A word too short for what it designates
When we say “second messenger”, we picture a small molecule running between two proteins. Bacterial reality is more interesting: an ordinary nucleotide (ATP, GTP, CTP, UTP) is cyclized by a dedicated enzyme—the CD-NTase—to produce a unique chemical signal. This signal diffuses, finds an effector protein sensitive to its exact shape, and triggers a cascade. This cascade can kill the entire cell to prevent phage propagation.
This is the exact definition of innate immunity. Structurally, it is also the mechanism of human cGAS-STING—the system that detects foreign cytoplasmic DNA and triggers the interferon response.
This homology is no evolutionary coincidence. It is an inheritance.
CBASS — the dominant branch
The CBASS system (Cyclic Oligonucleotide-Based Anti-Phage Signaling System) was characterized in 2019 by Rotem Sorek’s team (Weizmann). It groups hundreds of known variants, distinguished by:
- The CD-NTase used—each produces a cyclic of a different nature (cGAMP, c-di-AMP, c-tri-AMP, c-UMP-AMP, c-AAA…)
- The activated effector—endonuclease Cap4, phospholipase CapV, pore-forming NucC, NAD+ depletion, etc.
The architecture is modular. A CBASS system is a sensor (the CD-NTase) coupled with an effector. A chemical signature (the cyclic messenger) bridges them. This modularity is the very definition of a “learnable pattern” in Raph Koster’s sense. Once you have seen three CBASS variants, you understand the grammar. Each new candidate protein becomes a recognition puzzle.
See the CBASS Type II sheet for the list of proteins indexed in Bactaegion, and the cGAMP mimetics lead for transposition to human STING.
Pycsar — the other language
Where CBASS speaks in cyclic purines (A and G bases), Pycsar speaks in cyclic pyrimidines: c-CMP, c-UMP. It is a distinct dialect. The effectors are different. The sensors are too.
The translational interest is unique: cyclic pyrimidines are unknown to canonical human immunity. We have no STING equivalent that recognizes cCMP or cUMP. It is a blank chemical space—precisely where Bactaegion can bring out original hits. See the immunomodulatory cCMP/cUMP analogues lead.
Tal et al. (Cell, 2021) characterized Pycsar three years after CBASS. The family grows every year. Each deeply sequenced bacterial genome adds two to ten new variants.
Why it is playable
The human eye is excellent at spotting patterns. When reading several CD-NTases in a row—their sequence, the arrangement of charged residues around the active site, the length of flexible loops—you progressively feel what makes a true CD-NTase and what is a false friend. This is what experienced biochemists do without realizing it. Algorithms struggle to formalize this intuition. It depends on functional context as much as pure sequence.
The Bactaegion bet: give a thousand curious minds a horizontal Piano-roll of colored amino acids, with two sliders to isolate a segment, and the instruction “mark what seems coherent”. We recover 80% of expert annotation in a few weeks. The remaining 20%—the true subtleties—remain the experts’ work. This is the novice/expert asymmetry in the sense of Galaxy Zoo and iNaturalist.
Want to try? Choose a CBASS or Pycsar protein and see what your eye suggests.
To go further
- CBASS Library — 8 indexed proteins
- Pycsar Library — 6 indexed proteins
- Chapter 1: The arms race — the global context
- Piano-roll Workshop — annotate a sequence yourself