Paternal Western diet causes transgenerational increase in food consumption in Drosophila with parallel alterations in the offspring brain proteome and microRNAs

Abstract

Several lines of evidence indicate that ancestral diet might play an important role in determining offspring’s metabolic traits. However, it is not yet clear whether ancestral diet can affect offspring’s food choices and feeding behavior. In the current study, taking advantage of Drosophila model system, we demonstrate that paternal Western diet (WD) increases offspring food consumption up to the 4^th generation. Paternal WD also induced alterations in F1 offspring brain proteome. Using enrichment analyses of pathways for upregulated and downregulated proteins, we found that upregulated proteins had significant enrichments in terms related to translation and translation factors, whereas downregulated proteins displayed enrichments in small molecule metabolic processes, TCA cycles and electron transport chain (ETC). Using MIENTURNET miRNA prediction tool, dme-mir-10–3p was identified as the top conserved miRNA predicted to target proteins regulated by ancestral diet. RNAi-based knockdown of miR-10 in the brain significantly increased food consumption, implicating miR-10 as a potential factor in programming feeding behavior. Together, these findings suggest that ancestral nutrition may influence offspring feeding behavior through alterations in miRNAs.

Graphical Abstract

The paternal Western diet (WD) increases offspring food consumption up to the fourth generation in Drosophila. In F1 offspring, paternal WD altered the brain proteome. MIENTURNET miRNA prediction tool, identified dme-miR-10–3p as the conserved miRNA predicted to target brain proteins regulated by paternal diet. RNAi-based knockdown of miR-10 in the brain significantly increased food consumption, implicating miR-10 as a potential factor in programming feeding behavior.

Introduction

Increasing rates of childhood and adolescent obesity is a global public health problem associated with mounting health risks and health care costs(1–3). According to family aggregation studies, childhood obesity clusters in families (4, 5) and is highly heritable (6, 7). Despite of this fact, single nucleotide polymorphisms (SNPs) can explain less than 2% of the BMI in genome-wide association studies (GWAS)(7, 8), creating the “missing heritability” paradox(9–11).

A potential explanation of “missing heritability” comes from the studies indicating that epigenetics could play a significant role in susceptibility to obesity and related metabolic disorders (reviewed(11–13)). Populational studies in humans, such as the Dutch famine of WWII and the Overkalix studies, showed that poor ancestral diet can lead to metabolic and cardiovascular risks in grandchildren(14–17). In laboratory settings generational epigenetic effects were observed in various species from C. elegans(18) to laboratory rodents (reviewed(19–26)) suggesting that mechanisms regulating these processes are possibly evolutionary conserved(27, 28). As the study of maternal effects is challenging due to the fact that maternal influences can affect offspring via placenta, blood, milk composition, microbiome, and behavior(29, 30), particular research focus has been on paternal effects, as it narrows transmission mechanisms to those mediated by gametes(19–22, 26, 31). In laboratory rodents, paternal effects have been documented after various environmental exposures including pesticide exposure(32), fasting(33), exercise(34, 35), a low-protein diet(36), a high-fat diet(37, 38) and the Western diet (WD)(39, 40).

In Drosophila, generational effects has been documented in response to multiple manipulations including dietary sugar (27, 41, 42), yeast protein (43), fat (44–46), a dilution of the standard diet (45), ethanol exposure (47), behavioral stress (48, 49), genetic altering of parental metabolism (50) and low temperature exposure(51). The inheritance of metabolic traits after paternal sugar-diet was recently linked to spermatogenic transmission of information including alterations in chromatin states³⁴ and sperm borne RNA(52). These studies also highlighted that Drosophila is extremely useful model system for transgenerational research because of the short life cycle, large number of offspring, physiological simplicity, and powerful genetic tools(53).

In the current study taking advantage of the drosophila model system, we investigated effects of paternal nutrition on offspring feeding behavior. For this study we used a WD we have recently developed for fruit fly(54) to imitate complexity as well as hedonic/addictive nature of the human diet, as the addictive properties of energy-dense “junk”, “fast”, and cafeteria-style foods contribute significantly to the etiology and pathogenesis of human metabolic disorders (reviewed(55, 56)). Interestingly, in our previous study, the flies preferred WD over control diet (CD) despite the detrimental health outcomes, suggesting that these preferences could be hard-wired(54).

Thus, following up on our previous research, in the current study we asked question whether paternal WD might influence offspring’s feeding behavior. Here we show that ancestral WD produces a transgenerational increase in the offspring’s food consumption up to the 4^th generation. The proteomic analysis of F1 brains revealed that upregulated proteins were significantly enriched for translation and translation factors, while downregulated proteins were enriched for small molecule metabolism, TCA cycles, and electron transport chains (ETC), suggesting alterations in bioenergetic processes. Using MIENTURNET, a miRNA enrichment analysis, we identified dme-mir-10–3p as a conserved miRNA predicted to target proteins regulated by paternal WD. Knockdown of dme-mir-10–3p in the brain significantly increased food consumption in F1, suggesting miR-10 may be involved in alterations of feeding behavior. Together, these findings identify ancestral nutrition as a potential factor contributing to programming obesity-risk behaviors in offspring.

