Adaptive radiations in natural populations of prokaryotes: innovation is key

Abstract Prokaryote diversity makes up most of the tree of life and is crucial to the functioning of the biosphere and human health. However, the patterns and mechanisms of prokaryote diversification have received relatively little attention compared to animals and plants. Adaptive radiation, the rapid diversification of an ancestor species into multiple ecologically divergent species, is a fundamental process by which macrobiological diversity is generated. Here, we discuss whether ecological opportunity could lead to similar bursts of diversification in bacteria. We explore how adaptive radiations in prokaryotes can be kickstarted by horizontally acquired key innovations allowing lineages to invade new niche space that subsequently is partitioned among diversifying specialist descendants. We discuss how novel adaptive zones are colonized and exploited after the evolution of a key innovation and whether certain types of are more prone to adaptive radiation. Radiation into niche specialists does not necessarily lead to speciation in bacteria when barriers to recombination are absent. We propose that in this scenario, niche-specific genes could accumulate within a single lineage, leading to the evolution of an open pangenome.


Introduction
A centr al c hallenge in e volutionary biology and ecology is explaining why species richness patterns in the Tree of Life vary dr asticall y between differ ent taxa (Sc holl andWiens 2016 , Mooers andHeard 1997 ).Differences in species richness are evident in many plant and animal sister clades; compare for example the lone species of Hoatzin (Order Opisthocomiformes) with the 5000 + species of passerines (Order Passeriformes).In eukaryotic taxa, suc h v ariation in species richness has long been interrogated using analyses of phylogenetic tree shape.Ho w ever, whether similar heterogeneity exists in bacteria and archaea has received less attention (Dykhuizen 1998 ).This is partly because the study of bacterial biodiversity faces two major challenges .T he first challenge is that most taxa are undersampled, hindering accurate estimates of species diversity (Quince et al. 2008 ) and phylogenetic reconstruction (Heath et al. 2008 ).As a result, estimates of total bacterial diversity vary wildly, from ∼10 4 (Mora et al. 2011 ), via ∼10 6 (Yarza et al. 2014, Louca et al. 2019 ), ∼10 9 (Larsen et al. 2017 ) to ∼10 12 species (Locey and Lennon 2016 ).Of course, estimates of species richness at least to some extent rely on how species are defined in the first place .T he second challenge is that there is no one-to-one a gr eement between curr ent taxonomy, species delineated based on ov er all genomic distance, or operational taxonomic units (OTUs) based on clustering of 16S rRNA sequences (Parks et al. 2018 ).Differential sampling effort and inconsistent taxonomy must mean that some of the observed intertaxon differences in bacterial species richness must be artefactual.These caveats notwithstanding, it is clear that there are substantial differences in species richness when surveying either named species or 16S amplicon-based OTUs (Fig. 1 ).
Numerous explanations for differences in species richness have been put forw ar d but many of these, such as the effect of tr ophic le v el, body size, geogr a phic r ange, latitude, or temper atur e (Hutc hinson 1959, Rosenzweig 1995, Dykhuizen 1998 ), do not necessaril y tr anslate to prokaryotes (e.g.Bahram et al. 2018 ).Ho w e v er, r easoning fr om first principles, species ric hness, be it in animals , plants , or bacteria, is ultimately the product of speciation and extinction adding and subtracting species over time.Taxa with a higher net diversification rate (i.e. a higher rate of speciation than extinction) are expected to have higher species richness.Ho w ever, it is possible that different clades with identical div ersification r ates still differ in species ric hness, as older clades will have had more time to accumulate new species (Fig. 2 A).
