Evidence of increased hypoxia signaling in fetal liver from maternal nutrient restriction in mice

Intrauterine growth restriction (IUGR) is a pregnancy condition where fetal growth is reduced, and offspring from IUGR pregnancies are at increased risk for type II diabetes as adults. The liver is susceptible to fetal undernutrition experienced by IUGR infants and animal models of growth restriction. This study aimed to examine hepatic expression changes in a maternal nutrient restriction (MNR) mouse model of IUGR to understand fetal adaptations that influence adult metabolism. Liver samples of male offspring from MNR (70% of ad libitum starting at E6.5) or control pregnancies were obtained at E18.5 and differential expression was assessed by RNAseq and western blots. Forty-nine differentially expressed (FDR < 0.1) transcripts were enriched in hypoxia-inducible pathways including Fkbp5 (1.6-fold change), Ccng2 (1.5-fold change), Pfkfb3 (1.5-fold change), Kdm3a (1.2-fold change), Btg2 (1.6-fold change), Vhl (1.3-fold change), and Hif-3a (1.3-fold change) (FDR < 0.1). Fkbp5, Pfkfb3, Kdm3a, and Hif-3a were confirmed by qPCR, but only HIF-2a (2.2-fold change, p = 0.002) and HIF-3a (1.3 p = 0.03) protein were significantly increased. Although a moderate impact, these data support evidence of fetal adaptation to reduced nutrients by increased hypoxia signaling in the liver.


INTRODUCTION
Intrauterine growth restriction (IUGR) is a pregnancy condition where fetal growth is suboptimal, resulting in an infant born with a birth weight <10th percentile. 1 Individuals born from pregnancies complicated by IUGR are at increased risk for peri and postnatal complications and, as adults, are at increased risk for metabolic disorders. 2 Adverse intrauterine conditions may lead to adaptations to enhance fetal survival but contribute to aberrant metabolism as adults. 2 Hypoxia-inducible factor (HIF) pathway is a cellular response to enhance survival in adverse conditions, such as hypoxia and nutritional stress. Under normal physiological conditions, the a subunit (HIF-1a, HIF-2a, HIF-3a) is posttranslationally hydroxylated and/or phosphorylated in the cytosol by regulators such as von Hippel-Lindau (VHL)/U3 ligase complex, factor inhibitor of hypoxia (FIH), and glycogen synthase kinase 3β (GSK-3β). [3][4][5] These posttranslational modifications promote proteasomal degradation. [3][4][5] During cellular stress, reduced hydroxylation and/or phosphorylation at stabilization sites inhibit recognition by the proteasome. Stabilization in the cytosol results in increased dimerization with the HIF-1β subunit. HIF dimers then translocate to the nucleus and transcriptionally regulate genes containing hypoxia-responsive elements (HREs). 5 HRE-containing genes are involved in metabolism, cell cycle regulation, and angiogenesis. 4,6 Adaptations in HIF signaling in the placentas from growthrestricted pregnancies or kidney and liver of maternal nutrient restriction (MNR) IUGR animal models have been documented. 7,8 However, others fail to find differences in HIF signaling in these tissues. 9,10 Previously, we have shown that MNR results in male offspring with reduced fetal and adult liver size and increased hepatic insulin sensitivity in adulthood. 11 Here we aimed to investigate whether HIF signaling has a role in fetal liver adaptations. Our results suggest that MNR in mice induces hypoxia-inducible signaling in the fetal liver. Such adaptations to growth and metabolism in the developing liver may contribute to long-term metabolic impacts of IUGR.

