Contribution
It names and formalizes a route to metabolic dependency in which selection is the cause, not drift. Before it, genome reduction was explained mainly for endosymbionts, by drift in small, bottlenecked populations. Prochlorococcus and “Candidatus Pelagibacter” have reduced genomes and huge populations, so drift cannot be the cause there. The paper explains how selection on individuals produces dependence. It also says why the dependence stops short of losing the function from the community.
Key insight
A function is a candidate for loss when it is costly and leaky, so that performing it “inevitably benefits other cells in their vicinity”. Hydrogen peroxide (HOOH) crosses membranes about as easily as water, so any cell that detoxifies its own interior also lowers the peroxide outside. A cell that stops doing this saves resources and still gets the protection, as long as enough others keep doing it. The advantage is frequency dependent. When helpers become rare enough, the public good falls until “the benefit of any further loss is negated by its cost”. The result looks like a division of labour, but “in fact this bifurcation arises through classic Darwinian selection acting on individuals”. The helpers are “passive players”, in the paper’s words “perhaps the unlucky ones that were the slowest to change”.
Assumptions
- Free-living, well-mixed populations for the basic argument. Spatial structure is discussed and set aside as not changing “the fundamental possibility”.
- The benefit of loss is a resource saving. B is the fractional reduction in the cell quota of a limiting resource, and it raises the growth rate proportionally.
- Linear response. Growth rises linearly with the effective rate of the function between zero and the rate that allows maximal growth (equation 3). The authors say “the general conclusions should still hold with more complex interactions”. This is asserted, not shown.
- Large effective population size, so that selection, not drift, governs loss.
- That katG was ancestral and lost. This is inferred from gene–species tree discordance and the clade’s monophyly (Figure 1). The paper calls the losses “the simplest explanation”.
Key results
-
The katG case. No sequenced Prochlorococcus genome has katG. Most Synechococcus genomes do, and their KatG proteins are monophyletic, so the gene was lost in several lineages, including all of Prochlorococcus (Figure 1). Axenic Prochlorococcus is killed by sunlit filtered seawater within hours, and in nature co-occurring cells act as a peroxide sink (references 17 and 28, not read here). “Ca. Pelagibacter ubique” also lacks catalase.
-
Benefit (equation 1). B = (Q_A − Q_M)/Q_A, where Q is the cell quota of the limiting resource for the ancestor (A) and the mutant (M). For katG the saving is about 0.2% of cellular iron.
-
Speed. With B = 0.002 and one doubling a day (μ ≈ 0.69 per day), the mutant gains at about 0.0014 per day. In a global population of about 10²⁵ cells, it would reach 99% “in fewer than 125 years”. Recomputed here, with the population taken as 10⁴ cells per ml over 10 m of depth across a quarter of the Earth’s surface:
- the population is ≈1.3 × 10²⁵ cells;
- going from one cell to 99% needs a change in log-odds of ln(99) + ln(1.3 × 10²⁵) ≈ 62.4;
- at 0.00138 per day that takes about 45,200 days, or 124 years.
The figure holds. It treats the mutant’s advantage as constant all the way to 99%, while the paper’s own frequency dependence would slow it.
-
Cost (equations 2–3). The effective rate is k = k₀ + Σᵢ Nᵢ Lᵢ kᵢ, the cell’s own rate plus everyone’s rate weighted by density N and leakiness L. The cost of loss is C = (k_M − k)(μ_M − μ₀)/(μ_M k_M), which is (k_M − k)/k_M when the mutant cannot grow alone. This follows from the linear assumption, and was re-derived here. A loss-of-function mutant invades when B > C.
-
Outcomes. Against its ancestor alone, the mutant spreads until the ancestor’s density gives both equal growth: helper and beneficiary coexist on one limiting resource. In a mixed community where unrelated species supply the good, the mutant drives its ancestor extinct. Loss then “will proceed in every species in the community until B is less than or equal to C for all individuals”.
-
Not kin selection. Hamilton’s rule asks whether a producer’s cost is repaid through relatives (rB > C). The Black Queen asks whether a non-producer’s saving outweighs its loss of protection, “without regard to the relationship between the producer and user”.
-
Not necessarily cheating. Beneficiaries harm helpers only if both are limited by the same resource. Prochlorococcus and its heterotrophic helpers are on different trophic levels, and the heterotrophs may depend on Prochlorococcus for carbon. So the outcome can be commensal or mutualistic.
-
“Shooting the moon.” Since B is a fraction of the remaining quota, each loss makes the next saving of the same size proportionally larger for a beneficiary. Loss should therefore accelerate in beneficiary lineages and not in helpers. The predicted community is many streamlined beneficiaries and a few helpers, which become rare but indispensable “keystone species”. This is the paper’s speculation, as it says.
-
Other candidates, all suggestions. Nitrogen fixation (less than 1% of marine cyanobacteria are diazotrophs), siderophores, Pelagibacter’s loss of sulfate reduction and dependence on DMSP, and biofilm matrix.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Loss of a costly, leaky function can be favoured by individual selection, and stops short of community-wide loss by frequency dependence | strong as a model result | equations 1–3; it follows from the definitions once leakiness and a resource saving are granted |
| C2 | Prochlorococcus lost katG adaptively, protected by a peroxide-removing community | moderate | gene absence and phylogeny (Figure 1), the cited sensitivity and sink experiments; the size of the benefit (0.2% iron) is computed, not measured as fitness |
| C3 | The genome reduction of Prochlorococcus and Pelagibacter is selected, not drifted | moderate | effective population sizes far beyond any experiment; argued, not tested here |
| C4 | Nitrogen fixation, siderophores, sulfur assimilation and biofilm matrix are lost by the same process | weak | fit to the four criteria; no analysis |
| C5 | Loss accelerates in beneficiaries, so communities polarize into many beneficiaries and few helpers that become keystone species | weak | a property of equation 1’s ratio, with “all else equal”; offered as speculation |
| C6 | Beneficiaries are cheaters only when they share the helpers’ limiting resource | moderate as an argument | the definitions of cheating and of the interaction |
Concepts
- Leaky function: a function some of whose product becomes a public good available to other cells. L, the leakiness coefficient, is the share of activity that does so.
- Helper / beneficiary: a cell that still performs the function, and one that has lost it and depends on helpers.
- Black Queen function: a function that is costly to perform and yet indispensable to someone in the community, like the queen of spades in Hearts, which nobody wants and the game cannot be played without.
- Streamlining: genome and metabolic reduction that lowers the resource cost of making a cell.
- Shooting the moon: a lineage passively becoming a helper for many functions, with “job security” from the community’s dependence.
Connections
The paper sets itself against the Red Queen (antagonistic coevolution) and Hamilton’s rule (kin-selected production of public goods). It relies on earlier work for the drift account of symbiont reduction (its reference 9) and for experimental evolution with Salmonella bottlenecks and E. coli deletions (references 10–12), none read here. LIT-1570 (Athreya, Czuppon & Gokhale), read in NOTE-1211, cites it as the understood route to mutual dependence. It groups it with “environment-induced redundancy” as a driver of symbiont genome reduction other than the Red King effect. LIT-1225 (Cooper & West, unread) models extreme specialization within a group, where helpers are selected as helpers. Here helpers are a by-product.
Bearing on the record
- QUESTION-277, the first route. The reading confirms what the question said of this route, and sharpens it. Dependence arises with no collective at all: there is no co-transmission, no host and no shared fate. The dependence is directional, from beneficiaries to helpers, and helpers need not depend on beneficiaries. Function is distributed across the community’s genomes (G). Reproductive cohesion (R) and regulatory control (C) are absent by construction. So of the question’s five dimensions, the route moves D (one-way) and G, and none of the others. That is the clearest case in the record of dependence without individuality. The inference is mine.
- The scope correction in the curation entry of 2026-10-27 stands. The hypothesis is stated for free-living organisms, and the paper credits drift for endosymbiont reduction. Applying it to endosymbionts is LIT-1570’s move (“environment-induced redundancy”), not this paper’s.
- CLAIM-060. The conclusion says the division of labour “is often to the advantage of all”. Its own cheater section allows the opposite when helpers and beneficiaries share a limiting resource. So whether the community gains is contingent, which is consistent with CLAIM-060’s thesis that the whole’s success and its members’ welfare come apart. This is a further instance, not a ground. The comparison is mine.
- THEORY-141. Its condition for a stable collective, that selection between collectives trump selection within, does not apply here, because no collective is formed. The Black Queen is a case of interdependence that the transitions framework does not cover.
- Filed as THEORY-1048. No instruction for machine-learning practice.
Limitations
- One worked case. Everything general rests on katG. The fitness benefit is an iron-quota calculation, not a measured selection coefficient, and the peroxide sink is cited, not shown.
- Linearity is assumed. Equation 3’s closed form, and the claim that the conclusions survive nonlinearity, are not demonstrated.
- Constant advantage in the speed estimate. The 125-year sweep ignores the frequency dependence the hypothesis itself requires. It is an upper bound on speed, not a prediction.
- Ancestral state. That katG was present and lost is the “simplest explanation” of the tree. Horizontal gain into Synechococcus is not excluded by the paper’s argument.
- The generalizations are untested. The four other functions are matched to criteria, not analysed.
- Selection versus drift is argued from population size alone, without the dN/dS evidence the paper cites for symbionts.
Open questions
- Has a Black Queen loss been reproduced experimentally, invading and then stopping at the frequency equations 2–3 predict? LIT-1570 cites D’Souza and Kost (2016) for metabolic dependency evolving quickly in experiment. That paper is not in the record.
- Does “shooting the moon” happen? The prediction is that helper-for-many-functions lineages are rare keystone taxa, and that beneficiary genomes keep shrinking faster.
- Where the beneficiary does become an endosymbiont, does the Black Queen start the reduction and drift finish it? The paper does not say, and QUESTION-277’s routes may run in sequence.