Living material assembly of bacteriogenic protocells

Advancing the spontaneous bottom-up construction of articial cells with high organisational complexity and diverse functionality remains an unresolved issue at the interface between living and non-living matter. To address this challenge, a living material assembly process based on the capture and on-site processing of spatially segregated bacterial colonies within individual coacervate micro-droplets is developed for the endogenous construction of membrane-bounded, molecularly crowded, compositionally, structurally and morphologically complex synthetic cells. The bacteriogenic protocells inherit diverse biological components, exhibit multi-functional cytomimetic properties and can be endogenously remodelled to include a spatially partitioned DNA/histone nucleus-like condensate, membranized water vacuoles and a self-supporting 3D network of F-actin proto-cytoskeletal laments. The ensemble is biochemically energized by self-sustainable ATP production derived from implanted live E. coli cells to produce a cellular bionic system with amoeba-like external morphology and integrated life-like properties. Our results demonstrate a novel bacteriogenic strategy for the bottom-up construction of functional protoliving micro-devices and provide opportunities for the fabrication of new synthetic cell modules and augmented living/synthetic cell constructs with potential applications in engineered synthetic biology and biotechnology.


Introduction
Establishing true-to-life functionality in synthetic cells is a global grand challenge that traverses multiple elds including synthetic biology, bioengineering and origins of life research. [1][2][3][4][5] The transitioning from inert capsule-based technologies to dynamical micro-compartmentalized entities capable of autonomous cytomimetic behaviour requires breakthrough advances in functional integration and on-board energization of multiplexed micro-systems. To date, the engineering of synthetic cellular systems (protocells) has been mainly approached using self-assembled vesicles, [6][7][8] semi-permeable microcapsules 9-15 and membrane-less or coated coacervate micro-droplets. [16][17][18][19] These compartments provide a controllable medium for the encapsulation and exchange of biological and non-biological components that are experimentally selected to demonstrate single functions such as gene expression, 6,9,20,21 enzyme catalysis 22 and ribozyme activity 23,24 within the synthetic cell models.
Achieving high organisational and functional complexity in these constructs is methodological demanding due to di culties in establishing su cient compositional diversity and chemical complementarity by conventional methods of micro-compartmentalization under close-to-equilibrium conditions. These limitations restrict the structural and chemical complexity of current protocell models, inhibit the implementation of integrated componentry, and impede the development of energized cytomimetic systems.
To address these issues, herein we develop a radically new approach to engineer the rst generation of energized protocells based on prokaryotes as structural and functional building blocks. In brief, we implement a living material assembly process for the bottom-up endogenous construction of membrane-bounded molecularly crowded synthetic cells with multi-functional cytomimetic properties.
Our approach is based on the co-capture and on-site processing of two spatially segregated bacterial colonies (Escherichia coli (E. coli) and Pseudomonas aeruginosa (P. aeruginosa, PAO1 strain) that are coassociated with individual poly(diallyldimethylammonium chloride) (PDDA)/adenosine 5'-triphosphate (ATP) coacervate micro-droplets. In situ lysis of the captured bacteria spontaneously gives rise to lipid membrane-coated protocells enclosing an extensive repertoire of functional biological components. We demonstrate that the bacteriogenic protocells are capable of complex processing such as the endogenous production of ATP via proto-metabolic activity (glycolysis) and inherit a su cient complement of the bacterial gene expression machinery to implement in vitro transcription and translation. To increase the level of internal structural organization, we use a combination of endogenous polynucleotide liquid-liquid phase separation, ATP-driven supramolecular protein assembly and hypotonicity to augment the synthetic cells with a spatially partitioned nucleus-like DNA/histone condensate, 3D network of F-actin proto-cytoskeletal laments and osmotically responsive membranebounded water vacuoles, respectively. As a step towards self-sustainable energization, we construct living/synthetic hybrids in which we exploit implanted live E. coli cells as surrogate mitochondria to increase and extend the endogenous production of ATP for enhancing kinase activity, glycolysis, in vitro gene expression and cytoskeletal assembly within the bacteriogenic protocells. The protoliving constructs adopt an amoeba-like external morphology and decreased membrane permeability due to onsite bacterial metabolism and growth to produce a cellular bionic system with integrated life-like properties. Taken together, our results demonstrate a novel bacteriogenic strategy for the bottom-up construction of functional protocellular micro-devices and provide opportunities for the fabrication of new synthetic cell modules and augmented living/synthetic cell constructs with potential applications in engineered synthetic biology and biotechnology.