Diversification can proceed at a constant rate, but can also occur in pulses (or sporadic declines).Bursts in diversification ('r a pid cladogenesis') ar e commonl y ascribed to the exploitation of ecological opportunity (Schluter 2000 , Gavrilets and Losos 2009 )  (Fig. 2 B).Suc h ada ptiv e r adiations ar e contingent on two main conditions: first, man y nic hes m ust be av ailable (or one lar ge niche space that can be partitioned), and second, only few linea ges m ust be in a starting position to fill them (i.e.competition m ust be r elaxed).Labor atory experiments hav e demonstr ated that frequency-dependent competition for niche space can drive ada ptiv e r adiations in bacteria.In a seminal experiment, Pseudomonas fluorescens pr edictabl y div ersified into thr ee types ov er the course of only a few days when incubated in static flasks, with wrinkly spreaders inhabiting the broth-air interface, fuzzy spreaders occupying the bottom of the flask, and the ancestral smooth morph residing in the broth (Rainey and Travisano 1998 ).
Phylogenetic methods offer ways to uncover bacterial diversification on m uc h longer (geological) timescales .T hey often r el y on PCR amplification and sequencing of the conserved 16S ribosomal marker from environmental samples serving as proxies for species or on higher-resolution concatenated core genes se-quenced from isolated strains .T hese studies indicate that bacterial speciation rate is slightly higher than extinction rate (Loren et al. 2014, Marin et al. 2016, Louca et al. 2018 ) (but see Martin et al. 2004 ), consistent with results for multicellular organisms wher e turnov er of taxa is high and wher e most div ersity is now extinct (Louca et al. 2018 ).Some studies have uncov er ed bursts in diversification rate in (sections of) bacterial phylogenies (Morlon et al. 2012, O'Dwyer et al. 2015 ).As 16S-based datasets have limited po w er to detect diversification on shallo w er ev olutionary time scales (Louca et al. 2018 ) and studies using higher resolution markers gener all y surv ey onl y a r elativ el y limited number of taxa, such burst-like evolution could be present but overlooked in other studies.
The aim of this paper is to examine the evidence for bursts of ada ptiv e e volution in pr okaryotes and their e volutionary and ecological drivers, and how these compare to those in macroscopic species.We will discuss how differences in diversifica-tion rate between prokaryotes could affect other aspects of bacterial biology such as the evolution of pangenomes.Although highly insightful, lab experiments are generally performed on extr emel y short timescales that r el y solel y on mutation [and seldoml y incor por ate horizontal gene tr ansfer (HGT), a centr al driver of genomic and functional diversity in bacteria] and are based on pur el y artificial selection pr essur es in the absence of other community members .We , therefore , will focus on natural populations in this review and refer to other literature summarizing results on experimental ada ptiv e r adiations in bacteria (Travisano andRainey 2000 , MacLean 2005 ).We will r e vie w studies on isolates assigned traditional taxonomic labels, 16S amplicons, and closel y r elated clusters based on whole-genome sequences.

Key inno v a tions spur adaptive r adia tions in bacteria
In macrobes, the open niche space that forms a prerequisite for ada ptiv e r adiations is often pr ovided by r ar e colonization e v ents of remote localities such as mountains , lakes , or islands , where competing species are absent.Classic examples of suc h ada ptiv e r adiations include Darwin's finc hes in the Gala pa gos, Cic hlid fishes in East African Rift Lakes and Silversw or d plants in Hawaii (Schluter 2000 ).This scenario is not likely in bacteria, as they experience little dispersal-limitation due to their small size and high abundance, meaning niche specialists and niches will be efficientl y matc hed.This diminished r ole of biogeogr a phical barriers and allopatry in prokaryotes (and a corr espondingl y incr eased role for environmental filtering) is illustrated by many 16S-based studies (Lozupone and Knight 2005 ); for instance, most global soil diversity was found to be contained in an area as small as Central P ark in Ne w York City (Ramir ez et al. 2014 ).A r ecent lar ge-scale analysis of curated genomes from around the globe found that most prokaryotic clades on Earth's surface are globally distributed (Louca 2022 ).Consistent with an earlier housek ee ping gene-based study demonstr ating geogr a phical div er gence in a thermophile archaeon (Whitaker et al. 2003 ), thermophiles were found to be least dispersiv e, whic h makes sense as they live in relatively small, specialized habitats that are far apart (Louca 2022 ).Ho w ever, neither study could conclude that e v en extr emophile species displayed endemicity.There seems to be no bacterial equivalent of marsupials, and it is ecological opportunity-rather than geogr a phic isolation-that is most likely to drive bacterial diversification (Vos 2011 ).The oft-quoted adage 'everything is everywhere, the environment selects' thus seems to be vindicated by sequencingbased studies almost a century after it was first proposed (Baas Becking 1934 ).
