Interfacing bacterial microcompartment shell proteins with genetically encoded condensates

Abstract

Condensates formed by liquid–liquid phase separation are promising candidates for the development of synthetic cells and organelles. Here, we show that bacterial microcompartment shell proteins from Haliangium ochraceum (BMC‐H) assemble into coatings on the surfaces of protein condensates formed by tandem RGG‐RGG domains, an engineered construct derived from the intrinsically disordered region of the RNA helicase LAF‐1. WT BMC‐H proteins formed higher‐order assemblies within RGG‐RGG droplets; however, engineered BMC‐H variants fused to RGG truncations formed coatings on droplet surfaces. These intrinsically disordered tags controlled the interaction with the condensed phase based on their length and sequence, and one of the designs, BMC‐H‐T2, assembled preferentially on the surface of the droplet and prevented droplet coalescence. The formation of the coatings is dependent on the pH and protein concentration; once formed, the coatings are stable and do not exchange with the dilute phase. Coated droplets could sequester and concentrate folded proteins, including TEV protease, with selectivity similar to uncoated droplets. Addition of TEV protease to coated droplets resulted in the digestion of RGG‐RGG to RGG and a decrease in droplet diameter, but not in the dissolution of the coatings. BMC shell protein‐coated protein condensates are entirely encodable and provide a way to control the properties of liquid–liquid phase‐separated compartments in the context of synthetic biology.

1. INTRODUCTION

Cellular organelles have evolved to sequester macromolecules, substrates, cofactors, and enzymes to confine biological processes. Prominent examples of subcellular compartmentalization in bacteria are bacterial microcompartments (BMCs), metabolic organelles consisting of an enzymatic core encased within a selectively permeable protein shell (Kerfeld et al., 2018; Sutter et al., ²⁰²¹). In eukaryotic cells, compartments such as the nucleus and mitochondria are surrounded by a lipid bilayer, yet organelles such as the nucleolus, stress granules, P‐granules, and Cajal bodies provide compartmentalization without membranes (Brangwynne et al., 2009; Feric et al., ²⁰¹⁶; Hirose et al., ²⁰²³; Wasmeier et al., ²⁰⁰⁸). These “membraneless” organelles are biological condensates formed by liquid–liquid phase separation (LLPS) of RNA and/or intrinsically disordered proteins in the cell, driven by weak multivalent interactions such as ionic, cation‐π, and π‐π interactions (Banani et al., 2017; Courchaine et al., ²⁰¹⁶; Posey et al., ²⁰¹⁸). Biological condensates are dynamic structures that form and dissolve in response to stimuli, such as post‐translational modifications and changes in the intracellular concentrations of components, to regulate myriad biological processes. Aberrant formation or disruption of these structures has been implicated in diseases such as ALS and cancer (Ambadipudi et al., 2017; Boeynaems et al., ²⁰²³; Patel et al., ²⁰¹⁵). The self‐assembly of condensates and trafficking of proteins to specialized membraneless organelles are mediated by specific protein–protein or protein–nucleic acid interactions and nonspecific interactions, resulting in distinct chemical environments between different organelles that selectively recruit proteins and small molecules (Kilgore et al., 2024; Wei et al., ²⁰²⁰).

The principles underlying phase separation have been widely used to create artificial organelles under both in vitro and in vivo conditions (Abbas et al., 2021; Bracha et al., ²⁰¹⁹; Dai et al., ²⁰²³; Hilditch et al., ²⁰²³; Schuster et al., ²⁰¹⁸; Tomares et al., ²⁰²²). Formed by LLPS, these structures possess dense interiors that closely resemble the crowded environment of the cellular cytoplasm. A variety of functions have been engineered, including segregation of reaction components, regulation of gene expression, and controlled release of cargo on demand (Caldwell et al., 2018; Dzuricky et al., ²⁰²⁰; Garabedian et al., ²⁰²²; Gil‐Garcia et al., ²⁰²⁴; Lu et al., ²⁰²¹; Nakashima et al., ²⁰¹⁹; Reinkemeier et al., ²⁰¹⁹). The macromolecular composition of these phase‐separated droplets endows them with distinct physicochemical properties that define their unique microenvironment, which is markedly different from that of the surrounding dilute phase and can be tailored to specific functions (Kilgore et al., 2024). The surface behavior of biological condensates is governed by the interfacial tension between the condensed and dilute phases, which significantly affects their function and stability, particularly the size and structure of the coacervate. Over time, condensates undergo coarsening through Ostwald ripening and/or coalescence, leading to a reduction in the number of droplets (Brangwynne, 2013; Kelley et al., ²⁰²¹). The interfacial tension can be reduced by adsorbing onto the condensate surfactant‐like proteins or solid particles that stabilize the condensate and significantly reduce coarsening and coalescence (Abeysinghe et al., 2024; Folkmann et al., ²⁰²¹; Nakashima et al., ²⁰²¹). These results suggest that engineered coatings can be designed to control the interface between the condensed and dilute phases.

