De novo engineering of programmable and multi-functional biomolecular condensates for controlled biosynthesis

Abstract

There is a growing interest in the creation of engineered condensates formed via liquid-liquid phase separation (LLPS) to exert precise cellular control in prokaryotes. However, de novo design of cellular condensates to control metabolic flux or protein translation remains a challenge. Here, we present a synthetic condensate platform, generated through the incorporation of artificial, disordered proteins to realize specific functions in Bacillus subtilis. To achieve this, the "stacking blocks" strategy is developed to rationally design a series of LLPS-promoting proteins for programming condensates. Through the targeted recruitment of biomolecules, our investigation demonstrates that cellular condensates effectively sequester biosynthetic pathways. We successfully harness this capability to enhance the biosynthesis of 2'-fucosyllactose by 123.3%. Furthermore, we find that condensates can enhance the translation specificity of tailored enzyme fourfold, and can increase N-acetylmannosamine titer by 75.0%. Collectively, these results lay the foundation for the design of engineered condensates endowed with multifunctional capacities.

Fig. 1. Construction and characterization of synthetic condensate formed by SIDP1-GFP in vitro.

a Expected interactions promoting LLPS of 8-mer peptide FW1 into synthetic condensate, and the representative interactions are presented on the line below. b The SIDP1-GFP and SIDP1 protein at protein concentration of 50 μM formed condensates in vitro (150 mM NaCl, pH = 7.4, 25 ^oC, without PEG2000). GFP was used as a negative control. c Light microscopy images showing two individual condensates tend to fuse together. d Fluorescence recovery of a partially photobleached droplet (n = 3). The inset shows representative images of photobleaching on the droplet. e A regime diagram illustrates the formation of SIDP1-GFP condensates at different protein and salt concentrations. Scale bars = 10 μm. Data are presented as mean ± s.d. of three biologically independent replicates. Source data provided as a Source Data file.

Fig. 2. Liquid-like properties of cellular synthetic condensate.

a The fluorescence microscopy images of B. subtilis overexpression SIDP1-GFP fusion protein. All cells were cultivated at 37 °C and induced with 0.5 mM isopropyl-β-d-thiogalactoside (IPTG) for gene expression. b Characterization of the synthetic cellular condensate by urea perturbation after induction for 2 h. c Measurement of the ratio of synthetic condensates in B. subtilis under different SIDP1-GFP expression level and temperature (n = 3). Scale bars = 5 μm. Data are presented as mean ± s.d. of three biologically independent replicates. Source data provided as a Source Data file.

Fig. 3. Engineered intracellular condensates with intended function.

a Measuring recruitment efficiency of target fluorescent protein within condensates upon co-expression of RIAD-SIDP1-GFP and mKate-RIDD (n = 6 biologically independent cells). SIDP1-GFP protein without RIAD peptide was used as a negative control. b Measuring recruitment efficiency of mKate-RIDD (driven by xylose operon) within RIAD-SIDP1-GFP formed condensates (driven by IPTG). The first three columns indicate the dual-channel imaging and merging images of RIAD-SIDP1-GFP and mKate-RIDD. The fourth column represents the ratio of mKate fluorescence intensity of condensates to that of the whole cell (n = 6 biologically independent cells). The box plots in (a) and (b) show the mean value as a black cross, the center as a black line, box extending between the 25th and 75th percentiles, and whiskers indicating minimum and maximum data points. c Validation of the potential function of cellular condensates for controlling metabolic flux. The leftmost column represents the metabolic pathways of deoxyviolacein (DV) and prodeoxyviolacein (PDV). Cellular condensate programs flexibly redirect the metabolic flux of the DV/PDV biosynthesis pathway, resulting in a significant change in cell color. 1.5 mL culture were harvested by centrifugation. The “+” represents the gene being expressed in the strain, while the “-” indicates non-expression. Additionally, the “O” signifies that the protein is colocalized within the cellular condensate. Scale bars = 5 μm. Source data provided as a Source Data file.

