MICROBIOTA

Experimental evolution of a fungalpathogen into a gut symbiontGloria Hoi Wan Tso1, Jose Antonio Reales-Calderon1, Alrina Shin Min Tan1,XiaoHui Sem1*, Giang Thi Thu Le1†, Tze Guan Tan1, Ghee Chuan Lai1,K. G. Srinivasan1, Marina Yurieva1‡, Webber Liao1, Michael Poidinger1,Francesca Zolezzi1, Giulia Rancati2, Norman Pavelka1§

Gut microbes live in symbiosis with their hosts, but how mutualistic animal-microbeinteractions emerge is not understood. By adaptively evolving the opportunistic fungalpathogen Candida albicans in the mouse gastrointestinal tract, we selected strains that notonly had lost their main virulence program but also protected their new hosts against avariety of systemic infections. This protection was independent of adaptive immunity,arose as early as a single day postpriming, was dependent on increased innate cytokineresponses, and was thus reminiscent of “trained immunity.” Because both the microbeand its new host gain some advantages from their interaction, this experimental systemmight allow direct study of the evolutionary forces that govern the emergenceof mutualism between a mammal and a fungus.

Symbiotic relationships are ubiquitous innature and vary on a continuum from par-asitism to mutualism (1). Evolution plays acritical role in the establishment of theseinteractions and in moving them along

the parasite-mutualist axis. In the case of host-microbe interactions, the mode of host-to-hosttransmission dictates whether the microbe willevolve toward higher pathogenicity, commen-salism, ormutualism (2). Experimental evolutionvia serial passaging of pathogens in a new hostis a powerful strategy to study these dynamicchanges in real time. For instance, passagingparasites in new hosts via the systemic infectionroute most often selects for increased virulenceagainst the new host (3). However, experimentalsystems to study the evolutionary emergenceof mutualistic animal-microbe interactions arelacking (4).The mammalian gastrointestinal (GI) tract

harbors a large and diverse microbial commu-nity, whose members interact with the host pri-marily in commensal or mutualistic ways (5).This raises the question of how the mammalianhost establishes these neutral or beneficial inter-actions while purging potential pathogens. Wehereby hypothesize that the gut environmentselects microbes on the basis of their pathoge-nicity and moves them along the parasitism-mutualism axis via evolutionary processes.

To investigate how the gut environmentshapes the evolution of a microbe, we developedan experimental system, based on long-term GIcolonization of antibiotic-treated mice by thefungus Candida albicans coupled with serial fe-cal transplants from colonized to naïve hosts.Several variations of the protocol were tested(Fig. 1A), and all fecal transplants resulted insuccessful colonization of the recipient animals.Clonal isolates harvested after 8 or 10 weeklyserial passages (w8 or w10 strains), but not af-ter a 1-week passage (w1 strains), showed a sig-nificantly increased intra-GI competitive fitness(Fig. 1B), which was at least as high as that ofstrains deficient of the enhanced filamentousgrowth 1 (EFG1) gene (efg1−/− strains), whichwere previously known to be hyperfit in theantibiotic-treated mouse gut (6–8). Overall,these results demonstrate that a microbe canbe experimentally induced to increase its com-petitive fitness in the gut of an unnatural hostby means of adaptive evolution.Host adaptive immunity appeared to play little

or no role in this adaptation process, because allresults presented herein were qualitatively sim-ilar whether the evolution experiments wereperformed in wild-type (WT) or Recombinationactivating 1–deficient (Rag1−/−) mice, which lackfunctional T and B cells. To determine if antibi-otic treatment played any role in this evolution-ary process, we repeated the above evolutionexperiments in untreated mice. Without the useof antibiotics, however, and consistent with cur-rent knowledge (9), wewere unable to establish along-term colonization in the adult animals’guts, and the experiment had to be aborted after3 weeks (Fig. 1A). We thus resorted to an alter-nativemodel involving neonatalmice, which havelong been known to allow C. albicans GI colo-nization even in the absence of antibiotics (10).Specifically, we intragastrically inoculated 2-week-old infantmicewith the same ancestralC. albicans

used for the previous evolution experiments, withseven evolution lines maintained on oral anti-biotics and eight lines on control water (Fig. 1A).Notably, individual colonies isolated after 5weeksof evolution displayed a significantly increasedGI competitive fitness in infant mice only ifthey had been evolved in antibiotic-treated, butnot untreated, mice (Fig. 1B), suggesting thatthe presence of an intact bacterial microbiotalimits the adaptation of C. albicans in themouseGI tract.Because C. albicans strains that are hyperfit in