Animals
All animal procedures were approved by the Animal Use Subcommittee of the University Council on Animal Care at the University of Western Ontario and were described by Radford and Han, 2018. 11 Briefly, mice were housed in 12-h light and dark cycles. Virgin 8-week-old CD-1 females were bred, and the presence of a vaginal plug indicated E0.5. At E6.5, pregnant females received ad libitum standard chow (control, N = 5) or MNR (N = 5) (70% total calories) (#F0173, Bio-Serv, Flemington, NJ). At E18.5, dams were euthanized with CO 2 narcosis and pups with cervical dislocation. Right liver lobes were removed from each pup, snap frozen in liquid nitrogen, and stored at −80°C. Male offspring from MNR have been reported to be more sensitive than females to the long-term changes in glucose metabolism, 12 and only metabolic changes in males have been investigated previously in this model. 11 Male fetuses were identified with SRY PCR (forward primer-TGGACTGGT-GACAARGCTA, reverse primer-TGGAAGTACAGGTGTGCACTCT) and used for further analysis.
RNA isolations RNA for sequencing was isolated from ten control and ten MNR livers (two mice per litter). Additional RNA isolations from eight control and eight MNR littermates (one or two mice per litter) were used as a validation cohort for real-time PCR. Briefly, the frozen right liver lobe (30- 50 mg for sequencing and 10-15 mg  for real-time PCR)  Protein isolations In all, 10-15 mg of tissue was homogenized with 1 mL of ice-cold 1× Cell Lysis buffer (9803 S) from Cell Signaling (Danvers, MA) with protease inhibitor cocktail 1, 2, and 3 (P1860, P5726, and P0044, Sigma-Aldrich, St. Louis, MO) for 40 s with a Brinkman homogenizer Polytron® 3000. Lysates were sonicated (F550 Sonic Dismemberator, Fisher Scientific, Markham, ON) and then shaken at 4°C for 2 h. Cellular debris was removed from lysates by spinning at 12,000 × g for 10 min and the supernatant was collected for analysis. Protein concentrations were calculated with the Bio-Rad Protein Assay (Cat #500-0006, Bio-Rad Laboratories Inc., Des Plaines, IL) according to the manufacturer's protocol.
Western blotting Fifty μg of protein for HIF-2a blots and 25 μg for all other blots were loaded and run on an 8% polyacrylamide gel. Protein was transferred to a polyvinylidene difluoride membrane with the Transblot Turbo TM (Bio-Rad Laboratories Inc., Des Plaines, IL). Membranes were dipped in methanol and then blocked for an hour at room temperature in 5% milk in TBST (TBS+0.1% tween). After rinsing 3 times (5 min each) in TBST, primary antibodies were incubated in 5% bovine serum albumin in TBST overnight (Table 1). Blots were rinsed 3 times (5 min each) and incubated with secondary antibodies in 5% milk in TBST for 1 h at room temperature (Table 1). Blots were imaged with the Clarity TM Western ECL Substrate (1705061, Bio-Rad Laboratories Inc., Des Plaines, IL), and band intensities were quantified using the Imagelab software® version 5.1 BETA (Bio-Rad Laboratories Inc., Des Plaines, IL). As a loading control, all blots were rinsed 3 times with TBST (5 min each), stripped with 0.5 M sodium hydroxide for 15 min, and re-probed for actin according to the protocol above. Each target band was normalized to the respective actin and the control group was used to calculate relative fold changes. KDM3a was not assessed via western blot because we were not able to find an effective antibody for the mouse fetal liver samples. All blots were run in triplicate.
Statistical analysis Samples that contributed more variation than expected to each group were removed as outliers as described by Gierliński et al. with the mean+2 times the interquartile range as the threshold. 16 Differential expression was detected with a false discovery rate (FDR) <0.1 with consistency in ≥2 tools (EdgeR, 16 DESeq2, 17 19 ; and KEGG and NCI pathway enrichments were run on enrichR. 20 For quantitative PCR (qPCR) and western blots, unpaired t tests were used to compare delta CT values or band intensities on Graphpad Prism (version 5.2).

RESULTS
RNA sequencing data were assessed for sample outliers and gene distributions among remaining samples. 16 One control and one MNR sample contributed more variance to their respective maternal nutrient groups than other samples within the group and were removed from further analysis ( Supplementary Fig. S1).