How could ada ptiv e r adiations occur in sympatry?One pathway to ecological innovation that is not reliant on geogr a phical isolation was de v eloped by Miller , Mayr , and Simpson in the middle of the 20th century (Heard andHauser 1995 , Schluter 2000 ).These and other scientists posited that occasional evolutionary 'k e y innovations' gi ve rise to entirely new functional capabilities that allow the colonization of new 'adaptive zones' (Hunter 1998, Alfaro 2014 ) (Fig. 2 C).Such adaptations could provide a release from competition and access to niche space not available before.A w ell-kno wn example in animals is the radiation of Notothenioid fishes in the Antarctic Ocean.The evolution of antifreeze gl ycopr oteins that lo w er internal freezing point in their last common ancestor has allo w ed the invasion of compar ativ el y empty oceanic regions with subzero temperatures and the subsequent diversification into over 130 species (Matschiner et al. 2015 ).
It could be argued that pr okaryotes hav e an especiall y gr eat potential to e volv e k e y inno vations , as HGT allows the wholesale acquisition of entir el y nov el functional tr aits originating fr om other strains and species (Lawrence 2001 , Cohan andKoeppel 2008 , Hall et al. 2017 ).One population genomics study beautifull y uncov er ed a r adiation of bacterial nic he specialists driv en by HGT (Hehemann et al. 2016 ).In pr e vious work, the same group had identified multiple genetically distinct Vibrio clusters that were hypothesized to be ecologically differentiated, as they were enric hed in differ ent particle size fr actions in the same seawater samples (Hunt et al. 2008 ).Subsequent genome sequencing uncov er ed that the brown algal glycan alginate pathway had under gone extensiv e combinatorial c hanges mediated by HGT within and between these clusters as well as more distantly related species, leading to r a pid clade diversification.Subsequent gr owth r ate experiments demonstr ated that v ariation in enzyme type, copy number, and localization (on the cell wall or broadcast into the envir onment) tr anslated into physiological differences, which in turn could explain the differential association of different types with particle size (r epr esenting differ ent degr adable algal cell wall types) and season (Hehemann et al. 2016 ).This case bears all the hallmarks of an ada ptiv e r adiation mediated via a k e y innovation.
Another example of an ada ptiv e r adiation driv en by an HGTacquired k e y innov ation is offer ed b y the Thaumar chaeota, an abundant Archaeal phylum that plays a major role in the global nitr ogen cycle, specificall y via the oxidation of ammonia.Environmental pH is a major factor affecting the distribution of differ ent Thaumarc haeota clades (Gubry-Rangin et al. 2011 ).Phylogenetic methods could show that a r adiation occurr ed earl y in the evolution of the Thaumarchaeota, allowing niche expansion from neutral pH environments to acidic and alkaline environments (Gubry-Rangin et al. 2015 ).Interestingly, diversification r ate r emained high after this initial burst, whic h is not consistent with typical ada ptiv e r adiations, wher e an initial high diversification is follo w ed b y a slo wdo wn (a signatur e also observ ed in ada ptiv e r adiations inferr ed in bacteria (Morlon et al. 2012 )).pH ada ptation in Thaumarc haeota is at least in part mediated by Vtype ATP ase pr oton pumps (Wang et al. 2019 ).The phylogeny of acidophile V-type-like ATPase operons in Thaumarchaeota is incongruent with organismal phylogeny but is congruent with habitat, indicating that HGT is responsible for ATP ase-mediated nic he adaptation (Wang et al. 2019 ).
