Matching Dietary Amino Acid Balance to the In Silico-Translated Exome Optimizes Growth and Reproduction without Cost to Lifespan

1Diets should ideally match the nutritional needs of their consumers for important life history traits such as growth, reproduction, and lifespan. However, quantifying a balanced diet is challenging given the large numbers of nutrients involved. Among the major macronutrients, the proportion of protein is especially important, since relatively high levels that are important to sustain early life fitness can also incur a heavy cost to lifespan. Thus, establishing protein balance is critical for understanding how diets can be used to enhance lifelong health.

2Many organisms possess mechanisms to prioritize and maintain protein intake to a narrow range of values that are higher than those optimal for longer-term health. For example, when protein is relatively low in the diet, total food intake is elevated to maintain protein intake, causing overconsumption of other nutrients, a situation thought to contribute to obesity. By contrast, when dietary protein content is high, total food consumption is curbed such that energy may be underconsumed—a formulation effectively exploited for weight loss, but also associated with shortened lifespan in insects, mice, and humans. Thus, our evolutionary histories drive consumption of imbalanced foods in a manner that is associated with poor long-term health outcomes.

3Investigations into why these trade-offs exist, and how they might be reduced, form an intense area of research into how dietary restriction (DR) extends healthy lifespan. A long-held idea, derived from life history theory, is that DR improves lifespan by redirecting limiting resources away from reproduction toward somatic maintenance. However, supplementing the DR diet with methionine (M) in flies can improve reproduction without any cost to lifespan. Thus, enhancing dietary protein quality can increase early life fitness without compromising lifespan. However, understanding how to optimize dietary amino acid (AA) content is not trivial as it represents a 20-dimensional balancing problem. Whereas a theoretical framework is now established, a quantitative, evidence-based approach to optimal dietary balance design that does not rely on empirical data has so far proved elusive. The discovery of such a definition for major dietary components would be transformative: diets could be designed to match the requirements of the consumer without the need for lengthy trials. Here we report such a theory for dietary AA balance.

4We hypothesized that the requirement of an animal for each AA is encoded by its genome. Using the Drosophila melanogaster genome, we translated in silico its 19,736 predicted protein-coding genes and derived the proportional representation of the 20 AAs (Figure 1A). This fly “exome-matched” AA ratio (FLYAA) was substantially different from an AA proportion previously referred to as HUNTAA during the development of a holidic diet, referred to here as MM1AA (mismatched1 AA) (Figure 1B).

5We tested if the matched diet (FLYAA) was perceived by flies as preferable to MM1AA. In a two-diet choice assay, in which the foods on offer were identical except that the ratio, but not the total mass, of AAs differed, female flies pretreated on AA-deficient food spent more time on FLYAA than MM1AA (Figure 2A). To assess if this preference was specific to the MM1AA versus FLYAA choice, we designed another ratio (MM2AA) that was equally mismatched to the fly exome translation as MM1AA, but with a different proportion of all 20 AAs (Figure 1B). Again, flies preferred FLYAA over MM2AA (Figure 2A). Interestingly, the flies exhibited no preference for MM1AA or MM2AA. These data indicate that in recovering from AA deficiency, flies that selected a food were not exhibiting indiscriminate food neophilia, but instead specifically detected and chose the matched AA ratio over MM.

6To assess the appetitive values of each ratio, we pre-fed all flies on a yeast-based diet and then measured their food intake when restricted to holidic media containing each of the three AA ratios. By pre-feeding a nutritious diet different from any of the test diets, we ensured that subsequent feeding decisions were neither biased by previous experience of the holidic medium, nor redressing a gross nutritional deficit. Using the automated fly feeding monitor flyPAD to track consumption for 1 hr, flies on FLYAA ate significantly more than those on MM1AA, and also weakly exceeded consumption on MM2AA (p = 0.06 down to 0.02 after outlier removal; Figure 2B). MM1AA consumption did not differ from MM2AA. FLYAA thus had a higher appetitive value than either MM ratio. The rapidity with which this behavior appeared shows that decisions about food consumption precede any possible changes in egg laying (see below). Thus, consistent with and, the internal nutritional state of the fly sets egg laying and feeding decisions in parallel, rather than in series.

