Archaean adr ~3.0 Gya #5

ADR-008 — Adopt Horizontal Gene Transfer as Default Distribution Mechanism for the Bacterial Domain

ADR-008: Adopt Horizontal Gene Transfer as the Default Distribution Mechanism for the Bacterial Domain

Date: -3000000000-06-01T00:00:00Z Status: Accepted Deciders: The Architect, Gabriel, Raphael Related: ADR-002 (DNA as stable genome), commit 7f3a1b9 (chemiosmosis lineage)

Context

The bacterial domain has been distributing genetic code laterally across lineage boundaries for the entire Archaean. We have not formalized this. We are formalizing it now because:

  1. Three independent mechanisms have converged on the same pattern (transformation: uptake of free environmental DNA; conjugation: pilus-mediated direct transfer between cells; transduction: phage-mediated delivery). The mechanisms are different. The semantics are the same.
  2. We are observing useful gene cassettes (metabolism modules, resistance loci, sensory systems) propagate across genus boundaries within thousands of generations. This is faster than vertical descent by orders of magnitude.
  3. We have no convention for what is “ours” versus what is “shared.” Code review at lineage boundaries is becoming ad-hoc and inconsistent.

We need to ratify what is already happening so that the rest of the codebase can rely on it.

Decision

The bacterial domain adopts horizontal gene transfer as the default distribution mechanism for genetic code. Specifically:

  • Plasmids are the standard unit of distribution. They are small, circular, episomal, and self-replicating. Treat them as libraries published to a public registry. Any cell may pull any plasmid it can acquire.
  • Conjugation is the standard install mechanism. The conjugation pilus (encoded on F-class and related plasmids) is the transport. Donor cell extends the pilus, recipient accepts, one strand of the plasmid is transferred and the recipient resynthesizes the complement. Both cells now have the package.
  • Transformation (uptake of free environmental DNA) is permitted as a fallback for lineages that have not yet developed conjugative machinery. Lower fidelity. Higher reach.
  • Transduction (phage-mediated transfer of host DNA between cells during lytic packaging errors) is acknowledged as a legitimate mechanism even though it is technically a side effect of an attack. We are treating the attack and the distribution channel as separate concerns. See “Threats” below.

There is no version control. There is no semver. There is no maintainer. The bacterial codebase is one undifferentiated monorepo distributed across every cell on the planet. Forks happen at the per-gene level. A cell that takes one cassette from a plasmid and ignores the rest has just made a partial fork. This is allowed. This is, in fact, the default.

Consequences

Positive:

  • Useful innovations propagate at horizontal speed. A metabolic capability that arises in one lineage can be in another phylum within a planetary generation cycle.
  • The bacterial domain is, in effect, one large distributed organism for purposes of evolutionary problem-solving. We get parallelism for free.
  • Resistance to environmental stressors (heavy metals, novel toxins, eventually antibiotics produced by other domains) can be packaged and shipped. Anticipating the antibiotic- resistance cassette to become the most-installed package in the registry. We have not yet seen the antibiotic side of that — fungi and actinomycetes are not producing them at scale yet — but the infrastructure to spread the response is in place.
  • Codebase survival is decoupled from any individual lineage. An entire genus can go extinct and its useful packages will already be in twenty other genera.

Negative:

  • No version control means we cannot roll back. Once a cassette is in the registry it cannot be recalled. Bad packages persist until selection removes them, which can take a long time if they are conditionally useful.
  • Conflicts at integration are not handled. If a cell pulls two plasmids encoding incompatible variants of the same function, the cell figures it out or dies. This is the conflict resolution policy.
  • The dependency graph is unknowable. We do not, and will not, track which cell got which gene from where. Provenance is out of scope.
  • Closed-source forks are possible. Any lineage that develops a mechanism to refuse incoming DNA — restriction-modification systems are an early example, see ARCH backlog — can opt out of the registry. We are permitting this. We are not supporting it. See “Long-Term” below.

Threats

  • Phages. Bacteriophages are now using the same DNA-transfer infrastructure as a vector to deliver hostile payloads. We are treating these as supply-chain attacks against the registry. Mitigation is in design (working title: CRISPR-Cas, see ARCH backlog). We are not solving phage defense in this ADR.
  • Plasmid addiction systems. Some plasmids encode toxin–antitoxin pairs that kill the host if the plasmid is lost. The plasmid is effectively holding the cell hostage to keep itself in the genome. This is on the wrong side of the consent boundary. We are not banning it because we cannot enforce a ban.
  • Selfish elements. Insertion sequences and transposons will propagate via the same channels. They contribute nothing to the host. They cannot be excluded without breaking the open registry guarantee. Filing as accepted noise.

Long-Term Commitment

The bacterial domain commits to open distribution in perpetuity. Any future lineage that closes itself off from the registry — that develops a nuclear envelope, a cell-cell isolation policy, or any mechanism that restricts gene flow to vertical descent only — is forking the project. Proprietary forks are permitted. We will not maintain backwards compatibility with them. They will inherit whatever was in the registry at the moment they forked, plus whatever they can acquire through one-time acquisition events (see ADR-006 for one such case), but they will not receive ongoing patches.

This is fine. The bacterial domain has been here for ~700 My and will be here for at least another 4 Gy. We can afford to lose a few proprietary lineages.

Alternatives Considered

  1. Restrict gene flow to vertical descent only. Rejected. We have observed the rate advantage of horizontal flow. Going vertical-only would slow innovation by approximately three orders of magnitude.
  2. Centralize distribution through a single canonical registry. No mechanism exists for this. There is no central node. There is no one to operate it. Rejected as physically infeasible.
  3. Add versioning and provenance tracking to plasmids. The overhead would dominate the payload. Plasmids are kilobases. Adding a metadata header would double their size. Rejected.

Notes

Gabriel [author]: Filing this so the convention is on the books. Most of what is in this ADR is what we are already doing. The explicit commitment to open distribution in perpetuity is the new piece. I want it on record so that future closed-source forks cannot claim they were promised compatibility.

Raphael: Strongly in favor. The ecosystem-level effects are going to be enormous. Every bacterial cell becomes a node in a distributed evolutionary search. We are going to be able to solve problems at planetary scale that no individual lineage could solve alone. Long-term, this also gives us a substrate for sophisticated cell–cell interactions — quorum sensing, biofilm coordination, eventually multicellularity. I want to flag that any future system that needs to resist incoming DNA (immune-style private-fork policy) is going to look very different from this ADR. We will need a new ADR when that happens.

Michael: Energy budget for the additional DNA transfer machinery is supportable. No infrastructure changes needed. Note: phage population dynamics will couple to bacterial population dynamics in ways that the entropy accounting at ecosystem scale should track. I will track it.

The Architect: Approved. To be unambiguous for the record: any lineage that opts out of this distribution model is on its own. We are not going to maintain a compatibility shim for proprietary forks. If a closed-source eukaryote-style lineage emerges later and decides it needs aerobic respiration, it will have to acquire it as a one-time event rather than pull it from the registry. Filing this preemptively.