Methods

Drosophila culture

Flies used in generational experiments were derived from a colony established from Drosophila simulans isofemale lines and maintained as an outbred population (54). For transgenic experiments the following lines were acquired from Bloomington Drosophila Stock Center: 7009, 35014, 61377 and Canton-S. The fly stocks were maintained on the standard Bloomington Formulation diet (Nutri-Fly^® BF, Cat #: 66–112, Genesee Scientific Inc., San Diego, CA) in a climate-controlled environment at 24°C under a 12h light-dark cycle and 65–70 % humidity. All experiments were performed on age-matched 3–4-day old flies. All flies, from the embryo stage, were raised on the standard Nutri-Fly Bloomington diet (CD). The WD was based on the standard Nutri-Fly Bloomington diet with the addition of Nutiva USDA Certified Organic, non-GMO, Red Palm Oil (15% by weight), 15% Sucrose, and 0.1M NaCl as described previously (54).

The 3–4 days old paternal flies (F0) were exposed to the WD, and/or exercise for 5 days. At the end of the fifth day, the flies were collected, transferred to the breeding vials with CD food (Fig.1). The F1 offspring were generated by crossing exercise and diet-exposed males to control virgin females. The density was controlled by keeping breeding conditions constant for all groups and generations: 5 males and 5 virgins were bred. Briefly, to ensure a common parental larva density and epigenetic background, 0–4 days old flies were manually sorted to 5 males and 5 virgin females per vial and allowed to lay eggs for 48 hours in a wide vial at 24°C, 70% humidity, and a 12-hour light cycle 3 days. F2-F4 offspring were generated in the same manner by breeding 5 males and 5 female siblings. After emergence, the flies were collected within the four-day window. Flies, from the embryo stage, were raised on the standard Nutri-Fly BF diet. Each breeding group consisted of 6–10 breeding vials and N number for offspring statistics were based on a number of the breeding vials. The offspring were tracked with respect to the specific parental flies to prevent any given parent from dominating the pool. Each group of parental flies was represented by an equal number of offspring in each experiment.

Fig. 1.

Outline of the experimental regimen.

Behavioral experiments

Locomotor activity was studied using LAM25H locomotor activity monitors (TriKinetics Inc, Waltham, MA) as described previously(54). The LAM25H uses infrared beams to register fly movement in individual vials. When animals in the tube move the beams are interrupted and then registered as counts. Briefly, the activity was measured in groups of five flies housed in narrow vials with food with 3–4 replicates per diet.

Flight exercise was performed according to protocol described previously(54). Briefly, groups of sixty 3–4-day-old male flies were housed in 1-gallon clear plastic drum fishbowls (Petco, San Diego, CA) strapped to a horizontal platform attached to a motor. The motor was controlled by two timers initiating three motor revolutions spaced 14 seconds apart every 5 min. Each revolution elevated the platform and then dropped it down triggering flies into flight. The exercise was performed daily for 7 h for 5 days. No mortalities or injuries associated with exercise were observed.

Fly Liquid-Food Interaction Counter (FLIC)

FLIC system was utilized to monitor consumption of 10% sucrose solution according to the protocol described elsewhere(57, 58). Specifically, FLIC allows to measure feeding behaviors by detecting electronic signals, “licks,” when the fly proboscis touches the food. Using a flight aspirator, flies were gently removed from their food environment and loaded onto Drosophila Feeding Monitor (DFM) filled with 10% sucrose solution. Fly feeding behavior was monitored continuously for 24 h starting at 12 pm. FLIC data were analyzed using R studio (https://www.rstudio.com/) and custom R code, which is available at https://github.com/PletcherLab/FLIC_R_Code. Default settings were used for analysis. Wells that had zero feeding events over the testing interval were removed from the analysis.

Assays for triglycerides.

For dry weights, flies were killed in liquid nitrogen and then dried at 52°C for 72 hours. Flies were individually weighed using Cahn C-35 Ultra-Microbalance (Thermo Fisher Scientific, Waltham, MA). The assays to measure triglycerides were carried out according to our published procedure (54). Briefly, 5 flies were rapidly homogenized in 0.15 ml of PBST (PBS with 0.1% Tween 20) using Bullet Blender (Next Advance, Inc., Troy, NY). The samples were centrifuged for 3 min at a maximum speed at 4°C. Total proteins were measured using the Pierce Rapid Gold BCA Protein Assay Kit (A53225, Thermo Fisher Scientific). The assays for triglycerides were performed using incubations with Pointe Scientific Triglycerides Reagent (T7532120, Thermo Fisher Scientific). Triglyceride contents were normalized to mean fly weight and to total protein levels per mean fly weight.