Bacterial microcompartment shell proteins are candidates for the functional coating of LLPS droplets because of their higher‐order self‐assembly and selective molecular permeability. Furthermore, BMC shell proteins have been engineered to alter their electrostatics, colocalize other proteins with protein–protein interaction domains, enable electron transport, and undergo stimuli‐responsive higher‐order self‐assembly (Aussignargues et al., 2016; Ferlez et al., ²⁰¹⁹; Hagen et al., ²⁰¹⁸; Huang et al., ²⁰²⁰; Kirst et al., ²⁰²²; Lee et al., ²⁰¹⁸; Plegaria et al., ²⁰²⁰; Young et al., ²⁰²⁰; Zhang et al., ²⁰¹⁹). Indeed, our group and others recently reported that BMC shell proteins assemble at the liquid–liquid interfaces between condensates formed by polymers, such as PEG/PVA/dextran mixtures (Abeysinghe et al., 2024; Izri & Noireaux, ²⁰²⁴). However, these condensates lack genetic encodability and are formed through a distinct mechanism compared to condensates formed in biological systems.

Here, we examined the interaction of BMC shell proteins from Haliangium ochraceum (BMC‐H) (Sutter et al., 2016, 2017) with a genetically encoded condensate sequence used in synthetic biology; the sequence spans the N‐terminal domain of the RNA‐binding helicase LAF‐1, an intrinsically disordered region that contains several arginine–glycine–glycine (RGG) repeats and is responsible for phase separation in cells (Elbaum‐Garfinkle et al., 2015; Oliva et al., ²⁰²⁰). Constructs containing tandem repeats of the RGG domain, referred to as RGG‐RGG, undergo phase separation at lower concentrations, higher temperatures, and higher salt concentrations than those of the single domain (Caldwell et al., 2018; Schuster et al., ^{2018, 2020}). The WT BMC‐H protein forms hexamers with exposed negative charges on both surfaces in solution, and these hexamers self‐assemble into extended 2D sheet assemblies mediated by shape complementation and electrostatics, as shown by crystallography, atomic force microscopy, and transmission electron microscopy (Sutter et al., 2016). We hypothesized that the negatively charged surface of the hexamers would interact with the positively charged condensates and that the condensates could template the formation of coatings on their surfaces. We found that WT BMC‐H colocalized with the condensed phase; however, the shell protein was prone to uncontrolled self‐assembly and altered the morphology of the RGG‐RGG condensates. To facilitate compatibility with the droplets, we designed fusion tags of different lengths and compositions derived from RGG and encoded them at the C‐terminus of the shell protein. One of these fusion tags, T2, optimized the coating of the droplets and minimized the partitioning of BMC‐H into the droplets; the interface assembly was modulated by the shell protein concentration and pH. Once formed, the coatings on the droplets were stable and did not exchange with BCM‐H proteins in the dilute phase. The coatings also stabilized the droplets, prevented coalescence, and changed the zeta potential of the condensates. Coated condensates concentrated small proteins with basic isoelectric points. Treatment with TEV protease, which digests the tandem RGG‐RGG construct to two single RGG domains, led to shrinkage of the coated droplets to a more uniform size distribution centered around 1 μm. In summary, this study showed that genetically encodable BMC shell proteins and RGG‐RGG‐based condensates form scalable and stimuli‐responsive compartments capable of concentrating varied molecular cargos, suggesting general applicability for synthetic cell and/or organelle construction.

2. RESULTS

2.1. Engineered BMC‐H shell proteins coat RGG‐RGG coacervates

We first analyzed the interaction between BMC‐H shell proteins and droplets formed by RGG‐RGG, which forms condensates in physiological buffer and at room temperature at concentrations as low as μM (Schuster et al., 2018) (Figure S1), using confocal microscopy.

WT BMC‐H was labeled with rhodamine for visual tracking and mixed with RGG‐RGG to a total concentration of 5 μM BMC‐H and 15 μM RGG‐RGG, well above the critical concentration for phase separation in the conditions used. We found that WT BMC‐H was sequestered and formed higher‐order assemblies within the droplets and that the morphology of the droplets deviated from spherical (Figure 1a–c). To test whether modulating the self‐assembly properties of WT BMC‐H would affect its interaction with the condensates, we designed several BMC‐H variants with altered solubility and extensions (Figure S2). First, we assessed the behavior of a variant of BMC‐H containing a SUMO tag at the N‐terminus (SUMO‐BMC‐H), which solubilizes the protein and prevents higher‐order assembly of hexamers into 2D sheets (Folkmann et al., 2021). We found that SUMO‐BMC‐H preferentially partitions into the droplets, as indicated by an enrichment index (EI, ratio of fluorescence intensity inside to outside the droplets) (Schuster et al., 2018) of 14.5, with no evidence of higher‐order assembly, droplet distortion, or coating (Figure 1d–f).

