Microbial evolution and transitions along the parasite–mutualist continuum

Virtually all plants and animals, including humans, are home to symbiotic microorganisms. Symbiotic interactions can be neutral, harmful or have beneficial effects on the host organism. However, growing evidence suggests that microbial symbionts can evolve rapidly, resulting in drastic transitions along the parasite–mutualist continuum. In this Review, we integrate theoretical and empirical findings to discuss the mechanisms underpinning these evolutionary shifts, as well as the ecological drivers and why some host–microorganism interactions may be stuck at the end of the continuum. In addition to having biomedical consequences, understanding the dynamic life of microorganisms reveals how symbioses can shape an organism’s biology and the entire community, particularly in a changing world.


0123456789();:
shifting into novel host taxa is an important process, often forging novel associations on the continuum 30,31 . Transitions can also occur if a parasite's own selfish traits benefit a host as a by product 27 , or by hosts rewarding 32 or capturing 33 symbiont genotypes that confer benefits. Conversely, mutualisms can break down into parasit isms. This breakdown can occur owing to the spread of cheater symbionts, which exploit the benefits of host association without paying the cost of returning a benefit 27,34 . However, shifts from mutualism to parasit ism appear rare in nature 15,29,35 . More frequently, symbi onts leave the host association continuum by reverting to free living environmental lifestyles, as demonstrated by Actinobacteria abandoning ant hosts 15,36 .
In this Review, we discuss the evolutionary transi tions of host-microorganism symbioses along the para site-mutualist continuum, the mechanisms underlying evolutionary changes, the selective pressures involved and common empirical approaches for studying them (Box 1). We also briefly discuss context dependent transitions and the consequences faced by micro organisms when their symbioses are constrained to the extreme ends of the continuum. Moreover, we focus the Review on eukaryotic host-microorganism symbi oses; however, we note that microbial interactions with mobile genetic elements (MGEs) can be analogous to sym bioses (Box 2) given the ability of these elements to confer beneficial traits and cause harm to bacterial hosts 37,38 .
Mechanisms of evolution along the continuum The gradual emergence of microbial mutualists from parasitic ancestors 15,29,31,39 and the rapid leaps in sym biont phenotypes observed in real time [40][41][42][43][44] provide fascinating insights into the proliferation of microbial symbiotic diversity. The genetic changes involved in microbial evolution are key contributors to the forma tion of mutualisms and parasitisms and their transitions along the symbiotic continuum. Mechanisms that result in these changes include, for example, selection on existing genetic variation 45,46 , de novo mutations 40,43,[47][48][49] and genome rearrangements [50][51][52] . Genome rearrange ments include inversions, duplications, translocations and gene loss 50,53,54 (for further discussion of gene loss, see later section, Stuck at the end of the line). Horizontal gene transfer (HGT) events, whereby genetic material moves between organisms in a manner other than vertically, are also important factors in microbial evolutionary transitions 42,[55][56][57][58] . These events often involve MGEs -such as plasmids, transposons, insertion ele ments and phages -coding for traits that are beneficial or harmful to hosts during their interaction.
Shifts between microbial parasitism and mutu alism can involve selection on existing variation. Through experimental evolution of the bacterial sym biont Parachlamydia acanthamoebae and its protist host Acanthamoeba sp., one study 46 observed an evolutionary shift of the microbial symbiont towards parasitism under horizontal transmission conditions. The molecular basis of this transition was a pronounced increase in the fre quency of specific genetic variants within the original symbiont population, alongside marked changes in the expression of machinery necessary for manipulating host cells, such as the type III secretion system (T3SS).
Selection on de novo mutations in bacterial popula tions has also been detected in evolution experiments, resulting in movement along the continuum. In these cases, experiments are started by propagating a sin gle clone in hosts. In one study 40 , a clonal population of Enterococcus faecalis was introduced into nematode host populations, and mutations that arose favoured enhanced production of reactive oxygen species. This phenotype allowed E. faecalis to become highly bene ficial to hosts, as production of these antimicrobials suppressed infection by Staphylococcus aureus. A sim ilar direction of travel, but from parasite to commensal, has been observed in nematode host populations by evolving Pseudomonas aeruginosa from a single clone 49 . Conversely, within the guts of old mice, mutations aris ing in clones of commensal Escherichia coli may have resulted in evolution towards pathogenicity 59 . In com parison with evolution within young mice, mutational targets linked to stress related functions and associ ated with virulence were under strong selection in the inflamed guts of older mice. Mutation might have a Examples from nature of microorganisms transitioning from free-living to hostassociated lifestyles include the evolution of parasitic species in the Bacillus cereus group (for example, the causative agent of anthrax) from soil-dwelling ancestors 237 (part a), and environmental Pantoea bacteria evolving obligate mutualistic roles in stink bug growth and development 16  prominent role in transitions when symbionts have a low initial diversity upon colonization. This situation could occur naturally when symbionts have a low infec tious dose or when transmission causes population bottlenecks (see section on Transmission below).