Assembly and construction of bacteriogenic protocells
To implement the endogenous assembly of bacteriogenic protocells, we rst constructed a microscale building site comprising 5-30 μm-diameter pre-former coacervate droplets of spatially segregated live bacterial cells (Fig. 1a). Single populations of E. coli cells were sequestered at high concentrations into the interior of the membrane-free coacervate droplets within a few minutes to produce viable microcompartmentalized colonies ( Fig. 1b and Movie S1). In contrast, PAO1 cells were almost instantaneously adsorbed onto the droplet surface to produce a shell of living bacteria ( Fig. 1c and Movie S2).
Importantly, addition of a mixture of E. coli and PAO1 cells gave rise within a few minutes to single coacervate droplets comprising an internally segregated E. coli population surrounded by a thin continuous shell of PAO1 cells (Fig 1d and Movie S3). Quantitative FACS analysis of mixed suspensions of coacervate droplets and E. coli or PAO1 cells con rmed that the bacteria were captured at relatively high e ciency (typically 40-50%) after 5 min of incubation (Figs 1e,f). Similar experiments on droplets containing captured E. coli and PAO1 cells with expressed eGFP or mCherry uorescence, respectively, indicated that the co-loading e ciency was ca. 15 % (Fig. 1g). Corresponding time-dependent uorescence measurements on single coacervate droplets showed logistic growth rates in eGFP and mCherry uorescence intensities up to threshold values as the droplet interior became lled with bacteria and assembly of the bacterial shell was completed (Fig. 1h). Staining the captured bacteria indicated that the majority of the cells associated with the coacervate droplets were alive (Fig. S1).
Having assembled the living pre-former droplets, we employed on-site processing to recon gure and re-purpose the spatially segregated bacterial colonies into membrane-bounded, molecularly crowded protocells with complex microstructures and diverse bio-derived functionalities (Fig. 2a). To achieve this, we implemented the stepwise in situ lysis of the co-captured bacteria populations using a cell wall hydrolase (lysozyme) and antimicrobial peptide (melittin) 25 to release an extensive suite of bacterial membrane lipids/proteins and cytoplasmic components into the adjacent coacervate milieu (Figs. S2, S3 and S4). While short lysis times generated coacervate droplets containing a living E. coli colony bounded by a 300-500 nm-thick shell of PAO1-derived membrane constituents, extending the processing time so that lysis of both types of bacteria occurred gave rise to structurally and compositionally complex protocells that consisted of a spatially segregated membrane of PAO1 lipids, which enclosed a coacervate matrix consisting principally of incarcerated E. coli cytoplasmic components along with PA01membranized water vacuoles (Figs. 2b-e and Figs. S5-S7). Although proteins and polynucleotides did not readily leach from the protocells due to their strong a nity with the coacervate matrix (Figs. 2c,d), the PAO1-derived membrane was permeable to small molecule uorescent dyes as well as macromolecular solutes, which could be readily taken up from the external solution (Fig. S8). The high membrane permeability was consistent with scanning electron microscopy images of freeze-dried bacteriogenic protocells, which showed a disordered arrangement of cytomembrane fragments on the surface of the droplets (Fig. S9). FACS quantitative analysis of samples prepared from co-captured E. coli and PAO1 cells with expressed eGFP and mCherry uorescence, respectively, indicated that the percentage of bacteriogenic protocells produced after lysis was ca. 30% ( Fig. S10) Quantitative component analyses indicated that the bacteriogenic synthetic cells inherited protein, lipid, RNA and DNA contents of 16, 3, 3 and 1 wt%, respectively (Fig. 2f). Full protein typing was achieved by LC-MS proteomics, which indicated that approximately 78% (1359 proteins) of the combined library of bacterial proteins were retained by the protocells (Fig. 2g). Most (77%) of the retained proteins from the two populations were derived from the E.coli cells initially housed within the interior of the droplets. Of these, 92% of the E. coli protein types were associated with the protocells compared with only 49% of those in the PA01 proteome (Fig. 2g). In terms of molecular function, proteins with catalytic (43%), binding (32%) and structural (14%) roles were highly represented, whilst anatomical (48%) and intracellular entities (40%) along with protein-containing complexes (12%) were dominant as cellular components. These proteins were primarily associated with cellular (44%) and metabolic processes (40%), with the remaining proteins assigned principally to biological regulation, localization and sensing (Fig. 2h).