FIGURE 1.

Colocalization of bacterial microcompartments (BMC) shell proteins in an RGG‐RGG liquid–liquid phase‐separated droplet system. Addition of 5 μM WT BMC‐H protein to 15 μM RGG‐RGG droplets results in altered morphology and higher‐order assembly; transmitted light detector channel (TD) (a) rhodamine‐labeled WT BMC‐H (b) and merged channel (c). Addition of 5 μM SUMO‐BMC‐H in 15 μM RGG‐RGG shows no change in droplet morphology; TD channel (d) rhodamine‐labeled SUMO‐BMC‐H (e) and merged channel (f). Conditions: 20 mM Tris pH 9 and 150 mM NaCl. Scale bars: 10 μm.

Next, we analyzed BMC‐H constructs containing small unstructured C‐terminal fusion tags connected via a 10‐residue glycine/serine linker and a TEV protease recognition sequence (Figure S2). A 6x‐lysine (BMC‐H‐6xLys) and a single cysteine (BMC‐H‐Cys) variant formed higher‐order sheets, as assessed by negative electron microscopy staining (Figure S3A,C). When mixed with RGG‐RGG, both unstructured C‐terminal fusions formed a patchy coating at the interface, with larger aggregates visible, and these samples appeared to lack higher‐order assembly inside the droplets compared to the wild‐type protein (Figure S3B,D).

These observations suggested that the intrinsic tendency of BMC‐H proteins to self‐assemble into higher‐order aggregates could be controlled by fusion tags that modulate the solubility and higher‐order assembly of hexamers while maximizing interactions with RGG‐RGG droplets. To test this hypothesis, we designed constructs containing portions of the RGG sequences fused to the C‐terminus of the BMC‐H shell protein, separated by a TEV protease recognition sequence (Figure 2a). Truncations 1 (BMC‐H‐T1) and 2 (BMC‐H‐T2) include the N‐terminal sequence of RGG, which preserves the RYVPPHLRGG motif that is crucial for phase separation (Schuster et al., 2018). BMC‐H‐T2 is approximately half the length of BMC‐H‐T1 and is enriched in hydrophobic residues (Table S1). Truncation 3 (BMC‐H‐T3) is derived from the C‐terminus of the RGG domain, includes fewer hydrophobic residues, and lacks the critical phase separation sequence, while maintaining the same length as BMC‐H‐T2. None of the variants undergo liquid–liquid phase separation by itself, as assessed by turbidity assay (Figure S4).

FIGURE 2.

Purified bacterial microcompartments (BMC)‐H fused to RGG truncations interact with droplets: (a) LAF‐1 RGG truncation sequences. T1 and T2 preserve the critical RYVPPHLRGG sequence necessary for phase separation. T3 contains only one hydrophobic residue. T2 and T3 are approximately the same length. (b–d) Confocal fluorescence microscopy of rhodamine‐labeled BMC‐H variants; top panels show the transmitted light detector channel (TD), middle panels show the fluorescence channel, and bottom panels show the merged image. (b) BMC‐H‐T1 is sequestered and forms patchy coating. (c) BMC‐H‐T2 forms consistent coatings. (d) BMC‐H‐T3 partitions to the droplets and not at the interface. Samples prepared with 30 μM RGG‐RGG and 60 μM BMC‐H variants in 20 mM Tris pH 9 and 150 mM NaCl. Scale bars: 10 μm.

Rhodamine‐labeled BMC‐H fusion proteins (60 μM) were mixed with RGG‐RGG (30 μM) and equilibrated for 2 h. Confocal microscopy analysis revealed that BMC‐H‐T1 and BMC‐H‐T2 fusions formed coatings, whereas BMC‐H‐T3 was predominantly localized inside the droplets (Figure 2b–d). Although both N‐terminal RGG truncations coated the droplets, the shorter BMC‐H‐T2 showed preferential localization at the interface (Figure 2c, Table S2).

Interface assembly could occur via an equilibrium process that requires molecules to move from the dense droplet phase to the interface. To test this, we mixed labeled BMC‐H‐T2 with RGG‐RGG droplets and imaged them immediately upon addition to a microscope slide. Initially, BMC‐H‐T2 partitioned into droplets and then assembled at the interface after approximately 1 h when mixed in ratios equal to or greater than the RGG‐RGG concentration; addition of shell protein at concentrations lower than RGG‐RGG required longer equilibrations (Figure 3a). Coated droplets were smaller in size than the uncoated droplets, with average diameter of 1.5 ± 0.8 μm versus 2.2 ± 1 μm (Figure S5), and remained stable for hours with no evidence of coalescence.

FIGURE 3.