the antibiotic-treated or germ-free mouse gutare typically deficient in hyphal morphogenesis(6, 8, 11), we next examined cellular morphologiesand noted that over the course of 10 weeks ofevolution in infant mice, all C. albicans popula-tions progressively lost their ability to respondto hyphal-inducing stimuli, but only if they hadbeen evolved in thepresence of antibiotics (Fig. 1C).Screening several independent clones across allevolution lines, we confirmed that in the pres-ence of antibiotics, virtually all C. albicans cellslost their ability to form true hyphae by the endof each experiment (Fig. 1, D and E). Importantly,in vitro daily serial passaging of C. albicans for5 weeks in the presence of antibiotics failed toyield such phenotypes (fig. S1). Though exper-imental evolution in germ-free mice would bethe ideal model to definitively prove this point,these data strongly suggest that it was the ab-sence of microbiota and not antibiotics per sethat selected for hyphal-defective strains in ourin vivo evolution experiments. Finally, non-filamentous w10 strains and efg1−/− cells, al-though being hyperfit in the antibiotic-treatedgut, actually colonized the GI tract of untreatedmice less efficiently than WT cells (fig. S2). To-gether, these results indicate that, although thehyphal morphogenesis program is required forcompetition of C. albicans with resident gutbacteria, it otherwise represents a fitness burdenand is thus rapidly lost when the gut microbiotaare absent or chronically perturbed by long-termantibiotic treatment.In search of the genetic basis underlying these

evolved phenotypes, we performed high-depth,high-coveragewhole-genome sequencing on fourancestral C. albicans strains used as inocula forthe evolution experiments, four w1 strains, and28 w5, w8, or w10 strains selected from 16 inde-pendent evolution lines across three differentserial passaging protocols (table S1). A total of 34verified open-reading frames were identified ascarrying ≥1 de novo nonsynonymous mutationin ≥1 evolved w5, w8, or w10 strain (Fig. 2A andtable S2). On the basis of Gene Ontology en-richment analysis, 20 of these genes were re-lated to filamentous growth (P = 5.5 × 10−10,hypergeometric test, Bonferroni correction), 10encoded proteins located in the cell wall (P = 3.5 ×10−7, hypergeometric test, Bonferroni correc-tion), and seven functioned as transcriptionfactors (TFs) (P = 2.2 × 10−3, hypergeometric test,Bonferroni correction). Specifically, six genes(e.g., EFG1) encoded TFs involved in hyphal geneexpression regulation (table S3).

RESEARCH

Tso et al., Science 362, 589–595 (2018) 2 November 2018 1 of 7

1Singapore Immunology Network (SIgN), Agency for Science,Technology and Research (A*STAR), 8A Biomedical GroveImmunos #04, Singapore 138648, Singapore. 2Instituteof Medical Biology, A*STAR, 8A Biomedical Grove, Immunos#05, Singapore 138648, Singapore.*Present address: Foundation for Innovative New Diagnostics(FIND), Campus Biotech, Chemin des Mines 9, 1202 Geneva,Switzerland. †Present address: Singapore BioimagingConsortium (SBIC), A*STAR, 11 Biopolis Way, #01-02 Helios,Singapore 138667, Singapore. ‡Present address: The JacksonLaboratory for Genomic Medicine, 10 Discovery Drive,Farmington, CT 06032, USA.§Corresponding author. Email: norman_pavelka@immunol.a-star.edu.sg

on Septem

ber 21, 2020

http://science.sciencemag.org/

Dow

nloaded from

With 22 of 28 sequenced evolved strains (or 11out of 16 evolution lines) carrying ≥1 de novomutation in FLO8, which encodes a TF requiredfor hyphal development (Fig. 2C) (12), this genewas the most frequently mutated (Fig. 2A). In-terestingly, all identified FLO8mutations lead topremature stop codons or frame shifts (Fig. 2B),implicating these as loss-of-function mutations.In accordance, many stop codons were locatednear amino acid position 654, where truncationswere shown to result in null phenotypes (13). Totest if FLO8 inactivation is sufficient to increasethe fitness of C. albicans in themouse GI tract, wegeneratedheterozygous (flo8+/−) andhomozygous

( flo8−/−) FLO8-deletion strains not previously ex-posed to the mouse gut environment. Consistentwith our predictions, flo8−/−, but not flo8+/−, cellswere nonfilamentous (Fig. 2D) and outcompetedWT C. albicans as efficiently as the efg1−/− strainor any of the gut-evolved C. albicans strains(Fig. 2E). To test if the single-point mutationsidentified in thew10 strainswere indeed requiredfor increased competitive fitness, we restoredone FLO8 allele back to its ancestral sequencein strains w7 and L11.w10.c3 and observed re-version of filamentation (Fig. 2D) and a sig-nificant loss of intra-GI tract fitness in bothstrains (Fig. 2E). Further demonstrating these

point mutations to be fully recessive and notdominant-negative, ectopic expression of a WTFLO8 under its own promoter fully restoredfilamentation in the tested w10 strains, whereasa mutated FLO8 had no effect when ectopicallyexpressed in aWT strain (fig. S3). Consistent withthis notion, 19 of the 22 strains carrying FLO8-inactivating mutations carried them in homo-zygosity (Fig. 2B). These results show that adaptiveevolution in the antibiotic-treated mouse GI tractreproducibly selects for inactivating mutations incentral transcriptional regulators of the hyphalmorphogenesis program of C. albicans and, inparticular, in FLO8, yielding hyphal-defective