Bacteriogenic protocells as cytomimetic constructs
Given the extensive repertoire of biological components embodied within the bacteriogenic protocells, we sought to exploit the compositional complexity for the fabrication of synthetic cell modules with diverse cytomimetic properties operating at the level of individual enzymes, proto-metabolic pathways (glycolysis) and information networks (in vitro gene expression).
Retention of functional enzymes within the ensemble of released components was demonstrated by monitoring the activity of alkaline phosphatase, protease, lipase and β-galactosidase (β-gal) within the bacteriogenic protocells after exposure to appropriate uorogenic small-molecule substrates in the external environment. In each case, production of the uorescent outputs occurred homogeneously throughout the interior of the bacterially derived protocells and the products retained predominantly within the coacervate matrix (Figs 3a-f, Movie S4 and Figs. S11-13). We also tested whether the bacteriogenic protocells inherited a su cient complement of coordinated enzymes to implement the tenenzyme cascade required for glycolysis. 26 For this, we added glucose-containing Luria-Bertani broth to protocells containing lactate dehydrogenase and assayed the onset of proto-metabolic activity via conversion of pyruvate to L-lactate along with formation of the reduced purple dye (formazan). Timedependent spectroscopic measurements showed signi cant glycolysis activity in the protocell population ( Fig. 3g,h), along with a blue-purple colouration and formation of insoluble formazan within their protocell interior (Fig. 3i,j and Fig. S14), indicating that su cient levels of bacterially derived cell respiration components were released by lysis and retained within the cytoplasmic-like coacervate matrix.
As bacterial lysates can be used for in vitro translation/transcription, 27 we loaded selected components of an expression system (buffer, T7 promoters and polymerases, amino acids, nucleotides and t-RNA) and a plasmid (pEXP5-NT/deGFP) for deGFP expression into PDDA/ATP coacervate droplets.
Notably, the bacterial lysate accompanying the cell-free expression kit was not included. We then used the primed droplets to capture E. coli and PAO1 cells followed by in situ lysis, bacteriogenic protocell construction and storage on ice. As a consequence, subsequent endogenous processing of the incorporated plasmid by the protocells was dependent on inheriting a su cient complement of gene expression machinery (peptidyl transferase, RNA synthetases, energy generation, translation factors (IF1, IF2, IF3), ribosomal proteins etc) from in situ lysis of the bacterial cells. Gene expression was initiated by raising the temperature to 37 °C with additional feed buffer added after 30 min to sustain the system. Production of deGFP occurred speci cally within the bacteriogenic protocells over 5 hours to give a protein yield of 0.01 μM with maximum and minimum expression rates at 0.5 and 3 h, respectively ( Fig.  3k-m). In contrast, negligible green uorescence was observed for synthetic cells lacking the pEXP5-NT/deGFP plasmid (Fig. 3m). Gel electrophoresis of the extracted proteins showed a band at 35 kDa that was absent without the plasmid (Fig. 3n), and Western blotting images indicated that the 35 kDa band was stained on exposure to an anti-GFP antibody (Fig. 3o), con rming deGFP expression within the bacteriogenic protocells.