Equilibrated droplets with bacterial microcompartments (BMC)‐H‐T2 coatings are stable. (a) BMC‐H‐T2 requires equilibration upon mixing with RGG‐RGG. Confocal images of rhodamine‐labeled 15 μM BMC‐H‐T2 added to 30 μM RGG‐RGG. (b) Droplets equilibrated with BMC‐H‐T2 shells labeled with Coumarin‐343 (5 μM) for 1 h. (c) The equilibrated droplets were mixed with BMC‐H‐T2 proteins labeled with rhodamine (5 μM) and allowed to equilibrate for 1 h. Rhodamine‐labeled shell protein did not exchange with the C‐343‐labeled coatings and partitioned to the condensed phase. Conditions: 20 mM Tris pH 9 and 150 mM NaCl Scale bars: 10 μm.

We explored whether the coatings exchanged soluble species by adding shell proteins to the pre‐formed coatings. BMC‐H‐T2 labeled with Coumarin‐343 was equilibrated with the droplets to form coatings (Figure 3b), followed by the addition of rhodamine‐labeled BMC‐H‐T2, and the system was imaged over 1.5 h (Figure 3c). The rhodamine‐labeled protein partitioned into the interior of the droplets and did not exchange with the interface. This behavior indicates that the coatings are stable after formation and do not exchange with other BMC shell proteins in solution.

To test whether RGG domain truncation fusions promoted assembly at the interface of other proteins, we fused the truncations to iLOV, a soluble fluorescent protein that does not form 2D assemblies and partitions poorly to the condensed phase. We found that all three truncation fusions increased partitioning into the RGG‐RGG condensed phase compared with WT iLOV, as shown by larger EI values, but did not result in localization at the interface (Figure S6).

2.2. Interface assembly is controlled by environmental pH and concentration

The higher‐order assembly of WT BMC‐H is controlled by the pH, salinity, protein concentration, and interface environment (Abeysinghe et al., 2024; Faulkner et al., ²⁰¹⁹; Sutter et al., ²⁰¹⁶). When imaged on mica, pH and salt composition influence the size of the 2D BMC‐H protein sheets formed, where higher pH promotes larger patch sizes and slower dynamics (Faulkner et al., 2019). Therefore, we tested the influence of pH on the interfacial assembly of BMC‐H‐T2 in an RGG‐RGG condensate system. We found that at low pH (corresponding to a higher solubility of the BMC shell protein), BMC‐H‐T2 was sequestered inside the droplets; no assemblies or coatings were visible (Figure 4a). We observed increased partitioning to the interface as a function of increasing pH, while the concentration of BMC‐H‐T2 in the diluted phase decreased (Figure 4c). The mean fluorescence intensity of the corona area, used as a proxy for the concentration of BMC‐H‐T2 at the interface, increases above pH 7 (Figure 4d).

FIGURE 4.

Coating behavior as a function of pH and bacterial microcompartments (BMC)‐H concentration. (a) BMC‐H‐T2 shells (60 μM) mixed with RGG‐RGG (30 μM) at pH values ranging from 6 to 9; representative images (top) and fluorescence intensity plotted by distance for 10 droplets (bottom). (b) RGG‐RGG (30 μM) mixed with increasing concentrations of BMC‐H‐T2 in 20 mM Tris pH 9.0 and 150 mM NaCl (top). Fluorescence profiles shown for 10 droplets and normalized for size (bottom). Scale bars: 5 μm. (c) Fluorescence of BMC‐H‐T2 in the dilute phase versus pH (n = 100–130). (d) Fluorescence intensity of the coatings (calculated as difference in fluorescence between the whole droplet and the interior) versus pH and (e) BMC‐H‐T2 concentration. Each point is the average of 10 droplets shown with standard deviation.

We tested whether increasing BMC‐H‐T2 concentration in coating reactions resulted in more complete coverage by measuring the average fluorescence of the corona area. After equilibration, we observed emerging coating at concentrations above 5 μM BMC‐H‐T2 in solutions of 30 μM RGG‐RGG (Figure 4e).

2.3. Partitioning of protein molecules within coated BMC shell protein droplets

We tested whether cargo proteins could partition into the RGG‐RGG condensed phase by adding FITC‐labeled proteins to droplets coated with rhodamine‐labeled BMC‐H‐T2 at pH 9.0, corresponding to the optimal conditions for coating formation. We found that the coated droplets were permeable to a variety of proteins, including those with higher molecular weights (Figure 5). The proteins partition to the droplets with varying efficiency; partitioning is slightly reduced in coated droplets compared to the uncoated droplets, indicating that the coating does not impair the ability of the droplets to sequester cargo (Table S3).

FIGURE 5.

Bacterial microcompartments (BMC)‐H‐T2 coated condensates sequester protein cargos. Proteins of various molecular weights and isoelectric points were labeled with fluorescein isothiocyanate (FITC) (with the exception of iLOV and sfGFP) and added to equilibrated droplet/BMC‐H‐T2 systems (20 mM Tris–HCl, pH 9, 150 mM NaCl). Fluorescence profiles of the droplets show the intensity of both the rhodamine‐labeled BMC‐H‐T2 (red) and the FITC‐labeled or fluorescent protein (green). Charges of solvent‐accessible residues at pH 9 are shown in Figure S11. Scale bars: 5 μm.

