Recent reports account for altered metabolism in adult offspring from pregnancy subjected to abnormal photoperiod, suggesting fetal programming of liver physiology. To generate a pipeline of subsequent mechanistic experiments addressing strong candidate genes, here we investigated the effects of constant gestational light on the fetal liver transcriptome. At 10 days of gestation, dams were randomized in two groups (n = 7 each): constant light (LL) and normal photoperiod (12 h light/12 h dark; LD). At 18 days of gestation, RNA was isolated from the fetal liver and subjected to DNA microarray (Affymetrix platform for 28,000 genes). Selected differential mRNAs were validated by quantitative PCR (qPCR), while integrated transcriptional changes were analyzed with Ingenuity Pathway Analysis and other bioinformatics tools. Comparison of LL relative to LD fetal liver led to the following findings. Significant differential expression was found for 3,431 transcripts (1,960 upregulated and 1,471 downregulated), with 393 of them displaying ≥ 1.5-fold change. We validated 27 selected transcripts by qPCR, which displayed fold-change values highly correlated with microarray (r2 = 0.91). Different markers of nonalcoholic fatty liver disease were either upregulated (e.g., Ndn and Pnpla3) or downregulated (e.g., Gnmt, Bhmt1/2, Sult1a1, Mpo, and Mat1a). Diverse pathways were altered, including hematopoiesis, coagulation cascade, complement system, and carbohydrate and lipid metabolism. The microRNAs 7a-1, 431, 146a, and 153 were upregulated, while the abundant hepatic miRNA 122 was downregulated. Constant gestational light induced extensive modification of the fetal liver transcriptome. A number of differentially expressed transcripts belong to fundamental functional pathways, potentially contributing to long-term liver disease.
- constant gestational light
- liver gene networks
- whole genome
- metabolic pathways
- fetal development
chronic liver disease afflicts millions of patients and is among the 10 leading causes of death in the United States (39). Recent studies have revealed deleterious effects of circadian misalignment on hepatic metabolism, which contribute to overt metabolic dysfunction. Moreover, it has been shown that circadian desynchrony also accelerates liver carcinogenesis in mice, further suggesting that tight and proper control of circadian clocks is essential for hepatic integrity (2, 14). In humans, shift work may have important public health implications, given that it affects one-fifth to one-fourth of the workforce (22). Shift work is interlocked with chronodisruption, which in turn is a significant disturbance of the temporal organization of endocrinology, physiology, metabolism, and behavior (13). In fact, people engaged in night-shift work suffer from higher incidence of obesity, diabetes, and metabolic syndrome (10, 36, 38). Combined, these studies suggest that disruption of circadian rhythms affects not only hepatic metabolism, but subsequently also triggers the progression to nonalcoholic steatohepatitis. In line with this, genetic variants of canonical clock genes have been identified as risk factors for the development of nonalcoholic fatty liver disease (NAFLD) (43).
In this context, the impact of gestational chronodisruption on fetal development and postnatal physiology and health is a novel and challenging research venue. We have recently reported that exposure of pregnant rats to constant light from midgestation induces intrauterine growth retardation and marked effects on both fetal adrenal physiology and cardiac transcriptome (15, 32). Moreover, in the adult offspring that had been gestated under constant light, we also found increased basal plasma glucose and glucose intolerance (49). These results are in line with those reported by Varcoe and colleagues (48), who exposed pregnant rats to chronic phase shifts throughout gestation and the first week after birth. The offspring displayed poor glucose tolerance and increased insulin secretion in response to a glucose challenge, accompanied by hyperinsulinemia and increased adiposity. Taken together, these emerging data suggest that gestational chronodisruption modifies gene expression and functional readouts in different fetal tissues/organs, which seem to translate into altered postnatal physiology and metabolism. This is in keeping with extensive literature linking prenatal adverse conditions (e.g., hypoxia and/or undernutrition) with long-term detrimental effects, including adult onset of hypertension, Type 2 diabetes, and metabolic dysfunction (17, 35).
Using a microarray-based approach, here we investigated the effects of constant gestational light on fetal liver transcriptomics and functional genomics. Our aim was to identify fetal hepatic molecular components and pathways targeted by constant gestational light, which may contribute to the metabolic dysfunction reported by others and us in the adult offspring.
MATERIALS AND METHODS
The protocols were approved by the Local Bioethics Review Committees from both Faculty of Medicine, Universidad de Chile (CBA#0234) and Universidad Austral de Chile (CB#20/10), while animal handling and care were performed following the National Institutes of Health Guide for Animal Experimentation Care recommendations.
Timed-pregnant female Sprague-Dawley rats were obtained after mating (the presence of spermatozoa in the smear of the vaginal contents was considered day 0 of pregnancy) from Bioterio Central, Facultad de Medicina, Universidad de Chile. The dams were maintained in a 12:12 h light/dark cycle (light period: fluorescent light, about 400 lux at cage level; lights on at 0700) under controlled temperature (18–20°C), and food and water were available ad libitum. In brief, from day 10 of pregnancy, rats were randomly divided in two groups (n = 7 per group): control (LD; 12:12 h light/dark cycle, lights on at 0700) and constant gestational light (LL; 24 h under lights-on conditions). No signals of stress or behavioral changes were observed in the pregnant rats exposed to constant light (LL), as we found no hair loss, stereotyped movements, or altered water/food intake. Furthermore, at 18 days of gestation [embryonic day 18 (E18)] LL dams presented similar plasma corticosterone levels (as area under curve in 24 h) relative to LD mothers, displaying a circadian rhythm with only a slight phase delay of peak plasma corticosterone in the 24 h, supporting that maternal stress was absent in this protocol [for details see Mendez et al. (32)].
All fetal livers from Sprague-Dawley rats used in the present study were obtained from larger cohorts that we have previously reported (15, 32, 49). For RNA isolation from the fetal liver, dams were weighed at E18 (term is ∼21 days). The dams were euthanized with an overdose of sodium thiopental (150 mg/kg) between 2000 and 2400 (i.e., 1–5 h after lights off), the pregnant uterus was exposed via a midline incision, and the anaesthetized pups were euthanized by spinal transection. All fetuses, males and females, were weighed and their livers were immediately dissected out under sterile conditions.
For RNA isolation E18 fetal livers were pooled from every litter (males and females pups), and therefore RNA samples do not represent individual fetuses. Thus, any given pool consisted of eight fetal livers coming from one litter, with n = 5 pools for the LD and LL conditions, respectively. It should be stressed that RNA samples are often pooled in a microarray experiment to reduce the effects of biological variation, with the rationale being that differences due to subject-to-subject variation will be minimized, making substantive features easier to find. In a seminal paper, Kendziorski and colleagues (20) showed that pooled microarray designs do not perform more poorly than nonpooled designs.
RNA was extracted using the “SV Total RNA Isolation System” according to the manufacturer's instructions (Promega, Madison, WI). RNA quality was assessed with the Agilent 2100 Bioanalyzer and the associated RNA 6000 Nano and Pico LabChip kits (Agilent Technologies, Palo Alto, CA). The RNA samples with RNA integrity number > 5.5 were used for further microarray analysis.
Sample processing and microarray hybridization were performed by an external dedicated Core Facility, the KFB - Center of Excellence for Fluorescent Bioanalytics (Regensburg, Germany; http://www.kfb-regensburg.de). All steps of microarray analysis were carried out exactly as previously described (31) using Affymetrix Rat Gene 1.1 ST GeneChip arrays (containing 700,000 probe sets representing 28,000 rat genes).
Microarray data normalization and analysis.
All array data were normalized using robust multiarray average (RMA) (19) with the Expression Console software (Affymetrix, Santa Clara, CA). RMA utilizes the probe set annotation provided by Affymetrix to identify genes directly from the CEL files. Genes that were significantly up- or downregulated were identified by Significance Analysis of Microarrays (47). To achieve high stringency, we set analysis parameters to 700 permutations, while Delta was set to 0.892 to result in false discovery rates (FDR) ≤ 10%. The FDR of a test is conventionally defined as the expected proportion of false positives among the declared significant results. Because of this directly useful interpretation, FDR is a more convenient scale to work on instead of the P value scale.
Functional analysis of microarray data.
We applied different analytical approaches for functional genomics: Ingenuity Pathways Analysis (IPA), DAVID online bioinformatics resource, and miRWalk Database, exactly as we recently described in full detail for the fetal rat heart (15).
IPA computes a score for each gene network according to the fit of that network to the user-defined set of “focus genes.” The score is derived from a P value and indicates the likelihood of the focus genes in a network being found together due to random chance. A score of 2 indicates that there is a 1 in 100 chance that the focus genes are together in a network due to random chance. Therefore, scores of 2 or higher have at least a 99% confidence of not being generated by random chance alone. In addition, IPA compares the direction of change for the differentially expressed genes with expectations based on the literature (curated in the Ingenuity Knowledge Base) to predict an integrated direction of change for each function, using the z-score algorithm. It is designed to reduce the chance that random data will generate significant predictions. Z-scores ≥ 2 indicate that the function is significantly increased, and z-scores ≤ −2 indicate that the function is significantly decreased. Hence, the z-score indicates how much each ontology is overrepresented (positive z-score) or underrepresented (negative z-score) in a gene list.
We also used DAVID (http://david.abcc.ncifcrf.gov), which, despite its massive utilization, remains relatively outdated regarding IPA. However, in our hands, there was an excellent correlation in the over- and underrepresentation of pathways/processes between DAVID and IPA.
Finally, we used the miRWalk database (available at http://www.umm.uni-heidelberg.de/apps/zmf/mirwalk/) (12) to perform bioinformatics analysis of differentially expressed microRNAs (miRNAs) and their validated target sequences among genes displaying changes in their expression level under constant gestational light conditions. To this end, only miRNAs displaying ≥ 5 hits and P < 0.05 were taken into account.
RT-qPCR validation of microarray data.
Reverse transcription coupled with quantitative PCR (RT-qPCR) was used to validate differential expression of 29 mRNAs selected from the microarray analysis. To this end, we choose hypoxanthine phosphoribosyltransferase 1 (Hprt1) as housekeeping gene. In the Affymetrix Rat Gene 1.1 ST GeneChip, Hprt1 displayed a standard deviation = 0.164, as calculated from the RMA normalized data for all 5 LD and 5 LL chips. Therefore, we chose Hprt1 mRNA as a robust housekeeping gene for the present microarray dataset against which to analyze the same RNA samples subjected to the microarray experiment, following an RT-qPCR protocol previously reported by us (15). In the fetal rat liver RNA samples, the greater stability of Hprt1 rendered cycle threshold (Ct) values more consistent for all samples (each one measured in triplicate). qPCR primer details are shown in Table 1. The efficiency for the qPCR primers used for this purpose ranged between 90 and 110%. We calculated relative amounts of all mRNAs by the comparative Ct method using the equation 2−ΔΔCt (28).
In a parallel cohort, E18 fetal rat livers were collected every 4 h for 24 h from fetuses obtained from pregnant dams maintained under control LD photoperiod (n = 5 per time point). The fetal livers were subsequently subjected to RNA isolation and RT-qPCR to evaluate the 24 h expression pattern of both core clock genes [Period 2 (Per2) and brain and muscle aryl hydrocarbon receptor nuclear translocator like protein 1 (Bmal1)] and clock-controlled genes [early growth response 1 (Egr1) and melatonin receptor 1A (Mtnr1a)] according to an RT-qPCR protocol previously described by us (32).
P values and Z-scores were calculated as appropriate by integrated algorithms of the DAVID, IPA, and miRWalk software systems. RT-qPCR data for microarray validation purposes were expressed as means ± SE, and significant differences were assessed by Student's t-test (P < 0.05), while the values of 2−ΔΔCT for Per2, Bmal1, Mtnr1a, and Egr1 were analyzed by ANOVA and Newman-Keuls, and results were considered significant when P < 0.05. Correlation between fold change values by microarray vs. RT-qPCR was performed with best-fit linear regression to determine the r2 coefficient. Statistical analyses were performed using GraphPad Prism version 5.01 for Windows (GraphPad Software, San Diego, CA).
Daily expression pattern of clock and clock-controlled genes under control LD photoperiod.
To check whether the effects of constant gestational light on the fetal rat liver transcriptome may relate to changes in gene expression with clock time in the fetal hepatocytes, we used RT-qPCR to measure daily expression patterns of both clock and clock-controlled genes. At E18, no differences among clock time points were detected in the expression of the core clock genes Per2 and Bmal1 or the clock-controlled genes Mtnr1a and Egr1 in the fetal rat liver under control LD photoperiod (Fig. 1) as extensively reported by others (26, 34, 42).
Validation of microarray data by means of RT-qPCR.
As mentioned, we used the same RNA samples from the microarray analysis to perform RT-qPCR validation assays. An FDR threshold of 10% predicts that about one out of 10 transcripts differentially expressed in microarray might not be validated by qPCR. This figure agrees well with the results obtained for the 29 mRNAs analyzed so far, with 27 of them displaying differential expression in both microarray and RT-qPCR. Indeed, as shown in Fig. 2, there was a highly significant correlation between microarray and RT-qPCR (r2 = 0.9126, P < 0.0001). Interestingly, the transcript encoding for 18S ribosomal RNA was also validated by qPCR, but displaying a fold change markedly higher than the one found by microarray (20.85 and 1.41, respectively; data not shown). Therefore, these combined qPCR results confirm the microarray technically.
Quantitative and topological impact of constant gestational light on the active fetal liver transcriptome.
The Affymetrix Rat Gene 1.1 ST GeneChip contains over 28,000 genes covering nearly the complete rat transcriptome. In the liver of rat fetuses subjected to constant gestational light, 3,431 genes displayed changes in their expression level (i.e., meeting the significance criteria of FDR ≤ 10%), with 1,960 upregulated and 1,471 downregulated genes. Therefore, differentially expressed transcripts accounted for a striking 26.7% of the 12,853 unique genes transcriptionally active in the rat liver at E18 (Fig. 3A, see discussion for details).
No differences were found in further analysis to assess whether the altered genes were distributed asymmetrically on the chromosomes. DAVID analysis showed significant coverage on four chromosomes: chr 5 (218 genes, P = 8.6 × 10−4), chr 7 (236 genes, P = 9.0 × 10−4), chr 14 (131 genes, P = 9.3 × 10−4), and chr 10 (256 genes, P = 3.9 × 10−3). This analysis also showed that the cytoband 5q36 was the chromosomal region with higher amount of differentially expressed genes (108 genes, P = 5.3 × 10−5). The distribution of genes exhibiting changes in their expression level under LL conditions in the fetal rat liver, broken down by chromosome, is shown in Fig. 3B.
Identification of differentially expressed genes under constant gestational light.
Comparison of global transcription in the fetal liver from mothers kept under LL with those gestated in LD conditions revealed that most of the differential genes were overexpressed (1,960 out of 3,431 genes, i.e., 57.1%), while the remaining differential genes were downregulated (1,471 out of 3,431 genes, i.e., 42.9%). The top 40 upregulated genes are shown in Table 2, with the top seven transcripts displaying a consistent change of at least twofold. These upregulated genes have different biological functions; however, an important fraction has nuclear function and localization, such as Terc (4.1-fold change), Hmga1, Ndn, Cited1, Pole3, E2f5, Ssbp4, Med21, and Commd6. Among all overexpressed transcripts, there were several established markers of hepatic disease, particularly NAFLD, such as Ndn, Pparα, and Pnpla3. The miRNAs 7a-1 and 136 were identified among the top 40 upregulated transcripts, while the miRNAs 146a and 153 were also upregulated. On the other hand, the top 40 downregulated genes are shown in Table 3, with an overall range of 2.2- to 6.1-fold change. Therefore, constant gestational light induced a transcriptome response where the top downregulated genes displayed higher fold change relative to the top upregulated genes. In the downregulated group, we found several genes encoding for enzymes, for instance Adh1, Car12, Gnmt, Pdk4, and Hsd17b13. In addition, most of the proteins encoded by downregulated genes display cytoplasmic localization. Among all downregulated transcripts, there were several established markers of hepatic disease, including the NAFLD marker glycine N-methyltransferase (Gnmt). Notably, the extremely abundant hepatic miRNA 122 was downregulated by 2.3-fold.
Pathways and gene ontology processes modulated by constant gestational light.
The top three associated gene networks identified by IPA were energy production, lipid metabolism, small molecule biochemistry (score: 93; for a definition of score, see materials and methods); cell cycle, DNA replication/recombination/repair, cellular growth and proliferation (score: 83); and lipid metabolism, molecular transport, and small molecule biochemistry (score: 69). Considering the high IPA scores for these gene networks, together with the strong correlation of differential gene expression between microarray and qPCR, these findings clearly support notion that constant gestational light has a pronounced and pervasive impact on fetal hepatic physiology. In keeping with this, there was a high number of differentially expressed genes included in several essential biological pathways. Thus, in the Top 1 Interactome shown in Fig. 4, the expression level of almost every molecular term was up- or downregulated. In this particular interactome, several functionally relevant nodes were identified, namely, Eif4e, Hoxd10, Trim24, Sbds, Nkx2-1, and Wdr77.
On the other hand, the top seven overrepresented canonical pathways (i.e., displaying the strongest P values) are shown in Fig. 5A. All these pathways are important for liver function, particularly in the context of fetal physiology: acute phase response signaling (64 genes, P = 1.03 × 10−9), eukaryotic initiation factor 2 (EIF2) signaling (50 genes, P = 6.56 × 10−4), NF-E2-related factor 2 (Nrf2)-mediated oxidative stress response (54 genes, P = 7.86 × 10−5), and coagulation system (17 genes, P = 8.24 × 10−5), among others. Reckoning that a number of established and potential markers of disease were identified among all genes differentially expressed under constant gestational light, we next applied a selective analysis. To this end, the gene set was restricted to those experimentally demonstrated to be predictive biological markers for disease or malfunction, as to determine the state, activation, or inhibition of processes by the z-score (for a definition of z-score, see the footnote to Table 5). As shown in Fig. 5B, several key liver functions were identified as inactivated: lipid metabolism, carbohydrate metabolism (103 genes, P = 6.43 × 10−3), and immunological disease (178 genes, P = 2.34 × 10−4). In contrast, this analysis indicated activation for hematopoiesis (55 genes, P = 2.91 × 10−3) and infectious disease (173 genes, P = 1.10 × 10−3), among others. It is worth noting that, combined, the overrepresented canonical pathways (Fig. 5A) and disease/malfunction processes (Fig. 5B) include seven out of 17 (i.e., 41.2%) terms directly linked to immunity and defense mechanisms.
DAVID analysis, restricted to differentially expressed genes exhibiting a fold change ≥ 1.5, pointed out several gene ontology (GO) categories displaying significant FDR values. Thus, the principal GO biological process outlined by this analysis were oxidation/reduction (38 genes, FDR = 3.05−6), response to wounding (31 genes, FDR = 3.41 × 10−6), and defense response (29 genes, FDR = 1.48 × 10−5). The main molecular functions overrepresented were carboxylic acid binding (16 genes, FDR = 1.8 × 10−3), electron carrier activity (17 genes, FDR = 1.65 × 10−2), tetrapyrrole binding (14 genes, FDR = 2.1 × 10−2), and iron ion binding (19 genes, FDR = 3.44 × 10−2). We also used KEGG, which is another DAVID database resource. The top FDR values for KEGG pathways and biological terms were complement and coagulation cascades (12 genes, FDR = 5.58 × 10−4). Interestingly, other enriched KEGG functions featuring high FDR values included several pathways of intermediary metabolism (i.e., starch and sucrose metabolism, citrate cycle, glutathione metabolism, and pentose phosphate pathway).
Effects of constant gestational light on gene expression regulatory mechanisms.
An important mechanism by which gene expression can be regulated is through small noncoding RNAs or miRNAs, which repress translation by promoting degradation of mRNA by binding to specific sequences in the untranslated regions of the mRNA (12, 18). The microarray analysis filtered 21 miRNAs that were differentially expressed under LL in the liver (10 upregulated and 11 downregulated). Among them, we found miRNA 7a-1 and miRNA 136 (both upregulated by 1.7-fold, see Table 1), as well as miRNA 122 (downregulated by 2.3-fold, see Table 3), while miRNA 127 and miRNA 146 are included in the Top Network 1 (see Fig. 4). Next, we used miRWalk Database to link upregulated miRNAs with their validated targets among the whole set of 1,471 genes downregulated by constant gestational light. As shown in Table 4, our analysis identified seven validated targets for miRNA 7a-1 and 5 for miRNAs 431, 146a, and 153 (all P < 0.05). Conversely, an miRWalk database search for validated targets of the downregulated miRNA 122 was done on the full list of 1,960 upregulated genes. This analysis identified nine validated targets (all P < 0.05, Table 4).
Finally, we used IPA to predict if any master transcription factor may be modulated by constant light, based on the relative direction of change (up- or downregulation) of its known downstream genes. Therefore, this functional genomics approach accounts for putative posttranslational modifications that cannot be detected by microarray or qPCR. From this bioinformatics analysis, activation state was predicted for Nrip1, Foxm1, Srebf1, and Stat5, while inactivation state was predicted for Hnf1a, Rb1, Hnf4a, and Smarcb1 (Table 5).
The effects of altered photoperiod during pregnancy on fetal development and postnatal physiology are just beginning to be explored (15, 32, 44, 48, 49). Here we show that, under constant gestational light conditions, as many as 3,431 transcripts displayed changes in their expression level in the fetal liver. However, not all genes are expressed in the liver at any given time; indeed, Chapple and colleagues subjected the LD rat liver to RNA deep sequencing at E18 and found transcripts encoding for 12,853 annotated genes (Chapple RH and Taylor JF, personal communication; see Ref. 6). Therefore, the large set of differential transcripts identified in the present report collectively represents a highly significant fraction of all genes transcriptionally active in the LL fetal rat liver (26.7%). The functional consequences of such a broad modification of the hepatic transcriptome remain to be elucidated; however, a substantial fraction of genes differentially expressed under LL conditions is integral to important liver functions during both fetal and adult life (including many established markers of fatty liver, metabolic syndrome, and immunological disorders).
The multiple functions carried out by the liver are different during fetal development, lactation, and adulthood. In particular, the rat liver is the primary hematopoietic organ during the second half of gestation, playing a key role in development of the innate and adaptive immune system (23, 50). Based on global transcriptome changes, both hematopoiesis and infectious disease were predicted to be activated under constant gestational light. An important finding was that several mRNAs encoding for factors of the complement system were downregulated in the LL fetal rat liver (which was confirmed by qPCR for C3, C4a, Masp2, C8b, and C9). A further effect of constant gestational light on the innate immune system was repression of acute-phase response signaling, displaying the strongest P value of all overrepresented pathways, with 91.3% of the whole pathway being differentially expressed (156 genes). Given that acute-phase response proteins bind hormones such as cortisol and retinol, their downregulation might translate into increased hormone bioavailability. In addition, we found that 97% of the coagulation system components were altered in the fetal liver subjected to constant gestational light. A similar result was obtained for the prothrombin activation system, including most components of both intrinsic and extrinsic pathways. Ongoing experiments are aimed at elucidating the potential consequences of alteration of the complement, acute-phase response, and coagulation systems for the LL adult offspring.
An interesting finding was upregulation of Keap1 by constant light. KEAP1 is a key inhibitor of the canonical pathway Nrf2-mediated oxidative stress response. Upon exposure to oxidative stress, NRF2 dissociates from KEAP1, translocates to the nucleus, heterodimerizes with different transcriptional regulators, and binds to the promoter region of an array of antioxidant genes (52). We found that in this oxidative stress response pathway, 10 mRNAs driven by NRF2 were downregulated (including the major antioxidant enzymes Sod2, Nqo1, Cat, and Gpx2), while four transcripts were upregulated.
Carbohydrate and lipid metabolism are essential hepatic functions in the adult liver, which were predicted to be inactivated in the LL fetal liver. For instance, there was a significant downregulation of enzymes participating in both glycogen synthesis (Gbe1, Gys2, and Ugp2) and glycogen degradation (Agl, Pgm1, and Pygb). Regarding lipid metabolism, a fraction of differentially expressed genes is directly associated with hepatic steatosis and NAFLD, which refers to the nonphysiological accumulation of fat in the liver, with a concomitant effect on increased incidence of obesity and metabolic syndrome (29). Indeed, a rather large set of differentially expressed genes under constant gestational light involved established NAFLD markers (upregulated: Pnpla3, Ndn, Smad6, Smad7, Igf2bp2, and Adipor2; downregulated: Sult1a1, Mttp, Mpo, ApoE, Ghr, Lpin1, and Pparα). Downregulated transcripts associated with NAFLD also comprised Gnmt, Bhmt1, Bhmt2, and Mat1a, known to be essential to sustain 1-carbon metabolism. This latter effect may in turn decrease global DNA methylation and, thereby, dampen physiological gene silencing (30).
It has been estimated that more than one-third of all genes are regulated by miRNAs (18). Our analysis of four miRNAs upregulated under LL conditions identified several validated targets of interest, which in turn were downregulated as assessed by microarray. Interestingly, two liver processes might be affected by degradation of the validated targets of miRNA 7a-1, glycogen metabolism, with the relevant targets being Gys2 and Ppp1r3b, and steroid metabolism, with Cyp4f39 and Sult1a1. Meanwhile, this analysis also revealed upregulation of miRNA 153 and concurrent downregulation of its target genes Nqo1 and Dcn, which are related with pathophysiological mechanisms involved in Type 1 diabetes (24, 51). Conversely, the mammalian liver-specific miRNA 122 was downregulated by constant gestational light, while its nine validated target genes were upregulated. miRNA 122 is the most abundant miRNA in adult mouse liver, accounting for about 70% of the total miRNA population in hepatocytes (5). This miRNA is known to regulate metabolic pathways in the liver, including cholesterol biosynthesis (25). Moreover, miRNA 122 has been recently established as a marker for NAFLD (4, 27).
The substantial effects of constant gestational light on hepatic gene transcription could be also explained by changes in the activation state of master transcription regulators. Indeed, several master transcription factors were pointed out by the IPA algorithm for prediction of functional state (inhibited: HNF4A, XBP1 and HNF1A; activated: SREBF1). For instance, Pparα, δ, and γ are established downstream targets of HNF4A, which were differentially expressed under constant gestational light. In turn, coactivation of HNF4A by the nuclear factor CITED2 is essential for liver development (37). HNF4A also plays an important role in insulin signaling and body weight regulation (11), especially by regulating apolipoprotein promoters (3). In addition, the 2.95-fold increase of Hmga1 expression under LL conditions could have major functional consequences. The architectural HMG transcription factors are the most abundant nuclear proteins after histones. In particular, HMGA1 is a chromatin-binding protein that, upon binding to AT-rich sequences of DNA, facilitates the assembly and stability of the “enhanceosome,” which in turn drives transcription (7). In recent years, the Hmga1 gene has been associated with insulin resistance and Type 2 diabetes (7). Interestingly, we recently reported that Hmga1 mRNA expression was increased by 2.1-fold in the LL fetal rat heart (15), raising the possibility that upregulation of Hmga1 may be part of the response to constant gestational light in different fetal tissues.
Additional epigenetic mechanisms may help to explain the extensive modification of the fetal rat liver transcriptome by constant gestational light. In fact, the transcription level of a large number of genes encoding for epigenetic modifiers was altered in the LL fetal rat liver, including gene silencing (6 genes), chromatin modification (17 genes), DNA methylation (10 genes), and Swi/Snf chromatin remodeling complex (10 genes). Collectively, these genes amount to 20% of the molecules known to be associated with epigenetic features. Hence, at least part of the profound changes imposed by constant gestational light in the fetal liver might carry on into the offspring's adulthood through epigenetic mechanisms, in agreement with reported long-term effects of gestational chronodisruption, such as increased adiposity and poor insulin and glucose handling (48, 49).
18S rRNA is widely used as housekeeping gene because of its stability. Therefore, a striking finding was that its expression was increased by 20.85-fold in the LL fetal liver. Moreover, many components of the translational machinery were targeted by constant gestational light, including the Eif4e node in the functional network “energy production, lipid metabolism, and small molecule biochemistry,” as well as the EIF2 signaling pathway. In particular, the translation initiation factors 2A and 2B (Eif2a and Eif2b2), together with six components of the ribosomal 40S subunit (ribosomal proteins AS, 2S, 7S, 14S, 21S, and 27S), were upregulated in the LL fetal liver. Combined, these results suggest increased translation initiation.
Although the role played by oscillatory gene expression and circadian clocks during pregnancy remains unclear, gestational chronodisruption may be regarded as an adverse prenatal condition given that increased risk of miscarriage, preterm delivery, and low birth weight have been consistently reported in shift worker women (1, 8, 9, 21, 53). In fact, despite the fetal suprachiasmatic nucleus not yet being the master circadian clock (at least in the conventional way it works during adulthood), several fetal organs seem to operate as peripheral oscillators in the rat (see Refs. 40, 45, 46, 49), probably driven by maternal circadian signals, such as melatonin, reaching the fetus (40, 41, 46). In this context, the present data suggest no difference among clock time points in gene expression in the fetal rat liver under control LD photoperiod at E18. Although this in agreement with previous reports for the murine fetal liver (26, 34, 42), it must be kept in mind that further experiments are required to actually demonstrate the presence or absence of circadian rhythms of core clock and clock-controlled genes in the fetal liver. On the other hand, it is conceivable that the present findings of broad changes in global gene expression secondary to manipulations of gestational lighting may be a consequence of changes in crucial maternal circadian signals crossing the placenta to synchronize fetal peripheral clocks. For instance, using the same model of maternal exposure to constant light, we have previously shown that maternal melatonin plays a key role in the entrainment of the circadian clocks residing in the fetal rat adrenal and hippocampus (32, 46, 49).
At the same token, a plausible way for maternal plasma melatonin to drive the fetal circadian system is through regulation of the fetal adrenal corticosterone circadian rhythm, which may in turn act as a further downstream rhythmic signal for other fetal tissues (46). We have reported that constant light during the second half of gestation (LL condition) is associated with maternal plasma melatonin suppression, thereby precluding a plasma melatonin rhythm in the fetus (32). This abnormal endocrine prenatal environment translated into intrauterine growth retardation and changes in fetal adrenal, heart, and hippocampus in vivo gene expression, as well as low content and no rhythm of fetal adrenal corticosterone (15, 32, 49). All these changes were reversed when the mother received a timed dose of melatonin during the subjective night, supporting a role of melatonin as a zeitgeber for the fetal organs, before the suprachiasmatic nucleus acts as the master clock (for further details, see Refs. 15, 32, 49). On the other hand, despite the clear-cut link between circulating corticosterone and liver function in adults, there is scant evidence for this link in prenatal development. Actually, [H3]-dexamethasone-binding sites become detectable in the fetal rat liver just at E18 (16). In vitro evidence shows that hydrocortisone (corticosterone metabolite) increases the activity of glycogen synthase in hepatocytes derived from E18 fetal rat (33), while we have reported that in vivo the rat fetus at E18 possesses a strong circadian rhythm of corticosterone (46). Together, these findings raise the possibility that suppression of the fetal corticosterone rhythm might induce changes in the fetal liver metabolism, modifying gene expression as reported here. However, from the present results it is difficult to establish whether low fetal adrenal corticosterone actually contributes to the extensive transcriptomic changes found in the E18 fetal liver from LL relative to LD conditions.
Our work reveals that constant gestational light brought about extensive modification of the transcriptome landscape in the fetal liver. Interestingly, a number of differentially expressed transcripts are included in fundamental functional pathways, with many of them being established markers of liver disease. In particular, the present findings might be associated with increased risk of adult disease, such as fatty liver, metabolic syndrome, and immunological disorders, which warrants further investigation about the effects of light at night during pregnancy.
This work was funded by FONDECYT Grant 1110220 (to H. G. Richter) and CONICYT Grant ACT1116 (to C. Torres-Farfan and H. G. Richter), Chile.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: C.S., C.T.-F., and H.G.R. conception and design of research; C.S., H.A.G., N.M., and P.A.V. performed experiments; C.S., C.T.-F., H.A.G., and H.G.R. analyzed data; C.S., C.T.-F., H.A.G., and H.G.R. interpreted results of experiments; C.S. prepared figures; C.S. drafted manuscript; C.T.-F. and H.G.R. approved final version of manuscript; H.G.R. edited and revised manuscript.
- Copyright © 2015 the American Physiological Society