Tso et al., Science 362, 589–595 (2018) 2 November 2018 2 of 7

Fig. 1. Adaptive evolutionof C. albicans in the mousegut selects for hyphal-defective variants.(A) Schematic overview ofthe different evolutionprotocols. Individual strainsare named according to line(L), week (w), and colony (c).Some specific strains areadditionally identified by ashorter name (e.g., W2N orR24). (B) Similar to theefg1−/− (efg1/efg1) mutant,10-week (w10) and 8-week(w8) C. albicans strainsevolved in antibiotic-treatedmice achieved increasedcompetitive fitness in themouse GI tract comparedwith WT (SC5314) and1-week (w1) gut-evolvedstrains (W1 and R1).However, after 5 weeks ofevolution in antibiotic-freeinfants, C. albicansstrains failed to increasetheir competitive fitnesswhen compared withstrains evolved in antibiotic-treated infants. Data aremeans ± SD. n = 4 to 10 miceper group. Mann-Whitneytest: **P < 0.01, ***P <0.001, and ****P < 0.0001.(C) Whole-stool populationsfrom evolution lines L44to L58 were plated on Spideragar and scored for smooth(indicative of yeast) andwrinkled (indicative offilamentous) colonies.Smooth colonies eventuallyappeared and oftendominated the stool pop-ulations in all antibiotic-treated (Infant+), but notantibiotic-free (Infant−), lines. (D) Individual strains (clonal isolates) of each evolution line were scored for hyphal formation in response to serum.Percentages of hyphal-proficient strains were significantly lower at the end (week 8 or 10, depending on the protocol) than after 1 week of the evolutionexperiment. Gray indicates no data. (E) Representative images indicating that gut-evolved C. albicans strains are defective in hyphal formation inresponse to in vitro stimuli. Black scale bar, 200 mm; red scale bar, 20 mm.

RESEARCH | REPORTon S

eptember 21, 2020

http://science.sciencem

ag.org/D

ownloaded from

mutants with increased intra-GI tract fitness.Notably, a spectrum of hyphal-formation defects,in part associated with homozygous nonsensemutations in key hyphal morphogenesis TFs,such as EFG1, are reported to also occur in hu-man clinical isolates of C. albicans (8).Beyond single-nucleotide changes, C. albicans

is known to frequently undergo large-scale ge-nomic mutations, such as short- and long-rangeloss of heterozygosity (LOH) and segmental orwhole-chromosome aneuploidy, especially whenunder stress and during infection (14, 15). In ac-cordance, 24 of 28 w5, w8, and w10 evolvedstrains (15 of 16 evolution lines) underwent atleast one LOH or aneuploidy event, with chromo-some 6 (chr6) LOH (11 strains across six evolution

lines) and chr7 trisomy (12 strains across 10 evo-lution lines) being the most frequently observed(fig. S4). Interestingly, FLO8 is located on chr6, andall observed chr6 LOH events occurred in strainswith homozygous FLO8 mutations (Fig. 2A),suggesting that the LOH event followed thesingle-nucleotide change to allow expression ofthe recessive phenotype. Chr7 trisomy, instead, didnot appear to contribute to the adaptive process,because an unevolved, independently obtainedstrain carrying this aneuploidy displayed neitherfilamentation defects nor an increased com-petitive fitness in the mouse GI tract (fig. S5).As the hyphal morphogenesis program is a key

virulence factor of C. albicans (16), we hypothe-sized that gut-evolved strains would be less

damaging to host cells and be less pathogenicduring infection. To test this, we first measuredtheir cytotoxicity in cocultures with either mu-rinemacrophages or human colon epithelial cells.Similar to the efg1−/− or flo8−/− strain, we foundthat w10 strains induced significantly lower celldamage thanWT or w1 strains (Fig. 3A). We nexttested the virulence of gut-evolved strains in amouse model of hematogenously disseminatedcandidiasis. As expected, using a lethal doseof 5 × 105 colony-forming units (CFUs) of WTC. albicans, we observed formation of hyphae inthe kidneys ofWTmice 2 days postinfection (dpi),as well as severe necrosis, moderate perivascularedema, and moderate diffuse pyogranulomatousrenal capsulitis in most animals (Fig. 3B).