Smaller proteins with a pI above that of RGG‐RGG (9.43), such as lysozyme, cytochrome c, and aprotinin, partitioned into the condensed phase as shown by positive EI values (Table S1, Figure S11). Proteins with a low pI exhibit a markedly lower propensity for partitioning: BSA has an EI = 4.4 ± 1.2, and iLOV an EI = 1.5 ± 0.5; sfGFP (pI 6.23) did not partition into droplets, as shown by an EI of 0.27 ± 0.6.

2.4. Physical properties

We assessed whether coatings change the propensity to coalesce of RGG‐RGG droplets by measuring changes in the zeta potential, which indicates the electrostatic repulsion or attraction between particles dispersed in solution. We found that uncoated RGG‐RGG droplets had a Z potential of −1.48 ± 0.49 mV at room temperature using the diffusion barrier method. When the droplets were coated with BMC‐H‐T2 (5 μM), the zeta potential decreased to −9.58 ± 3.25 mV, suggesting an increase in stability (Welsh et al., 2022). The lower potential likely reflects the overall negative charge of the BMC‐H hexamer at pH 9 (Figure S7).

The interactions with cargo proteins suggest that the physical properties of the condensates control which solutes can partition to the droplets. We tested the polarity of the droplets using betaine Reichardt's dye, which absorbs at different wavelengths based on polarity (Reichardt, 1994). When added to RGG‐RGG droplets, the dye became purple, indicating that the interior was less polar than the aqueous solution and had a polarity similar to that of ethanol (Figure S8). The molar electronic transition energy (E _T (30)) of Reichardt's dye in the condensed phase is 51.1 kcal/mol, similar to the E _T (30) of 51.9 kcal/mol measured in ethanol, and lower than the E _T (30) of 55.4 kcal/mol measured in methanol.

2.5. Stimuli responsivity of BMC‐H‐T2 coated IDP droplets

The programmability of synthetic cells and organelles to respond to external environmental stimuli, which alter their shape, size, permeability, or other properties, is necessary for next‐generation applications. In our system, both RGG‐RGG and BMC‐H‐T2 had TEV protease recognition sites engineered into their primary sequences. In the case of RGG‐RGG, the cleavage site is located between the RGG domains; digestion with TEV results in two RGG domains and the dissolution of condensates, as previously reported (Schuster et al., 2018). We engineered the TEV cleavage site in BMC‐H‐T2 between the shell protein and RGG truncation sequence so that digestion revealed WT BMC‐H.

First, we verified that TEV protease could partition into condensates. Rhodamine‐labeled TEV was treated with iodoacetamide, which inactivates the enzyme, and was added to the RGG‐RGG droplets. We found that inactivated TEV protease accumulated in the condensed phase (Figure S1). As expected, uncoated RGG‐RGG droplets formed at 5 μM protein readily dissolved when treated with active TEV protease (Figure S9A) because the resulting single repeat RGG does not undergo phase separation at 10 μM in 150 mM NaCl buffer, pH 7.5 (Schuster et al., 2018). RGG‐RGG droplets coated with BMC‐H‐T2 and treated with functional TEV protease visibly shrunk upon the addition of TEV protease; the average diameter (n = 50) decreased rapidly from 3.3 ± 2.3 μm and stabilized at 0.84 ± 0.22 μm over 1 h of digestion (Figure 6b). Under the conditions used, the initial interface coatings appeared incomplete, and regular coatings were formed as the droplets collapsed. The concentration of BMC‐H‐T2 in the condensed phase decreased as the droplet shrank and the coating appeared more complete (Figure 6c). FITC‐lysozyme localized within the droplets throughout the experiment (Figure S10).

FIGURE 6.

Protease addition alters coated droplets: (a) Schematic of TEVp addition experiment. (b) Change of the average diameter of the droplets (n = 50) over time in the protease digest. Inset: The difference in fluorescence between the whole droplet and the interior increases as the digest proceeds. (c) Microscopy time course of TEVp added to droplet/shell protein mixture of 15 μM RGG‐RGG and 5 μM rhodamine‐labeled BMC‐H‐T2 with 1 μM of TEV protease in 20 mM Tris, pH 8.5, and 150 mM NaCl; scale bar: 5 μm.