7Flies are thought to derive their protein from microbes, in particular yeasts, on decaying fruit. We assessed the potency of each of our AA ratios to affect yeast consumption using two different assays that have been used to infer satiety. For both, flies were pre-fed one of the three AA ratios, and their yeast feeding was then assessed in either a no choice (using flyPAD; Figure 2C) or sugar/yeast (SY) choice situation (; Figure 2D). In both assays, flies pre-fed the yeast-based diet or FLYAA ate less yeast than those pre-fed MM1AA or MM2AA (Figures 2C and 2D). Because we measured consumption of a common protein source (yeast) in the assays, these results indicate that yeast consumption specifically responds to AA status. Furthermore, the exome-matched diet was apparently an effective substitute for the natural proportion of AAs found in their ecologically relevant context (i.e., yeast) because pretreatment with FLYAA or a yeast-based diet suppressed yeast intake to the same extent (Figures 2C and 2D). Finally, the two MM diets were similarly ineffective at suppressing yeast appetite (Figures 2C and 2D), indicating that their degree of mismatch, and not the relative abundance of any one AA, was critical for informing the flies’ perception of protein quality.

8Finally, we assayed food consumption when flies were maintained on the same diet as their pretreatment. During a 1 hr period in flyPAD, flies maintained on FLYAA ate significantly less (∼20%) than those maintained on MM1AA (Figure S1). Thus, the net effect of the enhanced phagostimulatory properties and enhanced satiety value of FLYAA was a small reduction in steady-state feeding relative to flies maintained on MM1AA.

9Next, we assessed the physiological value of each of the AA ratios for egg production. In previous work, we found that the egg laying of flies feeding on a yeast-based diet is limited by the essential AA (EAA) M. If this reflects a stoichiometric limitation, then it should be predictable by the most underrepresented AA in the food when compared with the in silico-translated exome. We focused on the EAAs and conditionally EAAs (see Experimental Procedures) because all others can be acquired by de novo synthesis. More formally, we propose Drosophila egg production should be limited by the EAA with value r, wherefor EAA i, mini is the minimum, and di and pi are the relative concentration in the diet and translated exome, respectively.

10In line with the experimental data, taking the AA content of yeast, we thus identified M as the limiting EAA (Figure 3A). Using the same technique for MM1AA, we identified arginine (R), which is essential for flies, as limiting (Figure 3A).

11We confirmed this prediction of R limitation for egg laying experimentally. Increasing or reducing R concentration alone in the defined medium caused egg production to increase or decrease to the same extent as when all AAs were altered by the same amount (Figures 3B and S2A) and no other EAA produced these responses. Interestingly, increasing R by half yielded a quantitatively matched proportional increase in egg laying, but with higher levels of supplementation, egg laying only increased by ∼1.7× (Figure 3C). This attenuated response could be explained by exome matching, which quantitatively predicted this limitation by the next most limiting EAA in the diet (M). Furthermore, the same calculations and tests for MM2AA correctly identified isoleucine (I) as limiting and predicted the proportional change in egg laying from I addition until the limit imposed by the next most restrictive AA (tryptophan, W) at 1.61× I addition (Figure 3D). Together, these data indicate that Drosophila egg laying is quantitatively governed by a limiting EAA in the diet, which can be identified by exome matching.

12Since egg production is constrained by the most limiting EAA in the diet, all others are consumed in excess and must be excreted, in flies largely as uric acid. Thus, the degree of AA mismatch in a diet should be reflected in uric acid production. We measured uric acid produced by flies after exposure to the three AA ratios and found that those consuming MM diets had equal amounts of uric acid, both significantly higher than flies that consumed FLYAA (Figure S2B).