Ecological opportunity for ada ptiv e r adiations can be provided by abiotic factors such as resource type or pH as in the case studies abo ve .But as pr okaryotes ar e gener all y embedded in highl y div erse and dense comm unities of competitors, par asites, pr ey, pr edators , hosts , symbionts , and mutualists , biotic factors m ust be highl y r ele v ant too.As differ ent or ganisms can coevolv e with eac h other, selection exerted by other or ganisms is not onl y likel y to be str ong, but also long lasting and potentially diversifying (Van Valen 1973 ).A meta-analysis on 16S diversity collected acr oss man y differ ent biomes found that the div ersity of specific lineages correlated positively with whole-community diversity (Madi et al . 2020 ).This observation is consistent with mor e div erse comm unities offering mor e av ailable nic he space thr ough mor e div erse biotic inter actions.It could also be shown that this relationship was weaker for the most diverse communities, indicating that when niches are increasingly filled, there is less opportunity for invading lineages to diversify (Madi et al . 2020 ).

Entry into novel environments: adaptive zones
High dispersal rates mean that available niches are generally filled by the a ppr opriate nic he specialists.Howe v er, it also means that ther e is fr equent immigr ation of taxa that are not (well) adapted to the local en vironment.T he vast majority of such immigrants are unlikely to persist, let alone diversify (Madi et al . 2020 ).Ho w ever, if an ecologically and genomically distinct migrant manages to take up a niche-defining gene from the local community, it could be in a position to occupy (or create) hitherto unexploited niche space and give rise to an adaptive radiation.An example of one of the most drastic environmental transitions for metazoans and prokaryotes alike is that between marine and terrestrial (including fr eshwater) envir onments (Cohan andKoeppel 2008 , Logar es et al. 2009 ).Salinity is a major determinant structuring microbial diversity, with distinct phylogenetic shifts observed over salt gradients (Dupont et al. 2014 , Fortunato andCrump 2015 ).Successful marine-terr estrial tr ansitions r equir e significant r e wiring of central metabolism and osmotic stress responses (Eiler et al. 2016 ), which could be aided b y HGT (Wisniewski-Dy e et al. 2011 ).Phylogenetic analyses indicate marine-terrestrial transitions occasionally occur in bacterial taxa (Zhang et al. 2019 ) and it can be argued these form an excellent model for the colonization of novel adaptive zones (Jurdzinski et al. 2023 ).
Another example of the colonization of novel adaptive zones is offered by pathogens switching host.Staphylococcus aureus infects a wide range of v ertebr ates (and e v en inv ertebr ates) (Matuszewska et al. 2020 ).Host jumps are frequent and result in distinct genetic clusters wher e str ains carry specific host-ada ptiv e genes, and evidence loss of host-adaptive genes associated with their pr e vious host (Matusze wska et al. 2020 ).Specifically, different host specialists are characterized by the carriage of different combinations of mobile genetic elements, including genes known to target specific host innate immune responses and antimicrobial resistance genes conferring resistance to antibiotics used in particular husbandry r egimes (Ric hardson et al. 2018, Matuszewska et al. 2020 ).This further exemplifies the perv asiv e r ole of HGT in opening up ne w nic hes, although it is not clear whether MGEs ar e gener all y acquir ed just befor e or after host-switc hing e v ents (Richardson et al. 2018 ).
