Surface Display For Phage Assisted Continuous Evolution: A Platform For Evolving / Screening Nanobodies In Prokaryote Systems

This paper introduces SurPhACE, a scalable synthetic biology platform that combines *E. coli* surface display with {Phi}X174 phage-assisted continuous evolution to rapidly generate and screen high-affinity nanobodies in prokaryotic systems without the need for traditional library construction or animal inoculation.

The Gist

Imagine you are trying to find a specific key that fits a very tricky lock. In the world of medicine and biotechnology, these "keys" are special proteins called nanobodies (tiny, powerful antibodies) that can grab onto disease-causing viruses or cancer cells and stop them.

Usually, finding these keys is like trying to find a needle in a haystack, but the haystack is the size of a mountain, and you have to do it by hand, one needle at a time. It takes months, requires expensive animals (like camels or mice), and is very slow.

This paper introduces a new, high-speed machine called SurPhACE (Surface Display for Phage Assisted Continuous Evolution) that automates this process. It's like turning a slow, manual search into a high-speed, self-driving factory that churns out the perfect keys in days instead of months.

Here is how it works, broken down with simple analogies:

1. The Problem: The Old Way is Slow

Traditionally, scientists make these keys by:

• Injecting an animal with the "lock" (the disease target).
• Waiting for the animal's immune system to make a few keys.
• Harvesting those keys and testing them one by one.
• If none are perfect, they try to tweak them manually.

It's like trying to find the perfect key by asking a thousand people to guess what it looks like, then carving one by hand based on their guesses.

2. The Solution: The SurPhACE Factory

The authors built a "biological factory" inside a tiny petri dish. Here are the three main parts of this factory:

A. The Workers: The Bacteria (The Factory Floor)

Instead of using animals, they use E. coli bacteria. They engineer these bacteria to wear a specific "target" on their outer skin. Think of these bacteria as mannequins dressed in the "lock" (the disease target).

• The Good Mannequins (Helper Bacteria): These wear the target and have a special "battery" (a helper gene) that keeps the factory running.
• The Bad Mannequins (Decoy Bacteria): These wear a fake target or no target at all. They act as decoys to trick the workers.

B. The Workers: The Viruses (The Key-Makers)

They use a tiny virus called ΦX174. Usually, viruses are bad, but here they are the heroes.

• They attach a "key" (the nanobody) to the virus's surface.
• Normally, this virus is broken; it can't reproduce on its own. It needs the "battery" from the Good Mannequins to survive.
• The Catch: The virus can only reproduce if its key perfectly fits the lock on the Good Mannequin. If it grabs a Decoy Mannequin or nothing at all, it dies.

C. The Engine: Continuous Evolution (The Assembly Line)

This is the magic part. The bacteria and viruses are in a tank that is constantly being drained and refilled with fresh food and new bacteria.

• The Filter: Every 30 minutes, the "weak" viruses (those with bad keys) are washed away. Only the viruses with the best keys stay behind to reproduce.
• The Mutation: Inside the bacteria, there is a "glitch generator" (a mutagenic plasmid) that randomly changes the virus's DNA. This is like a randomizer button that slightly alters the shape of the key every time it is copied.
• The Result: In one day, this system can test over 5 trillion different key variations. It's a "survival of the fittest" race where the best keys get stronger and stronger, minute by minute.

3. The Experiment: Did it Work?

The scientists tested this machine in two ways:

• Test 1: Tuning an Existing Key. They took a key that was almost good but not perfect. They put it in the SurPhACE machine. Within days, the machine evolved a version of the key that fit the lock much tighter.
• Test 2: Finding a Key from Scratch. They started with a library of random keys (a "junk drawer" of shapes) and asked the machine to find a key for a completely new lock.
  • The Challenge: The machine struggled a bit at first because the "junk" keys were too weak to survive the washing process.
  • The Fix: They added a "starter library" of slightly better keys to help the process get going.
  • The Outcome: The machine successfully found new keys that fit the new locks. However, they noticed that sometimes the viruses got "lazy" and stopped making the full key, just making a tiny, useless piece that still survived. This is a lesson for the future: the machine needs to be tuned to prevent these "cheaters."

