Adaptation to chronic malnutrition leads to reduced dependence on microbiota in Drosophila

Numerous studies have shown that animal nutrition is tightly linked to gut microbiota, especially under nutritional stress. In Drosophila, microbiota are known to promote juvenile growth, development and survival on poor diets, mainly through enhanced digestion leading to changes in hormonal signaling. Here we show that this reliance on microbiota is greatly reduced in replicated Drosophila populations that adapted to a poor larval diet in the course of over 170 generations of experimental evolution. Protein and polysaccharide digestion in these malnutrition-adapted populations became much less dependent on colonization with microbiota. This was accompanied by changes in at least some targets of dFOXO transcription factor, which is a key regulator of cell growth and survival. Our study suggests that some metazoans have retained the evolutionary potential to adapt their physiology such that association with microbiota may become optional rather than essential.


30
Nutrient availability is a major factor limiting survival, growth and reproduction of 31 many animal species 1 , resulting in natural selection for adaptation to cope with nutritional 32 stress. Yet, little is known about evolutionary adaptations that help juvenile animals not only 33 to survive, but also grow, develop and reach maturity under chronic nutrient shortage.

34
However, recent studies point to a particular importance of gut microbiota in coping with subsample of ambient bacteria growing on their food (decomposing fruits) 3 . Nonetheless, as "germ-free" (GF) flies were fed heat killed inoculum to control for potential effect of bacteria 85 as food.

86
Larvae of our Selected populations had previously been reported to develop faster 87 and survive better than Control larvae on poor diet (but not on standard diet) 10,11 , a 88 manifestation of their evolutionary adaptation to the poor diet; however, in those studies the 89 colonization of the larvae by microbiota was not controlled and not assessed. We 90 hypothesized that the improved performance of Selected larvae on the poor diet is at least in 91 part mediated by an improved ability to benefit from interactions with microbiota. If so, one 92 would predict that their superiority over Control larvae would diminish if they were deprived 93 of the help of microbiota, i.e., in a GF state. To test this prediction, we compared the length of 94 larval development and survival of Selected and Control populations in a GF state and when 95 experimentally colonized with microbiota collected from adult feces. On the poor food, while 96 Control larvae colonized with microbiota developed 40% faster and were three times more 97 likely to survive than their GF siblings, the corresponding effect of microbiota treatment on 98 Selected larvae was much smaller (Fig 1A, B). On the standard food, the effect of microbiota 99 on development and survival was markedly smaller (Fig S1A, B). Thus, while in the GF state 100 the Selected larvae took 30% less time to pupate on poor diet and were about three times as (which are clustered together in the genome and reported to have a very localized expression 140 in the gut 13 ) exhibited the opposite pattern, i.e., were downregulated by microbiota (Fig 2C).

141
Out of the 11 proteases, we identified two (Jon66Cii, CG18180) whose mRNA levels were 142 consistently higher in Selected populations compared to Controls; we also observed that 143 CG8299 had higher expression in Control than selected populations (Fig 2C). Trends for 144 differences between Selected and Control populations could also be observed for several 145 other proteases (all three Trypsins, CG18179, Jon65Ai, Jon44E, Jon99Ci, Fig 2C), but they 146 were not sufficiently consistent between time points or replicate populations to be statistically 147 significant. The digestive proteases are likely to some degree functionally redundant, and thus 148 it is conceivable that evolution would achieve functionally similar changes in digestion by 149 targeting different genes in different replicate populations, making detection of a signature of 150 evolution in a gene-by-gene analysis difficult. Therefore, we analyzed the entire protease 151 expression dataset with multivariate analysis of variance (MANOVA) and Principal 152 Component Analysis (PCA). The correlation circle clearly confirmed that the levels of 153 expression of the three Trypsins were positively correlated and well separated from other 154 proteases (Fig 2B right). This suggests that these two groups of proteases are regulated by 155 different processes and/or may have a different function within the gut. GF and microbiota-156 colonized larvae were clearly separated by the 1 st PC, with Selected and Control populations Given that our poor diet is low in carbohydrate as well as protein content, we next 168 asked if carbohydrate digestion is also different between Selected and Control populations 169 and if it is differentially influenced by microbiota. About 30 % of carbohydrates in both poor 170 and standard diet consist of polysaccharides (starch) from the cornmeal (the rest are sucrose 171 and glucose). Polysaccharide digestion occurs as a two-step process whereby starches are first 172 broken-down to disaccharides by amylases before being hydrolyzed to monosaccharides.