Tso et al., Science 362, 589–595 (2018) 2 November 2018 3 of 7

Fig. 2. Recurrent muta-tions in TFs regulatingfilamentous growthunderlie increasedcompetitive fitness inthe mouse gut.(A) Clustering of gut-evolved C. albicansstrains based on muta-tional pattern across34 verified open-readingframes carrying de novo,nonsynonymous substi-tutions (DNSs) revealsrecurrent mutations inFLO8 and other TFsrequired for filamentousgrowth. Genes areordered on the basis ofchromosomal location.(B) Convergent acquisi-tion of frameshift (fs) ornonsense (*) mutationsin the FLO8 gene acrossmultiple independentevolution experiments,most often in homo-zygosity. (C) FLO8 andEFG1 act downstream ofthe cyclic adenosine3′,5′-monophosphate(cAMP) pathway toregulate the expressionof hypha-specificgenes in responseto environmental stimuli,such as serum (28).(D) The flo8−/− (flo8/flo8)mutant and two w10evolved strains (W2Nand L11.w10.c3) harboringFLO8 loss-of-functionmutations were unableto filament in hyphal-inducing media. Bycontrast, the flo8+/− (flo8/FLO8) mutant and the w10 evolved strains, in which one functional FLO8 allele had been restored (flo8R/−), but not control strainstransformed only with the selection marker (flo8S/−), were able to form hyphae. Black scale bar, 200 mm; red scale bar, 20 mm. (E) Although flo8−/− cellswere significantly fitter than flo8+/− cells in the mouse GI tract, restoration of one FLO8 allele to its ancestral sequence significantly decreased the GI fitnessof strains W2N and L11.w10.c3. Data are means ± SD. n = 5 mice per group. Mann-Whitney test: **P < 0.01 and ***P < 0.001.

RESEARCH | REPORTon S

eptember 21, 2020

http://science.sciencem

ag.org/D

ownloaded from

Eventually, all mice infected withWT or w1 strainssuccumbedwithin 3 to 4 dpi (Fig. 3C). By contrast,at the same infection dose, w10 and flo8−/− strainsstrictly remained in the yeast form in kidneys at2 dpi, whereas a few hyphae were observed withefg1−/− cells, and only mild to moderate necrosiswas observed with the w10 strains, with onlysome animals displaying mild edema or capsulitis(Fig. 3B). Similar results were obtained when weinfected Rag1−/− mice; mice infected by efg1−/−,flo8−/−, or w10 strains survived significantlylonger than those infected by the WT or w1strains (Fig. 3C). Again demonstrating a keyrole of the gut microbiome in this process,C. albicans strains evolved in the absence of, butnot in the presence of, antibiotics retained theirability to kill mice (fig. S6). Notably, similar lossof virulence was observed both for w10 strainscarrying and for those not carrying FLO8 mu-tations (Fig. 3, A and C). However, the twow10 strains that were “cured” of their FLO8-inactivating mutations displayed a significantlyincreased virulence compared to their w10 coun-terparts (fig. S7), indicating that those point mu-tations were required for the loss of virulence inthose evolved strains. Hence, this experimentalsystem, unlike serial passaging via the systemicinfection route or in cell culture (17–19), reproduc-ibly yields C. albicans strains that are geneticallylocked in the yeast form and avirulent. And al-though FLO8 appears to be a mutational hotspotwith a clear implication in this adaptation mech-anism, convergent evolution toward similar phe-notypes was also observed in strains lackingFLO8-inactivating mutations.Because gut colonization by commensal fun-

gi protects hosts against infections (20), wenext asked if hosts colonized by gut-evolvedC. albicans strains carry a potential advantageover mice colonized with WT C. albicans. Wefirst tested this hypothesis under conditions inwhich the evolved strains would express theirincreased competitive fitness, that is, in antibiotic-treated adult mice. Notably, animals colonizedwith the w10 strain R24 (which coincidentallydoes not harbor FLO8 mutations), but not withthe ancestral WT strain SC5314, showed a sig-nificantly increased survival upon a systemicchallenge with a fully virulent C. albicans strain(Fig. 3D). Under these conditions, both theevolved microbe and its new host benefit fromtheir interaction.We then also repeated the colonization in non-

antibiotic-treated 2-week-old pups, which werethen systemically challenged with the fully vir-ulent strains at 6 weeks of age. Although R24 ismore rapidly lost than WT in untreated pups(fig. S2), >30% of R24-colonized mice surviveduntil 35 dpi, whereas all SC5314-colonized ani-mals died by ~20 dpi (Fig. 3D). We also testedgut-colonized pups for protection against thedistantly related fungus Aspergillus fumigatus.Whereas most uncolonized mice or SC5314-colonized mice succumbed ≤8 or ≤15 dpi, respec-tively, >20% of R24-colonized mice survived atleast until 35 dpi (Fig. 3E). Therefore, GI col-onization with C. albicans, and especially with

a gut-evolved strain, at a young age protects hostsagainst infections later in life, thereby conferringa potential advantage to the host.The above-described protection is reminiscent

of an innate memory-like mechanism termed

“trained immunity” (21). To test if gut-evolvedC. albicans strains are efficient immune trainers,we adopted a well-established mouse modelbased on intravenous priming of WT C57BL/6mice followed by systemic challengewith a lethal