3. DISCUSSION

Our results demonstrate that BMC shell proteins can form coatings on the surface of LLPS droplets formed by RGG‐RGG proteins. We observed that coating behavior was optimized by fusion tags: WT BMC‐H formed higher‐order assemblies within the condensed phase, whereas SUMO‐BMC‐H, which does not form 2D sheets in solution, was miscible with the condensed phase and did not assemble at the interface. A BMC‐H variant containing a positively charged fusion tag and capable of forming 2D assemblies in solution displayed a behavior intermediate between higher‐order assembly and partitioning. Fusion to truncated versions of RGG was necessary to yield robust interfacial assembly; as hypothesized, this behavior was maximized in truncations (T1 and T2) that preserved an N‐terminal sequence necessary for phase separation in RGG. The sequence of the fusion tag used modulated the interaction of the shell protein with the condensed phase: a long N‐terminal truncation (BMC‐H‐T1) resulted in a mixed behavior, with partitioning into the droplets and partial assembly at the surface of the droplets. A short N‐terminal truncation (BMC‐H‐T2) drove the preferential assembly of the shell proteins at the surface. A C‐terminal truncation (BMC‐H‐T3) resulted in partitioning to the droplets, with no interfacial assembly observed. These behaviors suggest that assembly at the interface results from a balance between the intrinsic self‐assembly of the WT BMC‐H protein and its solubilization in the RGG‐RGG condensed phase, mediated by the fusion tags.

The BMC‐H coatings formed a barrier on the surface of the condensates, which prevented coalescence of the droplets and changed their zeta potential, while also preserving sequestration of cargo proteins and enzymes. WT BMC‐H proteins formed higher‐order assemblies within RGG‐RGG droplets; however, the formation of regular coatings was enabled by fusion tags similar in composition to the condensed phase. The formation and stability of coatings were influenced by pH, protein concentration, and the ratio of shell proteins to RGG‐RGG (Figure 4). Once formed, the coatings were stable and did not exchange with soluble proteins. The coatings were permeable to cargo proteins, with partitioning efficiency depending on the pI and other properties of the proteins (Schuster et al., 2020). Treatment with TEV protease, which released the fusion tags and dissolved the condensed phase, led to a decrease in droplet size and an increase in interface coating.

Our findings expand on previous work on protein‐based compartmentalization and interface engineering (Abeysinghe et al., 2024; Izri & Noireaux, ²⁰²⁴; Kelley et al., ²⁰²¹). The ability of engineered BMC‐H variants to form coatings on RGG‐RGG droplets aligns with previous observations of BMC shell proteins assembling at aqueous liquid–liquid interfaces and expands the behavior to an entirely encodable system (Abeysinghe et al., 2024; Izri & Noireaux, ²⁰²⁴). The pH and concentration dependence observed for the formation of droplet interface coating is consistent with previous studies on the higher‐order assembly of BMC shell proteins for which lower than physiological pH interrupts higher‐order assembly and higher than physiological pH values promote larger assemblies (Faulkner et al., 2019). The coatings' stability and responsiveness to stimuli (e.g., proteolytic cleavage) open up possibilities for creating dynamic, reconfigurable compartments that sequester (bio)molecular cargo.

This genetically encodable strategy described here not only defines an approach forming stable biologically derived protein‐coated liquid–liquid phase‐separated droplets but also provides a framework to develop the design principles of protein self‐assembly in phase‐separated systems. For example, disordered‐predicted, C‐terminal extensions are frequent additions to natural BMC shell proteins and their role(s) are yet to be fully understood (Trettel et al., 2024). Follow‐up studies addressing how disordered extensions influence higher‐order assembly inside a condensate, at the interface, or the dilute phase will enable predictive engineering of protein assembly with condensates. We anticipate that continued advances in near‐atomic cryo‐EM imaging of protein/polymer complexes will play a significant role (Rizvi et al., 2024).

Overall, synthetic biology continues to push the complexity and diversity of strategies used to construct synthetic cells and organelles (Rothschild et al., 2024). Here, we show that shell proteins derived from bacterial microcompartments can functionalize protein IDP condensates, mimicking the function of membrane bilayers in organelles. Conversely, interaction with the condensates can serve as a template to form stable coatings around IDP droplets. These results set the stage for further functionalization of the interface by taking advantage of the functional flexibility of the BMC shell proteins, which can be engineered to accommodate metal centers, cofactors, and biotic and abiotic cargo.

4. EXPERIMENTAL SECTION

4.1. Expression of RGG‐TEV‐RGG

The RGG‐RGG construct was purchased from AddGene (#124940) and transformed into BL21‐DE3 E. coli (C2527I) purchased from New England Biolabs. Starter cultures were grown in Terrific Broth media with 1 mM kanamycin overnight at 37°C. Large cultures of Terrific Broth auto‐induction media were inoculated with 1 mL/L and grown overnight at 37°C. Cell pellets were resuspended in 30 mL of buffer (20 mM Tris pH 7.5, 1 M NaCl, 20 mM imidazole) and lysed by continuous sonication for 15 min at room temperature with a protease inhibitor cocktail (Roche cOmplete™, Mini, EDTA‐free Protease Inhibitor Cocktail). Lysate was diluted to 100 mL and centrifuged at 18,000 rpm for 90 min at 40°C to prevent phase separation. The cleared lysate was loaded onto a 5‐mL Ni‐NTA column equilibrated with 20 mM Tris, pH 7.5, 500 mM NaCl, and 20 mM imidazole. The column was washed at 4.0 mL/min for 9 CV. Protein was eluted with a gradient to 500 mM imidazole over 10 CV at 3.0 mL/min. Fractions containing protein as analyzed by SDS‐PAGE were dialyzed against 20 mM Tris pH 7.5 and 500 mM NaCl at 60°C to prevent phase separation. Aliquots of ~90 μM RGG‐RGG were flash frozen in liquid nitrogen and stored at −80°C. Experiments involving RGG‐RGG used aliquots thawed and heated to 70°C. Concentrations were determined by diluting small volumes into 8 M urea.