Major new microbial niches have originated throughout Earth's history, from the emergence of oxygenic habitats allowing aerobic r espir ation to the e v olution of animal and plant hosts (J affe et al. 2023 ).Such niches range from 'closed' with purely vertical transmission (as those occupied by endosymbionts) to 'open' with mainly horizontal transmission (as those occupied by planktonic marine bacteria).Dispersal needs to occur to allow the colonization of novel adaptive zones, but it is not clear whether migration r ates m ust be v ery high to allow r ar e k e y innovations to occur, or if they need to be at some intermediate le v el to pr e v ent establishment of the best curr entl y ada pted species, in turn pr e v enting the opportunity of a new best-adapted lineage to evolve.

Generalists as progenitors of adaptive r adia tions
Prokaryotes can be classified as specialists or generalists based on the broadness of their niche requirements (Bell and Bell 2021 ).Bacteria with larger genomes and higher metabolic versatility are associated with greater niche width (Barberan et al. 2014 ).Living in a wider range of microbiomes means that such generalist species will encounter more distinct selection pr essur es as well as interact with more species that could serve as donors of k e y adaptations thr ough HGT.A lar ge-scale meta-anal ysis of 16S sequence data found that 16S OTUs present across a greater number of distinct habitats (likely to be generalists) was found to have a 19-fold higher speciation rate than OTUs present in only a single habitat (likely to be specialists) (Sriswasdi et al. 2017 ).That generalist-tospecialist tr ansitions ar e mor e common than vice versa , is consistent with increasing specialization resulting in the closing of doors leading to other ecological lifestyles, which is consistent with results from lab experiments on bacteria (Buckling et al. 2003 ).

Are some taxa inherently more prone to adapti vel y r adia te?
Speciation rate is dependent on ecological opportunity, but also on the rate at which new niche-defining traits can arise.Taxa that ar e mor e e volv able (Díaz Ar enas and Cooper 2013 ), thus could be expected to be in a better position to radiate into novel types.Species-specific variation in factors such as mutation rate, generation time, and population size all influence the rate of adaptation to new niches, but a high frequency of HGT specifically can be expected to facilitate the evolution of k e y innovations (Lawrence 2001 ).
High rates of HGT mediated by gene transfer agents (GTAs; exa pted bacteriopha ges that function to secr ete host DNA) hav e been implicated in a well-documented case of a bacterial ada ptiv e radiation (Guy et al. 2013 ).Bartonella are vectorborne, intracellular pathogens of mammals comprising multiple species-level clades.Two clades with similar host range display evidence of increased diversification, and both could be shown to have independently taken up the VirB type IV secretion system (T4SS), which acts to inject virulence factors into host cells (Engel et al. 2011 ).All ancestr al str ains harbour ed a GTA ca pable of in vitro gene tr ansfer (Guy et al. 2013 ); inter estingl y, the GTA is colocated with the T4SS genes, whic h r esults in a higher-than-av er a ge c hance of being secreted and taken up by other cells (Tamarit et al. 2018 ).It has been proposed that his coupling of niche-defining genes and genes incr easing r ecombination has allo w ed the successful diversification of this pathogen genus (Guy et al. 2013 ).
It is important to stress that HGT transfers do not necessarily lead to adaptive radiations when they do not increase functional diversity or when ecological opportunity is absent.For instance, hybridization e v ents wher e donor DNA r eplaces up to 20% of the recipient genome have been observed in a variety of human pathogens (Chen et al. 2014 , Cr ouc her and Klugman 2014 ) without concomitant div ersification.Mor eov er, it is possible that high rates of HGT could impede, rather than promote adaptive radiations.One of the very few studies that has discussed the concept of k e y adaptations in the context of prokaryotes has argued that HGT hinders ada ptiv e r adiations, because it could r esult in k e y adaptations being transferred to many lineages rather than just a single one (Martin et al. 2004 ).

Adaptive r adia tions with and without specia tion: implica tions for pangenome evolution
HGT in bacteria, like meiotic sex in eukaryotes, is a doubleedged sw or d: on one hand it is centr al to cr eating genetic diversity, but on the other hand it can impede genetic divergence of nascent niche specialists.Without some ecological or genetic barrier to HGT, diversification cannot proceed to the species- ).As a consequence, there could be unappreciated links between ecological diversification, recombination barriers, and the evolution of pangenomes (Fig. 3 ).
