A reassessment of positive growth effects of expressed random sequence clones in E. coli

This study provides robust experimental evidence that a subset of random sequence clones genuinely confers fitness advantages to *E. coli* under specific growth conditions, thereby validating the capacity of non-coding sequences to generate adaptive functions and supporting the mechanism of *de novo* gene evolution.

The Gist

Imagine you are a chef in a massive kitchen, and you decide to test a wild theory: Can you create a delicious new dish just by throwing random ingredients into a pot?

In the world of biology, this is the question of "de novo" gene evolution. Scientists have long wondered if completely random strings of DNA (which usually do nothing) can accidentally turn into useful genes that help an organism survive.

A few years ago, a team of scientists (including the authors of this paper) claimed they found proof: when they put random DNA sequences into bacteria (E. coli), some of the bacteria actually grew faster. They said, "Look! Randomness can create superpowers!"

But, like any good scientific debate, other chefs in the kitchen weren't convinced. They said, "Wait a minute. Maybe the pot itself (the vector) was making the bacteria sick, and the random ingredients just happened to stop the sickness. Or maybe the bacteria were just fighting each other in a weird way, and the 'winners' weren't actually better, just lucky."

This new paper is the team's comeback. They went back to the lab, set up a much stricter, fairer test, and said, "We've checked the pot, we've checked the fighting, and we can prove: Yes, some random DNA really does create a superpower."

Here is how they did it, broken down with some everyday analogies:

1. The "Taste Test" vs. The "Blind Taste Test"

In their first experiment, they had a giant soup pot with thousands of different random DNA "ingredients." They watched which ones survived.

• The Skeptics' Argument: "If you have a soup with 100 bad ingredients and 1 good one, the good one might just look like a winner because the bad ones died off. It's not that the good one is better; it's just that the others were worse."
• The Fix: In this new study, they didn't use a giant, messy soup. They took a specific, small group of 64 "ingredients" (clones) that they knew were good, bad, or neutral. They put them in a controlled race. This is like taking 64 specific runners and timing them individually, rather than just watching a chaotic crowd and guessing who was fastest.

2. The "Sick Pot" Problem (The Vector Effect)

The scientists used a delivery truck (a plasmid/vector) to put the random DNA into the bacteria.

• The Skeptics' Argument: "Maybe the delivery truck itself makes the bacteria sick. If you put a random piece of paper in the truck that blocks the sickness, the bacteria looks healthy, but it's only healthy because the truck stopped hurting it. It's not a superpower; it's just relief."
• The Fix: They built "dummy trucks" that had the delivery mechanism but no random DNA. They also built trucks with "stop signs" that prevented the truck from making any noise or mess. They found that while the trucks did cause a little bit of trouble, the random DNA sequences that made the bacteria grow fast were still faster than the trucks could explain. The "superpower" was real, not just a fix for a broken truck.

3. The "Marathon" vs. The "Sprint"

Bacteria grow in cycles. Sometimes they grow until they are full and tired (stationary phase), and sometimes they are in the middle of a fast sprint (exponential growth).

• The Test: The scientists ran the race in two ways: long, slow marathons (24-hour cycles) and short, intense sprints (3-hour cycles).
• The Result: They found that the random DNA sequences that gave a "fitness boost" worked best during the sprints. This is crucial because in nature, bacteria are often in a sprint mode, trying to reproduce as fast as possible. The random DNA sequences acted like a turbocharger for the engine.

4. The "Background Noise" (Genetic Background)

Sometimes, a bacteria might have a secret mutation in its own DNA that helps it grow.

• The Test: The scientists took the same random DNA, put it into a fresh batch of bacteria, and ran the race again.
• The Result: Most of the time, the results were the same. The random DNA was the hero. However, for a few specific cases, the "hero" DNA stopped working when moved to a new bacteria. This taught them that sometimes, the "superpower" depends on the specific team it's playing with. But for the majority, the power was real and consistent.

The Big Conclusion

After all these careful checks, the authors concluded:

Randomness isn't just chaos.

Think of it like a lottery. Most tickets are losers. But if you buy enough tickets, you will eventually find a winner. This study proves that in the lottery of life, random DNA sequences can actually win the jackpot. They can create tiny proteins that help a cell grow faster, survive better, and eventually become a brand new gene.

Why does this matter?
It changes how we see evolution. We used to think genes had to be carefully built over millions of years. This paper suggests that nature can "hit the jackpot" much more often than we thought, using random noise to create new tools for survival. It's like finding a working key in a pile of junk; it happens more often than you'd think, and it can change everything.

Technical Summary

1. Problem Statement

The emergence of de novo genes from non-coding sequences is a recognized evolutionary mechanism, but the functional potential of truly random DNA sequences remains debated.

• Previous Findings: A prior study (Neme et al., 2017) suggested that expressing random sequence clones in Escherichia coli could confer a fitness advantage (enhanced growth) to the host cells.
• The Controversy: These findings were challenged (e.g., by Knopp and Andersson, 2018) with two primary arguments:
  1. Competition Artifacts: Enrichment in a mixed library might be a passive byproduct of competition with clones that have strong negative effects, rather than a true positive growth effect.
  2. Vector Artifacts: The expression vector itself might produce a toxic peptide. If random inserts simply reduce this toxicity (by disrupting the vector's open reading frame), the observed "growth advantage" is merely a relief of a negative effect, not a genuine positive selection.
• Objective: This study aims to rigorously test whether random sequences confer genuine fitness advantages independent of vector artifacts and competition dynamics, using controlled competitive growth assays.