With remaining samples, two distributions of genes were formed along PC1 ( Supplementary Fig. S2). The second distribution contained genes enriched in GO pathways in neural and epithelial cell development (Supplementary Table S1), since hepatocyte and hematopoietic populations are of primary interest in this study this second distribution of genes was removed from the analysis.
Overall gene expression was explored with PCA plots and cluster analysis. The optimal number of clusters with the gap statistic of the centered-log ratio and regularized-log ratio transformed data was 1. However, k-means clusters with k = 2 resulted in clusters separating based on counts per sample (Supplementary Fig. S1a and Supplementary Table S2). The PCA plot of the top 500 variable genes from the centered-log ratio (Fig. 1a) did not result in separation of samples based on maternal nutrition. Although the PCA of the top 500 variable genes of the regularized-log ratio transformed data indicate moderate separation of maternal nutrition along PC1, the first component only explains 8% of the total variance (Fig. 1b). These plots suggest that overall gene expression was similar between the control and MNR fetal livers.
Despite similar overall expression, 49 protein-coding genes were differentially expressed in MNR fetal livers relative to controls using a FDR cut-off of <0.1 and consistency of two or more tools (Supplementary Table S3). GO enrichment indicated negative regulation of transcription from RNA polymerase II promoter in response to hypoxia as the top pathway when ranked according to fold enrichment (Cited2 and Vhl) (Fig. 2a). Hypoxia signaling was also the top pathway enrichment according to combined scores with KEGG (Fig. 2b) and NCI (Fig. 2c). Additional genes included in these enrichments are Pfkfb3 and Hif-3a. Investigation into gene functions of differentially expressed genes indicate further involvement in cellular response to hypoxia (Fig. 3).
Genes involved in hypoxic regulation of metabolism, cell cycle regulation, and chronic hypoxia were selected to be validated by qPCR in a separate validation cohort (Fig. 4). Pfkfb3 transcript was increased 1.8-fold (p = 0.002) in MNR relative to controls, similar to the fold change in the RNAseq data of 1.5 (FDR = 1.2 × 10-6). Fkbp5 and Kdm3a transcripts were also confirmed to be increased by 1.3-and 1.5-fold (p = 0.03 and 0.02) and similar to sequencing fold changes of 1.6 and 1.2, respectively (FDR = 0.006 and 0.04). Lastly, Hif-3a was significantly higher by 1.4-fold in the MNR livers in the validation cohort (p = 0.01), similar to the 1.3-fold increased initial MNR cohort (FDR = 0.009). Although not significant, Ccng2, which was 1.5-fold higher in the RNAseq data (FDR = 0.0001), was relatively higher in the MNR fetal livers by 1.4-fold (p = 0.05). Btg2 and Vhl were not significantly different in the validation cohort (p = 0.1 and 0.7). Both Btg2 and Ccng2 had high variation among MNR offspring and a small fold change relative to controls (<2). Larger sample sizes could be required to confirm changes in Btg2 and Ccng2 transcripts. Hif-1a and Hif-2a transcripts were not significantly different in the initial MNR cohort, but Hif-1a was 1.3fold decreased in the validation MNR group relative to controls (p = 0.01).
Protein levels for genes differentially expressed in the validation cohort were determined by western blots (Fig. 5). Since the HIF alpha subunits can be regulated posttranscriptionally in response to cellular stress, the protein levels were also measured. HIF-1a was not detected in the fetal liver ( Supplementary Fig. 3) but HIF-2a was increased 2.2-fold in MNR fetuses (p = 0.002). HIF-3a was also increased by 1.3-fold (p = 0.03) in MNR relative to controls (Fig. 5a, b). FKBP5 (1.6-fold change, p = 0.09) and PFKFB3 (1.3-fold change, p = 0.5) protein levels were not significantly different between the maternal nutrient groups (Fig. 5c, d).