13In Equation 1 above, because r will predict the limiting EAA for egg laying on the current diet (diet 1), the proportional change in egg laying when flies feed on a diet with a different AA ratio (diet 2) should also be predictable according to r1/r2. We found that exome matching predicted egg-laying outcomes for flies on a range of concentrations of five different AA ratios (Figures 3E and 3F). In each case, both the identity of the most limiting AA and the concentration of all predicted non-limiting AAs were varied (Figure 3E), indicating that the most limiting EAA, and no other AA in the mixtures, determined egg-laying output. At intermediate and higher egg-laying rates, the data plateaued at a level that was less than predicted. This level of egg laying at the plateau corresponds to the maximum obtainable on our optimal SY diet (2SY), indicating that at this rate of egg laying, some other factor unrelated to nutritional ratios limits egg output (Figures 3F and 3G). Up to this threshold, exome matching predicts the identity of the most limiting EAA and the extent to which it modifies egg laying.

14We found it remarkable that exome information that was unweighted by transcriptome or proteome information could accurately predict egg laying. We therefore assessed the effect of a diet with AA proportions according to a published proteomics-derived measurement of Drosophila body composition (PROTEOMEAA; Figure S2C; ). Over a range of AA concentrations, the two diets had statistically indistinguishable effects, indicating that within the sensitivity of our assay, accounting for gene expression differences did not improve the accuracy of exome matching for egg laying (Figure S2D).

15The failure of proteomics-based prediction to improve on that from exome matching could have several explanations. On the technical side, experimental determination of the proteome is semi-quantitative and can be biased, while on the biological side, measurements based on the whole fly proteome may not detect the specific requirements for egg laying. For instance, exome matching could be accurate because the genes transcribed in the ovary use AAs in a ratio that is peculiarly representative of the whole genome. To assess this, we ranked all in silico-translated genes on the basis of how similar their predicted AA usage is to the average for the whole translated exome (Figure 4A). This revealed that the average ranking of the ovarian-expressed transcripts was smaller than expected by chance when compared with randomly selected gene lists of the same size (p = 0.09, Catmap;; Figure 4B). All other tissue gene sets were less similar to the whole genome. Thus, the ovarian transcriptome may have been constrained toward encoding AAs in the proportions required to maximize future biomass production. This could be particularly important within the confines of an egg, where nutrient and waste exchange with the environment is not possible and thus a balanced AA ratio facilitates development while minimizing the danger of toxicity from accumulated nitrogen catabolites.

16To test if exome matching is also relevant for growth, we reared Drosophila larvae on a variety of AA ratios and measured egg to adult development time. Development on the holidic medium is delayed when compared with SY food due to limitation for some unknown factor that is not related to the AA ratio. It is possible, however, to establish conditions in which growth is AA limited. Within this range, we again found that the exome-matched FLYAA ratio was superior to MM1AA, MM2AA, and a ratio based on the flies’ natural food source, yeast (YAA; Figure 1B) and was again no different from PROTEOMEAA (Figure 5A).

17We found that the major delay to development between ratios at 10.7 g total AA mass was caused by lengthening of the third instar phase (Figure 5B). Nutrient-mediated delay in pupariation can be attributed to a combination of reduced systemic insulin/IGF-like signaling and lowered TOR signaling in the fat body and prothoracic gland. These combine to reduce growth rate and delay the ecdysone peak that triggers pupariation, thus extending the duration of the third instar stage to allow for some degree of compensatory growth. Indeed, we found that the 4e-BP protein, whose transcription is elevated with decreased IIS and TOR, was lower in FLYAA-reared larvae than those on MMAA diets (Figures 5C and S3A), but the proportion of phosphorylated protein was not. We also found no differences in the amount or proportion phosphorylated for S6k (data not shown). The effect on total 4e-BP levels rather than phosphorylation status of either protein indicates a longer-term response to the higher nutritional value of FLYAA. We also found that at low AA concentrations, FLYAA significantly increased the proportion of larvae surviving to adulthood over that in MM1AA, MM2AA, and YAA, as well as the body mass and adult wing size of both adult males and females (Figures S3B–S3E). Thus, FLYAA improved several aspects of development by providing a more nutritious substrate, even improving on a ratio modeled on the flies’ natural food source.