2. Methodology

The authors employed a highly controlled competitive growth assay design to isolate variables and ensure reproducibility.

• Clone Selection: A defined subset of 64 random sequence clones was selected from a previous library (Neme et al., 2017), representing a spectrum of previously observed positive, neutral, and negative growth effects.
• Experimental Design:
  • Competition Assays: Clones were mixed in equal proportions and grown in competitive cycles.
  • Cycle Durations: Two regimes were tested:
    • 24-hour cycles: Resulting in growth saturation (stationary phase).
    • 3-hour cycles: Keeping growth within the exponential phase (closest to conditions used in previous critical studies).
  • Induction States: Experiments were run with and without IPTG induction to control the expression of the random sequences via the Ptac promoter.
  • Replicates: High technical replication (5 replicates per condition) across multiple independent experiments (Exp A–F).
• Controls and Variations:
  • Vector Constructs: Included specific vector modifications (from Knopp & Andersson, 2018) designed to suppress the expression of the vector-coded peptide: delPtac (promoter deletion), delRBS (ribosome binding site deletion), and lux (premature transcription terminator).
  • Genomic Background Check: Compared clones directly from the library against the same clones after plasmid extraction and re-transformation into fresh host cells to rule out host genome mutations.
  • Subset Experiments: A specific experiment (Exp F) competed only the "positive" clones (excluding negative ones) to test if positive effects persist without competition from deleterious clones.
• Data Analysis:
  • Sequencing: Illumina MiSeq sequencing of plasmid inserts to track clone frequencies via unique 20nt search motifs.
  • Selection Coefficients: Calculated using DESeq2 (linear trend function) to determine the slope of abundance change over time. Values were converted to natural log scales to represent selection coefficients ($s$).
  • Net Selection: The "net" effect of the random sequence was calculated by subtracting the selection coefficient of the uninduced (minus IPTG) condition from the induced (plus IPTG) condition to isolate the effect of the expressed peptide.

3. Key Results

• Reproducibility: The experimental design showed high technical reproducibility. Coefficients of variation (CV) between replicates were low (<0.2), and selection coefficients correlated strongly between independent replicate sets ($r > 0.4$, often much higher).
• Genomic Background Influence: While most clones showed consistent behavior, two specific clones reversed their phenotype (from positive to negative) after re-transformation. This indicates that host genomic background mutations can influence clone performance, necessitating the use of re-transformed plasmids for accurate assessment.
• Vector Effects Confirmed: The vector constructs designed to suppress the vector-coded peptide (delPtac, delRBS, lux) consistently showed higher fitness than the wild-type vector, confirming that the vector peptide has a slight deleterious effect.
• Genuine Positive Effects:
  • Several random sequence clones exhibited higher positive selection coefficients than even the lux vector construct (which completely abolishes vector peptide expression).
  • This finding is critical: if the "advantage" were merely relief of vector toxicity, no clone could exceed the fitness of the lux construct. The fact that they did proves a genuine positive fitness contribution from the random sequences.
  • Subset Validation: In the experiment excluding all negative clones (Exp F), the positive clones still showed positive selection coefficients relative to the vector, proving the effect is not driven by competition dynamics with deleterious clones.
• Magnitude of Effect: The observed selection coefficients for beneficial random sequences ranged from 0.05 to 0.21. These values are comparable to beneficial mutations observed in long-term experimental evolution studies (e.g., Lenski's LTEE, where $s$ ranges from 0.006 to 0.059).

4. Key Contributions

1. Resolution of Controversy: The study definitively refutes the hypothesis that previous positive growth effects were solely artifacts of vector toxicity or competition with negative clones.
2. Methodological Rigor: It establishes a robust protocol for measuring fitness effects of random sequences, emphasizing the need for:
   • Exponential growth phase measurements (3h cycles).
   • Correction for basal expression (minus IPTG controls).
   • Re-transformation to control for host background.
   • Competition against "clean" backgrounds (excluding negative clones).
3. Quantitative Evidence: It provides quantitative data showing that a subset of random sequences confers fitness advantages comparable to known beneficial mutations.

5. Significance

• Evolutionary Biology: The findings provide strong experimental support for the de novo gene birth hypothesis. They demonstrate that random DNA sequences are not just a source of deleterious noise but contain a significant reservoir of sequences capable of generating adaptive, beneficial functions immediately upon expression.
• Functional Potential: The study suggests that the "functional space" of random peptides is vast, and the emergence of new genes from non-coding regions may be a frequent and viable evolutionary process.
• Future Directions: The results underscore the importance of considering random sequence expression in evolutionary models and suggest that microproteins derived from intergenic regions may play a more significant role in bacterial adaptation than previously thought.

In conclusion, Künzel et al. demonstrate that while experimental artifacts must be carefully controlled, a genuine subset of random sequences can indeed drive positive selection in E. coli, validating the potential for de novo gene emergence from non-coding DNA.