Wide ranging genetic changes -HGT, gene loss and genome rearrangements -have had a profound role in Yersinia pestis becoming more virulent and adapting to new host species 50,60,61 . Y. pestis is the causative agent of plague in mammalian and arthropod hosts. It is thought to have diverged from its less harmful ancestor Yersinia pseudotuberculosis 1,500-55,000 years ago 62,63 . Sequencing of isolates of the two species revealed that both HGT and insertion sequence mediated genome rearrangements and deletions facilitated Y. pestis evolution 50,60,61 . The bacterium acquired two plasmids, namely pMT1 and pPCP1, making it more virulent compared with its Y. pseudotuberculosis ancestor. The former plasmid carries the ymt gene encoding Yersinia murine toxin, required for the colonization of the flea host 64,65 , and the capsular antigen fraction 1, which inhibits phagocytosis 65,66 . These acquisitions contributed to the evolution of Y. pestis towards greater virulence. Adaptation of the parasite to new hosts was mediated by genome rearrangements, particularly via insertion sequences and gene loss. Gene loss was crucial in reduc ing the toxicity of Y. pestis to the flea vector, allowing biofilm to develop in the flea foregut 67 . Gene disruption by insertion sequences, in combination with deletion events, point mutations and frameshifts, further cre ated an extensive number of pseudogenes within the Y. pestis genome 50,60,61 . Altogether, these genetic changes facilitated a shift in lifestyle, from a less harmful mam malian enteropathogen to systemic pathogen of both mammalian and arthropod hosts.
Infection by various phages (mostly lytic, λ like phages) along with other MGEs facilitated the diver gence of the highly pathogenic enterohaemorrhagic E. coli strain O157 Sakai from its ancestor. The com mensal E. coli strain K12 is also descended from this common ancestor 68 . In strain O157 Sakai, prophages and prophage like elements encode a variety of virulence related genes -adhesins, tellurite resistance genes and urease -contributing to the acquisition of virulence factors that have determined this bacterium's trajectory towards increased virulence in humans. One of these elements also encodes the major virulence fac tor, the locus of enterocyte effacement (LEE), which is responsible for bacterial attachment followed by devel opment of the disease causing effacing lesions in the intestine 69 . Lambda like phages on the Sakai chromo some also encode the destructive Shiga toxin, as well as proteins involved in serum resistance and cell adhesion. Having become integral to the organism's virulence in this way, the prophages themselves have transitioned from parasitic to mutualistic elements within the O157 Sakai genome (for further discussion of MGEs as symbionts, see Box 2).
How commonly do shifts across the continuum occur owing to de novo mutation or machinery acquired by HGT? Host environments with complex, often open, microbial communities, such as the mammalian gut, might generate more extensive opportunities for HGT [70][71][72] . For example, phage driven HGT from the resident community can dictate the evolution of invad ing strains 73 and instigate change more rapidly than is achievable by mutation accumulation 74 . HGT has had a considerable role in major evolutionary transi tions of living organisms; it is increasingly confirmed as a dominant force in the evolution of host-symbiont associations 20,29,54,58,65,[75][76][77][78][79][80] . Yet, for symbionts nested within simple microbial communities (for example, intracellular environments), scarce opportunities for HGT may mean de novo mutation is more likely to underpin shifts along the continuum. Studies reporting selection on de novo mutation during transitions 40,49,59 highlight the power of this genetic means to generate remarkable change on the continuum. These experi ments typically involve a small number of microbial species and/or low levels of initial genetic diversity upon colonization. When incorporating a host background

Virulence
The damage caused to the host due to infection by a parasite, often measured as a reduction in host fitness.

Box 1 | Two approaches to evaluating evolution along the parasite-mutualist continuum
Phylogenetic inference there are challenges to judging transitions in symbiosis because ancestral partnerships no longer exist for direct comparison. Interactions that now appear mutualistic may actually reflect the result of a long period of conflict resolution or the evolution of tolerance by the host. Phylogenetic inference can shed light on the evolutionary history of transitions on the parasite-mutualist continuum. techniques such as ancestral state reconstruction and its extensions infer characteristics of ancestral taxa based on traits exhibited by extant descendants 240 . In this way, symbiotic phenotypes of ancestors (for example, parasite, mutualist, commensal or free-living) can be recovered and used to infer the origins and breakdowns of associations on the continuum, in addition to the rate of such transitions 29 . Such approaches are heavily contingent on the quality of the underlying phylogenetic tree, and reconstruction accuracy declines with increasing evolutionary time 240 . However, for many lineages of bacterial symbionts this approach has been used powerfully to demonstrate the marked rarity of reversions from mutualism to parasitism over evolutionary timescales 15,29 .
Experimental evolution experimental evolution permits the direct testing of hypotheses related to the tempo and pattern of the evolution of species interactions. this approach allows for evolution to be observed in real time. An added advantage in some systems is an ability to cryopreserve the eukaryotic host (for example, Caenorhabditis elegans 241 and Paramecium bursaria 159 ) and associated microbial lineages for subsequent analysis. this characteristic allows the fitness benefit or harm for both species to be compared with past and future archived generations, for example, via time shift assays 242 .
