Combinatorial optimization of protein systems in synthetic cells

This study demonstrates a combinatorial optimization strategy for synthetic cells by screening large populations of vesicles with varied translation rates across multiple genes, successfully isolating high-performing variants for both DNA self-replication and phospholipid synthesis pathways while revealing the predictive limits of single-mutation data for complex multi-gene systems.

The Gist

Imagine you are trying to build a tiny, artificial factory inside a microscopic bubble (a liposome). This factory's job is to either copy its own blueprints (DNA replication) or manufacture a specific type of oil (phospholipids).

The problem? Just like a real factory, if you hire too many workers for one job and too few for another, the whole system grinds to a halt. In biology, these "workers" are proteins, and the instructions on how fast to hire them are written in the DNA code.

This paper is about a team of scientists who decided to stop guessing how to build these factories. Instead, they built thousands of slightly different versions of the factory at once, let them compete, and picked the winners. Here is how they did it, explained simply:

1. The Challenge: The "Goldilocks" Problem

In a synthetic cell, you need the right amount of every protein.

• Too little? The factory is slow.
• Too much? The factory gets clogged, or the proteins get in each other's way (like too many chefs in a small kitchen).
• Just right? The factory runs perfectly.

Finding the "just right" combination for multiple proteins at the same time is incredibly hard. It's like trying to tune a radio with 100 knobs at once. If you turn one knob, it might fix the static, but it might make the bass too loud.

2. The Solution: The "Massive Lottery"

Instead of testing one combination at a time, the scientists created a combinatorial library. Think of this as a lottery where every ticket is a slightly different version of the factory's instruction manual.

• They changed the "hiring speed" (called RBS strength) for the genes.
• They created over 11,000 different versions of a 4-enzyme oil factory and 156 versions of a DNA-copying machine.
• They put these instructions into millions of tiny bubbles (liposomes).

3. The Selection: "Survival of the Fittest"

Now, how do you find the best factory among millions? You let them compete.

Scenario A: The Self-Replicating Machine
For the DNA-copying factory, they used a clever trick: The factory that works best makes more copies of itself.

• They let the bubbles sit for 16 hours.
• The bubbles with the best instructions copied their DNA thousands of times.
• The bubbles with bad instructions didn't copy much.
• The scientists then took all the DNA out, looked at which instructions were most common, and found the winners. It was like a race where the fastest runners naturally ended up with the most fans.

Scenario B: The Oil Factory
For the phospholipid factory, the bubbles couldn't copy themselves. So, the scientists used a high-tech bouncer (FACS).

• They added a glowing tag that lit up when the factory made oil.
• They ran the bubbles through a machine that acts like a super-fast sorting line.
• The machine zapped the bubbles that glowed the brightest (the best factories) and threw away the dim ones.
• They repeated this "sorting" four times, getting stricter each time, until only the absolute best factories remained.

4. The Discovery: What Made the Winners?

After finding the winners, the scientists read their DNA to see what made them special.

• For the DNA Copier: They found that the "hiring speed" for the two main proteins needed to be strong, but not necessarily the strongest possible. Interestingly, the system was very predictable. If you knew how fast Protein A worked and how fast Protein B worked, you could guess how well they would work together.
• For the Oil Factory: This was messier. The scientists found that one specific enzyme (PlsC) didn't actually matter much. Whether it was hired fast or slow, the factory still worked. However, the other three enzymes were critical.
  • The Surprise: When they changed the instructions for one enzyme, it sometimes changed how fast the other enzymes were made, even though those instructions didn't change. This is called epistasis (or "team chemistry"). It's like how a new manager might accidentally slow down the whole team, even if the team members' skills haven't changed.

5. Why This Matters

This research is a huge step toward building a true artificial cell.

• Old way: Scientists tweak one gene at a time, hoping for the best.
• New way: This paper shows we can test thousands of combinations at once to find the perfect "team chemistry" for complex systems.

The Big Analogy:
Imagine you are trying to write the perfect recipe for a cake.

• Old method: You change the sugar, bake it, taste it. Then you change the flour, bake it, taste it. It takes forever.
• This paper's method: You bake 10,000 cakes at once, each with a slightly different mix of sugar, flour, eggs, and baking powder. You then feed them to a hungry crowd. The crowd eats the best cakes the fastest. You look at the empty plates, realize which recipe was the favorite, and now you know exactly how to bake the perfect cake every time.

The Takeaway

The scientists proved that by using synthetic cells and massive parallel testing, we can engineer complex biological systems much faster than before. They showed that while some systems are predictable, others have hidden "team dynamics" that only reveal themselves when you test them all together. This paves the way for designing artificial life that can think, move, or produce medicine on its own.

Technical Summary

1. Problem Statement

The reconstitution of complex biological systems (e.g., metabolic pathways, genetic circuits) in synthetic cells (liposomes) often requires extensive optimization to achieve functional performance. While in vitro directed evolution has been successfully applied to single genes, optimizing multivariate protein systems remains a significant challenge due to:

• Combinatorial Explosion: The vast sequence space created when mutating multiple genes simultaneously.
• Epistatic Effects: Unpredictable interactions between genes, such as competition for limited translation resources (tRNAs, ribosomes) in cell-free systems, which can lead to non-additive fitness outcomes.
• Lack of Holistic Strategies: Current methods often optimize components in isolation, failing to capture the interplay necessary for system-level functionality.

The authors aimed to develop a strategy to screen large populations of synthetic cells containing combinatorial DNA libraries to optimize multi-protein systems, specifically investigating whether translation rates could be tuned to regulate system functionality and how predictable these combinatorial effects are.

