Commensal bacteria differentially shape the nutritional requirements of Drosophila during juvenile growth

The interplay between nutritional and microbial environments is one of the decisive environmental inputs that determine juvenile growth trajectory. Nutritional deficiencies contribute to developmental delays, and an immature gut microbiota is a hallmark of pathologies related to childhood undernutrition. However, how commensal bacteria modulate the impact of nutrition on juvenile growth remains elusive. Here, using gnotobiotic Drosophila melanogaster larvae independently associated with two major commensal strains, Acetobacter pomorumWJL (ApWJL) and Lactobacillus plantarumNC8 (LpNC8), we performed a large-scale, systematic nutritional screen based on larval growth in 40 different and precisely controlled nutritional environments. We combined these results with in silico metabolic network reconstruction to define the biosynthetic capacities of Drosophila germ-free larvae and the two commensals. We first establish that ApWJL or LpNC8 differentially fulfills the nutritional requirements of the ex-GF larva and parsed such difference down to individual amino acids, vitamins, other micronutrients and trace metals. We found that Drosophila commensal bacteria not only fortify the host’s diet with essential nutrients but in specific instances, they functionally compensate for host auxotrophies even if the bacteria fail to synthetize the missing nutrient. Our systematic work reveals that beyond the molecular dialogue engaged between the host and its commensal partners, Drosophila engage into an integrated nutritional network with its facultative commensals centered around the sparing and utilization of nutrients. We thus uncover a novel facet of the facultative nutritional mutualism engaged between Drosophila and its commensal bacteria, which allows the juvenile host to better cope with changes in nutrients availability during the critical phase of post-natal growth; hence ensuring optimal host fitness.

), Lp NC8 is able to produce 211 most amino acids from glucose or inner precursors. Exceptions are Phe, sulfur-212 containing (Cys, Met) and branched-chain amino acids (BCAA, Ile, Leu, Val). Indeed, 213 synthesis of Phe seems to be precluded in Lp NC8 since the prephenate dehydratase 214 activity (4.2.1.51) is not encoded by its genome (Table S1). Moreover, Lp NC8 has lost 215 the capability to reduce sulfate in sulfite, a necessary step for the integration of sulfur 216 in Cys biosynthesis [65]. Biosynthesis of Met seems to be feasible in Lp NC8 , but is 217 probably limited by sulfite production rate and consequently by Cys availability. As 218 previously reported, Lp NC8 is unable to produce the three BCAA, as four enzymatic 219 activities of this pathway are lacking (1. were identified in the genes of the operon. This suggests that Lp NC8 can be an Arg 232 autotroph, but does not exclude the possibility that Arg may become limiting in specific 233 conditions. Lp NC8 has the minimal gene requirements to produce Ala and Asp (Table  234 S1), but the biosynthesis of Ala and Asp invokes secondary metabolic routes. 235 Therefore, when possible, uptake of Ala and Asp from the culture medium should be 236 favored instead of de-novo biosynthesis. Similarly, biosynthesis of Thr is directly linked 237 to Asp and Cys and is probably very limited in Lp NC8 . Accordingly, all the Thr 238 interconnecting pathways are lost, thus limiting its production or uptake from the diet 239 for protein synthesis. 240 Regarding vitamin biosynthesis, Lp NC8 is able to produce folate, riboflavin and 241 thiamine, as well as all DNA bases including uridine and inosine ( Fig 1B and Table  242 S2). Lp NC8 cannot produce thiamine through the canonical pyrimidine pathway 243 because it lacks the 4.1.99.17 enzyme, but it seems possible through the pyrimidine 244 salvage pathway (2.1.7.49). Unlike Ap WJL , pyridoxine biosynthesis seems feasible in 245 Lp NC8 using an alternative route involving Thr phosphorylation. Conversely, 246 biosynthesis of nicotinate seems to be impossible. Indeed, the two first steps of its 247 Asp can easily become limiting in the contexts of high nutritional demand. Thus, Lp NC8 254 is expected to extract these amino acids from the diet, or its growth would be altered 255 when they are absent from the diet. Met synthesis seems to be possible, but is 256 dependent on Cys abundance. Finally, Lp NC8 can produce DNA bases, folate, 257 riboflavine, thiamine and perhaps pyridoxine, but not biotin, nicotinate, pantothenate, 258 choline and myo-inositol. 259 Collectively, our in silico metabolic network reconstruction shows that 260 D. melanogaster and its commensals have differential biosynthetic capacities. Indeed, 261 some of the complete biosynthetic pathways are only present in one organism while 262 others are present in two or all the three partners. However, for incomplete biosynthetic 263 pathways we did not detect potentially shared pathways between a commensal and its 264 host as observed between obligate mutualistic partners (Fig 1-2). 265 266

