Bacillus subtilis biofilm extends Caenorhabditis elegans longevity through downregulation of the insulin-like signalling pathway

Beneficial bacteria have been shown to affect host longevity, but the molecular mechanisms mediating such effects remain largely unclear. Here we show that formation of Bacillus subtilis biofilms increases Caenorhabditis elegans lifespan. Biofilm-proficient B. subtilis colonizes the C. elegans gut and extends worm lifespan more than biofilm-deficient isogenic strains. Two molecules produced by B. subtilis — the quorum-sensing pentapeptide CSF and nitric oxide (NO) — are sufficient to extend C. elegans longevity. When B. subtilis is cultured under biofilm-supporting conditions, the synthesis of NO and CSF is increased in comparison with their production under planktonic growth conditions. We further show that the prolongevity effect of B. subtilis biofilms depends on the DAF-2/DAF-16/HSF-1 signalling axis and the downregulation of the insulin-like signalling (ILS) pathway.

D uring the last 30 years, human life expectancy has significantly increased worldwide http://www.who.int/ gho/publications/world_health_statistics/2015. Between 1990 and 2012, humans (males and females) born in low-and high-income countries have gained 9.05 and 4.25 years in life expectancy, respectively http://www.who.int/gho/publications/ world_health_statistics/2015, which has been attributed to a plethora of genetic and environmental factors 1,2 . In mammals, the beneficial bacteria residing in the host intestine (that is, gut or commensal flora) play a crucial role in immune system development, tissue morphogenesis and aging [3][4][5][6][7][8] . Beneficial effects of the gut flora on host physiology, including slowing of aging, might be improved with the incorporation of probiotic bacteria into the host diet. Probiotics are live organisms that have a beneficial effect on host health when they are administered or present in adequate quantities 9,10 . However, the mechanisms causing potential prolongevity effects of beneficial bacteria on different hosts remain poorly understood [10][11][12][13][14][15] .
Microbial biofilms are three-dimensional structured communities of adherent microorganisms encased in a self-produced extracellular matrix, containing networks of channels for nutrient supply and long-distance cell-to-cell communication 16,17 . Bacteria living in biofilms are physiological very distinct from their planktonic counterparts, and they function as a cooperative consortium more similar to that of multicellular organisms than a unicellular organism 16,18,19 . The crucial role of biofilms in the success of different pathogenic bacteria in infecting different hosts is well documented 20 , but whether biofilms influence the interaction between beneficial bacteria and their host is largely unknown 10,16 .
Bacillus subtilis is a model beneficial bacterium with the ability to display many distinct cell types under developmental control, including the ability to form robust and sophisticated biofilms [21][22][23][24][25][26][27] . Interestingly, a recent study showed that planktonic B. subtilis modulates the longevity of the bacteriovorus nematode and model organism Caenorhabditis elegans independent of its role as a food source 21,28 . Worms fed on planktonic B. subtilis cells live longer (an approximately 15% increase in lifespan) than worms grown in the presence of other food sources, that is, E. coli 28,29 . Because of the limited understanding of the effect of biofilm formed by beneficial gut bacteria on host health, we determined whether the biofilm of B. subtilis might represent a novel anti-aging agent for improving host longevity.
Results and Discussion C. elegans longevity is increased by undomesticated B. subtilis. The various studies that have reported interactions of B. subtilis with C. elegans utilized genetically modified laboratory B. subtilis (domesticated) strains derived from the wild (undomesticated) Marburg isolate NCIB3610. However, these laboratory strains (that is, strains 168, JH642 and PY79; refs 30-32 are known to harbour several initially inadvertent mutations that affect collective behaviours (that is, social-surface motility and biofilm formation) 31,32 that might play an important role during bacteria-host interactions.
To answer this question, we initially compared the lifespan effects produced by the wild strain NCIB3610 (ref. 31) and its isogenic derivate, the domesticated JH642 strain 33 (Supplementary Table 1) on C. elegans. It has been noted that in contrast to the susceptibility of vegetative cells, B. subtilis spores survive the transit through the pharynx and germinate in the worm intestine 34,35 . Therefore, for all the subsequent experiments, the sporulated form of B. subtilis was provided as the worm food source to correlate the effects on worm lifespan with the activity of live bacteria 10,21 . C. elegans fed on domesticated B. subtilis cells showed an extended longevity compared with the effect of OP50 E. coli cells when used as a food source ( Fig. 1a and Supplementary Table 2). The survival of worms fed on JH642 cells increased by an average of 23 Table 2). Interestingly, the lifespan effect on C. elegans was significantly more pronounced with a statistically right-shifted survival curve (P ¼ 0.00) by log-rank test when the worms were fed on undomesticated B. subtilis cells (strain NCIB3610) ( Fig. 1a and Supplementary Table 2). In this case, NCIB3610 cells increased the worm longevity by an average of 52.59% compared with the lifespan effect produced by OP50 E. coli cells (average value, 30.90 days; 95% CI: 29.14-32.67) (Supplementary Table 2). The lifespan of C. elegans also increased by an average of 59.51% (average value, 32.30 days; 95% CI: 30.68-33.90; equality, P ¼ 0.00 by log-rank test) when fed on another undomesticated B. subtilis strain (that is, the probiotic B. subtilis natto strain RG4365, Supplementary Table 1) 22,27 . The different effects of these bacterial strains on worm survival cannot be attributed to physiological differences between the domesticated and undomesticated strains because both types of bacterial cells displayed similar efficiencies of spore germination, vegetative growth and spore formation ( Supplementary Fig. 1a-c). However, the possibility exists that the observed differential effects on worm survival were due to changes in the mechanical/physical properties of both types of spores. In this case, the worm grinder may not be able to efficiently break the spores of undomesticated cells, leading to higher colonization levels and/or reduced nutritional availability, which might explain the observed longevity effects (Fig. 1a). To test this possibility, we analysed and compared the robustness and resistance of JH642 and NCIB3610 spores. As shown in Supplementary Fig. 1d, spores derived from domesticated and undomesticated cells are indistinguishable one from the other in their resistance to different physicochemical treatments such as sonication, low and high pH exposure, and lysozyme and protease treatments. Accordingly, several physiological parameters of C. elegans, such as its chemotactic behaviour (in response to the different bacterial food sources), growth rate, body size, postembryonic development and egg-laying proficiency were not affected when fed on any of the selected bacterial strains (that is, OP50, JH642 and NCIB3610 cells; Supplementary Fig. 1e-i). Additionally, we used FUdR in the lifespan assays which was previously shown to be a potential co-founder for C. elegans lifespan studies 36 . To test this possibility, we analysed the lifespan effects produced by B. subtilis NCIB3610 cells on N2 worms either in the presence or the absence of FUdR, as well as the effect of FUdR in B. subtilis biofilm formation proficiency. As shown in Supplementary  Fig. 2a Fig. 2b). Overall, these results rule out potential drug-bacterial strain interaction, variations in food preference and/or food availability between the different bacterial food sources as reasons for the observed differences in C. elegans lifespan ( Fig. 1a and Supplementary Table 2).