4.2. Expression of BMC‐H RGG truncation proteins

E. coli T7 Express BL21 (DE3) cells were transformed with 1 ng of pET11b‐based vector: pEJY140, pEJY141, or pEJY145. Single colonies from carbenicillin plates were transferred to 5 mL liquid media plus antibiotic starters. Overnight liquid cultures were inoculated with a starter culture (1 L of LB media with antibiotic), grown until OD₆₀₀ 0.6–0.8 at 37°C, and then induced with IPTG (100 μm) for ~16 h of growth at 22°C. Cell pellets were formed via centrifugation of cultures and stored at −20°C. Frozen pellets were then resuspended in buffer (50 mM Tris pH 8.0, 100 mM NaCl, and 10 mM MgCl₂) supplemented with lysozyme (≈5 mg), DNAase (≈0.2 mg), and RNAase (≈0.2 mg). French press treatment at 1500 PSI then ruptured the cells. Triton X‐100 (1%) was then added and allowed to mix for 10 min before fractionation by centrifugation at 20,000 × g for 20 min. The insoluble protein pellet was washed with 30 mL 1% Triton X‐100 in resuspension buffer for a total of three rounds of centrifugation separating the pellet and fresh solution added with each round. Two rounds of resuspension buffer centrifugation washes removed lingering Triton X‐100, and then, the protein was stored frozen at −20°C.

4.3. Expression of iLOV RGG truncation proteins

The iLOV truncation constructs were cloned into a T7 pET vector and transformed into BL21‐DE3 E. coli (C2527I) purchased from New England Biolabs. 500‐mL cultures of TB auto‐induction media with 1 mM ampicillin were inoculated with one colony from the transformation and grown overnight at 37°C. Cell pellets were resuspended in 15 mL of buffer (20 mM sodium phosphate pH 8.0, 100 mM NaCl, 10 mM imidazole) and lysed by sonication with 30‐s intervals for 10 min on ice with a protease inhibitor cocktail (Roche cOmplete™, Mini, EDTA‐free Protease Inhibitor Cocktail). Lysate was centrifuged at 18,000 rpm for 60 min at 4°C. The cleared lysate was loaded onto a 1 mL Ni‐NTA column equilibrated with lysis buffer. The column was washed at 1.5 mL/min for 15 CV. Protein was eluted with 50 mM sodium phosphate pH 8.0, 500 mM NaCl, and 400 mM imidazole without gradient. Fractions that eluted green were collected and analyzed by SDS‐PAGE. Fractions containing pure protein were buffer exchanged by PD‐10 column (Cytiva) into PBS buffer. Aliquots of ~200 μM were flash frozen in liquid nitrogen and stored at −80°C.

4.4. Expression of sfGFP

The sfGFP in a pET‐15b plasmid was transformed into BL21‐DE3 E. coli (find NEB number) purchased from New England Biolabs. Starter cultures were grown in TB media with 1 mM ampicillin overnight at 37°C. 1 L of TB media was inoculated with 2 mL of starter and grown to OD₆₀₀ = 0.9 at 37°C. The cells were induced with 1 mM IPTG and grown overnight at 37°C. Cell pellets were resuspended in 35 mL of buffer (20 mM Tris pH 7.5, 1 M NaCl, 20 mM imidazole) and lysed by sonication with 30 s intervals for 20 min at room temperature. Lysate centrifuged at 18,000 rpm for 30 min at 4°C. The cleared lysate was loaded onto a 5 mL Ni‐NTA column equilibrated with the lysis buffer. The column was washed at 5.0 mL/min for 8 CV. Protein was eluted without a gradient with 20 mM Tris pH 7.5, 500 mM imidazole, and 500 mM NaCl for 16 mL at 5.0 mL/min. Purity of the protein was analyzed by SDS‐PAGE and was buffer exchanged by PD‐10 column (Cytiva) into PBS buffer. Aliquots of sfGFP were flash frozen in liquid nitrogen and stored at −80°C.

