible formations, tunable dynamics, and selective enrichments in components, depending on the protein levels and the ratio of charged residues (Mushnikov et al., 2019; Wei et al., 2020; Yeong et al., 2020). In a recently engineered condensate comprised of small ubiquitin-like modifier (SUMOylation), enzyme activity increased approximately 36-fold in the droplets (compared with the surrounding bulk solution) (Peeples and Rosen, 2021). These studies have paved the way for the construction of synthetic membraneless organelles with designer functions in prokaryotes. These synthetic membraneless organelles have broad application prospects in biocatalysis, synthetic biology, and metabolic engineering (Deshpande et al., 2001). Multi-stimuli-responsive carriers (thermal or pH-responsive reversible coacervate droplets) can also be imbued with the ability to package and deliver drugs (Gabryelczyk et al., 2019). Furthermore, microfluidic techniques can be employed to create monodisperse coacervate droplets, making it possible to mimic diverse intracellular activities within uniform unilamellar lipid vesicles (Deshpande et al., 2001; Van Swaay et al., 2015; Deng and Huck, 2017; Love et al., 2020; Zhao et al., 2021). However, the intrinsic properties and functions of these coacervate droplets may differ dramatically as a function of size, and it remains unclear how large a condensate must grow before specific functions can arise (Lyon et al., 2021). One of the biggest challenges is spatiotemporal control over the time-programmed condensation and disassembly of the coacervate. The time-programmed phase behavior is currently available by changes in pH, temperature, ionic strength, light (UV), and more recently by enzyme-mediated catalytic activity (Shin et al., 2017; Schuster et al., 2018; Martin et al., 2019; Wheeler et al., 2019; Garabedian et al., 2021). Furthermore, by converting chemical fuels, the coacervate droplet could behave like a protocell capable of self-division, making it an ideal model for approaching the dynamic complexity of living cells (Zwicker et al., 2017; Donau et al., 2020; Karoui et al., 2021).

In summary, the number of liquid-like condensates identified in microorganisms has grown rapidly during the last few years. While the fundamental role of LLPS in membrane-less compartmentalization has drawn intense interest, new questions and hypotheses concerning the molecular mechanisms and biological processes associated with these microbial condensates have been raised. These gaps in knowledge may be filled through the development of multiscale and interdisciplinary approaches. As the field moves forward, new applications for microbial condensates will be explored.

GZ and GY contributed to conception, revision, and wrote sections of the manuscript. ZG and WZ wrote the first draft of the manuscript. GZ and ZG designed and made the figures and the table. RC and SZ wrote sections of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.