2. Methodology

The study utilized PURE system (Protein synthesis Using Recombinant Elements) encapsulated within liposomes to create synthetic cells. Two distinct genetic modules were engineered and optimized:

A. Library Construction

• Method: The authors used MOSAIC (Multiplex One-Step Single-Stranded DNA Annealing Protein-mediated plasmid diversification), a recombineering technique in E. coli, to generate combinatorial libraries.
• Strategy 1: RBS Variation. Ribosome Binding Sites (RBS) for multiple genes were mutated to vary predicted Translation Initiation Rates (TIRs).
  • System 1 (DNA Replicator): 2 genes (DNAP, TP from Phi29 phage). Library size: 156 variants (12 DNAP × 13 TP RBSs).
  • System 2 (Phospholipid Synthesis): 4 genes (PlsB, PlsC, CdsA, PssA from E. coli Kennedy pathway). Library size: ~11,583 variants (combinatorial RBSs).
• Strategy 2: Codon GC Content Variation. For the phospholipid pathway, the first six codons of each gene were synonymously mutated to vary GC content (affecting mRNA secondary structure and translation rates). Theoretical library size: ~391 million variants.

B. Screening and Selection

• DNA Self-Selection (System 1): Liposomes containing the DNA replicator library were incubated. Successful replication led to DNA amplification. The DNA was recovered, PCR-amplified, and re-encapsulated for subsequent rounds. Fitness was measured by the enrichment of variants via Nanopore long-read sequencing.
• Functional Screening (System 2): Liposomes expressing the phospholipid pathway were screened using Fluorescence-Activated Cell Sorting (FACS). The pathway produced Phosphatidylserine (PS), detected by a fluorescent reporter (LactC2-mCherry). The top 1% of liposomes (highest fluorescence) were sorted, and their DNA recovered for sequencing.
• Validation: High-performing variants were "reverse-engineered" into clonal constructs and validated using:
  • Quantitative PCR (qPCR) for DNA amplification.
  • Confocal microscopy (dsGreen staining for DNA).
  • Flow cytometry (PS production).
  • SDS-PAGE with fluorescent lysine labeling (GreenLys) to quantify protein stoichiometry.

C. Data Analysis

• Long-Read Sequencing: Nanopore sequencing allowed the mapping of full-length combinatorial variants without the need for barcoding or assembly.
• Epistasis Modeling: The authors compared the observed fitness of combinatorial variants against expected fitness calculated using a multiplicative model based on single-mutant data to quantify epistatic interactions.

3. Key Results

A. DNA Replicator Optimization (2-Gene System)

• Selection Outcome: The library successfully enriched variants with strong RBSs for both TP and DNAP.
• Predictability: The fitness of combinatorial variants was extremely predictable ($R^2 = 0.94$). The observed enrichment closely matched the multiplicative model derived from single-gene data, suggesting minimal epistatic interference in this specific 2-gene system.
• Stoichiometry: High-performing variants showed a slight shift in the DNAP:TP ratio (from 2.6 in WT to ~3.0), indicating that precise stoichiometric tuning is possible.
• Limitation: Increasing TIR beyond a certain threshold did not linearly increase protein yield or fitness, likely due to resource saturation or plateauing effects.

B. Phospholipid Synthesis Optimization (4-Gene System)

• Selection Outcome: FACS screening enriched variants with high TIRs for PlsB, CdsA, and PssA. Interestingly, PlsC (the enzyme converting PA to PS) showed a broad range of RBS strengths in top performers, consistent with previous findings that PlsC is not the rate-limiting step in this pathway.
• Predictability: Predictability was significantly lower ($R^2 = 0.38$ for RBS library, $0.64$ for GC library) compared to the DNA replicator. This indicates strong epistatic interactions where the effect of a mutation in one gene depends on the genetic background of the others (e.g., resource competition).
• GC Content Strategy: Mutating the first six codons revealed specific base preferences (e.g., a negative effect of 'G' at the 3rd position of the 2nd codon across all genes). PlsB and PssA mutations had the largest impact on PS synthesis, while PlsC mutations had the least.
• Protein Expression: Changing RBSs often led to non-linear changes in protein expression levels. Downregulating one protein sometimes increased the expression of others, highlighting complex resource allocation dynamics in the PURE system.

4. Key Contributions

1. Scalable Combinatorial Screening: Demonstrated a workflow to screen combinatorial libraries of multi-gene systems (up to 4 genes) in synthetic cells using FACS and DNA self-selection.
2. Long-Read Sequencing Integration: Successfully utilized Nanopore sequencing to resolve full-length combinatorial variants directly from synthetic cells, bypassing the need for complex barcode strategies.
3. Quantification of Epistasis: Provided empirical data showing that while small systems (2 genes) are highly predictable, larger systems (4 genes) exhibit significant epistasis, making "small but smart" combinatorial libraries necessary rather than testing single mutations in isolation.
4. Resource Competition Insights: Revealed that in cell-free systems, translation resources are a limiting factor where increasing the strength of one RBS can inadvertently suppress others, necessitating a holistic optimization approach.

5. Significance and Future Outlook

This work represents a critical step toward the evolutionary engineering of entire artificial cells. By moving beyond single-protein optimization to multiprotein systems, the authors demonstrate that:

• Holistic Optimization is Essential: For complex pathways, the interplay between components cannot be ignored; combinatorial testing is required to find optimal global solutions.
• Synthetic Cell Engineering: The ability to select for functional metabolic outputs (like lipid synthesis) within liposomes paves the way for building self-sustaining synthetic cells.
• Future Directions: The authors suggest integrating active learning (machine learning) to navigate large sequence spaces more efficiently and expanding these methods to include other regulatory elements (promoters, terminators) and larger genetic networks.

In summary, the paper establishes a robust framework for the directed evolution of complex genetic circuits in synthetic cells, highlighting both the predictability of simple systems and the necessity of combinatorial approaches for complex, epistatic networks.