Recombinant human GH was kindly donated by Pfizer laboratories (Madrid, Spain). All regents, drugs and compounds cited in this work were purchased from Sigma-Aldrich (St. Louis, MO, USA), unless otherwise indicated.

All animal experiments were conducted in accordance with Spanish and EU animal care guidelines and with approved protocol by the Animal Care and Use Committee of the University of Las Palmas de Gran Canaria (OEBA-ULPGC 2006/07824). Adult (2-3 months of age) male Sprague-Dawley rats (n=6 per group) used throughout the study were housed four or five to a cage, maintained under standard laboratory conditions (12-hour light/dark cycle and controlled temperature; 21-23 °C) with food (standard chow diet, A04 SAFE Panlab, Barcelona, Spain) and water provided ad libitum, at ULPGC Animal Unit. To generate the hypothyroid experimental model, rats were provided with 0.05% (w/v) methimazole (MMI) in drinking water ad libitum for 5 weeks (equivalent to a MMI dose of 50-60 mg/kg/day) starting on postnatal day (PND) 59 until sacrifice on PND94 (33). Calcium chloride (1%) was included with MMI in the water to ensure adequate dietary calcium intake because hypothyroidism decreases food intake by up to 40% (33). The MMI-containing water was replaced twice a week. Another group of rats were no MMI-treated during this period (INTACT). Two weeks after starting MMI administration, TX rats were orchiectomized or sham-operated to obtain TXOX or testis-intact hypothyroid (TXSO) groups, respectively. Bilateral orchiectomy using sterile procedures was performed under isoflurane anesthesia. Testicles were removed through a single medial raphe scrotal incision after transection of spermatic cords. In sham controls, the testicles were visualized but not removed. The INTACT group was also subjected to sham-surgery to provide testis-intact euthyroid group (INTACTSO) ( Figure S1A ). Four days after orchiectomy, we began hormonal replacement with TP (50 μg/kg; sc; 5 days per week) (TXOXTP) or vehicle (0.20 ml of corn oil; VEH) (TXOX) for 20 days before hormonal replacement with GH (TXOXTPGH and TXOXGH, respectively) or T3 (TXOXTPT3 and TXOXT3, respectively), T3 plus GH (TXOXT3GH) or VEH (TXOXTP) for 7 days ( Figure S1B ). GH (0.3 mg/kg/day) was administered as two daily subcutaneous (sc) injections in 12-h intervals (08:00h and 20:00h) to mimic the physiological male-specific GH secretion (5). T3 (10 μg/kg body weight) was administered as a single daily intraperitoneal (ip) injection. T3 was dissolved in a minimum volume of 0.01 N NaOH and was brought up to the appropriate concentration with sterile saline. Hypothyroidism status was corroborated by measuring serum levels of thyroxine (T4) and T3 at PND94 and monitoring body weight gain every 7 days after MMI administration was started. Body weight gain was calculated every week for each animal by using the formula BWy – BWx, where BWy is body weight calculated 7 days after last measurement (BWx). On PND94, body weight gain was calculated through the determination of body weight at PND59 and PND94.

Twenty-four hours (in the case of TP or T3) or twelve hours (in the case of GH) after the last injection, animals were euthanized by isoflurane overdose. On PND94, blood samples were collected, and serum stored at -80°C until analysis. Portions of the liver were snap frozen in liquid nitrogen and stored at -80°C until processed for mRNA analysis.

Serum free T4 and T3 concentrations were measured in duplicate by enzyme immunoassay (Beckman Coulter Inc., Pasadena, CA, USA), with a detection limit of 0.60 ng/dl and 88 ng/dl, respectively. Serum levels of glucose, total cholesterol (CHO), and triacylglycerols (TG) were quantified by using an Olympus AU2700 chemistry analyzer (Beckman Coulter Inc.). Total T (ng/dl) was determined by Access Immunoassay system test kit and Unicel DXI 800 auto-analyzer (Beckman Coulter Inc.). Serum levels of leptin and IGF-I were determined by using rat immunoassays (Quantikine, R&D systems, Minneapolis, MN, USA) according to manufacturer recommendations. The IGF-I and leptin assays included quality controls provided by the manufacturer, and the standard curves of the assays were performed using the samples provided by the manufacturer. All the samples were assayed together, and each sample was assayed in duplicate.