18In the food industry, there are economic and environmental benefits from improving the efficiency of biomass production. To assess the potential utility of exome matching for mammals, we searched the literature for examples of rodent diets that had been designed to be AA limiting and for which the identity of the limiting EAA had been verified experimentally. In each instance, exome matching successfully identified the growth-limiting AA: R for rats fed an R-limiting diet, threonine (T) for rats fed the T-basal diet in, L for rats fed the L-basal diet in, and M for mice fed the 0.15% M diet in (Figure S4).

19To test the ability of exome matching to enhance mouse growth under protein-limiting conditions, we designed iso-energetic diets containing a constant mass of purified AAs whose ratio was varied (Figure S5A). To ensure all treatment groups had equal AA intake, meal sizes were standardized across treatments to a mass that was entirely consumed before the next meal. Under these conditions, purified AAs supplied in the proportion found in casein (CASEINAA), the normal source of protein in mouse chow, supported growth just as well as whole protein (Figure S5B). During the initial linear growth phase (weeks 3–6.5), increasing CASEINAA concentration from 6% to 8% (by 33%) caused an ∼40% increase in growth rate (Figure S5C).

20When we modified the AA proportion from CASEINAA to that of the translated mouse exome (MOUSEAA), but maintained total protein equivalents at 6%, initial growth rate improved by ∼31% in one trial and ∼33% in a second (Figure 5D; Table S1) and resulted in body mass differences that persisted into adulthood (Figure S6A). At 24 weeks of age, we also found that differences in the free AA profile of the hepatic portal vein for MOUSEAA- versus CASEINAA-fed mice were positively correlated with differences in dietary content (Figure S6B). Although the initial growth rate improvement was somewhat lower than the prediction of 48%, these data indicate that, similar to flies, exome matching enhanced nitrogen source quality for growth. For these same mice, those fed CASEINAA voluntarily consumed ∼35% more water than those fed MOUSEAA (Figure 6A), consistent with a greater proportion of the AAs in CASEINAA being inaccessible for growth and so increasing the water demand for urinary excretion. Indeed, urinary nitrogen excretion of mice that had developed on the diets for 20 weeks (23 weeks of age) was greater for those fed CASEINAA than MOUSEAA (Figure 6B).

21The greater body mass of mice developing on MOUSEAA than on CASEINAA was in part attributable to greater lean mass, with both the rate of accumulation and absolute level of lean mass attained being higher (Figure 6C). Fat mass accumulation was similarly increased (Figure 6C). Organs removed from 24-week-old mice revealed a significant increase in mass for white adipose tissue, kidney, liver, and skeletal length, but not for tissues whose size has previously been observed to be refractory to dietary change, i.e., heart, thymus, quadricep muscle, or brain (Figure S6C). Rearing on CASEINAA and MOUSEAA yielded no differences at 24 weeks in fecal energy content, patterns of movement, or the respiratory exchange ratio (RER), but mice fed CASEINAA showed greater thermogenesis (determined via indirect calorimetry) than did mice fed MOUSEAA (Figures S6D and S6E). Thus, the enhanced energy storage in MOUSEAA animals appeared to be a consequence of reduced energy use for heat production.

22Finally, because of their importance for health, we measured several parameters of bone structure in femurs of 23-week-old mice and found that cortical structure thickness, trabecular bone mineral density, and trabecular volume of MOUSEAA-fed animals were significantly greater than those of mice fed CASEINAA (Figure 6E). Thus, MOUSEAA is a higher-quality source of AAs than CASEINAA for both growth and bone structure.