4.5. Labeling

BMC‐H proteins were labeled with NHS‐ester dyes (Rhodamine: Thermo Scientific 46,406, Coumarin 343). The proteins were diluted to 150 μM in 20 mM sodium phosphate pH 8 and 500 mM NaCl. The samples were sonicated in a bath sonicator for 15 min. The NHS dye was dissolved in DMSO (rhodamine, 25 mg/mL) or methanol (C343, ~10 mg/mL) and added to the protein at a 10:1 concentration. The reaction was carried out for 1 h at room temperature. Excess dye was removed by centrifuging the proteins at 14,000 rpm for 5 min at room temperature, removing the supernatant, and resuspending in the desired buffer three times.

Proteins used in the permeability assay were labeled with fluorescein isothiocyanate (FITC, Sigma‐Aldrich F3651). Powdered stocks were dissolved into 100 mM sodium bicarbonate pH 9 to 2 mg/mL. The FITC stock solution (1 mg/mL in DMSO) was prepared, and 50 μL was added to 1 mL of protein solution 5 μL at a time. Solutions were incubated at 4°C overnight. Excess dye was removed by gel filtration (Cytiva PD‐10 columns 17085101).

The enrichment index (EI) was calculated from confocal microscopy images using ImageJ. The mean fluorescence intensity (MFI) of multiple droplets was averaged, as well as multiple areas of similar size in the dilute phase. The average MFI of the droplets was divided by the MFI of the dilute phase. EI = (F _drop)/(F _dil).

4.6. Zeta potential

Zeta potential readings were done in Malvern DTS 1070 cuvettes using the barrier diffusion method. Cuvettes were rinsed with 5 mL ethanol followed by 5 mL Milli‐Q water before being rinsed with 5 mL of buffer (20 mM Tris, pH 9, 150 mM NaCl). The samples were loaded at 50 μL with gel‐loading pipette tips. Readings were taken with three measurements of 20 subruns each at room temperature. Voltage was reduced to 20 V, and the samples were allowed to cool for 180 s between measurements. Samples included 15 μM RGG‐RGG with and without 5 μM 5815‐T2 labeled with rhodamine in the buffer listed above.

4.7. Confocal microscopy

Droplets were imaged using a Nikon AX R laser scanning confocal microscope (Ti2‐E inverted microscope), x60/1.42 NA Plan‐Apochromat objective. Images were collected and processed with NIS Elements software (Nikon). Fluorescence images were recorded using 445, 488, and 561 nm lasers for C343 and iLOV, FITC and GFP, and Rhodamine, respectively. Coverslips were treated with PEG‐silane to prevent glass wetting. HybriWell™ adhesive chambers (Grace Biolabs) were used to equilibrate samples.

TEV digests, photobleaching, pH studies, and permeability experiments were done with 15 μM RGG‐RGG and 5 μM labeled 5815‐T2 in 20 mM Tris pH 9 and 150 mM NaCl. The samples were allowed to equilibrate for 2 h before imaging. Permeability studies added 5 μM of FITC‐labeled protein to equilibrated systems before immediately imaging. TEV Digests were performed by adding 1 μM TEV (New England Biolabs, P8112S) to the sample and imaging every 5–10 min over 90 min.

4.8. Assessment of polarity

A 10‐mg/mL stock of Reichardt's dye was made in methanol (Reichardt, 1994). To remove the methanol from solutions, 10 μL of stock was placed in a microcentrifuge tube and evaporated on a heating block. A sample of 50 μM RGG‐RGG was added to the dried dye and heated at 70°C to prevent phase separation. The supernatant was transferred to a new tube leaving behind the insoluble dye. The sample was kept at 70°C until imaged. For absorbance, the sample was allowed to cool completely and the supernatant removed. The concentrated droplets were heated to 70°C and pipetted hot onto a NanoDrop 3000. The solution was allowed to cool to room temperature for 5 min before scanning. E _T (30) was calculated using the following equation:

$$E_T\left(30\right)\left(\text{kcal}/\mathrm{mol}\right)=\mathit{𝒉}{\mathit{cṽ}}_{\max }{N}_{\mathrm{A}}=\left(2.8591\times {10}^{-3}\right){ṽ}_{\max }\left({\mathrm{cm}}^{-1}\right)=28591/{\mathbf{\text{𝛌}}}_{\max }\left(\mathrm{nm}\right).$$ (1)

AUTHOR CONTRIBUTIONS

Michele Costantino: Conceptualization; investigation; writing – original draft. Eric J. Young: Investigation; conceptualization; writing – review and editing. Abesh Banerjee: Investigation. Cheryl A. Kerfeld: Funding acquisition; supervision; writing – review and editing. Giovanna Ghirlanda: Conceptualization; investigation; project administration; writing – review and editing; funding acquisition; supervision.

Supporting information

Data S1. Supporting information.

Costantino M, Young EJ, Banerjee A, Kerfeld CA, Ghirlanda G. Interfacing bacterial microcompartment shell proteins with genetically encoded condensates. Protein Science. 2025;34(3):e70061. 10.1002/pro.70061

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.