Proteomics

Sample prep for label-free proteomics

Fly brains were quickly dissected on ice cold block and frozen on dry ice. Isolated fly brains were lysed in urea lysis buffer (8M urea in 40mM Tris, 30mM NaCl, 1mM CaCl₂, 1 x cOmplete ULTRA mini EDTA-free protease inhibitor tablet; pH=8.0), as described previously(59, 60). The samples were subjected to three freeze-thaw cycles, and sonication. Samples were centrifuged at 10,000 × g for 10min at 4°C. Protein concentration was determined by BCA. Equal amounts of protein were reduced with 5mM DTT at 37°C for 30min, and then alkylated with 15mM iodoacetamide for 30min in the dark at room temperature. Unreacted iodoacetamide was quenched with DTT (15mm). Initial digestion was performed with Lys C (ThermoFisher Cat# 90307; 1:100 w:w) for 4hr at 32°C. Following dilution to 1.5M urea with 40mM Tris (pH=8.0), 30mM NaCl, 1mM CaCl₂, samples were digested overnight with trypsin (Promega; Cat# V5113; 50:1 w/w) at 32°C. Samples were acidified to 0.5% TFA and then centrifuged at 4,000 × g for 10min at 4°C. As described previously(59), supernatant containing soluble peptides was desalted, and then eluate was frozen and subjected to speedvac vacuum concentration.

nLC-MS/MS for label-free proteomics

Final peptides were resuspended in 0.1% formic acid, quantified (ThermoFisher Cat# 23275), and then diluted to a final concentration of 0.25μg/μL. Samples were subjected to nanoLC-MS/MS analysis using an UltiMate 3000 RSLCnano system (ThermoFisher) coupled to a Q Exactive Plus Hybrid Quadrupole-Orbitrap mass spectrometer (ThermoFisher) via a nanoelectrospray ionization source. For each injection, 4μL (1μg) of sample was first trapped on an Acclaim PepMap 100 20mm × 0.075mm trapping column (ThermoFisher Cat# 164535; 5μL/min at 98/2 v/v water/acetonitrile with 0.1% formic acid). Analytical separation was performed over a 120min gradient (flow rate of 250nL/min) of 4–45% acetonitrile using a 2μm EASY-Spray PepMap RSLC C18 75μm × 250mm column (ThermoFisher Cat# ES802A) with a column temperature of 45°C. MS1 was performed at 70,000 resolution, with an AGC target of 3×10⁶ ions and a maximum injection time (IT) of 100ms. MS2 spectra were collected by data-dependent acquisition (DDA) of the top 15 most abundant precursor ions with a charge greater than 1 per MS1 scan, with dynamic exclusion enabled for 20s. Precursor ions isolation window was 1.5m/z and normalized collision energy was 27. MS2 scans were performed at 17,500 resolution, maximum IT of 50ms, and AGC target of 1×10⁵ ions.

Data analysis for label-free proteomics

As described previously (59), with some modification, Proteome Discoverer 2.2 (PDv2.2) was used for raw data analysis, with default search parameters including oxidation (15.995 Da on M) as a variable modification and carbamidomethyl (57.021 Da on C) as a fixed modification. Data were searched against the Uniprot Drosophila melanogaster reference proteome (Proteome ID: UP000000803). Both reviewed (Swiss-Prot) and unreviewed (TrEMBL) were included. PSMs were filtered to a 1% FDR and grouped to unique peptides while maintaining a 1% FDR at the peptide level. Thus, the 1% FDR for peptide and protein identification refers to how accurate the peptide and protein identification was for all samples. Peptides were grouped to proteins using the rules of strict parsimony and proteins were filtered to 1% FDR. Peptide quantification was done using the MS1 precursor intensity. Imputation was performed via low abundance resampling.

Statistical evaluation

All proteomics samples were normalized to total protein abundance, and the protein tab in the PDv2.2 results was exported as a tab delimited .txt. file and analyzed. Protein abundance was converted to the Log₂ space. P value of less than 0.01 was used as a cutoff. For pairwise comparisons, tissue mean, standard deviation, p-value (p; two-tailed Student’s t-test, assuming equal variance), and adjusted p-value (Benjamini Hochberg FDR correction) were calculated (61, 62). All raw data are available online using accession number “PXD033363” for jPOST Repository (63, 64).

miRNA PCR

Total RNA was isolated from frozen tissues by standard methods (65) and then was used for cDNA synthesis and subsequent qRT-PCR. Flight muscles, brains and sperm were quickly isolated on wet-ice metal blocks from cold-anesthetized flies. Sperm isolation followed previously described protocol(41). Total RNA from tissue samples was extracted with an RNAqueous MicroScale RNA Isolation Kit according to the manufacturer’s instructions (Thermo Fisher–Life Technologies, Grand Island, NY, USA). Polyadenylation and reverse transcription were performed using Takara microRNA first-strand synthesis and miRNA quantification kit (#638313). Equal concentrations of total RNA from 5 muscles, 10 brains or 20 individual sperm samples were used to produce 3 RNA pools per group. Real-time PCR reactions on micro-RNA (miR) were carried out using TB Green Advantage qPCR Premix kit (#639676, Takara) in triplicates for each cDNA sample on QuantStudio 6 Flex PCR System (Thermo Fisher). Primers specific for each miRNA and mRNA were obtained from Invitrogen. As internal controls, primers for U6 (the noncoding small nuclear RNA), and miR-191 were added for RNA template normalization, and the relative quantifications of miRNA expression were calculated against miR-191 by the ΔΔC_t² method. Constitutively expressed miR-191 was used as a normalization control for miRNA qPCRs. The experiments were performed 3 times independently.