23Interestingly, the enhanced quality of the MOUSEAA diet was apparently perceived by the mice since when with ad libitum access to food, young mice consumed ∼15% less food per gram body mass of the 6% MOUSEAA diet than of 6% CASEINAA, and this effect persisted through to adulthood (Figure S7A). Thus, similar to flies, the exome-matched diet was both more efficiently used and resulted in lower steady-state feeding.

24To further test whether exome matching predicted AA quality for mouse growth, we designed another MMAA ratio (mmMOUSEAA; mismatch MOUSEAA) that differed from both CASEINAA and MOUSEAA in the proportion of AAs, but was predicted to impose a similar growth limitation to CASEINAA. There was little difference in growth rate between mice maintained on CASEINAA or mmMOUSEAA (Figure 7A). Furthermore, exome matching predicted T to be growth limiting in mmMOUSEAA (Figure S7B), and when T was reduced by 30%, growth was reduced (Figure 7B). In contrast, reducing the concentration of M by 30% did not alter growth (Figure 7C), consistent with the exome-matching prediction that it was in 1.4-fold excess (Figure S7B).

25Finally, we compared growth of mice fed the exome-matched diet to a cohort fed the recommendations for dietary AAs from the National Research Council (NRC) or one matched to a profile based on AA analysis of whole mice (BODYCOMPAA). Both are based on empirical data and are used to inform optimal dietary AA composition. We found no difference in initial growth rate of mice fed MOUSEAA versus those fed NRCAA (Figure 7D), while those assigned the diet based on body composition showed significantly slower growth (Figure 7E). Although not different in its effect during early linear growth, the NRCAA diet did, however, support an ∼8% greater adult body mass (Figure S7C) that whole-body MRI revealed was due to the NRCAA-fed mice gaining fat, but not lean mass, at a faster rate than MOUSEAA-fed mice (Figure S7D). In further tests, we found no differences in movement, fecal energy content, RER, heat production, glucose tolerance, insulin tolerance, or any of the measured bone quality metrics at 23–24 weeks of age (data not shown). Thus, the enriched fat mass of the NRCAA mice may be due to differences in fat biosynthesis.

26In summary, exome matching provides an easily implemented framework for establishing a high-quality nitrogen source for mouse growth. We note that its quantitative accuracy in mouse and fly growth was not as precise as for fly egg laying, perhaps due to nutritional buffering by resident microbiota similar to that observed in and. Nonetheless, our data provide evidence that our completely in silico method can be used for dietary AA design, and that it performs equally as well as, if not better than, natural diets or others that have been developed empirically over decades.

27The relative concentration of protein in the diet has been shown to modulate both early life fitness and lifespan, and can account for the benefits of DR in flies and mice. We therefore assessed the response of Drosophila lifespan to varying concentrations of MM1AA, MM2AA, and FLYAA. Consistent with our previous observation, we found a marked increase in lifespan with decreasing relative concentration of total dietary AAs, an effect that was not modified by AA ratio (Figure 7F), unlike early life where the AA ratio strongly affected growth (Figure 5A) and egg laying (Figures 3E and 3F).

28Interestingly, considering the egg laying and lifespan phenotypes together, their response differences caused flies on FLYAA to exhibit a single dietary optimum for growth, reproduction, and lifespan at a relatively low AA concentration (10.7 g/L). In contrast, flies on the MM diets displayed separate optima because egg laying and development required higher levels of AAs (at least 21.4 g/L) than the optimum for lifespan (10.7 g/L). Thus, as we have observed previously, changing the dietary AA balance can establish a diet in which the apparent trade-off between reproduction and lifespan is avoided. Here we show that this balance can be established by matching the dietary AA ratio to the in silico-translated exome. In looking for correlated effects on nutrient signaling, we found that total 4e-BP levels, but not the proportion phosphorylated, were reduced in the ovaries of adult flies exposed to 10.7 g/L FLYAA when compared to those on the same concentration of AA in the MM diets (data not shown). In contrast, 4e-BP levels and phosphorylation were unaffected across the same treatments when measured in adult flies from which the ovaries had been removed. Thus, across all conditions we measured, we found that 4e-BP levels changed with growth and reproduction when measured in dividing tissue, and held steady with unchanging lifespan when measured in non-dividing tissue. These data are compatible with an explanation that tissues vary in their sensitivity to AA ratios for modulating IIS and TOR, and that these inter-tissue differences may be key to separating the regulation of lifespan and reproduction.