On-site augmentation of bacteriogenic protocells Bacterially derived synthetic cells containing both a spatially partitioned nucleus-like organelle and 3D structural network of proto-cytoskeletal laments were prepared by endogenous remodelling of the cytoplasmic-like interior (Fig. 4a). Spontaneous aqueous two-phase liquid-liquid separation of a single bacterial DNA/histone condensate within the bacteriogenic protocells was achieved within the cytoplasmic-like interior by uptake of carboxymethyldextran (CM-dextran; 70 kDa). The procedure, which was dependent on the presence of linear double-stranded (ds) DNA and initially developed using native PDDA/ATP droplets (Figs. S15-17), involved loading the coacervate droplets with a histone protein (type II-A) and a mixture of DNAase I and MnCl 2 -the latter used to cleave the native bacterial plasmid DNA into linear fragments. This was followed by capture of a mixed population of E. coli/PA01 cells and release of the bacterial plasmid DNA (and other cytoplasmic components) by lysis to generate a homogeneous distribution of a linear DNA/histone complex within the protocell interior. Subsequent addition of CM-dextran and diffusion of the polysaccharide through the outer PA01-derived membrane resulted in the nucleation of a single membrane-free DNA/histone condensate inside the individual protocells within a few tens of seconds (Figs. 4b,c). Essentially all the released bacterial DNA (99%) was con ned to the nucleus-like condensate while most (75%) of the released bacterial proteins were sequestered into the molecularly crowded cytoplasmic-like region of the protocells (Figs. 4d,e). Similar experiments involving bacterially derived protocells without pre-loaded DNAase I and MnCl 2 resulted in the nucleation of large numbers of small (0.3-2 μm) plasmid DNA/histone condensates throughout the CM-dextran-enriched phase ( Fig. 4e and Fig. S18). When exposed to hypotonic conditions, the remodelled protocells were further elaborated with membrane-coated water vacuoles to produce synthetic cells that exhibited multiple types of spatially segregated cytomimetic structures (Figs 4f,g).
Having established a protocol for the in situ condensation and spatial localization of bacterial DNA within the protocells, we sought to re-structure the molecularly crowded proto-cytoplasmic region by onsite energy-driven supramolecular assembly of a F-actin cytoskeletal-like lamentous network to produce self-supported cytomimetic models. Initial experiments using PDDA/ATP pre-former coacervate droplets indicated that G-actin and Mg 2+ ions were readily taken up by the bacteriogenic protocells and that in situ F-actin assembly was activated endogenously by the coacervate-derived ATP (Fig. S19). However, as only short laments of F-actin were produced due to the high native ATP concentration (ca. 200 mM), 28 we prepared the sub-compartmentalized bacteriogenic protocells using PDDA/uridine-5'-triphosphate (UTP) coacervate droplets (Fig. S20) and pre-loaded the droplets with an enzyme-based ATP-generating system based on pyruvate kinase (PK), phosphoenolpyruvate (PEP) and ADP (Fig. 4i). 29 Subsequent uptake of G-actin and Mg 2+ initiated F-actin assembly speci cally within the synthetic cells and transformed the bacteriogenic constructs into hydrogelled microstructures with retention of the spherical morphology (Fig. 4j). The results indicated that the on-board generation of ATP arising from the encapsulated PK/PEP/ADP cascade over a period of approximately 20 min was su cient to produce the controlled restructuring of the protocell interior. This was con rmed by confocal uorescence microscopy images of single protocells, which showed the presence of a localized low-density network of F-actin micro-laments speci cally dispersed within the proto-cytoplasmic region but not in the phase-separated DNA/histone sub-compartment (Figs. 4k-m and Fig. S21).
Live cell energization and recon guration of bacteriogenic protocells Given the potential for compositional, structural and functional complexity in the bacteriogenic protocells, we sought to couple these attributes to an internal ATP-based energy source as a step to increasing the autonomy and self-su ciency of the life-like constructs. To achieve this, we implanted E. coli cells into PDDA/UTP-based bacteriogenic protocells (Fig. 5a), and used the incarcerated living cells as surrogate mitochondria for the endogenous production of ATP. Confocal uorescence microscopy images indicated that the E. coli cells were sequestered speci cally into the cytoplasmic-like space; in contrast, no cells were observed in the DNA/histone-enriched proto-nucleus (Fig. 5b). Typically, 10-50 bacteria were initially captured within the cellular bionic system. Live/dead staining of the constructs indicated that many of the implanted E. coli cells remained viable after entrapment for 3 h (Fig. S22). Corresponding time-dependent FACS data indicated that the population of E. coli cells increased approximately 10 times over a period of 48 h with the percentage of live cells decreasing from 99 (1 h) to 60 % (48 h) (Fig. 5c and d and Fig. S23). Growth of the trapped bacterial colony produced a three-fold increase in the total protein concentration within the protocells over a period of 48 h (Fig. 5e).