The evolution and ecology of pangenomes, the total complement of genes within a species, which is usually much larger than the number of genes in any individual genome, is a topic of great interest in evolutionary microbiology (Bobay 2020 , Domingo-Sananes andMcInerney 2021 ).Se v er al distinct, nonm utuall y exclusiv e hypotheses hav e been put forw ar d to explain the existence of pangenomes.Some explanations invoke adaptive benefits where different gene repertoires correspond to differ ential nic he specialization (Domingo-Sananes and McInerney 2021 ).Other explanations invoke neutral processes; some species might be more prone to take up genes by HGT because their genomes ar e mor e accommodating to nov el genetic div ersity or because they are surrounded by a higher diversity of community members (Br oc khurst et al. 2019 ).Gr eater effectiv e population size is expected to result in greater pangenome diversity (Andreani et al. 2017 ), specifically via retainment of accessory genes with near-neutral fitness effects (Bobay and Ochman 2018 ).
Ho w e v er, ther e is another potential explanation of why pangenome size can vary among species, which is directly linked to div ersification.Ev ery time a ne w nic he specialist e volv es and r e-combination with the ancestor ceases, the niche specialists start with a 'minimal' pangenome (Fig. 3 A).Although this pangenome will increase in size during the lifetime of a species through adaptiv e pr ocesses (e.g.div ersifying selection), nonada ptiv e pr ocesses (e.g. the uptake of parasitic mobile genetic elements) as well as neutr al pr ocesses, it will be small initiall y.In contr ast, if r ecombination barriers are absent, for instance when different genotypes remain in close physical contact, new niche specialists still e volv e, but their core genes will remain tied together through continued recombination (Shapiro and Polz 2014 ).Clade-specific accessory genes will remain part of the pangenome, which will grow with the evolution of each new niche specialist (Fig. 3 B).Esc heric hia might fit this latter scenario: species numbers in this genus are low and E. coli has a famously large pangenome.In this scenario, E. coli displays an evolutionary 'shallow' ada ptiv e r adiation where niche specialists are unable to evolve into species (Fig. 2 D).

Discussion and conclusions
Ada ptiv e r adiations hav e been implicated in bursts of species richness in animals and plants, and multiple high quality case studies hav e demonstr ated that they also occur in bacteria.Ho w e v er, the stud y of adapti ve radiations in prokaryotes is still in its infancy and many questions remain to be answered.For instance, are certain genetic (e.g.restriction/modification systems) or ecological c har acteristics (e.g.type of metabolism or microbiome) especially conducive to the radiation of lineages?Are particular traits un-likely to give rise to adaptive radiations because they are especiall y pr one to horizontal spr ead and unlikel y to tr ansfer to a single lineage?Do k e y adaptations come as single genes or operons or can they be mor e complex, involving man y genes, suc h as in the evolution of cell walls (Cohan and Koeppel 2008 )? Could some radiations be started by purely mutational processes rather than HGT, as has been shown experimentally with the mutational evolution of citrate metabolism in E. coli (Blount et al. 2008 )? Are some clades species-rich because they are old rather than having undergone burst-like evolution?
Generalization of patterns and processes between very different organisms and lifestyles is a main challenge (Gillespie et al. 2020 ).We would argue that bacterial diversification does not differ qualitativ el y fr om that in macr obes but onl y quantitativ el y.In other w or ds: 'prokary otes also disperse, adapt, recombine and speciate, just to different extents'.HGT allows the uptake of complete operons from different species and could increase the likelihood of k e y inno vations .T his effect is likel y m uc h mor e perv asive but not wholly different from hybridization events preceding ada ptiv e r adiations in eukaryote species (Seehausen 2004 ).When genome-wide HGT remains ongoing between differ entiall y ada pted linea ges this means that ada ptiv e r adiations cannot pr oceed and will not result in increased species richness, but rather highl y div erse 'str ain floc ks'.The same pr ocess has been observ ed in stic klebac ks, wher e speciation also occurs along a continuum, including repeated and reversible specialization and re producti ve isolation (Hendry et al. 2009 ).Ar guabl y the most pronounced differ ence between pr okaryotes and multicellular organisms is that environmental filtering is much more important than dispersal limitation.