Tso et al., Science 362, 589–595 (2018) 2 November 2018 4 of 7

Fig. 3. Gut-evolved C. albicans strains are avirulent and protect gut-colonized hosts fromsystemic fungal infections. (A) Similar to the efg1−/− and flo8−/− mutants, C. albicans w10 evolvedstrains with (filled circles) or without (empty circles) FLO8-inactivating mutations had reducedin vitro cytotoxicity against J774A.1 mouse macrophages and HT-29 human gut epithelial cellscompared with WT (SC5314) and w1 evolved strains (W1 and R1). LDH, lactate dehydrogenase. Dataare means ± SD. n = 3 independent experiments. Unpaired Welch’s t test: *P < 0.05, **P < 0.01,and ***P < 0.001, ****P < 0.0001. (B) One representative Periodic acid–Schiff–stained kidney sectionsof 3 to 5 mice per group infected with WT (SC5314), efg1−/−, flo8+/−, flo8−/−+, and C. albicans w10evolved strains (W2N or R24). Black scale bar, 200 mm; red scale bar, 20 mm. Images on the right aremagnifications of the boxed areas to the left. (C) Similar to the efg1−/− and flo8−/− mutants, w10 evolvedstrains with (filled circles) or without (empty circles) FLO8-inactivating mutations are less virulent inWTand Rag1−/− mice compared with WT (SC5314) or w1 strains (W1 and R1). n = 5 to 10 mice per group.(D and E) Antibiotic-treated adults or antibiotic-free infants were colonized with either WT (SC5314)or a w10 strain (R24) and challenged systemically with either WT C. albicans (D) or A. fumigatus (E).In all tested cases, R24-colonized mice were qualitatively or quantitatively more protected from systemicfungal infections than SC5314-colonized mice. n = 8 to 25 mice per group. For (C) to (E), data wereanalyzed by log-rank test; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

RESEARCH | REPORTon S

eptember 21, 2020

http://science.sciencem

ag.org/D

ownloaded from

Tso et al., Science 362, 589–595 (2018) 2 November 2018 5 of 7

Fig. 4. Gut-evolved C. albicansstrains confer hosts withbroadly cross-protectiveinnate immunity against awide range of pathogens.(A) WTmice systemically primedwith w10 evolved strains aresignificantly protected from sys-temic candidiasis. Survival issignificantly higher in these micethan in mice primed with efg1−/−

or flo8−/− mutants (shown here)or with a sublethal dose of WTC. albicans cells (fig. S8A).n = 10 mice per group.(B) Rag1−/− mice primedwith most w10 evolved strainsare significantly protectedfrom systemic candidiasis. Inmost cases, survival is higherthan that of mice primedwith the efg1−/− mutant orwith a sublethal dose of WTC. albicans cells (fig. S8B). n = 5to 10 mice per group. In (A)and (B), w10 evolved strainsharbored (filled circles) or didnot harbor (empty circles)FLO8-inactivating mutations.(C) WTmice primed with aw10 evolved strain (R24) aresignificantly protectedfrom systemic candidiasis asearly as 1 dpp. n = 10 miceper group. (D) Serum IL-6 con-centrations of WTmice infectedwith live w10 strains (W2N orR24) strains are increased at7 dpp compared with miceinfected with a sublethal doseof live (104 CFUs) or a full doseof heat-killed (HK) WT (SC5314).n = 4 to 8 mice per group.(E) Kidney IL-6 amounts,normalized based on organweight, of WTmice areincreased both at 2 and 28 dpp.n = 4 to 17 mice per group.(F) Splenocytes extracted fromWTmice 28 dpp with a w10 strain(W2N or R24) produce higheramounts of IL-6 upon ex vivostimulation with HK-WT SC5314.By contrast, the efg1−/− mutantfails to significantly trainsplenocytes. Cytokines concen-trations were measured 48 hourspoststimulation. n = 6 to 10 miceper group. Data are means ± SD.(G) 50% of mice primed with aw10 strain (R24) succumbed after systemic challenge by WT SC5314 when treated with an IL-6 neutralizing antibody but not a control antibody. n = 7 to13 mice per group. (H to J) WTmice primed with a w10 strains (W2N or R24) are significantly protected from systemic challenge with A. fumigatus (H),S. aureus (I), or P. aeruginosa (J). n = 10 to 11 mice per group. In many cases, survival is significantly higher than that of mice primed with the efg1−/− mutantor a sublethal dose of WT SC5314. Data were analyzed by log-rank test [(A) to (C) and (G) to (J)] or Mann-Whitney test [(D) to (F)]. For (A) to (J),*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