To energize the living/synthetic cell construct, we initiated ATP production in the incarcerated E.
coli cells by addition of glucose (in Luria-Bertani broth) and exploited the extracellular secretion of the energy-rich mononucleotide 30 to chemically charge the co-located bacteriogenic protocells (Fig. S24).
Consequently, continuous biologically mediated ATP production in the PDDA/UTP-based bacteriogenic protocells was sustained for up to 36 h compared with 20 minutes in the presence of a sequestered PK/PEP/ADP enzyme-based pathway (Fig. 5f). Bioproduction of ATP was tted to an exponential decay curve that attained equilibrium after ca. 36 h with a maximum ATP concentration of 1.65 mM under the conditions typically employed (Fig. S25). Importantly, ATP bio-generation within the hybrid protocells was readily sustained by replenishing the nutrients in the external medium, whilst addition of new substrates (ADP, PEP) to the enzyme-based pathway disrupted the structural integrity of the protocells.
Bio-generation and extracellular secretion of ATP resulted in prolonged enzyme activity and gene expression, as well as increased levels of F-actin polymerization in the living/synthetic constructs. For example, kinase activity was approximately doubled when compared to PDDA/UTP bacteriogenic protocells containing released bacterial lysate (no live E. coli cells) and increased 1.3-fold in comparison with the implanted PK/PEP/ADP pathway (no live E. coli cells) (Fig. 5g and Fig. S26). In vitro gene expression of deGFP was extended from 3 to 24 h with a nal deGFP concentration of ca. 0.015 μM (Fig.  5h,I and Fig. S27) and pyruvate production was increased 1.7-fold due to sustained glycolysis in the implanted bacteria (Fig. S28). On-site assembly of F-actin after passive uptake of G-actin and Mg 2+ ions gave rise to an extensive network of protein laments that lled most of the proto-cytoplasmic space within 30 min (Figs. 5j,k and Fig. S29). Compared with supramolecular polymerization driven by the loaded PK/PEP/ADP cascade, living cell ATP production gave rise to a considerably higher density matrix of F-actin laments that unlike the enzyme-supplemented pathway lled the proto-cytoplasmic space after 0.5 h (Fig. S30). As a consequence, living bacteria, a single DNA/histone sub-compartment and water vacuoles were immobilized in the cytoskeletal-like framework to produce a self-supporting cellular bionic systems that remained structurally intact in water and salt solutions for at least 7 days (Fig. 5l and  Fig. S31). In contrast, living/synthetic constructs prepared without F-actin were only stable for up to 3 days in buffer and disassembled immediately under hypertonic conditions (Fig. S31).
Finally, we sought to exploit the continuous bio-generation of metabolic products within the bacteriogenic protocells to implement changes in membrane interfacial tension as a step towards producing living/synthetic cell hybrids with non-spherical morphology. In the absence of incarcerated E. coli cultures, mounting the F-actin containing spherical bacteriogenic protocells onto pegylated glass substrates and leaving the samples at room temperature in Luria-Bertani (LB) broth for 48 h resulted in minimal changes in structure and morphology (Fig. S32). In contrast, similar experiments with bacteriogenic protocells containing encapsulated live E. coli cells and a F-actin network gave rise to progressive changes in morphology (Fig. 6a). Within 48 h, most of the spherical microstructures adopted a more irregular cell-like form whilst maintaining their complex internal organization and outer membrane structure (Fig. 6b). During this period, the average number of E. coli cells per protocell increased approximately eight-fold within the rst 24 h, after which the entrapped population of live cells decreased by approximately 35% after 48 h (Fig. 6c). Growth of the bacteria was associated with a two-fold increase in the average protocell volume over 48 h (Fig. 6d).