DISCUSSION
Fetal growth restriction leads to increased risk for perinatal complication and changes to adult glucose metabolism in adults. 1,2 Currently, no treatments are available to address these perinatal and long-term concerns in IUGR infants. Identification of pathways altered in response to restricted growth in utero in metabolically important tissues, such as the liver, provide potential targets for therapeutic intervention. HIF signaling is a pathway that has gained recent attention in growth-restricted offspring; however, evidence of HIF induction has been conflicting and tissue specific. [8][9][10] The approach taken in this study was unbiased, investigating relative transcript levels of all protein-coding genes, and supports the concept that HIF signaling is induced in the MNR fetal liver of male offspring from moderate MNR. This underlines the importance of further studies to determine how hepatic HIF signaling influences fetal survival and long-term glucose metabolism in response to fetal undernutrition.
HIF-2a protein was increased in MNR offspring, but HIF-1a was not detected. Normal development in cells, such as endothelial cells, involve a switch from HIF-1a to HIF-2a as the primary HRE transcriptional regulator, 21 similar to the transition observed with chronic hypoxia in cancer cells. 22,23 In these tissues, this switch is thought to occur in part because oxygen tension increases as organ perfusion during development becomes more efficient. FIH and prolyl hydroxylases less efficiently hydroxylate HIF-2a allowing it to stabilize and accumulate longer than HIF-1a. 24,25 In addition, differential expression of miRNAs can lead to decreased Hif-1a mRNA stability but not Hif-2a. 26 HIF-1a and HIF-2a have both redundant and non-redundant gene targets that depend on cellular context and cell type. 27 The presence of HIF-2a but not HIF-1a is likely due to the late-gestational sampling in this study and duration of hypoxia signaling. After formation of the liver bud (~E10), hepatoblasts begin to differentiate into biliary cells and then immature hepatocytes. 28 When fetal livers were collected in this study (E18.5), 10 HIF-2-alpha transcription factor network_Homo sapiens_37358832-6193-11e5-8ac5-06603eb7f303 HIF-1-alpha transcription factor network_Homo sapiens_20ef2b81-6193-11e5-8ac5-06603eb7f303 Regulation of Ras family activation_Homo sapiens_397d91c7-6195-11e5-8ac5-06603eb7f303 IL12 signaling mediated by STAT4_Homo sapiens_72cf19b8-6193-11e5-8ac5-06603eb7f303 Signaling events mediated by HDAC Class III_Homo sapiens_a7191c08-6195-11e5-8ac5-06603eb7f303 Noncanonical Wnt signaling pathway_Homo sapiens_7ff5fe76-6194-11e5-8ac5-06603eb7f303 Signaling events mediated by HDAC Class II_Homo sapiens_a1da5607-6195-11e5-8ac5-06603eb7f303 TCR signaling in naive CD8+ T cells_Homo sapiens_15a017bb-6196-11e5-8ac5-06603eb7f303 Notch-mediated HES/HEY network_Homo sapiens_8ee56389-6194-11e5-8ac5-06603eb7f303 Hypoxic and oxygen Homeostasis regulation of HIF-1-alpha_Homo sapiens_4c0f3584-6193-11e5 8ac5-0660 3eb7f303 Glycerophospholipid metabolism_Homo sapiens_hsa00564 HIF-1 signaling pathway_Homo sapiens_hsa04066 Other glycan degradation_Homo sapiens_hsa00511 Propanoate metabolism_Homo sapiens_hsa00640 Allograft rejection_Homo sapiens_hsa05330 Bladder cancer_Homo sapiens_hsa05219 Fructose and mannose metabolism_Homo sapiens_hsa00051 Graft-versus-host disease_Homo sapiens_hsa05332 Non Fold change Fig. 4 Relative fold change for genes involved in hypoxia-inducible factor signaling validated in an additional cohort with qPCR. Data are plotted as mean ± SEM and asterisks represent p < 0.05 with an unpaired t test. Control N = 8 and maternal nutrient restriction N = 8, 1 or 2 pups/litter differentiation into hepatocytes and hepatic outgrowth would be occurring and would continue into the perinatal period. 28 Studies in human induced embryonic stem cell liver buds and in the livers from fetal mice suggest that HIF-1a expression is positively associated with differentiation into biliary cells and negatively correlated with differentiation into hepatocytes. 29 Conversely, HIF-2a regulates proliferation and hepatic outgrowth in zebra fish. 30 Given the distinct roles, disruption in the timing or protein level of HIFs could alter liver development. Models of MNR, including ours, 11 have demonstrated reduced liver size (Radford and Han, 2018), 11 reduced transcription of genes promoting proliferation, and increased transcription of markers for differentiation. 31 While further studies to characterize temporal changes of HIF signaling induced by MNR are needed, increased hepatic HIF signaling late in gestation could indicate altered timing of differentiation and proliferation during liver development.