Experimental validation of commensal bacteria auxotrophies using holidic diets 267
In order to test experimentally the metabolic potential of Dm and its 268 commensals, predicted by our in silico analysis (see above), we adopted the exome 269 based FLYAA holidic diet (HD) [40]. We systematically removed a single component 270 at a time to generate 39 different fly nutritional substrates (henceforth named HDX, 271 X being the nutrient omitted), plus one complete HD medium. This medium can also 272 be prepared in a liquid version by omitting agar and cholesterol from the recipe. Liquid 273 HD can then be used to assess bacterial growth in 96-well plates, increasing the 274 experimental throughput. 275 We first assessed Ap WJL and Lp NC8 growth in each of the 40 different liquid HD 276 for 72 hours, using maximal optical densities (ODMax) as a readout (Fig 3A and Table  277 S3). In the complete HD, both Ap WJL and Lp NC8 grow well (Fig 3A,  found in sugar-rich niches such as flowers and fruits, but also in poorer niches such as 286 soil and water, where they need to synthesize all nutrients required for their own growth 287 [70]. These findings are corroborated by our genome-based in silico predictions (Fig  288   1). Furthermore, the in silico reconstruction predicted that Ap WJL would not be able to 289 synthesize choline, myo-inositol and pyridoxine, but we observed that Ap WJL  confirmed in vivo. The identified Arg auxotrophy was not surprising because, as 328 mentioned above, Arg is often described as essential to L. plantarum in high metabolic 329 demanding conditions even though all genes necessary for Arg biosynthesis are 330 present. However, auxotrophies of Lp NC8 to Thr, Ala, Asp, were not expected (Fig 3A,  331 denoted by "*"), even though these amino acids were predicted to be limiting (see 332 above). As mentioned previously, bacterial growth in liquid HD was assessed in 96-333 well plates using a microplate reader (see methods). Every cycle includes an agitation 334 step to homogenize the solution to improve OD reading accuracy. This agitation step 335 may oxygenate the media and thus negatively affects Lp NC8 growth in scarcity 336 conditions since L. plantarum strains are aerotolerant but optimal growth is achieved 337 under microaerophilic or anaerobic conditions [79]. To retest these unexpected 338 auxotrophies, we assessed Lp NC8 growth in liquid HDThr, HDAla, HDAsp in 15-mL 339 closed falcon tubes without aeration. After 72 hours of incubation we determined CFU 340 counts in each media ( Fig 3B). As predicted by our genomic analyses, Lp NC8 was able 341 to grow in each of the three deficient media in static conditions to the same extent as 342 in the complete HD ( Fig 3B). Therefore, Lp NC8 auxotrophies observed for Thr, Ala and 343 Asp in 96-well plates are likely an artifact due to excessive oxygenation. This could 344 also explain the poor growth of Lp NC8  and HDMn at the same rate as on a complete HD (Fig 4A). The absence of sucrose, 379 Tyr, inosine, Ca, Cu and Fe did not prevent pupae emergence, but increased their 380 developmental timing significantly (Fig 4A first  albeit with a severe developmental delay ( Fig 4B). Therefore, a complete Asn 393 auxotrophy is specific to our yw strain, which explains why we did not detect this 394 auxotrophy in our in silico analyses that were based on the genome sequence of the 395 Drosophila Reference Genome strain (Bloomington stock #2057). We therefore 396 sequenced the coding region of the enzyme AsnS that converts Asp to Asn in yw flies, 397 and did not detect any non-synonymous mutation ( Fig S1). Further studies may thus 398 be required to determine the origin of the Asn auxotrophy in our yw line on HD. 399 However, these results indicate that Asn is not an EAA per se but remains a limiting 400 NEAA, an observation that also applies to Tyr. Bacteria were grown in rich medium before association with larvae. Therefore, they 463 may have accumulated nutrients that can be used by the larvae to fulfill their nutritional 464 requirements. To test for the nutritional input brought by the initial bacterial inoculum, 465 we associated GF larvae with 10X heat-killed (HK) bacteria and measured D50 in 466 