Lifespan and stress resistance are interrelated 1,2 . Extended longevity of C. elegans correlates with enhanced worm resistance against different biological, physical and chemical stressors 37,38 . In this sense, in accordance with the increase in the lifespan effect produced by undomesticated B. subtilis, we found a strong positive effect on the thermotolerance of C. elegans. As shown in Fig. 1b, C. elegans fed on NCIB3610 cells showed higher resistance and survival to heat shock compared with the protective effects produced by domesticated JH642 (using OP50 cells as a control). When C. elegans was shifted from its routine growth temperature of 20 to 34°C, worms fed either on undomesticated or domesticated B. subtilis cells lived 278.57% (average survival, 132.16 min) and 179.52% (average survival, 97.58 min) longer, respectively, than did worms fed on OP50 cells (average survival, 34.91 min; Fig. 1b and Supplementary Table 2). Accordingly, we did not observe any significant FUdR interference on the thermotolerance behaviour of C. elegans when fed on NCIB3610 cells (Supplementary Fig. 2c and Supplementary  Table 3).
To confirm the higher beneficial effects of undomesticated B. subtilis cells over domesticated B. subtilis cells on worm survival, we measured the lifespan of N2 worms subjected to injuries other than heat shock (that is, osmotic, heavy metal and oxidative stresses). In all the analysed situations, C. elegans fed on NCIB3610 B. subtilis showed a significant increase in survival (equality, Po0.001 by log-rank test) compared with the survival of worms fed on JH642 or OP50 cells (Fig. 1c-e). Taken together, these results indicate that wild (undomesticated) B. subtilis strains are significantly more beneficial and efficient than domesticated B. subtilis in protecting and prolonging C. elegans lifespan (Fig. 1, Table 1A and Supplementary Table 2).  To shed light on the reasons for the differences in the magnitude of the lifespan effects produced by domesticated and undomesticated B. subtilis, we analysed the ability of each cell type to grow and colonize the worm intestine 14,15,34 . Our initial approach was to measure the intestinal activity of a reporter gene harboured by the bacterium used as the worm food source. Therefore, we fed C. elegans on NCIB3610 and JH642 cells that harboured a reporter of b-galactosidase activity 30 driven by the bacterium promoter of the surfactin operon (Supplementary Table 1). As shown in Fig. 2a, cellular extracts obtained from worms fed on both types of B. subtilis cells showed b-galactosidase activity derived from the reporter p srf-lacZ fusion, a finding that strongly suggests that both types of spores (JH642 and NCIB3610) were able to germinate (spores are metabolically inactive) in C. elegans gut. Interestingly, the pattern of b-galactosidase expression obtained from the undomesticated cells was significantly higher than that obtained from the worms fed on domesticated cells; peaks of 200 and 1,000 Miller units (MU) of b-galactosidase activity were reached after 70 h of gut colonization with JH642 and NCIB3610 isogenic strains harbouring the lacZ reporter-fusion system, respectively (Fig. 2a). The difference in the level of expression of the p srf-lacZ reporter fusion could be due to a different spore germination or bacterial growth rate, to the differential expression of the surfactin promoter in domesticated and undomesticated cells and/or to a different level of bacterial gut colonization (number of B. subtilis cells/worm). To distinguish between these possibilities, we measured the germination rates, the outgrowth and the b-galactosidase expression driven by the P srf-lacZ fusion in domesticated and undomesticated B. subtilis cells inoculated in nematode growth medium (NGM). As shown in Supplementary  Fig. 3, the kinetics of spore germination, the vegetative growth rates and the b-galactosidase expression from the reporter p srf-lacZ fusion of cells grown in liquid and solid NGM did not significantly differ between the domesticated and undomesticated B. subtilis cells ( Supplementary Fig. 3a-c). These results suggest that the different levels of intestinal b-galactosidase activity derived from the two types of utilized B. subtilis cells (Fig. 2a) were due to the different levels of bacterial worm gut colonization. Similarly, the fluorescence signal originated from the gut of worms fed on undomesticated B. subtilis cells harbouring a reporter gene of GFP expression (bslA-gfp) (Supplementary Table 1) was significantly higher than the originated from worms fed on domesticated B. subtilis cells (Fig. 2b). The observed differences in GFP expression were not due to differences in the growth rate or other physiological parameter between both types of B. subtilis strains because the GFP expression levels displayed by both types of cells were indistinguishable from one another under planktonic growth conditions on NGM agar plates ( Supplementary Fig. 3d). Therefore, these results demonstrate the proficiency of B. subtilis, as a commensal organism, to complete an entire cell cycle in  the intestine of C. elegans as was previously reported for this bacterium in other animal models [39][40][41] . In addition, wild B. subtilis was hypothesized to be a better colonizer of the worm intestine than domesticated B. subtilis (Fig. 2a,b and Supplementary Fig. 3e). Supporting this hypothesis, the number of NCIB3610 cells (spores plus vegetative cells) recovered from the worm gut was significantly higher (4-fold higher) than the intestinal number of JH642 cells (spores plus vegetative cells) (Fig. 2c).
