Plasmids are nature's USB drives: small circular DNA molecules that shuttle genetic information between bacteria, spreading everything from antibiotic resistance to the ability to digest novel food sources. They're crucial for bacterial survival and evolution, and they're indispensable tools in biotechnology. But there's a problem: while plasmids transfer efficiently in nature, they often perform much less well in the lab. The reason probably lies in a biological arms race we're only beginning to understand.

Bacteria aren't defenseless against incoming DNA. Over billions of years, they've evolved an arsenal of immune systems to destroy foreign genetic material. The most famous one is CRISPR-Cas, which scientists have repurposed for genome editing. But CRISPR-Cas is just one example. Research over the past two decades has revealed over 150 distinct bacterial defense systems, from restriction enzymes that cut DNA to abortive infection systems that trigger cellular suicide rather than allow infection to spread. And this is likely just a fraction of what's actually out there.

Natural plasmids, shaped by the same billions of years of evolution, have developed countermeasures: molecular hacking tools that disable bacterial defenses. But while the defensive side of this arms race has received enormous attention, the counter-measures remain largely uncharacterized. There are tens of thousands of plasmids in nature, yet we've identified anti-defense functions for only a handful of bacterial immune systems.

This is precisely the gap that Prof. David (Dudu) Burstein of Tel Aviv University's Shmunis School for Biomedicine and Cancer Research aims to fill. His new ERC, DEFCONJ, will systematically map the landscape of plasmid-encoded anti-defense systems: what might be called the "hacking toolkit" that plasmids deploy to breach bacterial firewalls.

The Strategic Position

The story begins with connections. When a plasmid transfers between bacteria through conjugation (the process where one bacterium connects to another via a tiny channel and passes DNA through it), the circular double-stranded DNA doesn't arrive intact. One strand is nicked at a specific site called the origin of transfer, and from that point, single-stranded DNA threads into the recipient cell.

"We found that anti-defense systems are highly concentrated next to this cutting point and oriented such that they are the first genes to enter the new cell during the transfer process," explains Burstein. "This strategic positioning allows these genes to be activated very quickly, giving the plasmid a crucial advantage in overcoming the receiving bacteria's defense mechanisms."

This leading region, the first DNA segment to enter the recipient, contains the plasmid's assault team. The problem is that while we know this region is packed with anti-defense genes, most of them remain uncharacterized. Standard computational tools can't identify them because they share little sequence similarity with known proteins. They're small, diverse, and hidden in plain sight.

Reading Genes Like Words

One of Burstein's approaches to finding these hidden weapons draws on an unexpected source: linguistics. His lab has pioneered the use of genomic language models, artificial intelligence systems that treat genes like words in a sentence. Just as we can often guess the meaning of an unfamiliar word from its context ("The chef diced the vegetables with a sharp ____"), these models can predict the function of unknown genes based on their genomic neighbors.

If we see an unknown gene tending to appear near anti-CRISPR genes, near anti-restriction genes, in the leading region of plasmids, we can make informed predictions about its function. The approach has already revealed promising candidates: protein families with structural similarities to known anti-CRISPR and anti-restriction factors, despite sharing less than 30% sequence identity.

Another prong of the strategy involves working backward from targets. If we know what defense systems are present in a given bacterial population, we can predict which anti-defense tools a successful plasmid would need to carry. By analyzing clinical isolates of E. coli (bacteria from real infections), Burstein's team has already documented the diverse repertoire of defense systems these pathogens possess, providing a map of the "firewalls" that plasmids must breach.

The Timing Problem

But identification is only part of the puzzle. These anti-defense genes need to be expressed at precisely the right moment, during the narrow window after the plasmid enters but before bacterial defenses can mount a full attack. This timing is controlled by a peculiar type of genetic switch that remains mysterious.

Here's the challenge: when plasmid DNA first enters a recipient cell, it arrives as a single strand (the complementary strand hasn't been synthesized yet). Normal gene expression requires double-stranded DNA: the cellular machinery reads the double helix to produce proteins. Single-stranded DNA can't be read by the conventional machinery, which means the plasmid's genes are essentially silent during those critical first minutes.