complete and deficient HDs ( Fig 4A fourth and fifth columns). In most cases, the D50 467 of larvae in HK and GF conditions were similar, and they both show striking delay 468 compared to the mono-associated larvae. Therefore, bacteria need to be metabolically 469 active to fulfill the larval nutritional requirements on HDs. However, we found some His, Lys, Met, Thr, Asn and most vitamins (except pyridoxine). Nonetheless, these 488 results suggest that bacteria can actively supply the nutrients lacking in the HD to the 489 larvae. This phenomenon is reminiscent of previous observations using conventional 490 and gnotobiotic hosts, in which microbial sparing of riboflavin or thiamine by differential 491 pantothenate and Trp, respectively, to sustain their own growth in the depleted HD but not sufficiently or in a manner inaccessible to the larvae and thus fail to fulfill larval 498 requirements for these nutrients. 499 Case 2: the bacteria do not synthesize a nutrient and they cannot fulfill larval 500 requirements 501 Expectedly, we observed that when bacteria do not synthesize a nutrient, they 502 do not fulfill ex-GF larvae requirements for this nutrient. For instance, Lp NC8 cannot 503 produce the BCAA, nor grow in their absence, and thus it cannot fulfill larval 504 requirements for these amino acids. In some depleted diets such as HDCholesterol, 505 HDCholine and HDPyridoxine, bacteria were able to grow ( Fig 3A) even though they 506 cannot synthetize these nutrients (See above and Table S2), and they failed to fulfill 507 the larvae requirements of these specific nutrients. This is observed for Ap WJL and 508 Lp NC8 on HDCholesterol. The likely explanation is that cholesterol is an animal sterol 509 but is dispensable for bacterial growth [72,84]. Similarly, on HDCholine and 510 HDPyridoxine, Lp NC8 is also unable to fulfill larval requirements ( Fig 3A) as it cannot 511 synthetize these compounds. 512 Case 3: the bacteria do not synthesize a nutrient but they can fulfill larval requirements 513 In most cases, we observe growth rescue by bacteria provision of the missing 514 nutrients, but there are interesting exceptions. Ap WJL is unable to synthesize de-novo 515 choline and pyridoxine (See above and Table S2). Surprisingly, it fulfills larval 516 auxotrophies on HDCholine and HDPyridoxine. This could be achieved either by a 517 functional compensation mechanism or via alternative biosynthetic pathways of these 518 components that are undetected by our metabolic reconstruction. In the case of 519 choline, the former seems to be the most plausible explanation, as Ap WJL may 520 synthesize other compounds that Drosophila can use to replace choline. As stated membrane composition towards increased content of phosphatidylethanolamine (PE) 523 and phosphatidylglycerol (PG) [71]. PE  on HDCholine may capitalize on ethanolamine or glycerol produced by Ap WJL to 527 compensate for the lack of choline in their diet. In the case of pyridoxine, Ap WJL may 528 fulfill larval requirements by either a functional compensation or through pyridoxine 529 biosynthesis by non-canonical pathways (see above). We reach the same conclusion 530 regarding nicotinate. Both Ap WJL and Lp NC8 grow well on HDNicotinate and also can 531 fulfill the larval requirements in this vitamin even though our metabolic reconstruction 532 predicts that they cannot synthesize it (see above). 533 Moreover, Lp NC8 cannot grow in the absence of Phe (Fig 2A). The genomic 534 analyses point to the possible loss of the gene coding for the enzyme prephenate 535 dehydratase (4.2.1.51), the penultimate step on Phe biosynthesis. Yet Lp NC8 can fulfill 536 larval requirements for Phe ( Fig 4A). We wondered if the Phe auxothrophy we 537 observed in 96-well plates (Fig 3A) was due to the oxygenation generated by the 538 agitation through OD readings, as for Thr, Ala and Asp (Fig 3B). To test this, we set 539 cultures of Lp NC8 in HDPhe in static 15-mL closed falcon tubes and assessed bacterial 540 growth after 3 days of culture. In contrast to agitation, Lp NC8 grows in HDPhe ~1x10 2 541 times in static conditions (Fig 4C), whereas in the complete media (Fig 3B), Lp NC8 542 grows ~2x10 4 times. Therefore, these results indicate that the rescue of larvae 543 developmental timing by Lp NC8 in HDPhe is still mediated by bacterial nutrient supply. 544 However, the poor growth of Lp NC8 in HDPhe suggests the existence of an alternative 545 pathway for Phe biosynthesis in absence of the prephenate dehydratase ( Fig 1A). As 546 suggested by Hadadi et al. [87], Phe might be produced from L-arogenate using a 547 derivative catalysis through the 2.5.1.47 activity which is encoded in Lp NC8 by the cysD 548 gene (nc8_2167). 