To confirm that undomesticated B. subtilis is a better gut colonizer than domesticated B. subtilis, we measured the levels of bacterial gut colonization and the lifespan effect on worms in which one bacterial type (NCIB3610 or JH642) used as a food source for the initial colonization ('bacterial pulse') was switched ('bacterial chase') with other type of bacteria (OP50, JH642, NCIB3610 or a mixture of them; Supplementary  Fig. 4). Overall, the results of the food switching experiments confirmed the higher ability of NCIB3610 cells to colonize and persist in the worm intestine compared with OP50 and domesticated JH642 cells. In addition, each level of worm gut colonization produced by cells of the strains OP50 (Supplementary Fig. 4a,h), JH642 ( Supplementary Fig. 4c,i), NCIB3610 ( Supplementary Fig. 4b,j) or mixtures of them (Supplementary Fig. 4d-g,k-n) correlates with the lifespan effect produced by these strains when individually used to feed C. elegans during its entire lifespan. As observed in Supplementary  Fig. 4o,p, the presence of a higher number of NCIB3610 cells in the gut corresponded with the highest worm longevity (Supplementary Table 4). These results demonstrated that wild (undomesticated) B. subtilis is better suited to colonize and persist in the intestinal tract of C. elegans and to improve its longevity compared with domesticated B. subtilis and OP50 cells.
Interestingly, a previous report indicated that only a supply of vegetative B. subtilis, and not spores, was able to extend the NO-dependent longevity of C. elegans 28 . The failure of the dormant spores to increase the longevity of C. elegans was attributed to the inability of these spores to germinate and initiate the de novo synthesis of NO 28 . However, our study used sporulated B. subtilis that was able to germinate, grow, colonize the worm intestine, thus extending the worm longevity (Fig. 2). Consistent with our results, other reports have previously indicated that B. subtilis spores are better suited to survive the transit from the worm pharynx to the intestine than vegetative cells, which are primarily destroyed during this process 34,35,42 . We attribute this discrepancy with the mentioned work 28 to the presence of unnoticed genetic differences in the utilized B. subtilis strain (that is, differences in the efficiencies of spore germination and/or outgrowth). Supporting this interpretation, the existence of important genetic differences between domesticated B. subtilis strains used in different laboratories around the world has been previously reported 32,43 .
The biofilm is key to colonize C. elegans and extend its lifespan. The ability to adhere and form a multicellular structure called a pellicle or biofilm constitutes an attribute of bacteria that is key to their success in colonization and persistence in a particular niche [16][17][18][19][20] . In studies of host-bacteria interactions, the biofilm formation proficiency of both pathogenic and beneficial bacteria is believed to play a crucial role [44][45][46][47][48][49][50][51][52] . Because of the significant difference between domesticated and undomesticated B. subtilis strains in their biofilm formation proficiency 24,31 , we wanted to analyse whether the observed differences in worm longevity and gut colonization produced by B. subtilis were due to the different abilities of the bacterial strains to establish a biofilm in the worm intestinal environment. The tapA-sipW-tasA and epsA-G operons encode for two key components of the extracellular matrix of the biofilm, the TasA protein and the EPS exopolysaccharide, respectively 25,26,53 . In addition, the bslA gene encodes for an essential hydrophobin responsible for the formation of the hydrophobic surface layer that surrounds and protects the biofilm 54 . Therefore, to test our hypothesis, we introduced   Table 1). When each of the three undomesticated-derived B. subtilis mutant strains defective in biofilm formation (DtasA, DepsG and DbslA; Fig. 3a) were separately used to feed C. elegans, there was a significant decrease in lifespan (an average decrease of 30%) compared with C. elegans fed on the biofilm-proficient wild-type strain NCIB3610 ( Fig. 3b and Supplementary Table 5). For instance, while C. elegans fed on the wild-type NCIB3610 strain showed an increase in longevity of 58.73% (Supplementary Table 5  (DtasA, DbslA or DepsG) spores to germinate and/or to be more or less nutritious for C. elegans than the wild-type NCIB3610 cells because we did not observe any differences in the efficiency of spore germination or postembryonic worm development or worm chemotactic behaviour using any of the biofilm mutant strains (DtasA, DbslA or DepsG) compared with NCIB3610 spores as a food source. In addition, the spore robustness against chemical and mechanical injuries was indistinguishable between biofilm-proficient and biofilm-deficient strain spores (data not shown). Importantly, the decrease in worm longevity produced by feeding worms each biofilm-defective strain correlated with a decreased bacterial proficiency in gut colonization (Fig. 3c). Because EPS and the TasA and BslA proteins are secreted to the extracellular matrix of the biofilm 23 , mixtures of the DbslA, Deps and DtasA mutant cells complement each other to restore full biofilm formation proficiency 23 . Therefore, we fed C. elegans on a mix of spores of the three biofilm-deficient mutant strains and measured the lifespan effect produced by this mixture. As shown in Fig. 3d, C. elegans fed on a mixture of the three biofilm-deficient mutant strains restored its lifespan to levels comparable with the lifespan obtained when the worm was fed on wild-type NCIB3610 cells. In this case, the worm lifespan fed on a mixture of the three biofilm mutant strains (DbslA þ Deps þ DtasA) increased by an average of 52 Table 5). This lifespan effect correlates with the ability of the mixture of the three biofilm-defective strains (DbslA þ Deps þ DtasA) to recover full biofilm formation proficiency (see insert in Fig. 3d) and worm gut colonization (Fig. 3e). Neither JH642 cells that are poor biofilm makers 24,31,32 (Supplementary Fig. 1) nor the biofilm-defective NCIB3610 derivates (DtasA, DbslA or DepsG) affected the growth and postembryonic development of C. elegans (worms fed on each biofilm mutant strain required an average of 42.5 ± 2.0 h to reach the late L4/young adult stage). We observed no variation in food preference and/or food availability between the isogenic derivates of NCIB3610 cells as the reasons for the observed differences in the prolongevity effect of B. subtilis on C. elegans. Thus, the overall results indicated that biofilm formation proficiency constitutes a key positive property to fully understand the beneficial effect of B. subtilis on C. elegans lifespan (Table 1B).