In an evolution experiment, the source of selection can be hypothesized and manipulated. For example, this approach could be used to determine whether the presence or absence of an enemy could affect the position of a defensive symbiosis along the continuum 40 , as well as whether the evolution of the eukaryotic host or the microbial symbiont, or their coevolution was responsible for the shift 218 . Subsets of the population can be used to establish the next generation. one focal species can be evolved and others kept in evolutionary stasis by adding from an ancestral population each generation. Alternatively, additional community members can be reciprocally evolved, opening the arena for coevolutionary dynamics between two or more species 243 . the process continues for generations. At the end, phenotypic and genomic comparisons can be made between ancestral and evolved populations, and also across replicates, to assess convergence or divergence in transition outcome and the genetic basis.
candidate molecular targets in evolved lineages can be identified for manipulation and further experimentation. Moreover, follow-on genomic analysis can be powerfully combined with phenotypic assays across evolutionary time to identify the mechanism of relative benefit or cost for each species, as well as to confirm phenotypic traits under selection. one caveat is that experimental evolution might be less likely to yield increases in parasite virulence given the potential for breaking apart of the virulencetransmission trade-off at passage points 244 . with an ecologically relevant microbiota, HGT might be more dominant.
Drivers of evolution along the continuum Ecological sources of selection can drive microbial symbiont evolution towards increasing host benefits (TABle 1) or harm (TABle 2). Shifts occur across genera tions as microbial symbionts adapt to life in a new host species, encounter different transmission opportunities and face hosts that reciprocally evolve in response. The presence or absence of additional interacting species in the community can also drive evolutionary change in a host-symbiont relationship owing to changing distri bution of net benefits and costs across the community. Essentially, given a strong source of selection, genetic change can occur within just a handful of microbial gen erations. These transitions are often investigated using experimental evolution or over macro evolutionary timescales via phylogenetic comparisons (Box 1).

Novel hosts.
Microorganisms frequently encounter novel host environments. They can jump across species boundaries or colonize hosts from pools of free living environmental microorganisms. Novel infections can generate new diversity on the symbiosis continuum through divergence and speciation 81 . High profile cases of host shifts, such as the recent SARS CoV2 pandemic 82 , highlight the potential for investigating evolutionary changes in virulence upon emergence [83][84][85] . New associations are often maladaptive for both host and parasite 86 , and associations can move unpredictably on the continuum or burn out. This trajectory has been observed in emergences of avian influenza virus, where case fatality rates can be high but human to human transmission is low 87 .
Shifts between host species, possibly driven by HGT of virulence associated genes, appear to have been impor tant in the emergence of the Q fever parasite, Coxiella burnetii 30,88 . This proposed mutualist to parasite tran sition is a complex case for which the full evolutionary story remains unknown. However, phylogenetic analysis suggests that this highly infectious bacterium recently emerged from a clade of vertically transmitted mutu alistic endosymbionts of ticks 30 . C. burnetii may have evolved mechanisms to infect vertebrate cells, persist in the environment and be airborne transmitted. These

Box 2 | Mobile genetic elements as symbionts
Mobile genetic elements (MGEs) can cause genomic change in their microbial hosts. these changes can affect the position of the microorganism-eukaryotic host relationship on the parasite-mutualist continuum by coding for traits that harm or benefit microbial hosts. on a smaller scale, mGes are analogous to symbionts 37,38 as they are entities with their own evolutionary interests that can parasitize hosts or confer beneficial traits that promote innovation. the effects they have on microbial host fitness can change.
many nosocomial pathogens have acquired antibiotic resistance genes through horizontal gene transfer 245 , gaining a survival advantage in the presence of certain antibiotics. In the absence of the corresponding antibiotic, however, a resistance-conferring MGE can become costly to its host. For example, when large low-copy-number plasmids are cumbersome to their host, these plasmids force their maintenance through the action of resolution systems, partitioning systems and post-segregational killing systems. The latter of these includes toxinantitoxin systems, encoding both a stable protein toxin and a less stable, but more abundant antitoxin. If a plasmid fails to be inherited by a daughter cell, the antitoxin will rapidly degrade in the host, leaving it susceptible to being killed by the toxin 246 . the transition of mGes from beneficial elements conferring a survival advantage to parasites can take place over very short evolutionary timescales. In turn, in the face of antibiotic treatment and other clinical interventions, mGes can drive the evolution of their bacterial hosts towards higher virulence over an equally short period of time 58,247 .
mGes are not always maintained through natural selection. the genome of Wolbachia pipientis wmel, an obligate intracellular symbiont of the fruitfly Drosophila melanogaster, is highly streamlined from extensive gene loss during adaptation to its host; however, it is also overrun with mGes 248 . repeated population bottlenecks resulting in genetic drift and inefficient natural selection 248 likely contributes to the extensive maintenance of mGes in this genome and those of other heritable symbionts 249 . these elements may have contributed to the substantial phenotypic diversity among Wolbachia strains, fundamentally shaping Wolbachia evolution 248 . In this instance, mGes are parasitic elements maintained within the population effectively by accident via transmission of Wolbachia from one host to the next. ultimately, it is unclear whether these elements will cross the parasite-mutualist continuum and become permanent components of Wolbachia genomes.