RESEARCH | REPORTon S

eptember 21, 2020

http://science.sciencem

ag.org/D

ownloaded from

dose of a fully virulent C. albicans strain (SC5314)28days later (22). As expected, allmock-vaccinatedanimals died within 3 to 4 days postchallenge(dpc) (Fig. 4A). Consistent with previous reports(22), primingwith a sublethal dose (1 × 104 CFUs)of WT C. albicans SC5314 delayed host mortality,but eventually all animals succumbed to the chal-lenge (fig. S8). By contrast, 60 to 80% of miceprimed with a full dose of a w10 strain survivedthe secondary challenge (Fig. 4A). Importantly,afilamentous efg1−/− and flo8−/− strains also con-ferred a statistically significant protection, albeitnot as efficiently as the gut-evolved strains. Com-paring all strains at the lower dose, the evolvedstrain R24 protected WT and Rag1−/− miceagainst systemic candidiasis more efficientlythan SC5314 (fig. S8, A and B). This enhancedimmunity to a secondary challenge could be dueto persistence of the primary avirulent strains inthe mouse organs (fig. S9). However, the efg1−/−

strain protected its host less efficiently againstsecondary infections even though it colonizedthe mouse kidneys at least as efficiently as gut-evolved strains. Taken together, these resultssuggest that filamentation loss increases theability of C. albicans to boost host immunity butthat gut-evolved strains likely carry additionalmodifications that further enhance this ability.Because the increased protection observed in

w10-primed mice over those immunized withsublethal doses ofWT C. albicans correlatedwithincreased total as well as anti–C. albicans–specificimmunoglobulin G titers in the serum (fig. S10),we repeated above experiments in Rag1−/− miceto test the contribution of adaptive immunity tothis protective mechanism. Similar to WT mice,Rag1−/− mice immunized with w10 strains weresignificantly more protected from systemic can-didiasis than nonimmunizedmice. Again, immu-nization with afilamentous efg1−/− or flo8−/−

strains significantly delayed mortality, althoughnot as efficiently as some of the w10 strains(Fig. 4B). Also gut-colonized Rag1−/− pups weresignificantly protected against systemic candi-diasis once they reached adult age, with SC5314-colonized mice achieving ~30% survival andR24-colonized mice >40% survival at 35 dpc(fig. S11). Overall, these data indicate that gut-evolved C. albicans strains raise protective im-mune responses independently of B and T cells.Whereas adaptive immunity typically requires

a few weeks to mount long-lived memory re-sponses, innate immune responses are character-ized by a more rapid onset but shorter lifetimes(23). Consistent with an innate-like mechanism,significant protection against infection with afully virulent C. albicans strain was achieved asearly as 1 day postpriming (dpp) with the R24gut-evolved C. albicans strain (Fig. 4C) and at3months postpriming,whereasmost w10 strainsstill conferredWTmice with a significant protec-tion against systemic candidiasis, mouse survivalwas significantly reduced compared tomice chal-lenged 1 month postpriming with most w10strains (fig. S8C). Moreover, Rag1−/− mice showsome protection 3 months after immunization,although this effect was no longer significant for

some w10 strains (fig. S8D), suggesting that theinnate immune effect could last at least until thistime point.Cytokines play crucial roles in innate immune

responses (21). Consistently, w10 strains R24 orW2N induced a significant increase in circulatinginterleukin-6 (IL-6) concentrations 7 dpp, similarto that induced by efg1−/− cells (Fig. 4D). Unlikeefg1−/− cells, however, both gut-evolved strainsalso increased kidney IL-6 concentrations at28 dpp (Fig. 4E), as well as IL-6 and tumor ne-crosis factor–a (TNF-a) production capacity ofex vivo restimulated splenocytes (Fig. 4F andfig. S12). Importantly, kidney IL-6 amounts,albeit increased compared to mock-infected ani-mals, were at least 10-fold lower than those mea-sured using a lethal dose of the WT strain (Fig.4E), and IL-6 serum concentrations returnedto physiological levels at 28 dpp (Fig. 4D), therebylimiting the potential collateral damage of chronicsystemic inflammation.Moreover, injecting aneu-tralizing anti–IL-6 antibody significantly reducedR24-induced protection against systemic candidi-asis (Fig. 4G), demonstrating a crucial role for thiscytokine in this protective immune response.The non-antigen-specific nature of innate im-