The biofilm and nitric oxide produce synergic anti-aging effects. The beneficial effect of domesticated B. subtilis over OP50 on C. elegans longevity has been recently attributed to the bacterium's nitric oxide (NO) production proficiency 28 . NOmediated signalling plays key roles in all living organisms, and for an unknown reason, C. elegans is unable to produce its own NO but is able to incorporate the NO produced by B. subtilis 14,15,28,29 . Most organisms produce NO through aerobic conversion of L-arginine to L-citrulline in a reaction catalysed by the enzyme NO synthetaze encoded by the nos gene 55 . E. coli strains, several of which are routinely used to feed the worms (that is, OP50, HB101; refs 14,15,29), are not proficient in aerobic NO production because they lack a functional copy of nos 55 . However, E. coli can produce NO under anaerobic/microaerophilic conditions by a series of biochemical reactions associated with the anaerobic respiratory chain of the bacterium (Supplementary Table 6 and ref. 56). Under this scenario, E. coli might find permissive conditions for NO production in the oxygen-depleted environment of the worm intestine; therefore, the positive effect of B. subtilis on worm longevity 28 could be due to at least one as-yet unknown additional factor apart from the intestinal NO provision. First, we were interested in determining whether the biofilm proficiency-dependent lifespan effect of B. subtilis on C. elegans was due to NO production 28 . As shown in Fig. 4a, the mutations affecting biofilm formation (DbslA mutant) or NO synthesis (Dnos mutant) (Supplementary Table 1) produced a negative effect on C. elegans lifespan compared with the lifespan effect produced by the wild-type NCIB3610 strain. Interestingly, the negative effect on worm longevity was more pronounced when C. elegans was fed B. subtilis cells deficient in biofilm formation (but proficient in NO production; average value, 22.03 days; 95% CI: 20.13-23.94; equality, P ¼ 0.04 by log-rank test) compared with B. subtilis cells deficient in NO production (but proficient in biofilm formation; average value, 25.55 days; 95% CI: 23.31-27.79; equality, P ¼ 2.5 Â 10 À 7 by log-rank test; Supplementary Table 7). The average worm lifespan increases of 29.17 and 11.38% were observed when the worms were fed on B. subtilis cells defective in NO or biofilm production, respectively, in comparison with the 53.13% lifespan increase (related to OP50 cells) obtained when the worms were fed on wild-type B. subtilis (average value, 30.29 days; 95% CI: 28.26-32.33; equality, P ¼ 0.00 by log-rank; Fig. 4a (Fig. 4a and Supplementary Table 7). In addition, spore robustness and worm chemotactic behaviour toward Dnos (60.0 ± 2.0% of Dnos spore survival after 30 s of treatment at pH ¼ 2.0 and chemotaxis index of 0.95±0.2, respectively) and Dnos-DbslA (59.4 ± 1.9% of Dnos-DbslA spore survival after 30 s of treatment at pH ¼ 2.0 and chemotaxis index of 0.91±0.3, respectively) cells were not significantly different compared with the NCIB3610 strain. The different effects on C. elegans lifespan correlated with the capacity of each B. subtilis strain to form a biofilm (Fig. 4b), produce NO (Fig. 4c) and colonize the worm intestine (Fig. 4d). Thus, the positive lifespan effects of NO production and biofilm formation proficiency are different and synergistic.
If NO production and biofilm formation are distinct but complementary processes improving the longevity of C. elegans, then we hypothesize that a mixture of mutant B. subtilis cells, that is, cells deficient in biofilm formation (DbslA) and cells deficient in NO synthesis (Dnos), would complement each other and restore the full lifespan effect of wild-type NCIB3610 cells on C. elegans. As predicted by this hypothesis, a mixture of DbslA and Dnos cells (strains RG3610 plus RG3603, light blue curve in Fig. 4a) Fig. 4a and Supplementary Table 7).
Overall, these results demonstrate that biofilm formation proficiency constitutes a new attribute, different from NO production, for the anti-aging effect of B. subtilis on C. elegans (Table 1B). The discovery that NO production is not the only factor that explains how B. subtilis extends the longevity of C. elegans opens the possibility of the existence of other factors that positively contributes to increase worm longevity.
B. subtilis quorum sensing extends worm longevity. Intra-and interspecific quorum sensing (QS) constitutes a widely used strategy that bacteria use in nature to communicate with each other and with cells of different kingdoms 18,19,57,58 . Recent studies have shown the potential effect of bacterial QS on C. elegans physiology. C. elegans recognized Vibrio cholerae QS signalling molecules through the AWC on neuron 59 . Furthermore, infection of C. elegans with Candida albicans or Pseudomonas aeruginosa was repressed by bacterial QS molecules 48,60 , increasing the idea that the nematode can detect microbial signals and thus modulate each other (bacteria and host) in a symbiotic relationship 61 . B. subtilis QS pentapeptide CSF (Competence Sporulation stimulating Factor, also named PhrC) 62 was previously reported to contribute to intestinal homeostasis by activating key survival pathways of the host (p38 MAP kinase and protein kinase B) and by inducing cytoprotective heat shock proteins (Hsps) 21,63 . These effects of CSF 21 depend on its uptake by the protein OCTN2, a host cell membrane transporter of organic cations present in the apical face of epithelial cells 63 . Because an orthologue of octn2 is present in C. elegans 64 , we were interested in determining whether the CSF molecule affects the lifespan of C. elegans. As shown in Fig. 5 When the longevity of worms fed on the single, double and triple isogenic mutant strains RG4010 (Dcsf), LC3601 (DnosDcsf) and LC3611 (DnosDcsfDbslA) (Supplementary Table 1) was compared, a synergic decrease in the worm lifespan was observed (Fig. 5a). The average survival of worms fed RG4010 (Dcsf), LC3601 (DcsfDnos) or LC3611 (DcsfDnosDbslA) cells was    Interestingly, the B. subtilis triple mutant strain (deficient in biofilm formation and NO and CSF production, strain LC3611) produced a lifespan effect on C. elegans survival that was smaller than the effect obtained with OP50 cells used as a control (6.7% lifespan decrease, Fig. 5a). In addition, the lifespan effects of the single (Dcfs), double (DcsfDnos, DbslADnos) and triple (DbslADnosDcsf) mutant strains ( Fig. 5a and Supplementary  Fig. 5) correlated well with the proficiency or deficiency in biofilm formation and worm gut colonization of each strain, and there was no significant difference in spore robustness or worm chemotactic behaviour toward these mutant cells compared with the NCIB3610 strain ( Supplementary Fig. 1).