Hif-3a mRNA and protein were higher in MNR compared to controls. Hif-3a mRNA is induced by HIF-1a during chronic hypoxia and transcriptionally downregulates Hif-1a as a negative feedback mechanism. 32,33 In mice, Hif-3a has three isoforms, a full-length transcript and two variants, IPAS and NPAS. All three isoforms can inhibit the transcriptional activity of HIFs by binding with HIF-1β or HIF-1/2a, preventing dimerization and nuclear translocation. 34 Some HIF-3a splice variants can also weakly induce canonical HIF-1a target genes. 35 Consequently, increased HIF-3a protein could contribute to moderate differences between control and MNR HIFinduced transcription.
Fkbp5 was also increased at the transcript level and, although not significant, relatively increased at the protein level. FK506binding protein (FKBP5) is induced by chronic hypoxia in adipose tissue 36 and the induction is HIF-2a dependant in hepatocellular carcinoma cell lines. 37 It functions as co-chaperon to HSP90 and as a scaffold protein. Subsequently, FKBP5 has diverse functions, including reducing glucocorticoid receptor activity 38 and promoting adipocyte differentiation 36 and AKT dephosphorylation. 39 Other studies have been conducted in cancer cell lines or adipocytes, but the role in the developing liver in unknown. It is conceivable that increased FKBP5 transcription could be an attempt to adjust metabolism and/or cell growth but that is not sufficient to induce changes at the protein level due to posttranslational regulation. Posttranslational regulation of HIF target genes such as FKBP5 could also play a role in the maladaptive response to MNR in the developing liver.
Changes to hypoxia signaling in the liver could be indirect through placental adaptations. In guinea pigs, MNR results in increased hypoxyprobe-1 staining in both male and female liver and kidneys. 8 However, no differences were detected in the placentas. 8 Alternatively, MNR causes fetal circulation to distribute oxygenated blood received from the umbilical cord away from the liver to the brain. 40 MNR could increase HIF signaling in the developing liver through changes to fetal circulation and/or nutritional signaling. Further studies into the mechanism of HIF signaling need to be elucidated. However, increased HIF-induced transcripts support the concept that MNR results in increased hypoxia signaling in the fetal liver.
MNR resulted in fetal expression changes in hypoxia-inducible signaling pathways of E18.5 liver. Although expression changes were detected, the protein levels of genes induced by HIF transcription factors were not significantly different. Since tissue hypoxia declines during development, 21,29 it is possible that downstream changes were more evident earlier in liver development. Still, owing to their importance in regulating multipotency, cell differentiation, and proliferation, HIF-induced transcription could result in differences in liver maturity and cell populations. In addition, differential expression of epigenetic regulators such as KDM3a may prime HRE-containing genes to respond differentially to aging or nutritional abundance. Evidence of aberrant hypoxia signaling was evident in growth-restricted mice, but the impact was moderate that may relate to the duration of nutrient restriction and the timing of analysis. Differentially expressed transcripts support the concept that hypoxia signaling may play a role in growth restriction in response to fetal undernutrition.

Hif-2α
Hif-3α Hif-3α Hif-2α Actin Actin  FKBP5 (c) and PFKFB3 (d) were not significantly different in MNR relative to controls (p = 0.09 and 0.5, respectively). All blots were run in triplicates and representative blots are shown. Data are plotted as mean ± SEM and asterisks represent p < 0.05 with an unpaired t test. Control N = 6 and MNR N = 7, 1 or 2 pups/litter