549 A second such interesting case is larval developmental rescue by Lp NC8 in 550 HDCys. Lp NC8 is auxotroph for Cys (Fig 3A), even in static conditions (Fig 4C). Lp NC8 -551 associated larvae develop faster than GF larvae, though GF larvae are not auxotroph 552 to Cys (Fig4A). This beneficial effect of Lp NC8 on ex-GF larvae development in HDCys 553 is similar to what is observed on a complete HD (Fig 4A, first row) and therefore 554 probably reflects the basal nutrient-independent growth-promoting effect of Lp NC8 , that 555 we previously reported and which relies on the molecular dialog between Lp NC8 and its 556 host ( Fig S2) [23]. Taken together, our results indicate that Lp NC8 is able to grow in 557 HDCys only in the presence of Dm larvae. To test this hypothesis, we assessed Lp NC8 558 growth in solid HDCys in the absence (Medium Load) and the presence (Niche Load) 559 of Dm larvae (Fig 4D). Without larvae (ML), Lp NC8 grew one log above the inoculum 560 level in solid HDCys (Fig 4D, Medium load). This minimal growth in solid HDCys 561 could be due to the Cys reserves from Lp NC8 growth in rich media prior to inoculation, 562 or from contaminants in the agar and cholesterol added to prepare the solid HD. 563 Interestingly, in the presence of larvae in the HDCys (NL), Lp NC8 CFU counts 564 dramatically increased over time, reaching ~10 8 CFU/tube at day 6 (net growth: ~3x10 4 565 times inoculum. Fig 4D). These results suggest that in HDCys larvae support Lp NC8 566 growth, probably by supplying Cys or a precursor/derivative, which in turn promotes 567 larval development and maturation. This observation extends the recent demonstration 568 that Dm and L. plantarum engage in a mutualistic symbiosis, whereby Dm benefits the 569 growth of L. plantarum in their shared nutritional environment [30]. Here we discover 570 that Cys is a Dm symbiotic factors also referred to as "bacteria maintenance factors" 25 Case 4: mineral and metal traces 573 We observed that both Ap WJL and Lp NC8 could compensate for Cu, Fe and Zn 574 deficiencies (Fig 4A, second and third column). Requirements in Cu and Fe were also 575 fulfilled by HK bacteria (Fig 4A, fourth and  provide amino acids to their host by releasing them in the HD. To directly tested this 614 hypothesis, we cultured Ap WJL and Lp NC8 in liquid HD lacking each EAA Fly and 615 quantified the production of the corresponding missing EAA Fly . We focused on EAA Fly 616 whose deficiency could be compensated by the commensals in our DT experiments 617 (Fig. 4A). In these assays, Ap WJL was cultured under agitation and Lp NC8 cultures were 618 grown in both agitated and static conditions (See Methods). After three days, we 619 quantified the amino acids concentration from bacterial supernatants using high 620 pressure liquid chromatography (HPLC). We quantified amino acid production by HDMet, HDPhe, HDThr, and HDVal and observed accumulation all missing AAs 623 expect for Lys and Met (Fig 5A). For Lp NC8 , we analyzed the supernatants of HDs that 624 support Lp NC8 growth under agitation ( Fig 3A): HDHis, HDLys and HDMet. We also 625 analyzed supernatants from static conditions, HDHis, HDLys, HDMet, HDPhe 626 and HDThr. Surprisingly, from all tested conditions we only detected His 627 accumulation in the supernatant of Lp NC8 grown on HDHis under agitation (Fig 5B). 628 We did not detect Lys and Met in Ap WJL culture supernatant or Lp NC8 culture under 629 agitation supernatant nor His, Lys, Met, Phe or Thr in Lp NC8 static culture supernatants. 630 However, Ap WJL or Lp NC8 can both fulfill larval requirements in a HD lacking these 631 nutrients ( Fig 4A). We only analyzed supernatants after 72h of growth, it is therefore 632 possible that we missed the peak of accumulation of the targeted AA, which may have 633 taken place at another time point during the growth phase. Also, Ap WJL and Lp NC8 may 634 only secrete precursors or catabolites of these AA that we did not target in our analysis. 635 Such AA derivatives may also be used by the larvae to compensate for the lack of the 636 cognate amino acids in the diets. However, we detected Arg, His, Ile, Leu, Phe, Thr 637 and Val production by Ap WJL and His by Lp NC8 , a production which correlates with the 638 respective ability of Ap WJL and Lp NC8 to compensate for the lack of these AAs in the 639 respective depleted HD. Of note, the concentration of newly synthetized amino acids 640 accumulating in the supernatant is low compared to their concentration in a complete 641 amino acids to the larvae is probably a continuous process, which can be stimulated 643 upon uptake and transit through the larval intestine. Thus, amino acids are directly 644 supplied to the larvae and will fulfill its nutritional requirements without the need to 645 accumulate in the surrounding media. Altogether, our results show that Ap WJL and results confirm our hypothesis that Dm commensal bacteria Ap WJL and Lp NC8 produce 648 these EAA Fly while growing on HDEAA Fly . When associated to Dm larvae, Ap WJL and 649 Lp NC8 will therefore supply these amino acids to the larvae, allowing larval development 650 on these deficient media as observed upon mono-associations (Fig 4A). 651 652