For some microbial hosts, the acquisition of deleterious mGes can be partially rescued via compensatory evolution, leading to a type of host tolerance. In such cases, the association is maintained but the host ameliorates the cost, as shown for Pseudomonas fluorescens and a megaplasmid conferring mercury resistance. In low-mercury environments, the plasmid is costly, yet experimental evolution across a mercury gradient showed P. fluorescens consistently compensated via mutation in the gacA-gacS two-component system, downregulating chromosomal and plasmid gene expression and relieving translational cost 44 . Such compensatory evolution may also explain the persistence of contextdependent mutualisms in environments where they do not benefit hosts.
mGes can also become 'immortalized' in host lineages. once genomic parasites, they can become indispensable components of the host genome that are ultimately passed on to daughter cells. vestigial mGes in the form of cryptic phages, ancient regions of viral DNA and disrupted transposon sequences or pseudogenes can be found immortalized in the genomes of organisms throughout the tree of life 250 . bacterial chromosomes, for example, can contain as much as 20% phage DNA 251,252 . once parasites to their hosts, these mGes have infected the genomes of host organisms, maintained their stability as they coevolve with their host (forcibly in some cases, for example, toxin-antitoxin systems) and finally been irreversibly integrated into the genome. Integration can occur by accident during genome rearrangements, recombination, population bottlenecks and speciation events 248 , or by natural selection because of a fitness benefit on which the host has become dependent 68 . the ubiquitous presence of vestigial viral DNA in the cells of all organisms 250 is a prime example, demonstrating how mGes have been formative in the evolution of organisms, just like many eukaryotic host-microbial symbioses. mGes leave behind remnants of DNA in host genomes like partial segments of an ancient diary. mGes therefore possess the capability themselves to go from genomic parasites to mutualistic or commensal components of the genome. In many situations, this process can also drive the evolution of their bacterial hosts along the continuum. mGes have forcibly maintained their interaction with bacteria in some cases, while in others, their maintenance has been a by-product of environmental conditions or population bottlenecks. they represent fascinating examples of entities that can be both effectors and subjects of evolutionary transitions along the parasite-mutualist continuum. traits are unlikely to be found in the arthropod restricted ancestors 30 . Ticks feeding on vertebrates likely provided the ecological bridge. Similar transitions occurred within Sodalis allied symbionts, a group of host restricted bac teria common to insects including the tsetse fly vector. A free living Sodalis sp. was isolated after a person suffered a wound from a tree branch, and this seren dipitous finding provided evidence that symbiont lin eages emerged from environmental ancestors 31 . Early vectoring of these environmental strains by insects was likely pivotal in the evolution of the beneficial, heritable Sodalis endosymbionts observed today. Novel species interactions can drive rapid innovation. This might particularly be the case if a microorganism bears characteristics that can provide instant benefits. Microorganisms encoding functions of light generation, photosynthesis, nitrogen fixation or antimicrobials may provide such rapid benefits 15 . These characteristics may be  remodelled (or act as pre adaptations) for transitions in symbiosis 15 . Such repurposing may have occurred in the antifungal producing Burkholderia symbionts associated with Lagriinae beetles. Burkholderia symbionts appear to have transitioned from a plant parasite to insect mutual ist. In this context, secondary metabolites previously used as virulence factors against plants may have been repur posed for antifungal defence on beetle eggs 89  evidence comes from marine hosts, including within the bulbs of anglerfish and the Vibrio fischeri filled light organs of bobtail squid. These hosts benefit from these bioluminescent bacteria to lure prey and avoid predation, respectively, and the symbionts often retain the capacity to live freely, or persist in the environment 22,51 .

. Additional
Transmission opportunities. Transmission mode has been considered to predict the direction of a symbiont's evolution on the continuum. When horizontally trans mitted symbionts can move between unrelated host individuals, the fitness interests between species are uncoupled, a scenario thought to favour parasitism 7 . The degree of harm caused to hosts from infection is often framed by the virulence-transmission trade off 90,91 . The relationship assumes that virulence -the reduction in host fitness caused by parasite infection -is costly to the parasite as host resources are needed for replication 92 .
The cost of harming the host too much or too soon from replication might result in less transmission. Thus, it is predicted that transmission should be highest at inter mediate virulence, which balances the costs of withinhost replication and infectious period length 90 . This model is particularly pertinent for symbionts that rely on a mobile host for transmission (for example, socially transmitted microorganisms). Those that do not (for example, vector and water borne microorganisms) can bypass trade offs between virulence and transmission 91 . This conventional model goes some way to hypoth esizing on general patterns of virulence, yet several extensions and alternatives have been suggested [93][94][95] . It has been suggested that mutualists may evolve from parasitic ancestors when the frequency of hori zontal transmission routes is reduced or lost 7 . If verti cal transmission is the remaining dominant mode of transmission then the fitness of host and symbiont can become tightly coupled, reducing the arena for evolu tionary conflict and thereby favouring selection for mutual benefit 7,90,96 . Mutualisms involving symbiont inheritance are predicted to be stable on the continuum and unlikely to revert to parasitism 15,97 . But exclusively vertical transmission can endanger associations via genetic bottlenecking (see section on Stuck at the end of the line). Clearly, becoming inherited is not the sole route by which bacterial mutualists evolve. Comparative analysis has found no evidence for vertical transmission preceding the origin of mutualism 15 . Many mutualisms involve horizontal transmission such as conjugative plas mids in bacterial populations 98 and the vast networks of mycorrhizae that improve plant productivity 99,100 . In particular, evolution of defensive traits in symbionts are proposed to be facilitated by the genetic diversity and selection for innovation promoted by horizontal transmission 101 . Many horizontally transmitted micro bial symbionts are obligate for host fitness 16,22,102 , but many can be facultative 24 and confer costs in different environments.