In-depth genomic and ecological knowledge on species and ecotypes will be necessary to identify patterns of increased diversification, links to ecological niches, barriers to recombination and specific k e y inno vations .As in all fields of microbiology, the way we study bacterial diversification depends gr eatl y on tec hnological adv ances.Incr easing sequencing po w er allo ws for the routine use of metagenome assembled genomes (Parks et al. 2017 ).Ancient DN A (Wibo w o et al. 2021 ), HGT transfers (Davín et al. 2018 ), and bacteria-eukaryote associations (Wang and Luo 2021 ) all can help explicitly date radiations and aid in the reconstruction of ancestral states.Despite technical and computational challenges, it could be argued it is actually easier to stud y adapti ve radiations in bacteria, as vicariance is less important r elativ e to selection.In ad dition, genomic di v ersification is mor e tr actable compar ed to macrobes and experiments can be designed to test the ecological function of genes under controlled lab conditions.Experimental evolution studies incorporating multiple species and allowing HGT (e.g.Hall et al. 2016 ) are a crucial way forward to study diversification.We look forw ar d to more high-resolution genomic studies of natural populations examining the interplay between ecology, evolution, and genetics that ultimately leads to diversification of clades , genomes , pangenomes , and micr obial comm unities.

Figure 1 .
Figure 1.Variation in species diversity among bacterial genera.Rank abundance curves of total species diversity from all taxonomically recognized species (LPSN; https:// www.bacterio.net/) (black line) and 16S rRNA-based ASVs (amplicon sequence variants) assigned to genera in the Earth Micr obiome Pr oject (gr ey line).

Figure 2 .
Figure 2. Four scenarios leading to differences in species richness between taxa.(A) All else being equal, older clades should be of larger size .T he root ages for both sister clades ar e differ ent, suc h that the blue clade has had more time to diversify than the y ello w clade.(B) Clades might diversify when faced with multiple potential niches to exploit, demonstrated by partitioning and subsequent diversification into red, y ello w, green, and blue niches.(C) The capacity for diversification into multiple lineages might be mediated by the presence (or absence) of a k e y innovation, here indicated by the star.The clade on the right has acquired the capacity to exploit multiple niches into which it diversifies, while the branch on the left does not.(D)Ada ptiv e r adiations caused by k e y ada ptations (star symbols) in the pr esence of r ecombination barriers, allowing ne w nic he specialists to e volv e into distinct species (deep br anc hes, right clade), or in the absence of recombination barriers, leading to the evolution of many niche specialists that do not e volv e into species 'proper', with a shared core genome (shallow intermingled branches, left clade).

Figure 3 .
Figure 3. P angenome div ersification with and without barriers to r ecombination.(A) Div ersification coupled to speciation in a species with barriers to r ecombination.The ancestr al pangenome (1) acquir es differ ent k e y innov ations (2); eac h uniquel y ada pted linea ge ceases to r ecombine with the ancestor or with other ne wl y e volv ed nic he specialists because of r ecombination barriers (dashed lines).Ne w nic he specialists subsequentl y gr ow their pangenome through adaptive and nonadaptive processes (3).When a new k e y innovation occurs (4), the process is repeated.(B) Diversification of a species without barriers to recombination.The ancestral pangenome (1) grows progressively with each k e y innovation, de picted by red (2), green (3), and purple (4) stars as well as nonadaptive gene additions.