Abstract
The mechanisms by which the genotype interacts with nutrition during development to contribute to the variation of complex behaviors and brain morphology of adults are not well understood. Here we use the Drosophila Genetic Reference Panel to identify genes and pathways underlying these interactions in sleep behavior and mushroom body morphology. We show that early-life nutritional restriction has genotype-specific effects on variation in sleep behavior and brain morphology. We mapped genes associated with sleep sensitivity to early-life nutrition, which were enriched for protein-protein interactions responsible for translation, endocytosis regulation, ubiquitination, lipid metabolism, and neural development. By manipulating the expression of candidate genes in the mushroom bodies and all neurons, we confirm that genes regulating neural development, translation and insulin signaling contribute to the variable response of sleep and brain morphology to early-life nutrition. We show that the interaction between differential expression of candidate genes with nutritional restriction in early life resides in the mushroom bodies or other neurons, and that these effects are sex specific. Natural variation in genes that control the systemic response to nutrition and brain development and function interact with early-life nutrition in different types of neurons to contribute to the variation of brain morphology and adult sleep behavior.
Introduction
Nutrition is an environmental factor that plays a crucial role in the maturation and functional development of the central nervous system 1–3. In mammals, including humans, severe prenatal malnutrition negatively impacts neural development and complex behaviors such as sleep, memory, and learning 1, 4–7. At the population level, adults that were exposed to hunger in utero have an increased risk to develop schizophrenia, affective disorders, addiction and decreased cognitive function 8, 9. The origin of these disorders may be associated with defects in early brain development 10.
Little is known about the mechanisms by which individual genotypes (G) respond differently to a nutritional environment (E) during development. We define this type of genotype by environment interaction (GEI) as a genotype by early-life nutrition interaction (GENI) 11. Drosophila provides exceptionally powerful tools and approaches for exploring the mechanisms underlying GENI at the single gene and genome-wide level 12. The D. melanogaster Genetic Reference Panel (DGRP) 13 consists of sequenced inbred lines derived from a natural population that has been extensively used to chart the genotype-phenotype architecture of complex traits, including behaviors and brain morphology 12. DGRP lines reared under different nutritional conditions show changes in behaviors and metabolic and transcriptional profiles, revealing a key role of GENI 14–18.
Alterations in sleep behavior are a common symptom of many neurological and psychiatric disorders, including neurodegenerative dementias and schizophrenia 19, 20. Therefore, uncovering the genes and pathways underlying the contribution of GENI to sleep behavior variation may shed light on altered neurodevelopmental mechanisms that can lead to mental illness. Few studies have focused on the role of early-life nutrition in sleep behavior. Prenatally malnourished rats exhibit decreased sleep and increased waking activity 21, 22; however, whether genetic variation affects such responses remains elusive. The influence of genetic variation on the effects of restrictive early-life nutrition on adult sleep behavior and brain structure remains unknown.
A genome-wide association (GWA) study of sleep using the DGRP 23, allowed the identification of naturally occurring sleep behavior-related genetic variants. Many of these variants were within or near candidate genes with human orthologs that have been associated with sleep, which suggests that genes affecting variation in sleep are conserved across species 23. Several of the candidate genes associated with natural variation in sleep affect developmental processes and neural function 23, 24 supporting the idea that variation in sleep is influenced by variation in the brain structures that control it.
The Drosophila mushroom bodies (MBs), an associative memory center, play an important role in sleep regulation 25–27. Early work showed that chemical ablation of the MBs results in a significant decrease in sleep 25, 26. MBs contain two types of sleep- regulating neurons: those that promote sleep when cyclic-AMP-dependent protein kinase A (PKA) is increased and those that inhibit sleep under such conditions 26. Each MB consists of around 2000 Kenyon cells (KCs) whose axons are arranged in parallel arrays projecting into different lobular structures, the α and β, the α’ and β’ and the γ lobes 28. KCs are sequentially generated from four neuroblasts in each hemisphere 29 that start to proliferate in embryos at stage 13 and continue uninterrupted until adult eclosion 29, 30. Therefore, nutritional restriction during early-life stages is likely to impact the development of MBs 31. Interestingly, natural genetic variation in the length and width of the α and β lobes has been correlated with variation in aggression and sleep behaviors 32.
Here, we assessed the effect of GENI on adult sleep behavior and MB morphology in the DGRP and found significant effects on both. We used GWA analyses to identify genetic variants and top candidate genes underlying GENI in the variation of sleep traits. Many proteins encoded by candidate genes are expressed in the MBs and form conserved protein-protein interaction networks required for neural development, translation, endocytosis regulation, ubiquitination, and lipid metabolism. Since genetic variants may affect the function or expression of candidate genes, we asked whether decreased expression of candidate genes in the MBs or in all neurons modifies sleep and MB morphology in response to early-life nutritional restriction. We found that diminished expression of a group of genes required for neural development, transmission of the nerve impulse and splicing, in specific neuronal populations, modifies the response of sleep behavior and MB morphology to early-life nutritional restriction. Together, our results suggest that tissue-specific variation in the expression of genes controlling these processes underlie GENI in sleep behavior and MB morphology.
Results
To determine the consequences of nutritional restriction during early life on sleep behavior and brain morphology variation, we first evaluated whether reducing larval nutrients to 20% (restricted food, RF) of the standard culture medium (normal food, NF) affects development by analyzing adult size (S1 Fig). Flies reared in RF show a reduction in the notum length (NL) (S1C and S1F Figs), while the interocular distance (IOD) is not affected (S1B and S1E Figs) 33. These data confirm that nutrient restriction during development reduces the size of the thorax but not the head 31, which is evident in the increased IOD/NL ratio in flies reared in RF (S1D and S1G Figs).
We raised larvae of 73 DGRP lines on NF or RF and transferred newly eclosed adults to NF for three days to evaluate sleep behavior. We quantified nine sleep traits: total sleep duration, day and night sleep duration, day and night sleep bout number, day and night average sleep bout length, waking activity, and latency (i.e., the time it takes for the flies to start their first sleep bout after the lights are off) (Fig 1; S2 and S3 Figs; S1 and S2 Tables). We observed significant variation in sleep traits among the DGRP lines reared under both nutritional conditions (Fig 1; S2 and S3 Figs; Table 1; S1-S3 Tables).
Cross-sex genetic correlations (rFM), which represent the extent to which the same variants affect a trait in females and males, were significantly different from unity (NF, rFM = 0.48-0.91; RF, rFM = 0.62-0.88) (Table 1 and S3 Table). Thus, some polymorphisms affect sleep susceptibility to rearing diet in both sexes, while others will have sex-specific or sex-biased effects.
The differential responses of different genotypes to RF are evident from the complex pattern of crossing reaction norms, which is a hallmark of genotype by environment interaction (Fig 1; S2 and S3 Figs). To quantitate the contribution of GENI to the genetic variance, we estimated the interaction coefficient (i2) across food, which is calculated by subtracting the cross-environment genetic correlations from 1 (1- rNR). Estimates for i2 showed GENI contribution to night sleep traits genetic variation range from 9% in night bout length to 16% in night bout number in females, and from 7% in night sleep to 11% in night bout length in males. In contrast in day sleep traits, GENI contribution ranges from only 2% in day bout length to 10% in day bout number in females, and from 3% in day sleep to 13% in day bout number in males with no contribution to day bout length (Table 1 and S3 Table). Therefore, we decided to exclude day bout length from further analyses.
These data support a role of GENI in sleep variation, which may depend on variation in genes that act during development or whose adult expression is programmed during development in response to nutrient restriction.
GENI contributes to morphological variation of the mushroom bodies
To assess to what extent GENI contributes to variation of MB morphology, we raised 40 DGRP lines under both dietary conditions during larval stages and examined the MB morphology of adult females. These include 36 lines that were previously used to demonstrate the natural variation of MB morphology 32. First, we assessed the gross morphology of α and β lobes 32. We observed a variety of large morphological defects at a broad range of frequencies (5–80%) (Fig 2; S4 Table), including missing or very thin structures and lobe fusions. These gross defects have been attributed to mutations that have major effects on MBs morphology 32, or could be due to multiple variants in each line. Strikingly, a total of 14 (35%) and 6 (15%) DGRP lines exhibited a decrease of the α- and β- lobe defects, respectively, when reared on restricted food (Fig 2; S4 Table). In turn, a total of 9 (22%) and 17 (42%) DGRP lines showed an increase of the α- and β- lobe defects, respectively, when reared on restricted food (Fig 2; S4 Table). These data indicate that early-life nutrition affects gross MB morphology.
We next evaluated quantitative variation in MB morphology by measuring the length and width of the α and β lobes (Fig 3A; S5 and S6 Tables) 32 to reveal more subtle effects on the morphology of brain structures (Fig 3). Quantitative genetic analyses revealed substantial and significant genetic variation among the lines for length and width means of α and β lobes (Table 2; S7 Table). The contribution of genetic variation to phenotypic variation ranged from low for β-lobe length (H2 = 0.07 in NF and H2 = 0.13 in RF) to moderate for β-lobe width (H2 = 0.28 in NF and H2 = 0.30 in RF), α-lobe length (H2 = 0.23 in NF and RF) and α-lobe width (H2 = 0.31 in NF and H2 = 0.32 in RF) (Table 2; S7 Table).
We found a significant line by food interaction term (L x F) for all four traits, indicating that flies with different genotypes respond differently to RF (Fig 3A’-D’, S7 Table). The contribution of GENI to morphological traits, measured as i2, was low for α- lobe width (i2 = 0.06), moderate for α-lobe length (i2 = 0.27) and β-lobe width (i2 = 0.18), and high for β-lobe length (i2 = 0.65) (Table 2 and S7 Table), indicating that the contribution of GENI to genetic variation highly depends on the trait.
In summary, we found that GENI contributes considerably to variation of MB morphology, suggesting an essential role in development and structure of this brain region.
Genetic variants associated with GENI for sleep
To identify genetic variants underlying GENI in sleep behavior and MB morphology, we performed GWA analyses for the difference of phenotypic values for each trait between the two diets. We also excluded traits which Q-Q plots did not show larger than expected numbers of P-values below 10-5. These include day and night bout number in females, day sleep in males, and total sleep and waking activity in both sexes, and all morphology traits (S4-S6 Figs).
All variants associated with sleep traits from both sexes were pooled together for subsequent analysis as variants associated with sleep susceptibility to early-life nutrition. We found a total of 1,162 variants across all sleep traits (at a nominal reporting threshold of P ≤ 10-5) that mapped in or near to 611 candidate genes (S7 and S8 Figs; S8 Table). Among these, 11% of such variants were located within coding sequences, while 47% were in introns and the 5’ and 3’ untranslated region (UTR), and the remaining 42% were classified as intergenic (more than 1 kb away from the gene body) (S9 Fig; S8 Table). 95 (16%) candidate genes are highly enriched for GO terms (FDR < 0.05) associated with the function and development of the nervous system, including nervous system development, neurogenesis, neuron differentiation and development, and axonogenesis (S10 Fig; S8 Table). Twenty percent (122 out of 611) of the candidate genes are expressed in MBs either during larval stages or in adults (S9 Table) 34–37, 9 have orthologs associated with aspects of human brain size, and 41 genes have orthologs associated with human sleep traits (S11 Fig; S8 Table). These results suggest that the effect of early-life nutrition on sleep behavior in adulthood occurs in part through the regulation of neurodevelopmental mechanisms.
To identify potential cellular processes and molecular pathways underlying GENI, we generated protein-protein interaction networks with proteins encoded by candidate genes using the STRING database, which considers interactions based on direct (physical) and indirect (functional) associations 38. These proteins connect through processes that include translation (e.g., RpS27A and RpL23), vesicular trafficking (e.g., drongo), ubiquitination (e.g., Trim9 and Kel), lipid metabolism (e.g., LpR1 and LpR2), neural development (e.g., Fas2), and protease activity (e.g., Fur1 and Mmp2) (Fig 4). We found a significantly enriched network (PPI enrichment P-value = 1.06 x 10-5) using a high confidence score (score ≥ 0.700) (Fig 4).
Functional assessment of candidate genes associated with GENI for sleep
We asked whether early-life nutrition modifies the knockdown effects of 17 candidate genes. These were chosen based on their role in nervous system development, our network analysis, or both (Figs 4 and 5; S10, S12-S15 Figs; S10 Table).
To test the hypothesis that these genes affect GENI in MBs we reduced their expression from embryonic stages onwards in all MB neuroblasts by driving the expression of specific UAS-RNAi transgenes using the OK107-Gal4 driver (Fig 5; S12- S13 Figs; S10 Table) 35. We use the 201Y-Gal4 driver for genes that showed an effect with OK107-Gal4, as the latter is strongly expressed in other brain regions, including the optic lobe, pars intercerebelaris (PI), tritocerebrum (TR), and subesophageal ganglion (SOG) (Fig 5; S15 Fig; S10 Table). 201Y-Gal4 displays a more restricted pattern both in the MB, specifically in the γ and α/βc and neurons, and to a lesser extent in the PI, TR, and SOG 39.
In addition, to control for the specificity of the RNAi-mediated knockdown, we use a second RNAi for candidates that showed effects with OK107-Gal4 (S14 Fig; S10 Table). To distinguish whether the effects on sleep are specific to MBs, other regions of the nervous system, or both, we also drove the expression of UAS-RNAi transgenes in all neurons using the elav-Gal4 driver (Fig 5; S12-S15 Figs; S10 Table) 40. We reared flies expressing the RNAi and their specific control genotypes in NF or RF and quantified sleep behavior in adults (Fig 5; S12-S15 Figs; S10 Table).
All candidate gene RNAis tested affected at least one sleep trait in response to early-life nutrition in males, females, or both sexes, suggesting pleiotropy (S13 Fig). Interestingly, early-life nutrition modifies the effects of the expression of the RNAis with elav-Gal4 in 33 traits, while with OK107-Gal4 in 57 traits, supporting a key role of the MB in GENI (S13 Fig). The knockdown of 7 genes affects the same trait associated with the candidate genetic variant, supporting their role as specific modifiers of behavioral response to diet (Fig 5, S13 Fig). Among them Tenascin-a (Ten-a) 41, axotactin (axo) 42, guk-holder (gukh) 43, Multiplexin (Mp) 44, Protein tyrosine phosphatase 61F (Ptp61F) 45, and wunen (wun) 46, affected more traits than in the GWA analyses. On the other hand, early-life nutrition only modifies RNA-binding Fox splicing factor 1/ Ataxin 2-binding protein 1 (Rbfox1) 47 effect in night sleep, supporting their specific role in this trait.
In females, the expression of an RNAi targeting axo (axo-RNAi) with OK107- Gal4, a gene required for transmission of the nerve impulse 42, suppresses night average bout length increase in response to early life nutrition (Fig 5, S13 Fig, S10 Table). We found a similar response when axo-RNAi was driven with the 201Y-Gal4 driver, supporting its specific role in the MB. Here, RF suppresses the difference in phenotypic values between the control and the RNAi expressing flies observed in NF (Fig 5A). The expression of a second axo-RNAi (axo (2)-RNAi) showed similar but not significant effects (S14A and S15 Figs). These data suggest that variation in axo expression in the MB may modify the adult’s response to early-life nutrition restriction.
Unlike axo-RNAi, the expression of Ptp61F-RNAi in the MBs with OK107-Gal4 increases night average bout length in response to RF, and suppresses it when expressed with elav-Gal4, suggesting that its expression in different neuronal populations may affect in different ways sleep response to early-life nutrition restriction. Although Ptp61F second RNAi (Ptp61F (2)-RNAi) did not show similar effects in response to diet (compare Fig 5B with S14B and S15 Figs) we decided to continue our work using Ptp61F-RNAi since its functionality has been proven in hemocytes, where it promotes activation of Jak/Stat 48.
Analysis of night sleep shows that expression of Rbfox1-RNAi or Ten-a-RNAi under the control OK107-Gal4 slightly suppresses the night sleep increase in RF in males. In addition, expression of Ten-a RNAi under the control of elav-Gal4, decreases night sleep in NF but not in RF (Fig 5, S13 Fig, S10 Table). Expression of a second Rbfox1-RNAi (Rbfox1 (2)-RNAi) under the control of OK107-Gal4 phenocopies the Rbfox1-RNAi effect (S14C and S15 Figs, S10 Table). These data support the hypothesis that variation of the expression of Rbfox1 in cells of the MBs that only expresses OK107-Gal4 but not 201y-Gal4 underlies GENI in night sleep behavior. On the other hand, although Ten-a (2)-RNAi did not phenocopy the effect of Ten-a RNAi, we continue using the latter since the pan-neuronal expression of Ten-a-RNAi results in undetectable levels of Ten-a immunostaining, and that the function of Ten-a (2)-RNAi has not been tested 49 (S14D and S15 Figs, S10 Table).
Interestingly, pan-neuronal expression of and RNAi targeting gukh, a gene involved in cell polarity and synapse formation 43, increases night sleep in response to early-life nutrition restriction (S12A and S13 Figs, S10 Table). In contrast, its expression under the control of OK107-Gal4 increases sleep in both diets and suppresses the increase in sleep in response to RF observed in the control genotype. These data suggest that gukh expression in different neuronal populations affects the response to early-life nutrition restriction in different ways, similar to Ptp61F-RNAi (S12A and S13 Figs, S10 Table). On the contrary, knockdown of Mp 44 using elav-Gal4 suppresses the increase in night sleep observed in the control genotype (S12B and S13 Figs, S10 Table).
Finally, expression of an RNAi targeting the lipid phosphatidate phosphatase wunen (wun) under the control of OK107-Gal4 in females increases latency in response to RF, while in males, it increases latency in both diets (S12C, S12D and S13 Figs, S10 Table). Moreover, its knockdown using elav-Gal4 in females increases latency to the same extent in NF and RF, while in males, it decreases it in response to RF (S12C, S12D and S13 Figs, S10 Table). These data indicate that the interaction between wun knockdown and early-life nutrition depends on the sex and neuronal population affected and suggest that early-life nutrition modifies the effect on sleep of diminished expression of wun in non-MB neurons.
In summary, we show that reducing candidate gene expression in all neurons or specifically in MBs, can have different effects on sleep behavior upon early-life nutrition and that these effects can be sex-specific. Moreover, depending on their local expression in the nervous system, candidate genes may be involved in different aspects of sleep regulation in response to early-life nutrition. These data support the idea that a subgroup of candidate genes associated with variation in sleep behavior act at the MBs underlying GENI in sleep behavior.
Early-life nutrition restriction modifies the effect of PTP61F and Rbfox1 knockdown in MB morphology
Since early-life nutrition modifies the sleep effects associated with axo, Ptp61F, Rbfox1, and Ten-a knockdown in the MB, we asked whether they affect MB morphology and whether their effects are modified by early-life nutrition. We used the OK107-Gal4 driver since it is strongly expressed in all MB neurons 39.
We found that α-lobe width and β-lobe width increases in response to RF in flies expressing Ptp61F-RNAi under the control of OK107-Gal4, but not in control flies (Fig 5A and 5B, S16 Fig, S11 Table). On the other hand, the expression Rbfox1-RNAi decreases α-lobe length in response to RF, while the control genotype does not display a response to diet, resulting in a significant length reduction RF (Fig 5C, S16 Fig, S11 Table). In addition, Rbfox1 knockdown increases the α-lobe width in RF and β-lobe length in NF, however, these genotypes as the control did not show differences between diets (S16 Fig, S11 table). Finally, Rbfox1 knockdown increases β-lobe width in NF and RF (Fig 5H, S11 table), but both the control and Rbfox1-RNAi expressing flies decrease β-lobe width in response to RF, indicating plasticity but not GENI (Fig 5, S11 table). Together, these data indicate that variation in the expression of Ptp61F or Rbfox1 in the MB neurons interacts with early-life nutrition affecting MB morphology.
Finally, we found that the expression of Ten-a-RNAi and axo-RNAi did not show clear GENI effects. Their knockdown increases α-lobe width in both diets (S16 Fig, S11 table) and axo knockdown also increases α-lobe width in response to RF indicating plasticity (S16 Fig, S11 table). Ten-a knockdown increases β-lobe length in RF but not in NF suggesting weak GENI. Finally, we did not find differences between the axo knockdown and the control in the other traits in response to diet (S16 Fig, S11 table).
Discussion
The development and function of the central nervous system in animals, including humans, rely on proper nutrition during the prenatal period 22, 50. In humans, early-life nutritional restriction increases the risk of mental and neurological diseases associated with morphological brain defects 51, 52. To what extent the interaction between the genotypes of individuals interacts with early-life nutrition to sensitize or protect from developing pathological behaviors is unknown. The use of animal models and genetic reference panels in mammals and Drosophila, such as the DGRP, have been critical to determining the contribution of GENI in complex behaviors at the genome-wide level16, 17.
Adult flies display genetic variation in the sensitivity of sleep behavior and MBs morphology to early-life nutrition. Moreover, we found that gross morphological defects displayed by a group of DGRP lines can be strongly modified by nutrition, indicating that early-life diet plays an important role in modifying the phenotype associated with single or multiple mutations that strongly affect morphology.
We identified candidate genes and molecular processes whose variation may support sleep behavior variation in response to the environment. A total of 135 of the 611 candidate genes (22%) associated with GENI for sleep phenotypes were found in a previous analysis of sleep variation in the DGRP under standard nutritional conditions 23 (S16 Table), and only 12 out of the 611 have been previously linked to sleep phenotypes 53. Thus, the contribution of GENI to variation in sleep behavior involves a new set of genes.
Although the study of only 73 lines may lead to a higher rate of false positives 54 and to miss interesting candidate genes due to the low frequency of some variants, we were able to find a significant protein-protein interaction network that reveals biological processes that affect GENI 38. We identified a subset of cellular processes involved in proteostasis, lipid metabolism, and nervous system development and function, suggesting that perturbations in the proper expression of genes encoding these proteins may impact central nervous system function later on. Cytosolic and mitochondrial ribosomal proteins generate a subnetwork including Rps27A, a ribosomal protein fused to a single copy of ubiquitin 55, and RpL23 55, 56. Mutants of ribosomal proteins result in the Minute phenotype, characterized by delayed development due to reduced protein biosynthesis 57, 58. This condition decreases Drosophila insulin-like peptide (Dilps) secretion systemically by affecting the insulin-producing cells (IPCs) in a cell- autonomous manner, leading to reduced body size and delayed larval development 59. Our results showed that variation in the expression of Rps27A and RpL23 in the MB or all neurons modify sleep behaviors in response to early-life nutrition (S13 Fig and S10 Table), suggesting that modulation of protein biosynthesis can play a neuronal-specific role in GENI. Since neuronal knockdown Rps27A or RpL23 does not affect latency or night sleep, to whose variation the original genetic variants are associated, we support the hypothesis that they play a role in non-neuronal cells, such as in the fat body, contributing to GENI more systemically.
Rps27A is a central network node connecting the ribosomal proteins subnetwork with other subnetworks containing the ubiquitin ligases and the lipid receptor proteins LpR1, LpR2, and Megalin 60. LpR1 is required in a group of octopaminergic neurons to enhance starvation-induced hyperactivity by inhibiting the degradation of the adipokinetic hormone receptor 61. Here we found that LpR2-RNAi expression with OK107-Gal4 driver suppresses the increase in night sleep in response to early-life undernutrition in males and the decrease of night bout numbers in females (S13 Fig and S10 Table).
The network also contains a group of proteins involved in axon growth, targeting, synaptogenesis, and fasciculation, including PTP99a 62, Nrg 63, and Fas2 64, 65. Another group is enriched in protein degradation, highlighting the ubiquitin ligase encoded by Trim9, which mediates Netrin function in axon guidance 66. These data suggest that variation in the expression of genes required for neuronal terminal differentiation contributes to GENI in sleep behavior.
Recent GWA studies on sleep traits and insomnia in humans provide insight into the genetic basis for variation in sleep 67–70. We found that 41 genes identified in these GWA studies, including FURIN and RBFOX1, are human orthologs of Drosophila genes associated with different sleep phenotypes in response to early-life nutrition (S14 Table). For example, Furin convertase is an enzyme that processes the precursor of endothelin-1 (ET-1) 71 and is associated with insomnia 68. Furin convertase also cleaves the precursor form of brain-derived neurotrophic factor (BDNF) to generate mature BDNF 72. BDNF plays a role in homeostatic sleep regulation 73 and an essential role in brain development and synaptic plasticity 74.
Early-life nutrition restriction modifies the knockdown effects of two regulators of Insulin signaling in sleep behavior and brain morphology: Ptp61F 75 and Rbfox1 76. Ptp61F encodes a protein tyrosine phosphatase ortholog to the mammalian PTP1B, which dephosphorylates insulin receptor (IR), inactivating it 45. Since overexpression of insulin/IGF-like peptides in glia or PI3K/Akt activation in neuroblasts results in increased proliferation under protein restriction 77, we propose that knockdown of Ptp61F contributes to uncoupling neuroblast growth from systemic control. These effects may increase the number of adult neurons and α- and β-lobe width.
On the other hand, Rbfox1 encodes a conserved RNA-binding protein with nuclear isoforms that regulate tissue-specific alternative splicing 76, while cytoplasmic isoforms regulate mRNA translation 78, 79, suggesting that variation in these processes underlies sleep and neuronal morphology sensitivity to early-life nutrition. Rbfox1 targets alternative splicing of Tsc2 80, which antagonizes cell growth and cell proliferation induced by Insulin signaling in Drosophila 81. On the other hand, Insulin signaling inhibits Tsc1 and Tsc2 to promote axon growth and branching during metamorphosis 82. Accordingly, our results suggest that variation in Rbfox1 expression interacts with early-life nutrition to affect α-lobe length, which may be due to axon growth defects.
Cytoplasmic Rbfox1 regulates the expression of synaptic genes by binding to the 3’UTR of mRNAs that are targets of microRNAs, independent of its effect on splicing 79. Moreover, disruption of Rbfox1 in the central nervous system leads to neuronal hyperactivity, while the deletion of Rbfox2 results in cerebellum development defects in mice 83, 84. Thus, the mechanism by which variation in Rbfox1 expression affects sleep may also be linked to its role in synapse formation.
DGRP lines reared under different nutritional conditions showed changes in the metabolic phenotype and transcriptional profiles that rely on genotype by environment interactions 14–18, 85. These metabolic changes correlate with previously reported phenotypes on the DGRP 14. Therefore, polymorphisms in Ptp61F and Rbfox1 could significantly impact the response to environmental cues that affect nervous system development and function.
Our results suggest that gene expression variation in the MBs neurons and other groups of neurons can have different effects in sleep behavior and brain morphology depending on the nutritional environment during development. Our work supports the hypothesis that variants affecting local gene expression in the brain play an essential role in GENI in sleep behavior. It also may explain why few genetic variants associated with GENI in complex behaviors affect gene expression in the whole animal 16. Further evaluation of the effect of genetic variants in candidate gene expression 86 in specific developing and adult brain regions and neural lineages will be necessary to unveil the molecular and cellular mechanisms underlying GENI in sleep behavior.
Materials and Methods
Drosophila stocks and husbandry
We used 74 DGRP lines 13. The DGRP lines and GAL4 drivers were obtained from the Bloomington Drosophila stock center (http://flystocks.bio.indiana.edu/). UAS-RNAi transgenic flies were obtained from Vienna Drosophila RNAi Center (VDRC) (https://stockcenter.vdrc.at), and the Transgenic RNAi Project (TRiP) at Harvard Medical School (http://www.flyrnai.org). All fly lines used are listed in S17 Table. All flies were reared under standard culture conditions (25 °C, 60–70% humidity,12-hour light:dark cycle) and controlled density.
Drosophila culture media
The Normal Food (NF) diet for stock maintenance contains: 10% (w/v) Brewer’s yeast, 5% (w/v) sucrose, 1.2% (w/v) agar, 0.6% (v/v) propionic acid, 3% (v/v) nipagin 87. The Restricted Food (RF) medium contains 20% of Brewer’s yeast and sucrose of NF (2% (w/v) Brewer’s yeast, 1% (w/v) sucrose, 1.2% (w/v) agar, 0.6% (v/v) propionic acid, 3% (v/v) nipagin) 87. For each DGRP line, we used 35 females and 15 males to obtain comparable levels of offspring density. For the few DGRP lines that yield less offspring in restricted food than in normal food, we used 50 females and 20 males. Adult flies were discarded after three days of egg-laying on NF or RF. Newly eclosed adults were transferred to new NF vials for three days before any behavioral assay.
Drosophila morphometric body measurements
We evaluated two morphometric body traits in adult Oregon-R flies reared under normal or restricted food. We distributed approx. 100 eggs in bottles of NF or RF in two biological replicates. Flies were reared under standard culture conditions (25 °C, 60– 70% humidity,12-hour light:dark cycle). After eclosion, flies were placed under NF for 3- 5 days. Adult females and males were selected and fixed for 24 h in 70% ethanol. The flies were placed on Sylgard plates with 70% ethanol, legs and wings were removed avoiding damage. After immobilizing the sample, the ILUMINA software was used to take a picture of the head and thoraxes at 6X with a stereoscopic magnifier attached to an INFINITY Lumenal Photo Camera 1 88. All images were processed and assembled using Fiji software and Adobe Illustrator 2020. Depending on the number of flies available, between twenty and twenty-five flies per sex were measured. Interocular distance (IOD) was measured from eye edge to eye edge along the anterior edge of the posterior ocelli and parallel to the base of the head (S1A Fig) 33. Notum length (NL) was measured in the midline from the anterior edge of the thorax to the anterior edge of the scutellum (S1A Fig).
Sleep phenotypes
We evaluated sleep traits in 73 DGRP lines (S1 Table). We picked groups of 3 to 10 DGRP lines to grow them in bottles in the two diet conditions without any block design plan. After eclosion, adult flies (males and virgin females) were transferred to vials with NF for 3 days until the sleep behavior was assessed. All sleep measurements were performed under standard culture conditions (25 °C, 60–70% humidity,12-hour light:dark cycle). Sleep measurements were replicated three times for most lines. Eight flies of each sex and each diet (NF and RF) were measured in one Drosophila Activity Monitors (DAM2, Trikinetics, Waltham, MA) per replicate. To mitigate the effects of both social exposure and mating on sleep behavior, males and virgin females were collected from each line and retained at 30 flies per same-sex same-diet vial. Individual flies were loaded into DAM2 monitors and sleep and activity parameters were recorded for seven continuous days. To mitigate the effects of CO2 anesthesia or any other potential acclimation effects, the first two days of data recording were discarded. The DAM2 monitors use an infra-red beam to detect activity counts in individual flies as they move past it; five minutes without an activity count is defined as sleep 89, 90. Flies were visually examined after the sleep and activity recordings were completed; data from any flies that did not survive the recording period was discarded. All DGRP sleep behavior analysis was done with data collected from days 3 to 4 after flies were placed into the DAMs. PySolo 91 software was used to calculate nine sleep parameters: total, night and day sleep duration in minutes, night and day sleep bout number, night and day average sleep bout length; it also calculated sleep latency, the time in minutes to the first sleep bout after lights are turned off, and waking index, the average number of beam crossings within an active bout.
Immunohistochemistry
We used 40 DGRP lines based on lines previously tested for MB morphology (36 lines in common) 32, which were also used to evaluate sleep behavior (39 lines were contained in the 73 lines evaluated) (S17 Table). Adult brains from female flies reared on NF or RF under standard culture conditions (25 °C, 60–70% humidity,12-hour light:dark cycle) were dissected and processed for immunohistochemistry as described previously 32, 92. All flies were between 3 and 5 days old at the time of dissection.
Drosophila brains were fixed in phosphate-buffered saline (PBS)-4% formaldehyde for 25 min at room temperature, washed 3 times with PBS-0.3% Triton X (PBST), and blocked in PBST containing 5% Normal Donkey Serum for 30 min at room temperature. Brains were incubated overnight at 4°C with mouse monoclonal anti-Fasciclin 2 antibody (1D4) (1:10; Developmental Studies Hybridoma Bank, University of Iowa, IO, USA) to visualize mushroom body α and β lobes. After washing 3 times with PBST, brains were incubated with a 100-fold dilution of Rhodamine (TRITC) AffiniPure donkey anti-mouse IgG for 2 h at room temperature, followed by washing 3 times with PBST. Brain samples were mounted in Vectashield mounting medium (Vector Laboratories, Burlingame, CA). Immunostaining was documented with an Olympus Fluoview FV1000 confocal microscope. To avoid any effects of variation in Fas2 expression between DGRP lines, we adjusted fluorescence intensities so that unambiguous measurements could be made 32, 92.
Morphometric measurements
The length and width of the MB α and β lobes in adult female brains were evaluated using the straight-line measurement tool from Fiji/Image J software (version 2.1.0/1.53c) 93 and expressed as values relative to the distance between the α lobe heels as described previously 32, 92. This internal calibration controls for differences in brain size when assessing variation in morphometric parameters among genotypes. MBs were scored individually (i.e., per hemisphere). Values were obtained for 10-12 brains for all genotypes, thus allowing the analysis of 20-24 hemispheres. Images, diagrams, and figures were assembled using Adobe Photoshop 2020 and Illustrator 2020.
Quantitative genetic analyses of sleep in the DGRP
We partitioned the variance in each sleep parameter in the DGRP using mixed model analyses of variance (ANOVA): Y = µ + L + F + S + (L × F) + (L × S) + (F × S) + (L × F × S) + R(L) + F x R(L) + S x R(L) + F x S x R(L) + ε, where Y is the sleep parameter; µ is the overall mean; L and R are the random effects of line and replicate, respectively; F and S are the fixed effects of food (control, restricted) and sex (males, females), respectively; and ε is the error variance. In addition, we performed reduced analyses within: (i) each food condition using mixed model ANOVAs of form Y = µ + L + S + (L × S) + R(L) + S x R(L) + ε; and (ii) each sex condition using mixed model ANOVAs of form Y = µ + L + F + (L × F) + R(L) + F x R(L) + ε. All ANOVAs were performed using the PROC GLM function in SAS.
We calculated broad-sense heritability by sex across food as H2 = (σ2L + σ2LF)/(σ2L + σ2LF + σ2E), where σ2L is the variance component among lines, σ2LF is the line-by-food variance component, and σ2E is the residual variance. In addition, we calculated broad-sense heritability by food across sex as H2 = (σ2L + σ2LS)/(σ2L + σ2LS + σ2E), where σ2LS is the line-by-sex variance component. We calculated broad-sense heritability by food and sex as H2 = (σ2L)/(σ2L + σ2E). We calculated genetic correlations by sex across food rNR = (σ2L)/(σ2L + σ2LF), and by food across sex rMF = (σ2L)/(σ2L + σ2LS). We defined the interaction coefficient across food (i2) as a measurement of the genotype by early-life nutrition interaction (GENI) contribution to the genetic variation, i2 = 1- rNR.
Quantitative genetic analyses of morphometric measurements in the DGRP
We partitioned the variation in the length and width of the α and β MB lobes in the DGRP using mixed model ANOVA: Y = µ + L + F + (L × F) + ε, where Y is the morphometric parameter; µ is the overall mean; L is the random effect of line; F is the fixed effect of food; and ε is the error variance. We estimated broad sense heritability (H2) by food as H2 = (σ2L)/(σ2L + σ2E), where σ2L is the variance component among lines, and σ2E is the sum of all other sources of variation. In addition, we calculated broadsense heritability across food as H2 = (σ2L + σ2LF)/(σ2L + σ2LF + σ2E), where σ2LF is the line-by-food variance component. We calculated genetic correlations across food rNR = (σ2L)/(σ2L + σ2LF).
Genotype-phenotype associations
We performed GWA analyses using line means for all sleep parameters and morphometric measurements using the DGRP pipeline (http://dgrp2.gnets.ncsu.edu/). This pipeline accounts for the effects of Wolbachia infection status, major polymorphic inversions and polygenic relatedness 94 and implements single-variant tests of association for additive effects of variants with minor allele frequencies ≥ 0.05. We tested the effects of 1,901,174 DNA sequence variants on each trait. We focus our GWA analysis on the associations that represents the interaction of both diets, that is, the difference of phenotypic means between restricted and control food. All annotations that map within or nearby (<± 5,000 bp from gene body) are based on FlyBase release 5.57 (http://www.flybase.org).
Network analysis
We annotated candidate genes identified by the GWA analyses using FlyBase release 5.57 and mapped gene-gene networks through the genetic interaction database downloaded from FlyBase. We then constructed gene networks using STRING (version 11.0) 38 where candidate genes directly interact with each other. We used the following STRING settings: (i) Experiments and Databases as active interaction sources, (ii) high confidence (0.700) as the minimum required interaction score. For network visualization, we used the igraph R package to plot gene networks, where nodes correspond to genes and edges indicate the interaction between genes.
Gene Ontology analysis
We performed gene ontology (GO) enrichment analysis using PANTHER 17.0 (http://www.pantherdb.org/) 95 and STRING 11.0 (https://string-db.org/) 38. We used the
DIOPT–Drosophila RNAi Screening Center (DRSC) Integrative Ortholog Predictive Tool 9.0, with all available prediction tools and only retrieving the best match when there is more than one match per input gene or protein, to identify human orthologs 96.
Functional analyses
We performed tissue-specific RNAi-mediated knockdown of 17 candidate genes implicated by the GWA analyses using pan-neural elav-Gal4 and MBs specific OK107- Gal4 and 201Y-Gal4 drivers (Bloomington, IN, USA) (S17 Table). We crossed males from UAS–RNAi lines with virgin females of each of the Gal4 drivers and reared under standard culture conditions (25 °C, 60–70% humidity,12-hour light:dark cycle) in the two diet conditions. For each cross, we used 35 virgin females and 15 males to obtain comparable levels of offspring density. For the few RNAi lines that yield less offspring in restricted food than in normal food, we used 50 virgin females and 20 males. Adult flies were discarded after three days of egg-laying on NF or RF. After eclosion, adult flies were transferred to vials with NF for 3 days until the sleep behavior or MB morphology was assessed as described above, except that to get more robust results, the sleep data was collected from days 3 to 7 after flies were placed into the DAMs. Sleep measurements were replicated three times for each line. Eight flies of each sex and each diet (NF and RF) were measured in one monitor per replicate. We performed two- way ANOVA followed by Bonferroni’s post hoc test (*p≤ 0.05; **p≤0.01; ***p≤0.001; ****p<0.0001; ns non-significant) to compare the effect of control or candidate gene RNAi knockdown between normal and restricted food. We also performed Welch’s t-test or one-way ANOVA followed by Dunnett’s post hoc test (*p≤ 0.05; **p≤0.01; ***p≤0.001; ****p<0.0001; ns non-significant) to compare the effect of control versus candidate gene RNAi knockdown in each food condition.
Author Contributions
G.H.O., P.O. and R.A.V conceived and designed experiments with input from T.F.C.M. F.N.-V. and N.C. maintained fly stocks. G.H.O., F.N.-V., F.V.-M., and N.Z. performed sleep experiments. G.H.O., F.N.-V, F.V.-M., and N.C. performed immunohistochemistry experiments. G.H.O. and F.N.-V. performed morphometric measurements. PO performed adult size experiments. F.N.-V. performed RNAi validation for sleep and morphometric analyses with input from G.H.O. G.H.O. performed statistical analyses with input from R.A.V. G.H.O. and K.O. performed gene networks analyses. G.H.O. and C.M. performed bioinformatic analyses with input from R.A.V. Figures and tables were prepared by G.H.O. The manuscript was written by G.H.O. and P.O. with input from R.A.V. and T.F.C.M.
Data availability statement
The datasets generated for this study are available on request to the corresponding authors.
Competing Interests Statement
The authors declare no conflict of interest.
Acknowledgments
We thank Chad A. Highfill and Brandon M. Baker from T.F.C.M. lab (Clemson University) for insightful advice on statistical analyses, Cristian Yáñez from R.A.V. lab for bioinformatic advice, Jimena Sierralta (Universidad de Chile) for reagents and fly stocks, Carlos Oliva (P. Universidad Católica de Chile) for the immunohistochemistry protocol, and Kajan Ratnakumar for critical comments. This work was supported by ANILLO ACT-1401 to P.O. and R.A.V., ICM P09-015F BNI to P.O., UCH-VID INFRA 0440/2018 to R.A.V.
Footnotes
We demonstrate that our nutrition restriction protocol during development imposes a stressful nutritional environment affecting body size. We deepened the functional analysis by using different drivers and two different RNAis for a subset of candidate genes. Our data show that a particular candidate gene's knockdown in different neuronal populations may affect GENI or plasticity in sleep behavior in different ways. We revised our quantitative genetic and statistical analyses. We explained the design and methodology used in detail and provided all raw data. We propose a new ANOVA model that integrates the replicates, which leads us to correct the heritabilities and the estimates of the genetic correlations across environments. We perform GWAS on the traits that showed significant genotype by environment interaction based on quantitative genetics analyses. We also provided all Q-Q plot analyses and only performed GWAS in traits that showed enrichment of true positives above the -log10P-5 value threshold, leading to removing a group of sleep traits and all morphological traits GWAS. These corrections lead us to restructure the manuscript and generate a new protein-protein interaction network. We also tested new candidate genes based on the sleep network and others involved in neural development. Finally, we tested the knockdown effect of genes that show GENI in sleep behavior in mushroom bodies morphology. As in our previous functional analyses, we confirm that genes encoding regulators of Insulin signaling, specifically PTP61F and Rbfox1, affect the response of sleep behavior and MBs morphology to early-life nutrition restriction. In summary, the new version of our manuscript includes a new and accurate data analysis that allow us to conclude that GENI plays an essential role in sleep behavior and MBs morphology variation. It also supports the notion that gene expression variation in neurons and specific groups of neurons, including MBs neurons, underlies GENI in sleep behavior variation and brain morphology.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.
- 6.
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.↵
- 96.↵