munity predicts that priming with a w10 strainshould confer broad cross-protection against awide range of pathogens. To test this, we firstintravenously primed naïve mice with a w10 strainand then challenged them systemically with alethal dose of the unrelated fungus A. fumigatus;a Gram-positive bacterium, Staphylococcus aureus;or a Gram-negative bacterium, Pseudomonasaeruginosa. In all cases, animals immunizedwith a w10 strain were significantly protectedfrom infection compared with both naïve miceand those immunized with a sublethal dose ofWT C. albicans (Fig. 4, H to J). Moreover, W2Nand R24 protected mice significantly better thanefg1−/− cells from systemic A. fumigatus challenge(Fig. 4H), and R24 protected them more ef-ficiently than efg1−/− cells from P. aeruginosainfection (Fig. 4J). Confirming that this cross-protection was independent of adaptive immunity,Rag1−/− mice were also protected from systemicaspergillosis after immunization with R24 (Fig. 4H).Altogether, gut-evolved w10 strains were able

to raise protective immune responses character-ized by rapid onset, short lifetime, increasedinnate cytokine responses, pathogen-aspecificity,and independence from T and B cells; thus, thisprotection mechanism resembles several criticalaspects of trained immunity (21), although otherinnate immunity mechanisms cannot be dis-counted. This experimental system hence repre-sents a highly efficient method to generate fungalstrains that are not only less pathogenic but alsohighly immunogenic. These evolved C. albicansstrains could therefore lead toward safe andeffective universal vaccines with a broad spec-trum of cross-protection (24) and with activityalso in individuals with impaired adaptive immu-nity, such as HIV/AIDS patients or transplantrecipients.Taken together, these findings indicate that

adaptive evolution of a microbe in the mamma-

lian gut can alter a host-pathogen interactioninto a mutually advantageous relationship, inwhich both the microbe and the host gain somebenefit from their interaction.While themicrobeenhances its competitive fitness in the host GItract, the host augments (albeit temporarily) itsresistance against a wide variety of pathogens,though long-term consequences on host fitnesscould not be investigated here. A similar obser-vation was recently reported in laboratory worms(25); however, these organisms lack adaptiveimmunity and a complex gut microbiome. Ourdata demonstrate that, although adaptive im-munity does not influence this innate selec-tion mechanism, the presence of unperturbedbacterial microbiota interferes with this evo-lutionary process by shifting the balance be-tween selective forces: Selection by gut bacteriaoutweighs that by the host, and the filamen-tation program upturns from a fitness burdento a selective advantage. This model explainswhy, in its natural niche—the human GI tract,where it is heavily outnumbered by commensalbacteria—C. albicans cannot afford to lose thisimportant competitionmechanism andwhy the“virulence” of this pathobiont often emergesonly after antibiotic use (26), when this programis unintentionally redirected from the micro-biota to the host.

REFERENCES AND NOTES

1. D. A. Relman, Nat. Rev. Microbiol. 6, 721–724 (2008).2. P. W. Ewald, Ann. N. Y. Acad. Sci. 503, 295–306 (1987).3. D. Ebert, Science 282, 1432–1435 (1998).4. K. L. Hoang, L. T. Morran, N. M. Gerardo, Front. Microbiol. 7,

1444 (2016).5. G. P. Donaldson, S. M. Lee, S. K. Mazmanian, Nat. Rev.

Microbiol. 14, 20–32 (2016).6. J. V. Pierce, C. A. Kumamoto, mBio 3, e00117-12 (2012).7. K. Pande, C. Chen, S. M. Noble, Nat. Genet. 45, 1088–1091

(2013).8. M. P. Hirakawa et al., Genome Res. 25, 413–425 (2015).9. A. Y. Koh, Eukaryot. Cell 12, 1416–1422 (2013).10. L. M. Pope, G. T. Cole, M. N. Guentzel, L. J. Berry, Infect.

Immun. 25, 702–707 (1979).11. L. Böhm et al., PLOS Pathog. 13, e1006699 (2017).12. F. Cao et al., Mol. Biol. Cell 17, 295–307 (2006).13. J. Y. Liu et al., Microbiol. Res. 178, 1–8 (2015).14. A. Forche et al., mBio 2, e00129-11 (2011).15. A. Forche, P. T. Magee, A. Selmecki, J. Berman, G. May,

Genetics 182, 799–811 (2009).16. H. J. Lo et al., Cell 90, 939–949 (1997).17. A. Lüttich, S. Brunke, B. Hube, I. D. Jacobsen, PLOS ONE 8,

e64482 (2013).18. S. Brunke et al., PLOS Pathog. 10, e1004478 (2014).19. A. Wartenberg et al., PLOS Genet. 10, e1004824 (2014).20. T. T. Jiang et al., Cell Host Microbe 22, 809–816.e4

(2017).21. M. G. Netea et al., Science 352, aaf1098 (2016).22. J. Quintin et al., Cell Host Microbe 12, 223–232 (2012).23. S. Akira, S. Uematsu, O. Takeuchi, Cell 124, 783–801

(2006).24. H. S. Goodridge et al., Nat. Rev. Immunol. 16, 392–400 (2016).25. K. C. King et al., ISME J. 10, 1915–1924 (2016).26. M. S. Seelig, Am. J. Med. 40, 887–917 (1966).27. N. Pavelka, G. Tso, A. Tan, Raw data of ‘Experimental evolution

of a fungal pathogen into a gut symbiont’. Open ScienceFramework (2018).