Conversely, not all inherited microorganisms become mutualists 103 . Wolbachia, Spiroplasma and Arsenophonus species are common inherited parasites that manipulate host reproduction, maximizing resource allocation to the transmitting host sex (females) by feminizing hosts or killing their sons 104 . However, theory suggests that the spread of such reproductive parasites will be enhanced by the evolution of traits that benefit hosts 105 . A ben eficial trait (that is, defence) may even interact with a parasitic trait (that is, reproductive manipulation) to completely exclude a natural enemy 105 . Indeed, cryptic benefits are now found in several systems 106,107 , and there is evidence that some reproductive parasites may need to also transmit horizontally just to persist 108 .
Transmission as a determinant of the location of a symbiosis along the continuum is complex. There are numerous exceptions to classical theory. Nonetheless, experimental manipulation of transmission modes finds general support for the theory that horizontal transmis sion can select for parasitism and vertical transmission for reduced antagonism (TABle 1; TABle 2). In a symbiosis between a jellyfish and the alga Symbiodinium microadriaticum, cooperative traits, including growth enhance ment, were selected when transmission was restricted to heritable routes 109 . Such cooperative traits are funda mental for stable mutualisms, protecting against tran sitions to parasitism or abandonment events. In the reverse experiment, restriction of the alga to horizontal transmission selected for faster proliferation and disper sal (traits associated with parasitism), and declines in host fitness were detected 109 . Such findings are mirrored across terrestrial systems 46,110,111 . The common pill bug hosts a Wolbachia strain (wVulC) that feminizes genetic males 112 . Blocking the typical vertical route, and mim icking horizontal transmission, saw systemic increases in Wolbachia (wVulC) density and a drastic transition from a benign partner to a highly virulent one 110 .
The community. The drivers of transitions along the parasite-mutualist continuum can be complex and stem from the ecological and evolutionary movements of many different players. Defensive symbiosis 113,114 , whereby there are at least three interacting species (host, defen sive symbiont and an attacking enemy) is particularly dynamic along the continuum in response to commu nity composition changes. The absence of the symbi ont or enemy can have evolutionary consequences for other species in the community, even without direct interactions 115,116 . Co infections in hosts can also influ ence transitions in the symbiosis by providing new phenotypes via HGT of genetic material (for example, symbiosis islands, plasmids and phages) 78,80,114,117 .
The impact of community complexity is demon strated by the bacterium Hamiltonella defensa and its lysogenic phage, APSE. This association protects host aphids against parasitoid wasps 118,119 (Fig. 2). In this context, the fitness benefit afforded to the aphid host is contingent on parasitoid presence -in its absence, H. defensa with APSE phage is costly to the aphid 120 . The mechanism of protection (toxin production) hinges on the initial lateral transfer of phage from a co infecting symbiont 117,121 . Subsequent loss of the phage can move the interaction between H. defensa and aphids back towards parasitism 122 . Theory 105,116 , experi mental evolution 40 and field studies 123 have captured how microorganisms, even parasitic ones, can evolve rapidly to protect their hosts when collectively threatened, often

Defensive symbiosis
An interaction in which the symbiont protects the host (via direct or indirect mechanisms) against natural enemies, such as microbial parasites and eukaryotic parasitoids.
crossing the parasite-mutualist continuum in the pro cess. In Caenorhabditis elegans nematodes, a mildly par asitic gut bacterium was shown to evolve an enhanced ability to protect against infection by a more virulent parasite 40 . In the parasite's absence, the gut bacterium did not emerge as a microbial line of host defence.