Given that the amoeba-like morphology was also observed after 24-48 h for living cell/protocell constructs prepared without a F-actin network (Fig. S32), we ruled out the possibility that the changes in form were associated with anisotropic mechanical forces induced by formation of the cytoskeletal-like matrix. In addition, no morphological changes were observed after short periods (1 h) when the bacteriogenic protocells were loaded with extremely high loadings of freshly washed E.coli cells (Fig.  S32), suggesting that prolonged metabolic activity of the incarcerated bacterial cells, rather than an increase per se in the population density within the protocells, was responsible for the transition to the amoeba-like form. Although, time-dependent increases in bulk wettability were observed for single droplets of concentrated suspensions of the living/synthetic hybrids mounted on pegylated glass slides (Fig. S33), mounting the same constructs onto hydrophobic plastic or non-modi ed glass surfaces also showed amoeba-like morphologies after 48 h incubation, indicating that speci c interactions with the substrate were not critical in determining the shape change (Fig. S34). Indeed, low resolution confocal microscopy images of the living/synthetic constructs recorded in aqueous suspension also showed the non-spherical morphology (Fig. S35), in agreement with an endogenous mechanism associated with prolonged activity of the entrapped E. coli cells. Interestingly, the permeability of the outer membrane to macromolecular diffusion decreased markedly with change of shape such that dextran above a molecular weight of 10 kDa, BSA-FITC (66 kDa) and single-stranded DNA (8 kDa) were no longer taken up into living/synthetic hybrids that were biologically active for more than 20h (Figs. 6e-h and Fig. S36). Based on these observations, we speculate that the continuous production of bacterial lipids facilitated partially sealing of defects in the outer membrane, as well as increasing membrane surface area and elasticity.

Discussion
An ensemble of cytomimetic features is integrated into a synthetic cell construction pathway by using a living material assembly process in which prokaryotes are employed as on-site repositories of key structural and functional building units. Depending on the extent of the construction process, a total of seven distinct cytomimetic attributes can be integrated into the bacteriogenic protocells (Figs. 6i,j).
Implementation of the basic processing strategy gives rise to complex synthetic cells with a bacterially derived semi-permeable outer membrane and internal molecularly crowded proto-cytoplasm containing a PDDA/ATP (UTP) coacervate, an extensive suite of bacterial proteins and polynucleotides, a functional metabolic network and intact translation/transcription machinery. Further elaboration of the synthetic cells can be achieved endogenously by structural augmentation with a single membrane-free DNA/histone nucleus-like condensate; osmotically responsive membrane-bounded water vacuoles and a primitive F-actin cytoskeletal-like network; in each case, the additional design features are implemented without compromising the multiple cytomimetic functions. Higher-level operations are introduced by using implanted live E. coli cells as surrogate mitochondria for energization, which prolongs ATP bioproduction and transforms the spherical bacteriogenic protocells into a cellular bionic system with amoeba-like cell morphology.
A critical aspect of the living material construction sequence is the spontaneous coacervate dropletmediated capture and spatial segregation of E. coli and PA01 cells that together enable on-site processing and retention of diverse bacterial components for synthetic cell elaboration. The methodology opens the possibility of high levels of programmability by using engineered bacteria designed to deliver specialized components and biological processes for establishing robust metabolic networks and genetic circuitry in the bacteriogenic protocells. Additionally, the spatial arrangement of live E. coli and PA01 cells in the initial pre-former coacervate droplets could be modulated by changes in the coacervate composition (Fig. S37), suggesting that the construction sequence could be adapted in the future to fabricate complex synthetic cells with alternative types of bacteriogenic organization.
Finally, we expect our new living material assembly approach to provide opportunities for the bottom-up construction of highly integrated synthetic cells and augmented living/synthetic cell constructs. For example, the possibility of spatially coupling F-actin polymerization to the membrane of the bacteriogenic protocells or embedded vacuoles could lead to life-like micro-devices with mechanically adaptive properties, while information processing derived from the DNA/histone proto-nuclei could be a rst step to a rudimentary "proto-eukaryote" cell. From a cellular bionics perspective, the potential for symbiosis in living/synthetic cell hybrids constructed from bacterially derived construction pathways could offer more complex modules for development in diagnostic and therapeutic areas of synthetic biology as well as in biomanufacturing and biotechnology in general.