4. Why This Matters

• Speed: What used to take months now takes days.
• Cost: It uses a tiny amount of liquid (less than a cup) and standard lab equipment, no expensive animals needed.
• Versatility: It can be used to find keys for almost any lock (virus, cancer cell, toxin).

The Big Picture Analogy

Imagine you are trying to find the perfect pair of shoes for a marathon.

• The Old Way: You buy 10 pairs, try them on, run a mile, throw them away, buy 10 new ones, and repeat for a year.
• The SurPhACE Way: You put 1,000,000 pairs of shoes on a treadmill. Every 30 seconds, the treadmill speeds up. If the shoes slip, they are thrown off the belt. The shoes that stay on get "mutated" (a tiny bit of rubber is added or removed). Within an hour, the shoes on the belt have evolved into the perfect, custom-fit marathon shoe, and you didn't have to try a single pair on your own foot.

Conclusion

This paper proves that we can build a self-sustaining, automated evolution machine in a test tube. While there are still some kinks to work out (like stopping the "lazy" viruses), SurPhACE is a massive step forward in making life-saving drugs faster, cheaper, and more accessible. It turns the slow, manual art of drug discovery into a high-speed, automated science.

Technical Summary

1. Problem Statement

The generation of high-affinity protein binders (such as nanobodies, scFvs, and DARPins) is critical for biotechnology and therapeutics. Current methods face significant bottlenecks:

• Animal Immunization: Traditional methods require animal facilities, take months, and involve ethical concerns.
• In Vitro Screening (Phage/Yeast Display): While faster, these methods rely on pre-built libraries with limited diversity ($>10^9$ variants) and require labor-intensive, iterative "biopanning" rounds.
• Directed Evolution (PACE): Phage-Assisted Continuous Evolution (PACE) allows for massive, unbiased mutational libraries ($>10^{12}$ variants/day) but typically relies on M13 phage. M13 infects via bacterial flagella, making it difficult to couple with bacterial surface display for simultaneous positive/negative selection on the cell surface.
• AI Limitations: While AI tools (e.g., AlphaFold) are advancing, they still struggle with modeling specific binding motifs (like CDR3 regions) and lack the ability to experimentally validate and evolve binders in a wet-lab setting without initial structural data.

The authors aim to create a scalable, benchtop platform that combines the massive mutational space of PACE with the specificity of bacterial surface display to evolve binders without animal immunization or pre-built libraries.

2. Methodology: SurPhACE

The authors developed SurPhACE (Surface Display for Phage Assisted Continuous Evolution), a system integrating three core components:

A. Engineered Phage (ΦX174)

• Selection of Vector: Instead of M13, the team used ΦX174 (a microvirus), which infects E. coli by binding directly to the cell surface (LPS) via its spike protein (Protein G).
• Genetic Engineering:
  • A nanobody (VHH) was inserted into the G gene (spike protein) at residue C312.
  • The H gene (essential for DNA injection) was deleted from the phage genome, rendering the phage replication-defective.
  • To maintain genome size, the H gene was replaced with random "filler" DNA.
  • Result: The phage can only replicate if it infects a host bacterium that supplies the missing H gene and if the displayed nanobody binds to a specific target on that host's surface.

B. Bacterial Surface Display (Helper & Decoy Strains)

• Helper Bacteria: Engineered E. coli containing a plasmid that expresses:
  1. The H gene (complementing the defective phage).
  2. The Target Protein displayed on the outer membrane via a BAN tag (derived from Bacillus anthracis exosporium).
  3. Note: The authors found the BAN tag superior to OmpA or OmpX for surface display stability.
• Decoy Bacteria: Engineered E. coli containing a plasmid with the BAN tag but no target protein.
  • Function: Provides negative selection. Phages binding non-specifically to the bacterial surface (off-target) will infect decoys but cannot replicate (no H gene), leading to their elimination.