173
Alpha-amylase activity is under direct negative regulation by glucose concentration in 174 Drosophila larvae, which occurs at the transcriptional level. Amylase activity is therefore 175 expected to be lower in larvae with higher glucose concentration 14,15 . We quantified amylase 176 activity rates in Selected and Control larvae in both colonized and GF states. Microbiota had a 177 striking effect on how amylase activity (again normalized to total larval protein content) 178 changed over time: while it declined between an early and late 3rd stage in the microbiota-179 colonized larvae, it increased sharply during the corresponding developmental period in GF 180 larvae (slope difference p < 0001, Fig 3A). Because no such increase is observed for protease 181 activity (Fig 2A), it implies that GF larvae upregulate their investment in polysaccharide 182 digestion relative to protein digestion towards the end of their development. Irrespective of 183 these temporal changes, GF Selected larvae consistently showed three-fold lower amylase 184 activity than GF Control larvae of the same stage (blue symbols in Fig 3A); this difference is 185 much smaller and non-significant in microbiota-colonized larvae (orange symbols in Fig 3A).

186
Thus, we again observed a pattern of interaction such that the difference due to evolutionary 187 history was more pronounced in germ free than in microbiota-colonized state. However, fast 188 development and high survival on poor diet (Fig 1) were associated with lower amylase 189 activity. This implies that increased amylase activity is a sign of nutritional stress. Given the 190 negative regulation of amylase activity by glucose concentration 14,15 , these results suggest 191 that Control larvae may have lower glucose levels than Selected larvae under GF conditions.

192
To verify if the pattern we observed is regulated at the transcriptional level we 193 quantified amylase transcript levels in the guts. We analyzed expression of two amylases. C). Under GF condition, Amy-P levels were higher in Control populations than in Selected 196 populations, but no significant difference was detected in Amy-D levels (Fig 3C). Given that 197 relative expression abundance of Amy-P is much higher than Amy-D (roughly 20 times, Fig   198   3C), Amy-P is likely to be the major gene contributing to the amylase activity pattern that we 199 observed earlier (Fig 3A). Even though Amy-P expression is reduced by microbiota and is 200 expressed at lower levels in Selected than Control populations, the expression pattern does 201 not fully explain what we observe for amylase activity, and other regulatory mechanisms (e.g.

202
cAMP levels 14 ) may also play a role in regulating amylase activity.

203
In the gut, glucose is generated through the hydrolysis of maltoses by maltases. If 204 amylase activity is lower in Selected populations and upon microbiota colonization because 205 of glucose concentration in the gut and/or hemolymph, maltase activity is predicted to be 206 higher in these conditions. To check this we also analyzed expression of four maltase genes.

207
In agreement with this prediction, we observed a high expression of maltases in Selected 208 populations for Mal-A1, -A3 and -A4, although not for Mal-A8 (Fig 3C).  Mal-A1, -A8 (Fig 3C). Mal-A4 exhibits this trend only at late 3 rd instar but this is not 211 statistically significant due to high variation among populations (Fig 3C). Mal-A3 expression 212 is rather induced in Selected populations upon colonization, and remains unchanged in the 213 Control ones (Fig 3C). To spot the general trend among these carbohydrate-digesting 214 enzymes we performed multivariate analyses. We observed a clear separation between the 215 evolutionary regime, colonization status and developmental stage (Fig 3B left, Table S3).