The  Table 8). Accordingly, the use of FUdR did not interfere with the effects produced by the DbslA, Dnos and Dcfs B. subtilis mutant strains on C. elegans survival (Supplementary Table 9).
To confirm that CSF production constitutes a novel prolongevity factor for C. elegans, we supplemented the worm diet with pure CSF (100 nM) in Petri dishes containing OP50 (Fig. 5b) or LC3611 (DbslADnosDcsf; Fig. 5c) bacterial cells as a worm food source. Interestingly, the added CSF increased the lifespan of C. elegans fed on OP50 cells, with average survival values of 18.68 days (95% CI: 17.44-19.92 days) and 23.98 days (95% CI: 22.14-25.85 days; equality, P ¼ 0.0001 by log-rank test) in the absence or presence of CSF supplementation, respectively ( Fig. 5b and Supplementary Table 8). Similarly, the addition of CSF to C. elegans fed on the B. subtilis mutant strain LC3611 (DbslADnosDcsf) increased the worm lifespan to levels even higher than those obtained when DbslADnos cells (strain RG3611, proficient in CSF production) were used as a food source (Fig. 5c) Table 8). In sum, these results confirm the QS molecule CSF as the third factor that contributes to the lifespan extension effect of B. subtilis on C. elegans (Table 1B). In addition, these results strongly suggested that the positive lifespan effect of B. subtilis on C. elegans is primarily, or exclusively, dependent on the three properties studied here (that is, NO, CSF and biofilm production), which produced positive synergic effects on worm survival.
Using Cox proportional hazard model (Supplementary  Table 10), we confirmed that the biofilm formation proficiency (absence or presence of bslA activity) constitutes the more important B. subtilis property responsible for extending worm longevity. Worms fed on biofilm-deficient B. subtilis have a risk    Table 10). Interestingly, this statistical analysis showed that the effects of NO and CSF on lifespan were not additive and independent one from the other but showed the existence of a genetic dependence between nos and csf. Up to now, we do not know the molecular reason for this bacterial gene interaction in the worm gut environment but it shows up the complexity of the mechanism responsible for the extended host longevity. We are currently investigating this interesting finding.
The B. subtilis biofilm enhances the production of NO and CSF. The cell-to-cell communication and division of labour that occurs inside the biofilm produce significant differences in gene expression compared with the gene expression pattern of cells grew under planktonic conditions 16,19,23,58 . In this sense, the levels of NO and CSF produced in B. subtilis cultures have been measured only under planktonic growth conditions 62 . To determine whether NO and/or CSF production is affected in B. subtilis cells that develop as a biofilm, we grew NCIB3610 cells under planktonic and biofilm supporting conditions and measured the production levels of the two prolongevity molecules. As shown, the production levels of NO (Table 2) and CSF ( Table 2) were significantly enhanced when B. subtilis was grown as a biofilm compared with the levels of NO and CSF produced under planktonic conditions. These results ( Table 2 and Figs 2 and 3) reinforce the importance of the biofilm formation proficiency in the longevity of C. elegans.
The results shown in Table 2 also suggest that B. subtilis viability inside the worm gut would be an important prerequisite for the prolongevity effect. Only alive (and not dead) vegetative B. subtilis cells, originated from worm gut germinated spores, are able to form a biofilm and enhance production of the antiaging molecules NO and CSF. If this interpretation is correct, there should be a difference in the magnitude of the prolongevity effect between worms fed on alive or dead B. subtilis cells. To test this hypothesis, we fed C. elegans on dead and alive B. subtilis spores to measure the magnitude of the produced lifespan. Interestingly, B. subtilis must be alive in order to produce its beneficial prolongevity effect (Supplementary Fig. 6 Fig. 6 and Supplementary Table 11).
Accordingly, dead pathogenic bacteria (that is, killed P. aeruginosa) used to feed C. elegans increase worm survival in comparison with live pathogenic bacteria used as a food source. In addition, E. coli strains, including the 'non-virulent' OP50 strain, are mildly pathogenic to aging C. elegans 14,15,29 . We found that dead OP50 cells effectively prolonged C. elegans survival (21% lifespan increase), with average survival of 24.73 days (95% CI: 23.23-26.23 days; equality, P ¼ 2.4 Â 10 À 7 by log-rank test) in comparison with live OP50 cells (average survival, 20.51 days; 95% CI: 19.35-21.67 days) used as food sources. Overall, these results confirmed that B. subtilis constitutes a beneficial flora for C. elegans and showed that the prolongevity effect of this bacterium depends on its viability inside the worm gut.
B. subtilis downregulates the insulin-like signalling pathway.
The lifespan of C. elegans is subject to regulation by conserved signalling pathways and transcription factors that sense stress, environmental cues and nutrient availability 29 . Dietary restriction (DR) and the insulin-like signalling (ILS) pathway are central for the regulation of longevity in different animal models, including C. elegans 1,2,65-68 . These longevity regulatory pathways Biofilms of the wild-type undomesticated strain NCIB3610 were developed in MSgg and NGM as indicated in the 'Methods' section. After incubation for 48 h at 25°C, the developed biofilms were disrupted as described previously 23,24 and assayed for their NO-derivate species (nitrites and nitrates) content 28 and CSF production levels, which are measured as levels of b-galactosidase activity driven by the promoter of the csf gene 62   (DR and ILS) converge on the positive and negative regulation of the transcription factors DAF-16 and HSF-1, repectively [68][69][70] .