Additional symbionts, with previously unknown effects, are increasingly being identified even in iconic 'two player' symbioses, such as corals 124 and lichens 125,126 . It is thus not surprising that the complexity of a host's whole microbiota (which often includes a diverse reper toire of bacteria, fungi and viruses) can interact to pro duce new outcomes for individual strains, species and the community as a whole. Members of the microbiota compete and cooperate in a myriad of ways 127 , influenc ing the virulence of one another via processes such as the suppression of public goods 128 or the facilitation of biofilm formation 129 and epithelial translocation 130 . The passage of Candida albicans in mice lacking gut micro biota has highlighted the role of communities in deter mining fate on the parasite-mutualist continuum. In the absence of a gut microbiota, C. albicans mutants emerge that are defective in hyphal formation, no longer requir ing it for competition against other microbiota members. When compared with the wild type ancestor that coex ists with a microbiota, these C. albicans mutants are less virulent and protect their hosts against Aspergillus fumigatus infection in a manner independent of host adaptive immunity 43 . This transition from pathobiont to condi tional mutualist in this context appears to hinge on the absence of competing microorganisms. However, given a gradient of increasing microbiome diversity, it would be valuable to understand when the selective advan tage of the transition disappears. Other recent work, in microbiota free mice, noted that when E. coli is a lone colonizer of the gut, it is consistently selected to increase metabolism of amino acids serine and threonine. A small increase in microbiome diversity (the addition of a single competing species) alters the evolutionary trajec tory of E. coli substantially, instead favouring mutations associated with anaerobic metabolism 131 . This outcome suggests that bacteria may have low fidelity in metabolic function even within a single host generation 132 . Such a finding suggests host-microbial symbioses may not adhere to the idea of the 'holobiont' being a cohesive unit of selection 133 . This idea relies on high fidelity between partners 134 , which may easily be disrupted by changes to the surrounding microbial community.
If we can selectively drive the evolution of micro organisms and their communities, applications may improve on the already promising use of faecal micro biota transplants in medicine 135 , symbiont mediated vec tor control 136,137 and the manipulation of crop parasites 42 . There is, however, a pressing need to understand the long term response of microbial communities to the engineering of symbionts. Recently, theoretical models have treated virulence as a cost shared by all symbionts coexisting in a host 138,139 . These models find that defence by a symbiont often drives reduced viru lence across the microbial community (including in attacking parasites), an outcome dependent on the cost of defence being low and the shared cost of virulence also being low 139 . However, defensive microorganisms may also select for resistance mechanisms (for exam ple, toxin production and inflammatory stimulation) in the parasites they protect against, causing collat eral damage to hosts and driving increased parasite virulence 140 . This is akin to established predictions for co infecting parasite species, whereby competi tion selects for increased virulence [141][142][143] . Promisingly though, and in line with some theory 138,139 , selection for reduced parasite virulence has been revealed in response to microorganism mediated protection 144 . Others also report long term efficacy of protection mechanisms despite an evolving pathosphere 145 .
Host control. Beyond microbial symbiont evolution, hosts can affect the position of the symbiosis on the continuum 146 . Hosts can be resistant (that is, reducing symbiont colonization) and tolerant (that is, coping with Pathobiont Any organism that can cause harm to its host, but normally lives as a harmless symbiont. H evolves resistance/tolerance to P , reducing benefit of DM

MGE loss renders DM parasitic and P virulent
Greater protection by DM alleviates cost of P to H Greater virulence of P strengthens relative benefit of DM to H P absence or loss of virulence renders DM parasitic Loss of DM leaves H exposed to full virulence of P For example, if a MGE encodes key protective functions, then its loss (move 2) will shift the DM's position towards parasitism (all cost and no benefit to host). Meanwhile, the costs of P to H will increase now that H is no longer protected by the DM and its MGE. Transitions here can also alter the coevolutionary patterns and processes between players and species. symbiont associated damage without limiting coloniza tion) 147 , which reduces any negative impacts of the hostsymbiont interaction. Evolving control mechanisms (for example, sanctions and rewards, and microbiome mod ulators) 146,148 , or acquiring symbiotic function from an alternative source (for example, symbiont switching and HGT) 100 can also limit or cause a change in the position of the interaction along the continuum.
Resistance to symbiont infection is observed ubiq uitously across evolving host-parasite associations 149,150 . Mutations associated with membrane transporters in the bacterium Actinomyces odontolyticus coincided with a reduction in the negative effects of its ectoparasite (Nanosynbacter lyticus) 151 , perhaps indicating an adap tive host response to block resources to the ectopara site or prevent its attachment 151 . As host resistance and tolerance strategies can affect parasitic symbiont fitness, they can counter adapt 152,153 . This process may lead to a repeated back and forth along the continuum.
Hosts can also have key roles in restraining symbiont driven shifts along the continuum. They may act to prevent the emergence of cheating symbionts, which exploit the benefits of host association without paying the cost of returning a benefit 27,34 . Alternatively, hosts may maintain the association at a position opti mal for their own fitness. Sanction and reward strate gies, spatial segregation of symbionts and partner choice mechanisms have evolved to promote and maintain cooperation 27,154,155 . For instance, legumes may sanction defective nitrogen fixing bacteria by blocking resources to the respective root nodule 32,154 , and plants reward helpful mycorrhizal fungi with extra carbohydrate 156 . These mechanisms protect the host from investing in symbionts with net costs and avoid trajectories towards antagonism.