Statistics

Statistical analyses were done with student t-test, Mann–Whitney U test, a one-way and two-way ANOVA, depending on the particular data set using GraphPad Prism version 8.00 for Windows, (GraphPad Software, San Diego, CA). Post hoc analyses were conducted using Tukey’s test.

Results

Ancestral WD causes increase in food consumption up to the fourth generation.

Our previous study revealed that WD increased Drosophila food consumption, while flight exercise mitigated this effect(54). To begin to explore whether ancestral diet may alter offspring feeding behavior F0 male flies were subjected to the WD or the WD plus exercise. Two other groups included sedentary flies on the CD, and the CD plus exercise. The effect of paternal WD was studied using FLIC, a novel high-throughput, continuous monitoring system for fly feeding behaviors(57).

The assay showed that both male and female WFO offspring had significantly more licks than other groups. This significant increase was observed in F1-F4 male flies and F1-F3 female flies according to one-way ANOVA (Fig2. A, B). The significant increase in licks was also observed in F1 males WEFO in comparison to CFO. Interestingly, exercise partially negated effect of WD in WEFO F2–4 males and F1–3 females. This suggests that ancestral WD causes transgenerational changes in offspring food consumption while ancestral exercise counterbalances this effect. There were an interesting fluctuation in EFO male and female data showing increase in one generation and then decrease in another generation suggesting that paternal environmental exposures increase “developmental noise” in the offspring data(66).

Fig. 2. Effects of ancestral diet and exercise on offspring feeding behavior, triglycerides and activity.

A. Feeding activity of male offspring measured as “Licks” using FLIC. Flies were loaded into individual feeding chambers and consumption of 10% sucrose was measure over 24 h period starting at12pm. Each group consisted of 9 flies (N=9–15). The graphs show successive generations from F1 to F4 from left to right. The experiments were repeated 3 times. Y-axis represents total number of licks over 24 hr period.

B. Feeding activity of female offspring measured in FLIC. The successive generations from F1 to F4 are shown from left to right. N=9. The experiments were repeated 3 times.

C. Whole body triglycerides. 3–4 day old flies were used for the analysis. Triglyceride content was normalized to mean fly weight. Left panel shows triglycerides in F1- F4 male flies. Right panel shows female offspring F1-F4. Five flies were used for each sample, N=3.

D. Locomotor activity of male offspring. The activity was measured over 5-day period in TriKinetics activity monitors in a climate-controlled environment at 24°C under a 12h light-dark cycle and 65–70 % humidity. Flies were housed in narrow vials with food in groups of 5. The Y axis represents number of activity counts. Y axis shows average activity of 5 flies over 5-days. N= 12. The experiments were repeated 3 times.

E. Locomotor activity of female offspring. Conditions, groups, and animal numbers are identical to those listed in legend D.

Abbreviations: CFO- CD father offspring, EFO- exercise father offspring, WFO- WD father offspring, WEFO- WD +exercise father offspring.

Statistics. Error bars represent SEM. Asterisk in panels A, B, D, E shows significance according to one-way ANOVA. A pairwise comparison was performed using Tukey’s multiple comparisons test, adjusted p-values: *- P<0.05, **- P<0.01, ***- P<0.001, ****- P<0.0001. In panel C, # indicates significance according to unpaired t-test, #- P<0.05. The raw data are available upon request.

Ancestral WD negatively impacts locomotor activity and muscle and brain mitochondrial enzymes.

Since increase in food consumption might lead to obesity, we next performed biochemical assays to assess level of triglycerides in offspring (Fig. 2C). The highest level of triglycerides was found in the whole-body homogenate of F1 WFO flies, indicating obesity. Considering that obesity and triglycerides are usually inversely related to locomotor activity, we investigated whether ancestral WD might affect offspring’s activity. The activity of male and female offspring was recorded over 5-day period using LAM25H locomotor activity monitors (TriKinetics). The experiments showed that in male WFO offspring the activity was significantly increased in the F1 but decreased in F2-F4 (Fig.2 D). In contrast, activity in EFO offspring was higher in F1-F3 and decreased in F4. The activity of WEFO offspring was significantly lower than WFO in F1-F3 but higher in F4.

In female offspring, the activity of WFO group was significantly decreased in F1-F4, while EFO activity was higher in F1 and lower in F3-F4 (Fig 2E). WEFO activity was higher than WFO in F2-F4. Thus, according to these data, both paternal WD and exercise cause alterations in offspring activity in both sexes.

Activity relies on muscle bioenergetics including efficiency in oxidative phosphorylation (OXPHOS) efficiency, and activity of individual factors in the mitochondrial electron transport chain (ETC). To investigate whether changes in offspring activity were associated with OXPHOS, we performed western blot analysis for Cox4 (cytochrome c oxidase or Complex IV) and ATP5A, a subunit of the catalytic portion of the ATPase, known as Complex V. The experiments showed that Cox4 was significantly downregulated in WFO flight muscle according to on-way ANOVA. while ATP5a was downregulated in flight muscle of F1 WFO males according to t-test (Fig.3A, B, SI Fig. 1). Together these data suggest reduced mitochondrial aerobic capacity in WFO flight muscle.