The single-stranded state does offer some protection, as most bacterial defense systems target double-stranded DNA. But this protection is incomplete and temporary. Once complementary strand synthesis begins and double-stranded DNA forms, the full arsenal of bacterial defenses can spring into action.

The plasmid faces a deadly race: it needs to produce its anti-defense proteins before its DNA becomes the double-stranded substrate that defenses can destroy, yet protein production normally requires that very same double-stranded DNA.

Natural plasmids solved this paradox through an elegant hack: single-stranded DNA promoters. These are sequences that, when single-stranded, fold into flexible hairpin structures (loops where the DNA base-pairs with itself, creating short pseudo-double-stranded regions). These artificial double-stranded segments can be recognized by the cell's gene expression machinery, initiating gene expression directly from the single-stranded template.

The system is self-regulating. The moment complementary strand synthesis converts the DNA to its normal double-stranded form, the flexible hairpin collapses (the sequence that formed the stem is now paired with its true complement, not with itself). The single-stranded promoter vanishes, and expression from it stops. Meanwhile, conventional promoters elsewhere on the plasmid switch on, activating genes needed for replication and maintenance.

It's a built-in timer: anti-defense genes fire immediately upon entry, from single-stranded promoters, producing proteins that neutralize host immunity. By the time the defenses would normally strike, the plasmid's countermeasures are already in place.

Yet despite their importance, single-stranded promoters remain almost completely uncharacterized. Only a handful have been experimentally validated, all from just two plasmid types. Burstein's preliminary computational analysis has identified over 10,000 potential single-stranded promoters across diverse plasmids, but only a small portion shows sequence similarity to known examples. The vast majority represent unexplored regulatory territory, detected only through their predicted hairpin features, not their sequences.

"Understanding these promoters is crucial for two reasons," explains Burstein. "First, they tell us about the timing and coordination of the plasmid's attack on bacterial defenses. Second, if we want to engineer better plasmids for biotechnology, we need to be able to program this timing ourselves."

Complex Attacks, Complex Defenses

The final piece of the puzzle involves understanding how multiple anti-defense systems work together. Bacteria don't rely on a single immune system; they layer multiple defenses, each targeting different aspects of foreign DNA. A successful plasmid needs to overcome them all, simultaneously or in sequence.

Burstein's team has found intriguing patterns in how anti-defense genes are organized within the leading region. Some combinations appear together far more often than chance would predict, suggesting functional synergies. Others show conserved positioning relative to the origin of transfer, hinting at organizational constraints related to expression timing.

By dissecting these combinations (testing which anti-defense genes enhance each other's effects, which are redundant, and which might even interfere), the project aims to understand the logic of the plasmid's coordinated assault.

From Understanding to Engineering

"Our study will eventually help design more efficient plasmids for genetic manipulation of bacteria for industrial processes," says Burstein. "Plasmids are already a common tool for such purposes, but their efficacy of transferring genetic material via conjugation in industrial settings is much lower than in nature."

The gap between natural and laboratory plasmid performance is striking. Natural plasmids carry sophisticated hacking toolkits honed by evolution; synthetic plasmids are essentially sent in without the code needed to bypass host firewalls. By cataloging anti-defense systems, characterizing their regulatory elements, and understanding how they work in combination, the DEFCONJ project could provide the toolkit needed to engineer plasmids that actually work in real-world bacterial populations.

"Another possibility is to design effective plasmids for genetic manipulation of natural bacterial populations," adds Burstein, "whether it is to neutralize antibiotic resistance genes in bacteria in hospitals, 'teach' bacteria in soil and water to remove pollutants or fixate carbon, or manipulate bacteria in our gut to improve our health."

The bacterial world is engaged in a constant arms race between defense and counter-defense. By mapping the hacking tools on one side and the firewalls on the other, we might finally learn to tip the balance: whether to stop the spread of dangerous plasmids, or to harness their remarkable transfer machinery for our own purposes.