To determine how B. subtilis affects the signalling pathways involved in the regulation of C. elegans longevity, we fed daf-16 (mu86) and hsf-1 (sy441) mutant worms (strains CF1038 and PS3551, respectively, Supplementary Table 1) with wild-type B. subtilis and isogenic mutants showing a decreased effect on wild-type N2 worm lifespan (DbslA, Dnos and Dcsf cells). Interestingly, the lifespan extension effect of wild-type B. subtilis on daf-16 and hsf-1 worms was approximately half of the lifespan extension effect on wild-type N2 worms (Fig. 6a,b and Supplementary Table 12). While the average lifespan extension effect of NCIB3610 and JH642 cells (compared with OP50 cells) on wild-type (N2) worms was 55.54 and 27.13%, respectively (Table 1A), in daf-16 worms, these percentages were 23.79 and 12.47%, respectively ( Fig. 6a and Supplementary Table 12). Similarly, the lifespan extension effect on hsf-1 worms fed on NCIB3610 or JH642 cells (Fig. 6b) was partial in comparison with the lifespan extension effect on wild-type worms fed the same type of B. subtilis cells. In this case, the average lifespan extension effect produced by NCIB3610 and JH642 cells (compared with OP50 cells) on hsf-1 worms was 35.76% and 15.71%, respectively ( Fig. 6b and Supplementary Table 12). Because HSF-1 and DAF-16 are active in daf-16 and hsf-1 worms, respectively 69 , our results suggest that the prolongevity effect of B. subtilis on C. elegans depends on the activity of DAF-16 and HSF-1 (Table 1C,D).
How does each B. subtilis prolongevity property depend on DAF-16 and/or HSF-1 activities? To answer this key question, we measured the survival of N2, daf-16 and hsf-1 worms fed on NCIB3610-isogenic mutants defective in biofilm (DbslA), NO (Dnos) and CSF (Dcsf) production (Fig. 7). The abolishment of each of the B. subtilis prolongevity properties was responsible for partial lifespan decreases of 26.35, 12.52 and 10.96% for wildtype N2 worms fed on DbslA-, Dnosand Dcsf-B. subtilis cells, respectively, compared with the use of wild-type NCIB3610 cells ( Fig. 7a and 16 and hsf-1 worms, respectively, when fed on B. subtilis cells defective in NO synthesis (Dnos) in comparison with feeding daf-16 and hsf-1 worms wildtype NCIB3610 cells (Fig. 7b,c and Supplementary Table 13). Finally, survival decreased by 6.72% (average value, 16.37 days; 95% CI: 15.20-17.54 days) and 4.61% (average value, 16.55 days; 95% CI: 15.35-17.75 days) in daf-16 and hsf-1-worms, respectively, when fed on Dcsf B. subtilis cells compared with feeding wild-type NCIB3610 cells to daf-16 and hsf-1 worms (Fig. 7b,c and Supplementary Table 13). In addition, we confirmed that the lifespan effects produced by B. subtilis NCIB3610 and its isogenic derivates (Dnos, DbslA and Dcsf) on wild-type daf-16 and hsf-1 worms were not affected by the inclusion of FUdR in the assays (Supplementary Table 9). In sum, these results indicated that each of the three anti-aging properties of B. subtilis require the activity of HSF-1 and DAF-16 (ref. 69; Fig. 7 and Supplementary Tables 12,13) and reinforced the former conclusion suggesting that the biofilm formation proficiency of B. subtilis is the most important property causing the beneficial effect on C. elegans longevity.
As mentioned, host longevity is mainly regulated by the signalling pathways of DR and the nutrient-sensing pathway ILS that converge in the regulation of DAF-16 and HSF-1 activities [68][69][70] . Interestingly, ILS could mediate part of the DR prolongevity effect because DR downregulates the activity of ILS 1,2 . Therefore, we tested the putative participation of the ILS pathway in the prolongevity effect of B. subtilis on C. elegans. DAF-2 is the C. elegans homologue of the insulin-like receptor   . These lifespan values represent a 9.49% increase in the longevity of NCIB3610-fed daf-2 worms (Fig. 8a) versus a 55.54% average increase in the longevity of NCIB3610-fed N2-worms compared with OP50 cells used to feed daf-2 and N2 worms, respectively (Table 1A,E). These results strongly indicated that the anti-aging effect of B. subtilis on C. elegans is mainly transduced through the ILS pathway.
Aging is a multifactorial and poorly understood process characterized by progressive impairment of the host response to stresses and general cellular deterioration of key metabolic pathways 1,2 . Intensive work in different animal models (that is, yeasts, worms, flies, mice and monkeys) has identified a number of factors that promote longevity. Restricting food intake (DR), decreasing insulin/IGF-1 signalling (IIL), slowing mitochondrial respiration, reducing germline function or lowering temperature can all extend lifespan 1,2 . In addition, probiotics are associated with a broad spectrum of positive effects on host health, including positive effects on host longevity [9][10][11][12][13][14][15]21,22 .
To our knowledge, the present study is the first to formulate and address the question of how the bacterial biofilm might affect the prolongevity effect of beneficial gut bacteria on the host. Our evidence directly implicates the biofilm formation proficiency as the primary cause of the increased lifespan and stress resistance (that is, healthy longevity) of C. elegans when fed B. subtilis (Figs 1-3 and Table 1).
In the case of the host (C. elegans), the benefits of the biofilm are evident: lifespan extension, higher stress resistance and probably competitive exclusion of pathogens 2,21,22,42 . In addition to our results, it has been reported that B. subtilis strains proficient in biosurfactan production (that is, fengycin) are able to protect C. elegans from infection by bacterial pathogens 42 . In addition, the ability to establish a biofilm in connection with the host cells would allow for a stronger and closer interaction between the bacterium and the host 9,44 . The finding that exogenously added CSF (Fig. 5) and NO 28 extend the lifespan of C. elegans fed E. coli cells supports the notion of the underlaying phenomenon of worm-bacteria QS. This dual microbial-worm interaction would allow the bacterium to colonize and establish a multicellular biofilm in the friendly environment of the worm gut mucosa, a similar scenario to the formation of stable biofilms during the beneficial bacteria-plant interaction of some Bacilli (that is, B. subtilis and B. amyloliquefaciens) with PGPR (Plant Growth Promoting Rhizobacteria) activity on plant roots (rhizospheric biofilm) 23 . In addition, the bacterial properties and behaviour in a multicellular biofilm differ from the behaviour and properties of bacteria grown as individuals under planktonic conditions [16][17][18] . Specifically, our results demonstrated the importance of the biofilm for the success of B. subtilis to healthy C. elegans gut colonization (Figs 2 and 3), increase the production of the anti-aging molecules NO/CSF (Table 2), and increase worm longevity (Figs 1,4 and 5). Future studies would address how the production of the anti-aging molecules NO and CSF is enhanced in B. subtilis cells grown as a biofilm (Table 2) and whether this higher NO and CSF production might be influenced by signals originating from the host (Supplementary Table 10).