29Consuming a diet with a relatively low proportion of protein is critical for lifelong health. However, the costs to early life vigor, reproduction, and low satiety value are major detractions. We show that an exome-matched AA composition can reduce voluntary food consumption, and that its enhanced value for growth and reproduction means it can be supplied at low enough levels so as to avoid any cost to lifespan. Given that these are all beneficial outcomes, it will be interesting to determine if there is a physiological cost that we have not yet measured.

30It should be noted that the proportions of protein in our experimental diets are particularly low when viewed in the light of an average American diet containing ∼15% protein. While exome matching may be useful for screening supplements used to treat protein-energy malnutrition for AA imbalance, it is unlikely to enhance human growth or reproduction in developed countries where protein is generally non-limiting. However, because we have found that exome-matched diets represent high-quality protein that suppresses steady-state feeding at low concentrations, it is possible to envisage that diets to reduce total food intake to enhance long-term health, or even specifically to reduce the proportion of protein in the diet, e.g., for the management of kidney disease, could be better achieved using exome matching.

31Thus, the principle of exome matching provides a theoretical template for dietary AA design that can enhance the biological efficiency of food for the lifelong health of the consumer.

32Additional details in the Supplemental Experimental Procedures.

33Except where indicated, all experiments were conducted with our female Drosophila melanogaster (Dahomey). Stocks are maintained outbred and experiments were conducted in controlled conditions: 25°C, 65% humidity, and 12 hr:12 hr light:dark. Except where indicated, flies were reared on 1× SY food at standard density. Holidic media were made according to with appropriate substitutions for each of the different AA ratios (Table S2; Data S1).

34For each mouse experiment, 20 C3B6F1/J females per treatment were used, housed in four groups of five under specific-pathogen-free (SPF) conditions. Our parental mice were two inbred strains: C57BL/6J and C3H from the Jackson Laboratory.

35Except for ad libitum-fed animals, mice were pair fed. Water intake was measured for each cage separately twice per week. Diets were manufactured by Ssniff.

36Mouse experiments were performed in accordance with the recommendations and guidelines of the Federation of the European Laboratory Animal Science Association (FELASA), with all protocols approved by the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Germany.

37A total of 21,070 proteins were retrieved from FlyBase (FB2008_05, released May 30, 2008). From this, 821 proteins with length less than 100 AAs and 513 with length greater than 2,000 AAs were removed to generate a set of 19,736 non-extreme proteins. This trimming procedure was followed for all animal genomes used. For mouse, Ensemble v 54 (May 2009, downloaded July 2, 2009) was used and for rat, Ensemble v 56 (September 2009, downloaded January 19, 2009). The sum of each AA used in each protein was used to generate its proportional AA usage, and this was combined for all proteins to find the average AA usage for the predicted proteome.

38To predict the most limiting EAA in a diet, the proportion of AAs in the food was divided by the proportional representation of AAs in the translated exome of the consumer. The EAA with the lowest value after this transformation was considered limiting. If the requirement for either of the conditionally EAAs Y or C exceeded the available supply, their requirement was met by subtracting a mole of F or M, respectively, for each mole of AA required.

39To design MM2AA, we determined the Euclidian distance from MM1AA to FLYAA in 20-dimensional space (1 dimension per AA). We then found another point, MM2AA, that was equidistant from FLYAA as MM1AA, but as far away as possible from MM1AA. This procedure was also performed to generate mmMOUSEAA, using CASEINAA and MOUSEAA as starting points.