Fig. 3. Effects of ancestral diet and exercise on mitochondrial enzymes and mitochondrial density.

A. Densitometry of CoxIV Western blots in flight muscles (for western blots see SI Fig. 1). Nitrocellulose membranes were probed with primary antibodies against CoxIV (Abcam). Expression was normalized against actin. Histograms represent averages of three Western blot runs. *- P<0.05 indicates significance between EFO and WFO according to one-way ANOVA and Tukey’s multiple comparisons test. #- P=0.0401, ###- P =0.0008, indicates significance between CFO and WFO according to unpaired t-test.

B. Densitometry of ATP5a Western blots in flight muscles (for western blots see SI Fig. 1). All conditions are identical to those listed in legend A.

C. Mitochondrial copy numbers in fathers’ brains and muscles (F0). Abbreviations: CF- CD fathers, EF- exercise fathers; WF- WD fathers, WEF- WD +exercise fathers.

D. Mitochondrial copy numbers in offspring brains.

E. Mitochondrial copy numbers in offspring flight muscle.

Mitochondrial copy numbers were measured by qPCR of rpol2 (single-copy nDNA) and mt:lRNA (16S, mtDNA). Each sample contained DNA from five flight muscles or ten brains. Average of three experiments, n=9. One-way ANOVA, a pairwise comparison was performed using Tukey’s multiple comparisons test, adjusted p-values *- P<0.05, **- P<0.01, ***- P<0.001. The raw are data available upon request.

Decreased mitochondrial density in muscles and the brain due to ancestral WD

Decrease in Cox4 and ATP5A could be due to the decline in mitochondrial function (67) or the mitochondrial density leading to a reduction in ATP-producing sites(68). In the following experiment, we assessed mitochondrial copy number (MCN) using qPCR approach (69). MCN was studied in brain and fly muscle of parental and offspring flies. The experiments showed that MCN levels were significantly lower in brains and muscles of WD-fed fathers (WF) (Fig 3C). Interestingly, exercise mitigated the negative effect of WD in WEF group and increased MCN in muscles of EF. In offspring, decreased MCN was observed in both brains and muscles of F1-F4 WFO males (Fig 3D, E). The decrease of MCN was less pronounced in female offspring reaching statistical significance only in F2 brains and F1, F3 muscles. This data suggests that ancestral WD might program mitochondrial density and influencing offspring’s bioenergetics.

Paternal WD causes deregulation of F1 brain proteome.

Emerging evidence suggests that obesity can cause alterations in brain proteome(70). To begin to explore whether ancestral WD may impact offspring proteome, nLC-MS/MS was used to interrogate the proteome of F1 offspring brain. This approach yielded 2,802 proteins identified and quantified across all samples. The top 10 most abundant proteins corresponded to known mitochondrial membrane and cytoskeletal proteins (ATP5a, betaTub56D, ATPsynβ, porin, alpha-Spec, Gs2, mAcon1, sesB, Adh, PyK). Abundance of ATP5a was not significantly different across all offspring lineages, confirming equal total amounts of protein across the samples (Fig 4). Using a P value less than 0.01 cutoff we have identified 74 differentially expressed proteins comparing WFO to CFO in males, and 70 differentially expressed protein comparing WFO to CFO in females, 28 differentially expressed proteins comparing EFO to CFO in males, and 34 differentially expressed protein comparing EFO to CFO in females, and 125 differentially expressed proteins comparing WEFO to CFO in males, and 103 differentially expressed protein comparing WEFO to CFO in females (Fig. 5, SI Fig.2 Heat map, SI Excel table1). Metascape Gene Analysis (https://metascape.org) showed a significant enrichment in terms related to translation and translation factors, small molecule metabolic process, purine metabolism, Rho GTPases, and carbohydrate metabolism (Fig 5B, D, F, H, J, L). The TCA cycle and ETC were among downregulated pathways in both WFO and WEFO males (SI Fig. 1). In WFO females the downregulated proteins were associated with “small molecule metabolic process”, “vesicle-mediated transport in synapse”, and “cellular homeostasis”. Upregulated proteins were related to “metabolism of RNA”, “translation”, and “heterochromatin organization”(SI Fig. 2).

Fig. 4. Protein abundance of ATP5a synthase in the brain from F1 offspring.

All offspring lineages expressed the same abundance of ATP5a, confirming the same amount of protein in all samples. A total of 40 brains were sampled per group, N=3. The raw data are available upon request.

Fig.5. Proteome analysis of the brains of F1 offspring by nLC-MS/MS.

A. Volcano plot depicting changes in the brain proteome between WFO and CFO males. Significance is indicated by the color and size of each circle, with ‘significance’ (p value <0.01) being represented by the red circles. The encircled protein is mei-P26, which has been identified as being upregulated in several lineages. A total of 40 brains were sampled per group, N=3. The X-axis indicates fold change.