The present work also indicated that the prolongevity effect of B. subtilis on C. elegans requires the activity of FOXO/DAF-16 and HSF-1 (Figs 6 and 7 and Table 1). At the top of the regulatory pathways that modulate the activity of HSF-1 and DAF-16 in C. elegans are the ILS pathway (homologous to the insulin/IGF-1 pathway present in humans) and the process of DR 1,2,29 . We showed that the prolongevity effect of B. subtilis on C. elegans primarily depends on the functionality (downregulation) of the ILS pathway (Fig. 8a). This result points to the ILS pathway as the target of the B. subtilis biofilm to increase host longevity. Interestingly, it has been observed that healthy human centenarians likely have IGF-1 receptor genetic variants associated with reduced function, an intriguing observation that correlates with our findings. As ILS pathway is part of the nutrient signalling pathway, it is also (negatively) controlled by DR a scenario that open the possibility that ultimately DR might sense and control the ILS-dependent prolongevity effect of B. subtilis (Fig. 8b) Table 1). The OP50 E. coli strain was obtained from the Caenorhabditis Genetic Center (CGC). As indicated, bacteria were grown in Luria-Bertani (LB) broth and Schaeffer's sporulation medium (SM) 30 . Sporulation was induced by the starvation method: B. subtilis strains were grown in SM medium at 37°C for 48 h (ref. 30). After this incubation period, the culture was heat-treated for 20 min at 80°C to kill vegetative cells 30 . To obtain pure spores, the heat-treated culture was treated three times with lysozyme (25 mg ml À 1 ; Sigma Co.), washed each time with cold water and centrifuged until 100% of the culture consisted of phase-bright spores. When indicated, the pure spore solution was autoclaved twice to kill spores. Wild-type N2 (Bristol) and mutant strains (obtained from CGC) worms were maintained at 20°C on Nematode Growth Medium agar (NGM) 28 plates seeded with OP50 E. coli or B. subtilis cells with or without supplementation with ampicillin (100 mg ml À 1 ), respectively. If necessary, the antifungal amphotericin (25 mg ml À 1 ; Sigma Co.) was also added to the NGM medium. The QS pentapeptide (ERGMT) was obtained from Thermo Scientific Co. (formerly Invitrogen Co.) and conserved at À 20°C until use.
Lifespan analysis. Lifespans were monitored at 20°C as described previously 28 . Briefly, to obtain a synchronized population, embryos were isolated by treating adult worm hermaphrodites with alkaline hypochlorite and allowing them to develop. In all the cases, late L4/yound adult stage worms were used at t ¼ 0 for lifespan analysis, and 16 mM 5-fluoro-2 0 -deoxyuridine (FUdR, Sigma Co.) was used to inhibit progeny growth. During the quantification of the number of dead/live worms, the investigator was blinded to the group allocation. All the experiments were repeated at least three times. Worms were considered dead when they ceased pharyngeal pumping and did not respond to prodding with a platinum wire. Worms with internal hatching were removed from the plates and excluded from lifespan calculations. Given that some temperature-sensitive mutations in some genes lead to the formation of dauer larvae, the lifespan experiments involving daf-2(e1370) were performed shifting the temperature from 15 to 20°C, the non-permissive temperature, at late L4/young adult stage 1 . For lifespan analysis with no FUdR, a synchronous population was obtained by sodium hypochlorite treatment of gravid hermaphrodites to obtain eggs that were then raised on standard NGM plates. Late L4/young adult stage worm were transferred to fresh plates in the complete absence of FUdR every 2 days, while any of them were fertile. Survival was scored as described above.
Lifespan after different stress assays. Thermotolerance assays were performed as described 28 . The worms were allowed to develop and grow on NGM agar plates at 20°C until they reached late L4/young adult stage. Then, the worms were shifted to 34°C as 5-day-old adults. In a typical experiment, NGM plates containing B50 worms per plate were used. Single plates were removed (by triplicate) from the high temperature at the time intervals shown in the figures, and the worms were scored for signs of life as described in the 'Lifespan Analysis' above. Osmotic, metal and oxidative stress assays were performed as described. The worms were allowed to develop and grow on NGM agar plates at 20°C until they reached late L4/young adult stage. Then, we transferred B70 worms as 5-day-old adults from agar plates to each 24-well plate containing 200 mM NaCl (for osmotic stress), 50 mM cadmium (for metal stress), 25 mM H 2 O 2 (for oxidative stress) and incubated the plates at 20°C. At different time intervals, the worms were scored for signs of life as described in the 'Lifespan Analysis' above. To calculate the median survival time, each data set was fitted to a Boltzmann sigmoid curve; the average±s.e.m. is presented in the figures.
Culturing bacteria from worms. The N2 C. elegans eggs were isolated using a solution of 10% commercial bleach and 1 N NaOH followed by four washes with M9 buffer (22 mM KH 2 PO 4 , 42 mM Na 2 HPO 4 , 85 mM NaCl and 1 M MgSO 4 ). Approximately 500 eggs were transferred to a 60 mm plate with NGM agar and incubated overnight at 20°C with agitation to allow L1 larvae to emerge. Then, approximately 500 L1 larvae per experiment were grown for 48 h on NGM plates (a time that allows worm development to reach the L4 larvae stage) seeded with OP50 E. coli cells (1 Â 10 5 cells per plate) or spores (1 Â 10 5 spores per plate) of each B. subtilis strain. At different incubation times, 50 worms were transferred to Eppendorf tubes containing M9 buffer, and 1% Triton X-100. The worms were treated with 25 mM levamisole to induce temporal paralysis, superficially sterilized with 3% commercial bleach for 15 min and washed three times with M9 buffer. After the worms were surface sterilized, worms devoid of outside bacteria were disrupted using a pellet pestle (Sigma Co.), centrifuged and resuspended in 500 ml M9 buffer. Finally, 50 ml of each cell suspension was used to prepare serial dilutions of the bacteria before counting. To this end, 100 ml of the appropriate serial dilutions was spread with a Drigalski scraper on LB Petri dishes. The number of colony-forming units (CFUs) was determined after 24 h of incubation at 37°C. To determine the number of spores, samples were heat-treated at 80°C for 20 min before spreading. This treatment kills the vegetative cells, while the spore cells remain alive 33 . The number of CFUs obtained before and after the heat treatment represents an estimation of the number of vegetative and spore cells, respectively, inside the worm.