C. Continuous Evolution Setup

• Mutagenesis: The system utilizes the MP6 plasmid, which induces a high mutation rate in the host genome, generating diversity in the phage population continuously.
• Selection Pressure: Cultures are maintained in a semi-continuous flow (chemostat-like) where fresh helper bacteria and media are added, and old culture is removed.
  • Positive Selection: Only phages with high-affinity binders to the target on Helper bacteria can infect, receive H protein, and replicate.
  • Negative Selection: Phages binding to Decoy bacteria are washed out or fail to replicate.
• Library Strategy: The authors tested two approaches:
  1. Optimization: Starting with a single nanobody sequence and evolving it for higher affinity.
  2. De Novo Discovery: Starting with a synthetic CDR3 random library ($10^6$ variants) to discover binders for arbitrary targets.

3. Key Contributions

1. Novel Phage Vector: Successful engineering of ΦX174 for surface display, overcoming the limitations of M13 regarding cell-surface interaction.
2. Dual-Selection Platform: A robust system enabling simultaneous positive selection (target binding) and negative selection (decoy binding) within a continuous evolution loop.
3. BAN Tag Validation: Identification of the BAN tag as the most effective surface display anchor for this application compared to native E. coli tags.
4. Mathematical Modeling: Development of an ODE-based model to predict population dynamics, highlighting the importance of dilution rates and the risk of "clonal interference" in high-density cultures.
5. Proof of Concept: Demonstrated the evolution of nanobodies for both optimizing affinity (FcγRIIA) and discovering binders for novel targets (IL20RA, PAEP).

4. Results

• System Initialization:
  • The replication-defective phage (EΦ1) failed to propagate in decoy-only cultures (confirming no non-specific replication).
  • Propagation was successful in Helper cultures but required the MP6 mutagenesis plasmid and optimized dilution rates to sustain the population, indicating that initial affinity was insufficient for immediate selective sweeps.
• Optimization (FcγRIIA):
  • Starting with a nanobody against FcγRIIB, the system evolved variants against the homologous FcγRIIA.
  • NGS Analysis: After 168 hours, CDR3 mutations were rare (<1% of the population), suggesting that the initial affinity was already sufficient for survival, preventing a strong selective sweep. However, AlphaFold modeling predicted two specific mutants (FC_2.1 and FC_6.8) with significantly improved binding scores (ipTM > 0.79) compared to the parent.
• De Novo Discovery (IL20RA & PAEP):
  • Using a synthetic CDR3 library, the system successfully enriched specific variants for both targets over 168 hours.
  • Convergence: Hierarchical clustering revealed that enriched variants converged on specific amino acid motifs (e.g., 6 identical amino acids for IL20RA), confirming selection rather than random drift.
  • Challenges: The system was prone to clonal interference (multiple beneficial mutations competing) and the enrichment of loss-of-function indel sub-variants (truncations) that acted as "parasites," reducing the efficiency of the selective sweep.
• Modeling Insights:
  • The ODE model confirmed that if the initial population has moderate affinity, rare high-affinity mutants cannot easily dominate the population without a "normalization" step or starting with a diverse library where the initial binders are weak.

5. Significance and Future Directions

• Scalability: SurPhACE can process $>5 \times 10^{12}$ unique mutants per day in less than 100 mL of culture, far exceeding the capacity of traditional screening.
• Accessibility: The platform uses standard molecular biology equipment and plasmids available on AddGene, making directed evolution accessible to non-specialist labs.
• Complement to AI: While AI predicts structures, SurPhACE provides the experimental engine to evolve binders for targets where structural data is incomplete or where AI predictions fail (e.g., CDR3 flexibility).
• Future Improvements: The authors identify the need for:
  • Dynamic Selection Pressure: Tapering stringency to allow selective sweeps.
  • Normalization Steps: To enrich rare mutants and overcome clonal interference.
  • Optimized Insertion Sites: To prevent capsid instability and the accumulation of truncation mutations.

In conclusion, SurPhACE represents a significant step toward a fully automated, animal-free, and high-throughput platform for generating therapeutic protein binders, bridging the gap between rational design, AI prediction, and experimental evolution.