Fig. 4. Functionalization of cellular condensates for spatially separated orthogonal translation.

a Spatial separation of the necessary components to enable orthogonal translation to control language of protein translation. Without the synthetic condensate in vivo, the selected gene GFP and other gene such as mKate containing TAG both can be transcribed and translated into a complete fluorescent protein normally. When the genetic code expansion (GCE) machinery was applied within a synthetic condensate, the selected gene GFP containing TAG that targeted to the condensate can still express normally. In contrast, the corresponding stop codon in gene mKate that are not targeted to the condensate should not get translated. TyrRS, aminoacyl-tRNA synthetase; ncAA, noncanonical amino acid; SIDP, synthetic intrinsically disordered protein. b For all experiments, the indicated constructs and dual reporter GFP^3TAG::6RTBA and mKate^3TAG were co-expressed. The orange bars (normalized to cytoplasmic TyrRS) represent the fold change in the ratios of the mean fluorescence intensities of GFP versus mKate for all the systems tested. The gray bars represent the relative efficiency as defined by the mean fluorescence intensity of GFP for each condition divided by cytoplasmic TyrRS control. The J-2, JD-2, and JDM-2 systems were constructed by using one RTBA based on J, JD, and JDM1 systems, respectively. Data are presented as mean ± s.d. of three biologically independent replicates. Source data provided as a Source Data file.

Fig. 5. Controlling 2′-fucosyllactose biosynthesis by co-localization of targeted cargo enzymes.

a Illustration of the synthetic condensate platform of de novo biosynthesis of 2′-fucosyllactose (2′-FL) in B. subtilis. futC, α1,2-fucosyltransferase; manB, phosphomannomutase; manC, α-D-mannose 1-phosphate guanylyltransferase; gmd, GDP-mannose 6-dehydrogenase; wcaG, GDP-L-fucose synthase. b High-performance liquid chromatography (HPLC) quantification of 2′-FL produced by different recombinant strain with SIDP1 condensate co-localizing diverse pathway enzymes. The “O” signifies that the enzyme is colocalized within the cellular condensate, while “X” represents enzymes that are not recruited. c Fluorescence microscopy images of SIDP2-GFP, SIDP4-GFP, FUSN-GFP, and 3RGG-GFP driven condensates within B. subtilis. d Effects of condensates with virous physical properties on 2′-FL production. Additionally, trends of 2′-FL concentration of the all recombinant strains are shown. Data are presented as mean ± s.d. of three biologically independent replicates. Significance (P value) was evaluated by two-sided t-test. *: p < 0.05, **: p < 0.01. Scale bars = 5 μm. Source data provided as a Source Data file.

Fig. 6. Improving translation specificity of AGE^OMeY involving N-acetylmannosamine biosynthesis.

a Shake flask fermentation of ManNAc by strains MA1 and MA2. The gene GNA1 and age were integrated into the genome of the engineered strain S5, generating strain MA1. Besides, the original aaRS/tRNA_OMeY system was expressed in MA1, yielding the strain MA2. GlcN6P glucosamine-6-phosphate, GlcNAc N-acetylglucosamine, GlcNAc6P GlcNAc-6-phosphate, ManNAc N-acetylmannosamine, GNA1 GlcN6P N-acetyltransferase, AGE N-acylglucosamine 2-epimerase, yqab, hydrolase-like phosphatase, nagP phosphotransferase system GlcNAc-specific transporter subunit EIICB, OMeY O-methyl-L-tyrosine. b The Rosetta-predicted combined mutant AGE^OMeY with GlcNAc bound is shown. The blue dotted line means hydrogen-bonding, and the yellow dotted line means pi interactions. c Illustration of synthetic functional condensates to increase translation specificity of enzyme AGE^OMeY for improving the ManNAc biosynthesis. d Trends of cell growth, glucose, GlcNAc, and ManNAc concentration of the strains MAO1 and MAO2. Data are presented as mean ± s.d. of three biologically independent replicates. Source data provided as a Source Data file.