There is mounting theoretical and empirical evidence that many putative mutualisms may actually be a prod uct of hosts exploiting symbionts [2][3][4]33 . Interactions can benefit the host, but with no reciprocity to the symbi ont whose fitness is markedly reduced within the walls of host confinement 1 . These may be viewed as cases of inverted parasitism 5 . The host is the parasite of its smaller guest. This phenomenon is exemplified by zoox anthellae in which replication rates are severely compro mised by host association 4 , rising from 3 days outside of coral hosts 157 to around 70 days within 158 . Another example comes from Paramecium bursaria and photo synthetic Chlorella symbionts. Chlorella species provide fixed carbon in return for organic nitrogen, but the host tightly controls symbiont density in response to light conditions, ensuring the best nutrient trade for itself 159 . Control of the symbiont potentially occurs via digestion of Chlorella cells 160 . The host may win twofold, paying the workforce only when required and acquiring nutrition via digestion of surplus symbionts. The growth rate for Chlorella remains consistently better outside the host 159 , but inside, this symbiont avoids algal competitors 161 and may be protected against its own parasites 162 . Research on exploitation by hosts is in its infancy, with the greatest evidence coming from interactions with photosynthetic symbionts 4,159,163 . Many questions remain, including the ubiquity of the phenomenon and whether some classes of symbiont are more vulnerable to exploitation than others.
Although considered relatively rare over evolutionary time, hosts may also eschew parasitic 164 and mutualistic associations 100 . Fleeing the infectious environment is one strategy. Spatiotemporal escape by asexual rotifers prevents them interacting with fungal parasites consist ently over evolutionary time. By drying up and blow ing away in the wind, these animals are protected from infection, which allows them to maintain their asexual reproductive strategy 164 . Mutualistic associations can be abandoned via the recruitment of new symbionts 100 . As the Hodgkinia endosymbionts of cicadas teetered on the edge of genomic collapse, Ophiocordyceps fungi (com monly parasites) began to take over the essential roles in amino acid synthesis for the host 165 . Abandonment can also occur via exploitation of an alternative resource 100 . For example, the evolution of carnivory in plants led to several plant species deserting arbuscular mycorrhizal fungal symbionts, as the plant now gains nutrients directly from prey 100 . These cases chime with a grow ing debate over whether hosts can have the upper hand in symbioses, despite generally being the species that evolves more slowly (known as the Red King effect 166,167 ), exploiting and imprisoning their microorganisms to gain disproportionate control and benefit [2][3][4]33,159,168 .

Context-dependent shifts
The outcome of many microbial interactions with hosts are context dependent 14 . Both facultative and obligate symbioses can make shifts along the parasite-mutualist continuum that do not involve evolution, often occur ring within a generation and driven by ecological change or opportunity (TABle 3). Abiotic factors such as temperature 169 , resource availability 170 , environmental toxicity 171 and the biotic composition of the surrounding community 119 or host ontogeny 172,173 can all affect the dis tribution of costs and benefits incurred by the host and microbial symbiont. The position on the continuum can also change if the microbial symbiont becomes infected with its own symbionts (for example, phages and myco viruses) 42,122 . Here, we focus on short term disruptions to host-symbiont associations, but note that sustained alterations to context will feed back to evolutionary change for the interacting species.
Generally, theory predicts that nutrient limited envi ronments, or other harsh environments, can foster ben eficial interactions between compatible players 27,174 via mechanisms such as cross protection and cross feeding. This outcome has been substantiated by empirical work [175][176][177] . For symbionts that have nutritional roles (for example, vitamin synthesis and nitrogen fixa tion), abundant resources can substantially undermine the net benefit gained by the host. The provisioning of mineral nitrogen from fertilizer erases the benefit Bradyrhizobium symbionts provide to legume hosts (Lotus strigosusas) as this acquisition route is less ener getically costly for the legume than its symbiont fixed equivalent 178 . Some hosts evade context dependent costs by divesting themselves of associations when ecological conditions change, such as the phytoplank ton that abandon their nitrogen fixing cyanobacteria when environmental nitrogen is abundant 179 . For hostparasite systems, there is no evidence for a one way effect of nutrient availability to hosts on the harm caused by infection 180 . One study 180 suggested that the level of parasite virulence in a given environment is likely the result of a balance between the effect of host nutrition on the immune system and on parasite resources.
Temperature can affect symbiont phenotypes 181,182 , which directly impact symbiont virulence or benefit, such as the regulation of toxin production 183 or mole cules required for nutrient scavenging 184 . Some obli gate mutualists can constitute thermally 'weak links' for hosts, becoming non functional or even lost from hosts outside adapted temperature ranges, which can have cata strophic consequences for host fitness 185,186 . Interactions can occur between abiotic and biotic factors. For instance, a 5 °C increase in temperature diminishes H. defensa mediated defence against parasitoids 187,188 . This temperature dependent reduction in defence may be ameliorated if co infection with an additional bacterium, known as pea aphid X type symbiont, occurs 187 .
In other cases, community composition alone can temporarily cause transitions. Defensive symbioses present a clear demonstration of community context dependent shifts, whereby benefits to the host are con tingent on the presence of an enemy species 113,114 . In the absence of the enemy, the host pays the cost with no detectable benefit, and the association moves towards one that is parasitic 114,189 . Infection of a symbiont with its own symbionts (that is, hyperparasitism 190 ) can also generate transitions. Recent work found that the devas tating effects of a fungal parasite on rapeseed crop are significantly reduced if the fungus becomes infected with mycovirus SsHADV1 (reF. 42 ). The presence of the mycovirus appeared to affect the expression of a suite of both fungal and crop genes, including those encoding plant cell wall degrading enzymes and crop signalling pathways 42 .