216
However, we observed only a marginally significant interaction between the evolutionary 217 regime and developmental stage, and no interactions between other factors (Fig 3B left, 218 Table S3). Furthermore, PCA correlation circle on carbohydrate digesting enzymes shows 219 that amylase and maltase expression patterns are uncorrelated (Fig 3B right). Altogether,

243
Could then the differences between Selected and Control populations in the effects of 244 microbiota inoculation on larval performance and digestive enzymes be mediated by 245 differential colonization of their guts by this dominant Acetobacter strain? To address this 246 question, we mono-colonized freshly hatched GF larvae of all twelve populations with this 247 strain, allowed them to develop on the poor diet, and estimated the amount of bacteria inside 248 the larval gut at the end of larval development. This was done by using qPCR to quantify 249 bacterial DNA (using primers specific to Acetobacteraceae 16S rRNA gene) relative to host experimentally colonized Selected and Control larvae in the amount of bacterial DNA relative 252 to host DNA (Fig 4B orange symbols), nor in the absolute Ct values for the bacterial DNA 253 (Fig S1). The latter indicates that the amount of bacterial DNA in these samples was about 254 1000-fold above the detection threshold; based on preliminary data (not shown) this roughly 255 corresponds to 600-900 CFUs per larvae. Analogous Ct values for GF larvae were 256 comparable to what was observed in a mock sample only containing sterilized water, which 257 sets the detections limit (black line in Fig S1). This assures that our procedure of generating 258 GF animals was effective.

259
The above results indicate that Selected and Control populations become similarly 260 colonized by the dominant Acetobacter strain upon experimental inoculation followed by 261 development on the poor diet. This implies that adaptation of Selected populations to poor 262 diet did not cause any changes in the gut that would affect its colonization by commensals.

263
However, this does not preclude a difference in the amount of bacteria they normally harbor 264 under their respective evolutionary regimes (in their "conventional" environment), given that 265 the regimes differ in diet and does not involve experimental inoculation. To address this issue, 266 we used the same approach to quantify bacterial colonization by Acetobacter in the main 267 cultures used to propagate these populations under the experimental evolution that is ongoing 268 in the lab (i.e., on poor diet for Selected and on standard diet for Control populations).

269
Interestingly, despite the difference in diet, these larvae reared in their respective 270 conventional environments were colonized with comparable levels of Acetobacteraceae ( Fig   271   4B green symbols). This suggests that the ability of Selected lines to become largely 272 independent of microbiota (i.e. their ability to cope with being GF) is a physiological result of 273 being adapted to malnutrition and not of being maintained GF by coincidence.

276
Acetobacter pomorum has been shown to promote larval growth through induction of 277 Insulin/IGF-like signaling (IIS) by acetic acid secretion, evidenced by cytoplasmic retention 278 of dFOXO in larval fat body 8 . Acetobacter sp. in our system is also likely to secrete acetic acid since we observe a clear reduction from pH 3.5 to pH 2.0 in the media of all 12 280 populations upon colonization. We thus hypothesized that microbiota would promote larval 281 growth in Control populations, but less so in Selected populations. However, adult size is thus 282 not a good proxy for larval growth rate in these populations: because Selected populations 283 evolved a smaller critical size for metamorphosis initiation, they reach a smaller adult size 284 than Controls despite growing faster on the poor diet 10,11 . Therefore, we combined adult body 285 size (dry weight) of freshly emerged adults (Fig S2) with developmental time data (Fig 1A) 286 to estimate mean larval growth rate of each population under both microbiota conditions, 287 following the approach described in 10 . As expected, we found that inoculation with 288 microbiota increased larval growth rate, but this effect was significantly greater in Control 289 than in Selected populations (Fig 5A), suggesting that IIS and/or target of rapamycin (TOR) 290 pathways may respond differently to microbiota (Fig 5A).