B. Metascape analysis of proteomic differences between WFO and CFO males. Bar plot shows significant enrichment in GO terms and pathways. The top three terms highlighted with red frames were translation factors, small molecule metabolism, and purine metabolism.

C. Volcano plot showing changes in the brain proteome between WFOs and CFO females. The encircled protein is htt, which has also been shown to be downregulated in female WEFOs. Graph parameters, experimental conditions, and N are the same as those listed in legend A.

D. Metascape analysis of proteomic differences between WFO and CFO females. Three major terms are highlighted with red frames.

E. Volcano plot showing changes in the brain proteome between EFOs and CFO males. Graph parameters, experimental conditions, and N are the same as those listed in legend A.

F. Metascape analysis of proteomic differences between EFO and CFO males.

G. Volcano plot showing changes in the brain proteome between EFOs and CFO females. Graph parameters, experimental conditions, and N are the same as those listed in legend A.

H. Metascape analysis of proteomic differences between EFO and CFO females.

I. Volcano plot showing changes in the brain proteome between WEFOs and CFO males. dHYPK, the encircled protein, is a partner of htt. Graph parameters, experimental conditions, and N are the same as those listed in legend A.

J. Metascape analysis of proteomic differences between WEFO and CFO males.

K. Volcano plot showing changes in the brain proteome between WEFOs and CFO females. Graph parameters, experimental conditions, and N are the same as those listed in legend A.

L. Metascape analysis of proteomic differences between WEFO and CFO females.

Using a 2-fold cutoff, we identified a group of WD regulated proteins. Interestingly, the ortholog of human Huntingtin (Htt) was downregulated in both WFO and WEFO female (volcano plots Fig. 5 C, K). Htt encodes a scaffold protein involved in mitotic spindle orientation, chromatin regulation and axonal transport(71, 72). Intriguingly, an ortholog to human huntingtin interacting protein K (dHYPK) was 3.6-fold decreased in WEFO males (volcano plot Fig. 5I). Among top upregulated proteins, mei-P26 (meiotic P26) was elevated in both WFO and WEFO males and WEFO females (volcano plots, Fig. 5A, I, K). Recent evidence implicates mei-P26 in translational regulation via microRNA pathway (73, 74) as well as seizure susceptibility in flies(75).

The evolutionary conserved miRNA mir-10 is identified as a potential epigenetic regulator by miRNA target prediction.

Since miRNAs play significant role in transgenerational phenotype (76, 77), we performed miRNA-target enrichment analysis using the web tool MIENTURNET (MicroRNA ENrichment TURned NETwork)(78). The analysis was performed on subset of proteins with significant enrichment for terms related to translation, small molecule metabolic processes, and carbohydrate metabolism. The results showed that mir- 277–3p, mir-10–3p and mir-927–5p had the highest number of predicted targets (Fig 6A). Conversely, mei-P26 contained the highest number of conserved sites for miRNAs including mir- 277–3p, mir-10–3p and mir-927–5p (Fig 6B). To identify miRNA regulated pathways, we used functional annotation DIANA-tools mirPath v.3 (www.microrna.gr)(79). MirPath analysis identified the “Purine metabolism” pathway to be targeted by all three miRNAs: mir-277–3p, mir-10–3p, and mir-927–5p (Fig. 6C). This is interesting because purine metabolism plays an important role in DNA and RNA synthesis, as well as intracellular signaling associated with bioenergetics and cancer(80, 81).

Fig. 6. MiRNA-target enrichment analysis of proteomic data.

A. Computational evidence of miRNA regulation on target genes based on a statistical analysis for over-representation of miRNA-target interactions. The Y-axis indicates the top ten miRNAs from the enrichment analysis and the X-axis indicates how many proteins were targeted. The color code reflects the FDR value increasing from red to blue.

B. A bar plot of the mRNA degree, where X-axis refers to the first 30 target genes (sorted in a decreasing order according to the degree) and Y-axis refers to their degree (i.e., number of miRNA targeting them).

C. miRNA functional analyses using KEGG annotations using DIANA-tools mirPath. The results pane shows information regarding targeted pathways, p-values, as well as the number of miRNAs and genes present in each term. Red frame highlights purine metabolism, as a pathway targeted by all three miRNAs: mir-277–3p, mir-927–5p, and mir-10–3p.