Pathobionts provide an excellent example of context dependent transitions from neutral to harm ful agents 191 . In a host with a functional immune sys tem and healthy microbiota, pathobionts can exist as commensals [191][192][193] . Pathobionts are well adapted to pro liferate beyond their normal niche. During dysbiosis (for example, compromised immunity, disruption of the microbiota or introduction of medical devices such as catheters or surgical implants) pathobionts can cause disease in a wide variety of forms, from minor infec tions to more serious chronic or invasive disease 194 . This ability to transition from harmless to harmful in different contexts makes pathobionts hard to place on the continuum. They are neither consistent parasites nor consistent commensals, with the state of the host generally determining their transition from one to the other.
Stuck at the end of the line At either end of the continuum lie the extremes of host killing (or castration) and mutual dependence. What maintains an association here, and what is its future? The ability to shift along the continuum for some para sitic microorganisms could depend on transmission route. Some infectious agents may stay hypervirulent owing to a high degree of environmental transmission or a lack of reliance on hosts to transmit and propagate. The 'curse of the Pharoah' hypothesis 195 posits that micro organisms able to 'sit and wait' in the environment can be perpet ual killers, whereas others suggest that traits that enable persistence in the environment will be traded off with virulence 196 . There may also be constraints of the parasitic life cycle that prevent a transition. Microbial para sites that must lyse host cells to transmit (for example, lytic phages and Plasmodium species in mammals) or steal resources in a way that castrates the host (Pasteuria bacterial parasite infecting Daphnia magna 197 ) are systems in which transitions away from antagonism are unlikely.
At the opposing end of the continuum lie inherited, obligate endosymbionts, which often have nutritional roles. Although many of these associations are ancient and bestow mutual benefits, they can be risky, particu larly for the endosymbiont 3,53,198 . The genomes of these symbionts can gradually decay as transmission bottle necks allow deleterious mutations to become fixed by genetic drift, and mutational bias towards deletions removes genes [199][200][201][202] . Genomic decay can lead to extinc tion, unless heightened genetic and cellular support is provided by the host 203 or other symbionts 78,204,205 . For example, leafhoppers show gene expression patterns that appear tailored to the deficiencies of each of their endosymbionts' highly degraded genomes 203 . In rare cases, symbionts may transition to organelle status 206 , notoriously achieved by mitochondria and plastids, but this does not guarantee shelter against further gene loss or extinction 207,208 . Hosts may also avoid extinction alongside an endosymbiont by exploiting alternative nutritional resources or gaining new symbionts 158,159 .
Conclusions and future perspectives Plants and animals, including humans, are colonized by innumerable microorganisms. This observation has sparked a revolution in studying the impacts of those microorganisms on host biology and health. Many more examples of microbial evolution causing tran sitions across the parasite-mutualist continuum will emerge through further research using experimental evolution and investigating the microbiome in an evo lutionary context. The potential evolution of species in the human microbiome from good to bad 209,210 , and the degree to which beneficial interactions could be upset by microbiome perturbation 211 , are of clinical relevance for individuals vulnerable to infectious disease. In the future, such individuals may benefit from engineering of the microbiome or symbiont communities, via either direct genetic modifications to key transition loci in microbiome members, or exposure to selection sources with known outcomes. This approach has recently been achieved for honeybees, with the genetic modification of a core gut bacterium improving resistance to viral infection 212 . These are exciting applications, but we must strive to understand the evolutionary consequences for the parasites targeted too.
More fundamentally, understanding causes of tran sitions will provide insight into the dynamics of how an organism's biology and its community are shaped by microbial inhabitants. The ecological and evolutionary transitions of other species, as well as environmental change, can alter the scope for conflict in symbioses involving microorganisms. Interest has grown in think ing of host-microorganism symbioses as holobionts with highly aligned selective interests 134 . Many associ ations may be also viewed in an ecological community context 13,146 in which constant shifts occur back and forth on the parasite-mutualist continuum. The degree to which the host and symbiont, or both, have control over those shifts remains relatively unexplored. Research in the field has focused on the propensity of symbionts to invade unwilling hosts or cheat reciprocal arrange ments. Yet an exciting new avenue is emerging, one that is exposing hosts as exploiters and imprisoners of microorganisms 33,198 . The extent to which microorgan isms are able to evolve to counter or take advantage of that exploitation is also unclear.
Moreover, environmental changes have the poten tial to substantially alter selection in symbiotic interactions 213 . In addition to altering established symbi oses, marked changes to abiotic variables can also move the boundaries of environmental constraint, fostering the evolution of new interactions on the continuum that were previously impossible or profitless. How will the collectively growing impact of humans affect the stabil ity of beneficial associations and the emergence of para sites globally (for example, see reFS 214,215 )? This question is particularly timely given the COVID19 pandemic. Undoubtedly, as environmental perturbations increase in magnitude and frequency, and as the use of antimicro bials grows, understanding the effects on the real time evolution of host-symbiont interactions will become more and more valuable.
Published online 19 April 2021