291
In Drosophila, TOR and IIS pathways control systemic larval growth and dFOXO is 292 the key mediator of IIS in regulating ribosome biogenesis and cellular growth 12 . dFOXO is a 293 transcription factor that has >900 direct or indirect targets, a part of which respond to nutrient 294 sensing 16 . To study IIS-dFOXO activity, we analyzed the transcription of three established 295 dFOXO targets, namely d4EBP, dInR, and l(2)efl [17][18][19] in late third stage whole larvae upon 296 mono-association with the Acetobacter strain isolated from our populations. We observed a 297 significant reduction in InR mRNA levels upon colonization by bacteria in Control 298 populations but not in Selected ones (Fig5B). Since InR is negatively regulated by insulin-299 like peptides 19 , this suggests that Control populations do switch from "low nutrition" to "high 300 nutrition" mode physiologically whereas Selected populations are rather insensitive to 301 inoculation with Acetobacter and keep their metabolic state as it is. However, we observed no 302 significant difference in d4EBP (Fig 5B), indicating that differential dFOXO activity in 303 Selected and Control populations does not occur on all dFOXO targets.

304
Interestingly, we also saw that the stress response gene l(2)efl, known to be involved 305 in lifespan regulation 17 , was reduced in all populations when they were GF and significantly 306 induced in Acetobacter colonized larvae only in Selected populations (Fig 5B). This might suggest that Selected populations perceive colonization by Acetobacter as a stress signal; 308 however this gene might also play a hitherto unknown role in larval development or nutrition.

309
Taken together these data suggest that dFOXO activates the transcription of a selection of its 310 target differentially in Selected and Control populations in response to colonization by 311 Acetobacter.

314
We set out to study physiological bases of experimental evolutionary adaptation to 315 chronic juvenile malnutrition, expecting that they will involve an improved ability of the 316 animal host to exploit its microbiota. Instead, we found that our experimentally evolved     presumably in many other insects with diverse diets) this host-microbiota relationship is less 394 intimate than in mammals or in insects feeding on unbalanced or hard-to-digest diets, such as 395 blood, plant sap or wood 4,5 . Rather than relying on transmission of specialized gut microbes 396 from mother to offspring, Drosophila acquire their gut microbiota from the microbial 397 community living on the food substrate 3 . However, microbiota still exerts its beneficial effect 398 by supplementing food with vitamin B in poor environments and regulating sugar metabolism 399 in high glucose environments 28 . In addition, Drosophila microbiota also stimulates the host 400 immune system, interferes with pathogens, and provides signals to key pathways to which 401 regulate growth and tissue homeostasis 29,30 . And, given that the natural food for D.

402
melanogaster is decomposing fruit, the larvae are likely never deprived of those beneficial 403 microbes in nature. It is thus remarkable that the species retained the potential to rapidly 404 evolve a markedly reduced dependence on gut microbiota for fitness under nutritional stress.

435
To colonize larvae with microbiota, fecal transplantation was used. Adults (10 males and 10 436 females) were collected from all populations and kept on standard food for five days. They 437 were transferred on a petri dish with a slice of medium and allowed to defecate for 48 hours.

478
Reverse transcription was performed as described in 7 .  beads. An index PCR was carried out on the purified fraction using a Nextera XT Index Kit 526 (Illumina #FC-131-1001) to produce sequencing libraries. Libraries were again verified by All steps of sequence analysis were performed using the QIIME 1.8.0 bioinformatics software 530 32 . Raw 300 bp paired-end reads were filtered by size (minimum 100 bp overlap between 531 paired ends) and quality (phred-scores ≥ 20). Chimeric reads were eliminated using the

552
Multivariate analysis was done using "ade4" package in R 39 . Evolutionary regime (Selected 553 or Control) and microbiota treatment (germ-free or colonized) were fixed factors; time point 554 was also a fixed factor except for enzyme activity assays, where more than two time points 555 were included. Replicate populations were treated as a random factor nested in evolutionary 556 regimes. A priori pairwise contrasts were performed within the framework of the GMM (using the Slices option of Proc Mixed). Detailed output of all analyses can be found in 558 Supplementary Tables S1-S5.       Evolutionary Regime: