The term 'symbiosis' is often used in Chinese discourse as a warm metaphor: mutual benefit, cooperation, win-win outcomes. Yet in biology, the conceptual development of symbiosis over the past six decades has followed a different path: each substantive advance has coincided with a forced redrawing of the boundaries of the 'individual.' What counts as one organism? At what point does a group of partners constitute a single individual? This article reads together eight summary-level materials spanning 1967 to 2025; they form four acts of boundary revision: Act I brings the question of 'the individual' inside the cell; Act II adds a previously overlooked party to lichens; Act III centers on the unresolved debate over whether an animal plus its microbiome constitutes a single individual; Act IV is a systematic retraction of a popular forest narrative. After reading all four acts, three criteria for determining 'one individual' emerge—criteria also applicable to collaborative systems beyond biology, including human–machine systems.
## Cells May Be Chimeras
In March 1967, the *Journal of Theoretical Biology* published a paper from the Department of Biology at Boston University proposing a bold hypothesis: three fundamental eukaryotic organelles—the mitochondrion, the photosynthetic plastid, and the (9+2) basal body of cilia and flagella—were themselves once free-living prokaryotes. The paper claims to present evidence from cytology, biochemistry, and paleontology; lists experimentally testable predictions; and discusses implications for the systematics of lower organisms.
If this hypothesis holds, 'a cell' ceases to be a unit built once and for all, becoming instead the historical product of multiple mergers: the boundary of the individual must be redrawn within the cell (this article’s summary). How each part of this hypothesis was tested in subsequent decades falls outside the scope of the materials covered here; no claim is made about its current status. What can be confirmed is the structure of argument it exemplifies: a bold claim supported by three lines of evidence and a concrete list of testable experiments (this article’s judgment).
## Lichens Are Not Two-Party Systems
In July 2016, a paper in *Science* revised a definition that had stood for over 140 years. For over a century, lichens had been viewed as symbiotic associations between a fungus (typically an ascomycete) and a photosynthetic partner. Although other fungi had long been known to occur occasionally as parasites or endophytes, the 'one lichen, one fungus' paradigm was rarely questioned. This paper revealed that many common lichens are tripartite: the known ascomycete, the photosynthetic partner, and—unexpectedly—certain specific basidiomycete yeasts. These yeasts reside embedded in the cortex; their abundance correlates with previously unexplained phenotypic variation; these basidiomycete lineages maintain tight associations across vast geographic distances and with specific lichen species, having been found on all six continents. The cortex—a structurally critical layer long assumed to consist solely of differentiated cells of the ascomycete—turns out to have consistently contained two unrelated fungal species.
This act of revision does not overturn any symbiosis theory; it merely adds a previously omitted party: the number of parties within the boundary shifts from two to three (this article’s summary).
## Intimacy Does Not Imply Wholeness
Act III remains ongoing: does an animal plus its microbiome constitute a single individual? The concept of the 'hologenome' pushes this question to its extreme: it refers to the collective genome content of host plus microbiome, and proposes using it to explain evolution.
In December 2015, a commentary in *PLoS Biology* directly asked whether this concept is useful or hollow. It points out several easy-to-overlook confusions: microbes may be intimate with, ancient alongside, and mutually beneficial to their hosts, yet they have never undergone natural selection *for the host’s benefit*; observing host-specific microbiota composition—or finding that phylogenetically closer hosts harbor more similar microbiota—does not imply coevolution between symbiont and host, let alone evolution *for the host’s benefit*; selection at the community (hologenome) level does occur in some cases, but should not be accepted as the null hypothesis explaining host–symbiont association traits. The authors of this paper declare no competing interests.
## The Fidelity Loophole
In March 2016, a perspective article in *mBio* placed negation right in its title: 'Host–Microbiota Symbioses Are Not Holobionts' (title transcribed). The article argues that a host plus its microbiome is better viewed as an ecological community whose internal interactions span a spectrum—from parasitism to mutualism—and whose transmission spans a spectrum—from horizontal to vertical—with varying degrees of partner fidelity. The 'holobiont–hologenome' paradigm posits that host and microbiome evolve as a single unit of collaborative selection, which requires high partner fidelity: achievable for individual host–microbe pairs, but unlikely for entire microbiomes—thus making it unlikely that communities evolve as hologenomes; moreover, abundant empirical evidence shows fitness conflicts among community members, meaning evolution occurs not only at the 'whole' level. The article also provides background: the recent explosion of research in this direction stems precisely from the advent of relatively inexpensive tools for characterizing microbial communities. Partners appear increasingly numerous—not because they were previously invisible, but because seeing them became cheaper (this article’s summary).
## Formalizing the Debate
In April 2023, a paper in *The American Naturalist* built a mathematical theory for this debate. It models horizontally transmitted microbiomes as a genetic system exhibiting 'collective inheritance': the environmental reservoir of microbes corresponds to the gamete pool for nuclear genes, and Poisson sampling from that reservoir corresponds to binomial sampling from the gamete pool. Yet the model simultaneously shows: selection acting at the holobiont level on the microbiome does *not* yield an analogue of the Hardy–Weinberg law, nor does it necessarily fix microbial genes conferring highest holobiont fitness. A microbe that strikes an optimal balance between lowering its own within-host fitness and raising holobiont fitness will be replaced by another microbe identical in all other respects but contributing *nothing* to holobiont fitness—unless the host mounts an immune response against non-beneficial microbes, in which case this replacement is reversed. The paper concludes that host-driven species filtering combined with microbial competition—not coevolution or multilevel selection—is predicted to drive microbiome–host integration.
In July 2025, a forward-looking article in *Cell Host & Microbe* summarizes the current theoretical state as still lacking three pieces: nested ecology within hosts or holobionts, holobiont population dynamics, and symbiont-mediated speciation. The authors of this outlook declare no competing interests; their author list overlaps with that of the 2023 formalization paper (as stated in the abstract; this article draws no inference about stance change). A decade of debate has not hardened into a stalemate of fixed positions, but rather solidified into a clear inventory of theoretical tool gaps (this article’s judgment).
## When Metaphor Outpaces Evidence
Act IV is a retraction. Mycorrhizal fungal hyphae underground connect the root systems of multiple plants—of the same or different species—forming common mycorrhizal networks (CMNs). Recently, this concept has attracted broad audiences in discussions of forest function and forest management. Yet a review published online in *Nature Ecology & Evolution* in February 2023 points out: popular media claims about forest CMNs are disconnected from evidence, and even within the scientific literature, a bias toward citing positive-effect findings has taken hold. The authors systematically examine three popular claims. CMNs are ubiquitous in forests: field studies show excessive variability, alternative explanations exist, or samples are too limited—support is insufficient. Resources transferred via CMNs improve seedling performance: support is likewise insufficient. Mature trees preferentially transmit resources and defense signals to offspring via CMNs: there is no peer-reviewed published evidence. The authors further examine citation practices in this field and find unsupported claims doubled over the past 25 years; biased citation of positive effects may obscure understanding of CMN structure and function. The paper concludes: current knowledge about CMNs is too sparse and uncertain to guide forest management. In the same journal and same month, a correction by the same author group appeared (the abstract does not report correction content; this article makes no claim about it).
This act delivers not new knowledge about symbiosis, but a lesson about how cooperative narratives propagate (this article’s summary).
## What Turns Partners Into One Individual
Laid side-by-side, the four acts yield three criteria for judging 'one individual'—all independent of intimacy. First, transmission fidelity: do partners reliably pass across generations with the host? This was the fulcrum of the 2016 perspective article rejecting the hologenome. Second, mode of transmission and source-pool structure: does the microbiome arise via vertical transmission or environmental sampling? That determines whether it forms a 'collective inheritance'–like genetic system. Third, fitness alignment: are internal conflicts of interest suppressed by mechanisms—for example, host immune recognition and consequent species filtering? Only when all three are satisfied can one speak of 'one individual'; satisfying only one or two leaves the system an ecological community, a tool relationship, or a market relationship (this article’s judgment).
Discussions of human–machine symbiosis frequently borrow biological metaphors, yet the criteria offered by the four acts are cold: who drives filtering, where information is sampled from, and whether interests are mechanistically aligned. For any knowledge community relying on citation and dissemination, Act IV offers a directly transferable lesson: the gap between claims and evidence requires systematic replication and quantification—doubling over 25 years is the empirically measured cost of *not* conducting such systematic replication (this article’s judgment).
## Three Judgments
First, behind all four acts of boundary revision lie new observational or verification methods. The 1967 paper appealed to evidence from cytology, biochemistry, and paleontology; basidiomycete yeasts in the cortex were discovered only after the definition had stood for over 140 years—and linked to phenotypic variation; the explosion of microbiome research stemmed from cheapened characterization tools; the CMN retraction arose from systematic replication and quantification of citations and claims. Thus, the 'individual' question shifts from philosophical speculation to an evidentiary question (this article’s synthesis).
Second, what determines 'one individual' is not intimacy, but the triad of transmission fidelity, transmission mode, and fitness alignment. Two negative papers target precisely the two leaps: from intimacy to interest congruence, and from local fidelity to global fidelity; the 2023 formalization does not elevate wholeness further, but shifts to a mechanistic path: host filtering plus microbial competition. Metaphors do not generate evidence; mechanistic models do (this article’s judgment).
Third, symbiotic narratives carry inherent inflationary gravity. Cooperative stories spread well, and citation systems amplify positive results; unsupported claims doubling over 25 years has already been empirically observed in the CMN literature. The more we rely on cooperative visions—whether in biology or human–machine systems—the more we need systematic replication under uniform standards, not more elegant metaphors (this article’s judgment).
## What Different Readers Can Take Away
For biological researchers: the current shape of the hologenome debate is defined by three theoretical gaps (nested ecology, population dynamics, symbiont-mediated speciation); the gap inventory *is* the topic inventory. For practitioners of human–machine collaboration and multi-agent systems: before applying the term 'symbiosis' to a system, first check three things—whether partners transmit reliably, how the information source pool is sampled, and whether interests are mechanistically aligned; if all three cannot be verified, the system is closer to a tool relationship or market relationship than to 'one individual.' For science communicators and editors: before disseminating cooperative narratives, verify whether claims rest on peer-reviewed evidence—the popular CMN assertions provide a ready-made recent example. For general readers: your body and the forest beneath your feet are both more crowded and more loosely knit than the metaphors suggest; crowded because the number of parties within the boundary exceeds what older paradigms acknowledged, and loosely knit because most of those parties have not yet been proven to 'be one' with you.
Scope and Boundaries: All eight sources cited herein are summary-level materials; full texts were not obtained. Dates reflect PubMed-recorded online or issue dates; the Act IV review is dated February 2023 (online), though its issue date is April 2023. Author names do not appear in the main text; papers are cited by journal (original English name) and year. The 1967 paper is attributed to its institutional affiliation (Department of Biology, Boston University); its subsequent authorial circulation is not elaborated here. The later fates of the endosymbiosis hypothesis components (which parts supported, which revised) fall outside the coverage of this set of materials; the main text reports only the original 1967 abstract’s claims, without asserting their modern status. The phrase 'over 140 years' reflects the age of the lichen paradigm as given in the 2016 paper’s abstract. 'Unsupported claims doubled over the past 25 years' is quoted verbatim from the abstract; this article makes no inference about its counting methodology. The author correction accompanying the Act IV review, published in the same journal and month, is noted only for its publication; its content is not interpreted. Conflicts of interest: The abstracts of the *PLoS Biology* and *Cell Host & Microbe* papers contain declarations of no competing interests, which are reproduced verbatim; other abstracts lack disclosure sections, so this article makes no assertion about presence or absence. Overlap between the author lists of the 2023 formalization paper and the 2025 outlook is reported as stated in the abstracts—only as fact, with no inferred stance change. Translations: 'holobiont' rendered as 'holobiont' (first occurrence with English), 'hologenome' as 'hologenome' (first occurrence with English); 'common mycorrhizal network' introduced with both Chinese and English abbreviation (CMN), thereafter abbreviated.
Sources: *Journal of Theoretical Biology*, March 1967 (PMID 11541392, https://doi.org/10.1016/0022-5193(67)90079-3); *PLoS Biology*, December 4, 2015 (PMID 26636661); *mBio*, March 31, 2016 (PMID 27034285); *Science*, online July 21, 2016; print July 29, 2016 (PMID 27445309); *Nature Ecology & Evolution*, online February 13, 2023; print April 2023 (PMID 36782032); *Nature Ecology & Evolution* author correction, April 2023 (PMID 36922655); *The American Naturalist*, online April 17, 2023; print June 2023 (PMID 37229712); *Cell Host & Microbe*, July 9, 2025 (PMID 40639335)