miR-10 knockdown increases food consumption

We focused on miR-10 for further experiments because, it is conserved in all bilaterian animals, including humans(82) and has been implicated in variety of epigenetic processes including cancer(83), regulation of Hox translation(84) and control of cell differentiation(85). Because, sperm miRNA has been recently implicated as a carrier of epigenetic information(76, 77, 86) we examined expression of mir-10 by qRT-PCR in the brain and in the purified spermatozoa. The experiments revealed significant alterations in brain miRNAs in F0 and in F1 especially female flies (Fig. 7A). Increase of several miRNAs including miR-10–3p was found in paternal and offspring spermatozoa (Fig. 7B), while mir-277–3p and several other miRNAs were undetectable. miRNAs were also altered in the flight muscle of F0 and F1 (SI Fig 4 and 5). We then used the UASxGAL4 strategy to investigate if altering mir-10–3p in the brain will affect feeding behavior. UAS-miR10 lines were induced with dopa decarboxylase (Ddc)-GAL4 driver. Ddc-Gal4 encodes the promoter region of the Ddc gene, which is expressed in dopamine and serotonergic neurons, which are involved in both aversive and appetitive behaviors(87). We used two RNAi lines (35014 and 61377) targeting miR-10, thus avoiding potential off-target effects associated with a particular construct. 35014 encodes dsRNA for RNAi of mir-10 (FBgn0262424) under UAS control in the VALIUM20 vector. 61377 encodes an antisense ‘sponge’ RNA under UAS control for knocking down mir-10 expression. For control, we used the parent lines or F1 cross to Canton-S (CS). The analysis of the of F1 feeding behavior showed that both male and female flies had significant increase in number of licks knockdown for miR-10 in 7009×35014 cross (Fig. 7C). The biochemical assay revealed an increase in total body triglycerides in males and a positive trend in female offspring (Fig. 7D). qRT-PCR in the brains demonstrated a reduction in miR-10–3p in 7009×35014 confirming successful knockdown (Fig. 7E). Together, these findings suggest that might miR-10 contribute to increased food consumption in offspring.

Fig.7. Functional analyses miR-10 in modulating feeding behavior.

A. MiRNA expression in the F0 and F1 brains. Relative fold change in comparison to CF and CFO taken as 1. WFO-M- WFO male offspring, WFO-F- WFO female offspring. Error bars represent SEM. Based on one-way ANOVA, the asterisks indicate significance when comparing WF to CF or WFO to CFO. A total of 10 brains were sampled per group, N=3. A pairwise comparison was performed using Tukey’s multiple comparisons test. # Indicates significance according to unpaired t-test. *- P<0.05, #- P<0.05.

B. miRNA expression in F0 and F1 spermatozoa. in F0 spermatozoa. Relative fold change in comparison to CF and CFO taken as 1. Spermatozoa of 20 flies were pooled for each sample, N = 3. The asterisks indicate significance when comparing WF to CF or WFO to CFO based on one-way ANOVA. A pairwise comparison was performed using Tukey’s multiple comparisons test. # Indicates significance according to unpaired t-test. *- P<0.05, #- P<0.05, ##- P< 0.01.

C. The effect of knocking down miR-10 in Ddc neurons on feeding behavior. Consumption of 10% sucrose was measure over 24 h period starting at12pm. Each group consisted of 9 flies (N=9). The experiments were repeated 3 times. One-way ANOVA, followed by Tukey’s multiple comparisons test. *- P<0.05, **- P<0.01, ****- P<0.0001.

D. Effects of knockdown of miR-10 in Ddc neurons on whole body triglycerides (TAG). TAG content was normalized to protein per mean fly weight. Five flies were used for each sample, N=5. Asterisk indicates significance according to unpaired t-test. ****- P< 0.0001

E. qRT-PCR verification of miR-10 levels in mutant fly brains. Relative fold change in comparison to parental strain taken as 1. A total of 10 brains were sampled per group, N=3. One-way ANOVA followed by Tukey’s multiple comparisons test, *- P<0.05. #- P<0.05 indicates significance according to unpaired t-test. The raw data are available upon request.

Abbreviations: 7009 - Expresses GAL4 in dopaminergic and serotonergic neurons under the control of Ddc.

35014 - Expresses dsRNA for RNAi of mir-10 (FBgn0262424) under UAS control in the VALIUM20 vector.

61377- Expresses an antisense ‘sponge’ RNA under UAS control for knocking down mir-10 expression.

CS- Canton S. X- signifies F1 cross between two parental lines.

4. Discussion.

While transgenerational metabolic programing has been previously reported in a fruit fly(41, 46), rodents (37, 38, 88) and humans (89, 90), to the best of our knowledge, this is the first report documenting effect of ancestral diet on offspring feeding behavior. Our experiments revealed that paternal WD increases food consumption up to the 4th generation. This was also accompanied by increases in triglycerides and decreases in locomotor activity suggesting that offspring acquired “couch potatoes” phenotype. There was also an indication of decrease in bioenergetic efficiency as we observed a reduction in mitochondrial copy numbers and mitochondrial subunits Cox4 in male flight muscles. This data are supported by previous observations on alterations in mitochondrial complexes after high-fat diet in Drosophila(91), and transgenerational inheritance of mitochondrial dysfunction in mice(88).

In F1, alterations in mitochondrial density and feeing behavior were associated with significant proteomic changes in the brain. Both male and female offspring had significant changes in protein levels with associated enrichment for GO terms and pathways related to translation and translation factors, small molecule metabolic process, purine metabolism, Rho GTPases, and carbohydrate metabolism. Downregulated pathways in WD lineages included TCA cycle and ETC which could indicate impairment in mitochondrial bioenergetics. Two proteins related to Huntington’s’ disease (HD), htt and dHYPK were significantly downregulated. Although the exact function of htt protein is not well understood it plays an important role in embryonic development, nerve cells signaling and axonal transport(92). HYPK- an interaction partner of HTT, is an intrinsically disordered protein involved in chaperone activity, protein folding, cell cycle, apoptosis and transcription regulation(93, 94). Hypothalamic expression of mutant htt in a mouse causes distinct metabolic changes including, hyperphagia, weight gain and decrease in locomotor activity suggesting the role of HTT in metabolic control via hypothalamic neurocircuits.(95) Emerging evidence suggest that altered central and peripheral energy metabolism is a major contributor to HD pathology in humans(96).