28. P. E. Sudbery, Nat. Rev. Microbiol. 9, 737–748 (2011).

ACKNOWLEDGMENTS

We thank A. Tan, Y. S. Lim, and N. Tay for technical assistance;Singapore Immunology Network (SIgN) Mouse Core forproviding mice; IMCB Advanced Molecular Pathology

Tso et al., Science 362, 589–595 (2018) 2 November 2018 6 of 7

RESEARCH | REPORTon S

eptember 21, 2020

http://science.sciencem

ag.org/D

ownloaded from

Laboratory for histology service; C. B. Ong for veterinarypathology evaluation; J. Berman for C. albicans strains;and M. G. Netea, Y. Wang, and G. De Libero for scientificdiscussions and feedback on the manuscript. Sequencing wasperformed by the Genome Institute of Singapore GenomeTechnology and Biology Group, Singapore. This study wassupported by Agency for Science, Technology, and Research(A*STAR) Investigatorship awards JCO/1437a00117 to N.P.and JCO/1437a00119 to G.R. and by core funding fromSIgN. Author contributions: G.H.W.T. and N.P. designed thestudy; K.G.S. and F.Z. prepared sequencing libraries; G.H.W.T.,X.S., J.A.R.-C., A.S.M.T., T.G.T., G.T.T.L., and G.C.L. performed

all other experiments; G.R. designed and supervised cloningexperiments; M.Y., W.L., M.P., G.H.W.T., and N.P. analyzedsequencing data; G.H.W.T., J.A.R.-C., and N.P. analyzed all otherdata and wrote the manuscript; and N.P. supervised theproject. Competing interests: G.H.W.T., X.S., and N.P. areinventors on Singapore patent application no. 10201702472Tand International PCT application no. PCT/SG2018/050142submitted by A*STAR that cover methods of generatingattenuated fungi and uses thereof. Data and materialsavailability: The gut-evolved C. albicans strains are availablefrom N.P. under a material transfer agreement with SIgN,A*STAR. Sequencing data has been deposited at NCBI under

SRA accession number SRP116719. All other raw data and codecan be accessed at (27).

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/362/6414/589/suppl/DC1Materials and MethodsFigs. S1 to S12Tables S1 to S6References (29–39)

19 January 2018; accepted 12 September 201810.1126/science.aat0537

Tso et al., Science 362, 589–595 (2018) 2 November 2018 7 of 7

RESEARCH | REPORTon S

eptember 21, 2020

http://science.sciencem

ag.org/D

ownloaded from

Experimental evolution of a fungal pathogen into a gut symbiont

Norman PavelkaGhee Chuan Lai, K. G. Srinivasan, Marina Yurieva, Webber Liao, Michael Poidinger, Francesca Zolezzi, Giulia Rancati and Gloria Hoi Wan Tso, Jose Antonio Reales-Calderon, Alrina Shin Min Tan, XiaoHui Sem, Giang Thi Thu Le, Tze Guan Tan,

DOI: 10.1126/science.aat0537 (6414), 589-595.362Science 

, this issue p. 589; see also p. 523ScienceHowever, if an intact microbiota was present, only the virulent hyphal forms persisted.stimulated proinflammatory cytokines and conferred transient cross-protection against several other gut inhabitants.

gene, resulting in low-virulence phenotypes unable to form hyphae. Nevertheless, these phenotypesFLO8around the and were thus lacking gut bacteria (see the Perspective by d'Enfert). Passage accelerated fungal mutation, especially

under evolutionary pressure by serial passage in mice that were treated with antibioticsC. albicans put et al.there. Tso , are found in the mammalian gut, but we know little about what they are doingCandida albicansFungi, such as

Gut microbiota selects fungi

ARTICLE TOOLS http://science.sciencemag.org/content/362/6414/589

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2018/10/31/362.6414.589.DC1

CONTENTRELATED http://science.sciencemag.org/content/sci/362/6414/523.full

REFERENCES

http://science.sciencemag.org/content/362/6414/589#BIBLThis article cites 37 articles, 11 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2018 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on Septem

ber 21, 2020

http://science.sciencemag.org/

Dow

nloaded from