Total lipids were extracted with chloroform/methanol (2:1 v/v). The organic solvent was evaporated under a stream of nitrogen, and the lipid content was determined gravimetrically and stored in chloroform/methanol (2:1 v/v) containing 0.01% butylated hydroxytoluene as an antioxidant (34). Lipid classes were separated from a fraction of total lipids by one-dimensional double-development high-performance thin-layer chromatography using methyl acetate/isopropanol/chloroform/methanol/0.25% (w/v) KCl (5:5:5:2:1.8 by vol.) as the developing solvent system for the polar lipid classes and hexane/diethyl ether/acetic acid (22.5:2.5:0.25 by vol.) as the developing solvent system for the neutral lipid classes (34). Lipid classes were quantified by charring with 3% (w/v) aqueous cupric acetate containing 8% (v/v) phosphoric acid followed by calibrated scanning densitometry using a Shimadzu CS-9001PC dual-wavelength spot scanner (Shimadzu Europe, Duisburg, Germany). Total and neutral lipid fractions were subjected to acid-catalyzed transmethylation for 16 h at 50°C using 1 ml of toluene and 2 ml of 1% sulfuric acid (v/v) in methanol. The resultant FA methyl esters were purified by thin-layer chromatography and visualized under spraying with 1% iodine in chloroform (34). Fatty acid methyl esters were separated and quantified by using a Thermo gas chromatograph equipped with a flame ionization detector (250°C) and a fused silica capillary column Supelcowax™ 10 (30 m x 0.32 mm I.D.) (Sigma Aldrich, St. Louis, MO, USA). Individual fatty acid methyl esters were identified by referring to authentic standards. Equal amounts of total lipids (10 µg) were used in all analyses (34).

Total RNA was isolated by homogenization of 50-100 mg piece of frozen rat liver by using a polytrone PT-2000 (Kinematica AG, Malters, Switzerland) followed by the phenol-chloroform protocol supplied by the TRI Reagent-based method (Sigma-Aldrich). All samples were treated with RNAse-free DNAse set (Promega, Madison, WI, USA) and RNA was further purified by using the RNeasy Micro Kit (Qiagen, Valencia, CA, USA) following manufacturer’s recommendations. RNA yields were measured by UV absorbance and the quality of total RNA was analyzed by using a 2100 Bioanalyzer (Agilent, Palo Alto, CA, USA). A microarray containing 27000 rat 70-mer oligo probe sets produced at the KTH Microarray Center (www.biotech.kth.se) was used to evaluate the effects of hypothyroidism (TX), orchidectomy on TX (TXOX) and hormonal replacement in TXOX animals on liver transcriptome. Five μg of high-quality total RNA from the liver were reverse-transcribed and labelled with cyanine 3 (Cy3) and -5 (Cy5) using the ChipShot™ Direct Labeling System (Promega). After RNAse treatment, labelled cDNA was purified utilizing the QIAquick PCR Purification Kit (Qiagen). Hybridization was performed at 42°C for 16h and then, slides were washed following the Pronto™ Plus System protocol (Promega). Finally, slides were scanned using GenePix Scanner (Axon Instruments, Union City, CA, USA). To correct for the biological variability, 4 independent hybridizations were performed comparing individual animals from the different experimental groups for a total of 4 analyses (35).

Image analysis was performed using the GenePix Pro 6.0 software as previously described (33). The LOWESS method was used to normalize the raw intensity data (36). If the measured probe sets were not present in at least 3 of the 4 chips, they were assumed to contain no information and therefore were eliminated to reduce data complexity. Differentially expressed genes (DEGs) were identified by using the Significance Analysis of Microarrays (SAM) statistical technique (37). A q value was assigned for each detectable gene in the array. This value is similar to a p-value, measuring the lowest false discovery rate (FDR < 0.05) at which differential expression of a gene is considered significant. In addition, a mean ratio of log2 > |0.32| value was chosen to describe up- or down-regulated genes. A completed list of DEGs is available as supplementary files and they have been deposited in Gene Expression Omnibus (GSE111718; www.ncbi.nlm.nih.gov/geo) (38).

The Database for Annotation, Visualization and Integrated Discovery (DAVID; version 6.8), a widely used web-based tool for functional and pathway enrichment analyses (39), was used to perform Gene Ontology (GO) set enrichment analysis (GSEA) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of DEGs. The GO terms and KEGG pathways with p<0.05 and a corrected q value (Benjamini) < 0.1 were selected.

The biological database and web tool for the retrieval of interacting genes STRING (40) was used to identify known and predicted protein-protein interactions and make a network of the overlapping DEGs. The results were visualized using Cytoscape software (41).