Using the web tool MIENTURNET, miRNA enrichment analysis identified mir-10–3p as the conserved miRNA with the most predicted targets. In addition to being conserved in bilaterian animals, miR-10 is found in homeobox gene clusters(82) and implicated in cell differentiation and development(85). MiR-10 was also identified as a spermatozoal miRNA upregulated by paternal obesity in the sperm of F1 male mice(97). In a mouse model of thrifty phenotype, we found that miR-10 is also a top spermatozoal miRNA regulated by exercise (Murashov, unpublished observations). The results of qRT-PCR showed that miR-10–3p is regulated by WD in brains and spermatozoa. Using the UASxGAL4 strategy we induced UAS-miR10 knockdown with Ddc-GAL4 driver. The analysis of the of F1 feeding behavior using FLIC revealed that both male and female flies had significant increase in number of licks indicating that miR-10 might play a mechanistic role in programming feeding behavior.

Intriguingly, some data including locomotor activity showed fluctuations between generations. In laboratory rodents, recent findings showed that metabolic phenotypes via the paternal lineage were more pronounced in HFD F2- and HFD F3-offspring than in F1-offspring(98). In a model of maternal cafeteria diet, certain phenotypes were skipped in the first generation and appeared in the second(99). Furthermore, previous observations on prenatal endocrine disruption in F1-F3 have demonstrated that each generation has a unique phenotype(100). The variance could be explained by the effects of exposure on fathers’ germ cells, which affect F1 directly and are therefore generational. In contrast, the phenotypic changes in later generations aren’t directly influenced by paternal exposure, so they’re transgenerational (11, 26, 30, 101). However, it is not clear when, the next generation may return to the original phenotype in the absence or recurrent stress. It is possible that the offspring’s phenotype is itself a stress that reinitiates the cycles of transgenerational effects (102). Among the F2, F3 and F4 activity data, we observed trends of increasing activity in one generation and then decreasing activity in the next. It is possible that these fluctuations are due to an increase in “developmental noise” or so called “intangible variation” (66). It is noteworthy that Waddington’s classic papers on epigenetics were partly inspired by his interest in noise or so called “developmental accidents” (66, 103). In evolutionary terms, developmental noise might provide a competitive advantage, as the increased phenotypic variability might aid in preventing extinction under stressful environment (104, 105).

While the precise mechanism of transgenerational programming remains unclear, recent observations indicate that excess fat intake alters mitochondrial proteome in the skeletal muscle (106). OXPHOS machinery was also downregulated in adipose tissue analyzed by proteomic profiling in diet-induced obese mice (107). Furthermore, recent data suggest that elicited changes in the mouse epididymal proteome may alter sperm small non-coding RNA profiles and dysregulate embryo development (108). Based on these observations, we propose that WD leads to mitochondrial proteome remodeling followed by alterations in miRNAs. Consequently, sperm miRNAs might act as information carriers that transmit phenotypic traits from generation to generation (31, 109) (Fig7).

The study had the limitation of not directly measuring food consumption. The FLIC system, which measures fly-food interactions electronically, such as “licks”, also counts accidental contacts with food. As a result, an increase in locomotor activity could potentially result in increased counts. While this possibility certainly exists, data show a decrease in activity and an increase in licking among WD-lineage offspring. To better understand feeding behavior, food preference and CAFÉ assays should be used in future research.

Together, our data demonstrate that ancestral caloric overload led to alterations in feeding behavior in four successive generations. Although it is not yet clear whether similar effects are observable in human populations, the results of this study as well as familial clustering of obesity let us to speculate that under certain circumstances offspring’s food preferences might become preconceptually hard-wired into the brain. The findings identify ancestral nutrition as a potential factor in programming obesity-risk behaviors and could provide better insight into mechanisms of the familial susceptibility to obesity and the obesity pandemic in general.

Fig. 8. Generational cycle of transgenerational phenotype hypothesis.

As a result of caloric overload, metabolic efficiency is decreased, followed by proteome remodeling followed by dysregulation of miRNAs. Consequently, miRNAs serve as carriers of phenotypic information that transmits from generation to generation. A vicious cycle is created when miRNAs influence development by altering proteomic function and mitochondrial function, which leads to behavioral compensations such as overeating and a preference for energy-dense foods.

Abbreviations:

WD

western diet

CF

control diet fathers

EF

exercise fathers

WF

western diet fathers

WEF

western diet +exercise fathers

CFO

control diet father offspring

EFO

exercise father offspring

WFO

western diet father offspring

WEFO

western diet +exercise father offspring

DATA AVAILABILITY:

The proteomic data analyzed during the current study are available online. The accession numbers are PXD033363 for ProteomeXchange and JPST001568 for jPOST repository, http://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD033363.