Conclusion 653
In this study, we have unraveled the interactions between the microbial and the 654 nutritional environment of Drosophila, as well as the functional importance of these 655 interactions for Drosophila juvenile growth. We systematically characterized, both in 656 silico and in vivo, the biosynthetic capacities of growing GF larvae and two 657 representative commensal strains of bacteria (Ap WJL and Lp NC8 ). We show that both 658 commensals, each in its unique manner, alleviate the nutritional constraints in the 659 environment to accelerate host growth and maturation in diets depleted of essential 660 nutrients (Fig 6). The capacity of a bacterium to fulfill the larval requirements in a 661 specific nutrient correlated with its metabolic activity and in most cases its ability to 662 produce the missing nutrient. In contrast to obligate symbioses, our results highlight 663 the clear separation between the metabolic pathways of the host and its commensals 664 and reveal a particularly integrated nutritional network between the insect and its 665 facultative commensals around the sparing and utilization of nutrients. 666 Importantly, we further demonstrate that active nutrient provision by metabolite 667 biosynthesis not only evoke the canonical pathways, but also non-canonical pathways 668 such as the ones described here, for example nicotinate by both Ap WJL and Lp NC8 , 669 pyridoxine by Ap WJL , and Phe by Lp NC8 . Interestingly, we also detected two cases 670 where nutrient compensation is not explained by a direct nutrient supply: (1)  In some cases, the in silico prediction of bacterial biosynthetic capabilities were 692 incongruent with our in vivo assessment of bacterial auxotrophies (Table S6). Such 693 seeming discrepancy served as an entry point for us to discover novel phenomena and 694 interactions that would have been missed had we only adopted a single approach. One 695 such interesting example is Asn auxotrophy unique to the Dm yw line in GF conditions. 696 Another one is the larval provision of Cys (or its derivatives) to Lp NC8 to maintain a 697 mutualistic nutritional exchange between host and commensal. Previously, 698 combination of in silico and in vivo approaches has been successfully used for bacteria 699 After autoclaving at 120°C for 15 min, the solution was allowed to cool down at room 747 temperature to ~60 °C. Acetic acid buffer and stock solutions for the essential and non-748 essential amino acids, vitamins, nucleic acids and lipids were added. Single nutrient 749 deficient HD was prepared following the same recipe excluding the nutrient of interest 750 (named HDX, X being the nutrient omitted). Tubes used to pour the HD were 751 sterilized under UV for 20 min. HD was stored at 4°C until use, for no longer than one 752 week.  Table S4. Since larvae are cannibalistic and 815 can find missing nutrients by eating their siblings [108,109] therefore we excluded 816 replicates with low egg-to-pupa survival (<25%,i.e n<10). Moreover, we considered 817 that larvae failed to develop in one condition if the mean egg-to-pupa survival of the 818 five replicates was inferior to 25% (for details in egg-to-pupae survival, see Table S5). shown to support their growth (Fig 1A, B) and in which they fulfill larval requirements. replicates by assay. Each dot represents an independent replicate. We performed two-951 ways ANOVA followed by post-hoc Sidak's test. ns: non-significant, **: p-value<0,005. 952    Table S1 and Table  1000 S2). Can partner A grow in the absence of nutrient X?: auxotrophy observed in vivo 1001 (from Fig 3A-B). Can bacterial partner A promote larval growth on HD ΔX?: in vivo 1002 complementation of ex-GF larvae requirements (from Fig 4A)

Figure 3
Growth X-times inoculum (Log)    aa= Dm non-essential nutrient, can be produced by the bacteria.
aa= Dm non-essential nutrient. Compensation by the bacteria without synthesis= Functional compensation.
aa= Dm non-essential nutrient. Possible synthesis by the bacteria through undetected pathways.