40For each trial of each feeding assay, a fresh generation of flies was reared to avoid the possible confounding effects of memory-based decision making.

41A four-arm choice apparatus modified from was used. Mated females were AA deprived for 72 hr and acclimatized in the chamber, and their subsequent location was counted hourly for ∼8 hr. The food preference index (FPI) was calculated as (n flies on surface of food A − n flies on surface of food B)/(n flies on surface of food A + n flies on surface of food B). Total AA mass in food was fixed at 21.4 g/L.

42Groups of 3- to 5-day-old flies (15 females and 5 males) were maintained on yeast-based food or holidic medium. After 72 hr, the flies were tested for nutrient choice as described in.

43w Dahomey flies were reared in the same medium as for the SY choice assay. Mated adult flies were then maintained on either holidic or yeast-based medium for the pretreatment period (3 days) and assayed using flyPAD, as described in.

44First instar larvae were picked onto test media and each developmental stage scored at 25, 138, 258, and 330 hr after egg laying. Adult eclosion was scored daily at 24 hr intervals. Body mass was measured for pairs of newly emerged flies. Wings were measured from the edge of the distal tip to the edge of the alula.

45After 16 hr on holidic medium, flies and medium were removed from the vial, which was washed with 2 mL of 0.1 M sodium glycinate buffer (pH 9.2). Uric acid was quantified spectrophotometrically using the Amplex Red Uric Acid kit (Life Technologies). For TAG determinations, we used the Triglyceride Infinity reagent (Thermo Scientific) and normalized levels to total protein.

46The following antibodies were used: 4EBP1 (CST #4923), phospho-4EBP1 (CST #4923), S6K (custom-made, courtesy IHA, UCL), and phospho-S6K (CST #9209).

47Lifespan assays were performed as described in. Replicates 1–3 were performed preparing the holidic medium as described in, but for the fourth replicate, the three AAs (I, L, and Y) were added after autoclaving.

48Body fat content was determined by in vivo magnetic resonance tomography imaging (time domain [TD] NMR).

49Indirect calorimetry and movement were monitored over 48 hr for singly housed mice in purpose-built cages (Phenomaster, TSE systems) maintained at 22°C –23°C.

50Urea in urine from a 24 hr period was measured using the Urea Assay kit (Sigma) from mice housed in metabolic cages (Tecniplast).

51Portal vein blood samples were harvested from pair-fed female C3B6F1/J mice at 23 weeks of age following a 1 g meal and 2 hr food deprivation. Mice were euthanized and dissected, and portal vein blood was collected in EDTA tubes. Analysis was performed by the Finnish Institute for Molecular Medicine (FIMM).

52At 23 weeks of age, mice were euthanized using CO2 and organs were immediately harvested and weighed. For bone density measurements, right femur bones from 23-week-old mice were collected and scanned with a high-resolution μCT scanner (SkyScan 1176, Bruker). Trabecular and cortical bone regions of distal femurs were selected with reference to the growth plate. Bone mineral density was determined based on calibration with two phantoms of known density (Bruker), which were scanned under the same conditions as the bone samples.

53R (v3.2.0) and JMP (V11) were used for all statistical analyses. Generally, error bars represent SEM.

54Conceptualization, M.D.W.P. and L.P.; Methodology, M.D.W.P., G.A.S., E.B., I.A., S.J.S., C.R., and L.P.; Formal Analysis, M.D.W.P., G.A.S., E.B., S.L.H., I.A., C.R., and L.P.; Investigation, M.D.W.P., G.A.S., A.M., S.L.H., P.J., X.H., H.S., and M.Y.; Writing – Original Draft, M.D.W.P. and L.P.; Writing – Review & Editing, M.D.W.P., G.A.S., S.J.S., C.R., and L.P.; Funding Acquisition, M.D.W.P., C.R., and L.P.