b-Galactosidase assay. b-Galactosidase activity was assayed in B. subtilis cells harbouring lacZ fusions, and the specific activity was expressed in Miller Units (MU) 30 . To determine gut-derived b-galactosidase activity, 1-day-old L4 worms were separately fed on NGM plates with domesticated and undomesticated B. subtilis that harboured the p srf-lacZ reporter fusion. At different times, approximately 100 worms were taken from the NGM plates containing each NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14332 ARTICLE NATURE COMMUNICATIONS | 8:14332 | DOI: 10.1038/ncomms14332 | www.nature.com/naturecommunications B. subtilis type, washed, disrupted and processed as indicated in 'Culturing bacteria from worms' to measure the b-galactosidase expression level of each worm extract.
Microscopic visualization of bacterial cells colonizing the nematode intestinal tract. Day 5 adult worms fed B. subtilis strains carrying an integrated epsA::gfp or bslA::gfp reporter fusion were pipetted out from plates using M9 buffer and 0.25 mM sodium azide (Sigma Co.) and covered with a glass coverslip. Bacterial colonization of the nematode digestive tract was observed using an Olympus FV1000 laser confocal scanning microscope. Images were analysed with Olympus software and Fiji software.
Bacterial pulse-chase experiment. Day 5 adult worms fed different B. subtilis strains (JH642 or NCIB3610) were washed and transferred to NGM plates containing the indicated new bacterial cells (that is, OP50, JH642 and/or NCIB3610) at a titre of 1 Â 10 6 CFUs per plate as a new food source. At different times (indicated in the figure legends), the number of bacteria colonizing the worm intestine was calculated as described in the 'Culturing bacteria from worms'.
Determination of NO and CSF. B. subtilis strains were grown under planktonic or biofilm-supporting conditions (with or without shaking, respectively) in MSgg 31 or NGM 28 broth at 25°C for 36 h. Then, the supernatants of each culture were taken and filter sterilized. Appropriate aliquots of each cell-free supernatant were assayed for the presence of NO-derived metabolites (NO 2 À and NO 3 À ) as reported previously 28 . For determining the b-galactosidase activity driven from the csf-promoter, the cells were grown as a biofilm or planktonically, concentrated by centrifugation and treated as reported previously 30 . The production levels of NO and CSF are expressed as mmol and MU, respectively, per 1 Â 10 8 CFUs counted from the cultures developed under planktonic and biofilm-supporting conditions.
Self-brood size and egg production rate. These experiments were performed essentially as described 28 . The eggs from strain N2 nematodes were isolated, treated with hypochlorite and incubated for 20 h at 20°C in S-buffer (0.5 M KH 2 PO 4 , 0.5 M K 2 HPO 4 and 0.1 M NaCl). A synchronized population of L1 arrested worms was then placed on NGM agar plates seeded with OP50 E. coli or NCIB3610/JH642 B. subtilis cells. Five late L4/young adult stage animals were picked manually and transferred to a new plate. The worms were transferred twice a day to prevent overcrowding until egg laying ceased. The progeny was counted 3 days after removal of the parents. The experiment was performed in triplicate.
Postembryonic development and adult size. These experiments were performed as described previously 28 . Strain N2 worms were grown on NGM agar plates seeded with OP50 E. coli or NCIB3610/JH642 B. subtilis. Unstaged eggs were placed at 20°C and allowed to hatch for 4 h. Larvae that hatched during this period were placed singly on fresh plates and monitored every 5 h until they reached late L4/young adult stage. To measure the worm size, B20 randomly picked worms fed OP50 E. coli or NCIB3610/JH642 B. subtilis cells were anaesthetized with sodium azide, straightened and photographed. Their length in pixels was then compared with a 1 mm scale bar.
Chemotaxis assay. Batches of 20 age-synchronized worms (late L4/young adult stage) were transferred to a plate containing lawns of OP50 E. coli or JH642 or NCIB3610 B. subtilis with a diameter of B0.5 mm. The number of worms on each bacterial lawn was counted after 12 h of incubation at 20°C and the choice index was calculated. Chemotaxis of an individual animal was measured by watching the tracks that they left on the agar. The choice index was calculated as follows: Chemotaxis index CI: No. of worms in tested bacteria À No. of worms on OP50/Total number of used worms. If CI ¼ ( À 1.0) represents complete preference for OP50 E. coli, ( þ 1.0) represents complete preference for the test bacterium and (0.0) represents an equal distribution.
Statistical analysis. All assays were performed in at least duplicate or triplicate. Mean survival days, standard error, intervals of mean survival days with 95% confidence and equality P values to compare averages were calculated by logrank and Kaplan-Meier tests using OASIS programme. Cox' regression model was used to estimate the relative risk of worm death under the different B. subtilis treatments compared with the control group: h i (t) ¼ h 0 (t)exp(b 1 x i1 þ b 2 x i2 þ ... þ b k x ik ), where h i (t) and h 0 (t) denote the conditional hazard and baseline hazard rates, respectively; b i is the unknown parameter for treatment group; and x i takes values 0 and 1, being an indicator variable for two samples, the control and treatment group. Figures concerning the obtained quantitative information were prepared by Sigma Plot 10.0 or OriginPro 8.0 programmes. Data availability. The authors declare that all data supporting the findings of this study are available within the article and its Supplementary Information files, or are available from the corresponding author on reasonable request.