Briefly, 2 µg of total RNA were treated with RNase-free DNase I (Promega) and reverse transcribed using the iScript kit (Bio-Rad, Irvine, CA, USA) according to the manufacturer`s instructions. Two µl of cDNA served as a template in a 20 µl qPCR reaction mix containing the primers and SYBR Green PCR Master Mix (Diagenode, Seraing, Belgium). Quantification of gene expression was performed using the ABI PRISM^® 7000 SD RT-PCR System (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA). A relative standard curve was constructed with serial dilutions (1:1, 1:10, 1:100) using a pool of the cDNA generated from all animals used in the study. The amplification program consisted of 1 cycle of 95°C for 10 min, followed by 45 cycles of 95°C for 15 sec, annealing for 10 sec, and 72°C for 30 sec. The fluorescent intensity was measured at a specific acquisition temperature for each gene. Data were extracted and amplification plots generated with ABI SDS software (Applied Biosystems, Thermo Fisher Scientific). All amplifications were performed in duplicate, and C_t scores were averaged for subsequent calculations of relative expression values. The level of individual mRNA was normalized to the level of the housekeeping gene cyclophilin by using the Pfaffl method (42). Exon-specific primers ( Table S1 ) were designed by the Primer 3 program (43).

The significance of differences between groups was tested by one-way ANOVA, followed by post hoc comparisons of group means according to the GraphPad Prism 5 program. Statistical significance was reported if P<0.05 was achieved. For graphing purposes in the qPCR analysis, the relative expression levels were scaled so that the expression level of the INTACT group equaled one. Lipids classes and main FAs were additionally submitted to factor analysis by means of Principal Component Analysis (44). Factor scores were then analysed by two-way ANOVA to evaluate the combined effects of both hormonal treatments and their interactions. Multivariate analyses were performed using the SPSS package (version 15.0, SPSS Inc, Chicago, IL, USA).

Upon sacrifice on PND94, all MMI-treated groups showed lower or undetectable circulating T3 (ng/dl) [36.17 ± 5.43 (INTACTSO); 16.16 ± 2.73 (TXSO); 7.92 ± 4.84 (TXOX); 0,00 ± 0.00 (TXOXGH); 8.88 ± 3.07 (TXOXTP); 9.48 ± 0.96 (TXOXTPGH)] and T4 (ng/dl) [(1.90 ± 0.17 (INTACTSO) vs. 0)]. However, circulating levels of total T (ng/ml) in TXSO rats (10,14 ± 2,90) did not show differences compared to INTACTSO (10,66 ± 2,77) controls. Therefore, castration in TX rats was performed to explore the effects of TP replacement in the TXOX group. Consistently, 11-fold reduction of circulating T (ng/ml) (0,91 ± 0,12) was observed in the TXOX group when compared to non-castrated TXSO rats, whereas TP replacement led to 4-fold increase of T (4,10 ± 0,15) in the TXOXTP group. In contrast, circulating E2 (pg/ml) in TXOXTP rats (18,83 ± 8.45) was not significantly increased when compared to TXOX (12,50 ± 7,66). Next, as a first approach to assess the efficacy of TP replacement on somatotropic-liver axis, male phenotypic and functional biomarkers of GHR-STAT5b signaling were measured. Castration of TX rats (i.e., TXOX group) reduced body weight gain ( Figure 1A ), serum IGF-I ( Figure 1B ) and hepatic Igf-1 mRNA levels ( Figure 1C ) to a greater extent than non-orchiectomized but hypothyroid rats (i.e., TXSO). These changes were partially restored upon hormonal replacement with TP or pulsatile GH administration ( Figures 1A–C ). Interestingly, body weight gain and serum IGF-I were further increased when TP replacement was performed in the presence of GH ( Figures 1A–C ) which supports a positive cooperative interaction between T and GH on the somatotropic-liver axis.

To validate this cooperative interaction, in addition to hepatic Igf-1 mRNA levels ( Figure 1C ), several functional biomarkers of GHR-STAT5b signaling in the liver were also explored (e.g., Socs2, Cis, Cyp2c11, Cyp2c12). Castration of TX male rats (i.e., TXOX group) caused 80% decrease of Socs2 mRNA levels ( Figure 1D ). However, unlike TP, pulsed GH replacement induced Socs2 ( Figure 1D ) and Cis mRNA expression ( Figure 1E ). Furthermore, TP replacement partially prevented the positive effects of GH on Socs2 mRNA expression ( Figure 1D ), whereas GH-induced Cis mRNA levels remained unaffected ( Figure 1E ). Remarkably, TP and GH positively cooperated to regulate Socs5 gene expression ( Figure 1F ) but, conversely, Socs3 mRNA expression was increased in TXSO and TXOX ( Figure 1G ). As expected, the hepatic mRNA expression levels of gender differentiated genes were affected by TP or pulsatile GH replacement. First, the male specific Cyp2c11 gene was downregulated by 50% in TXSO rats, and it was abolished after castration (i.e., TXOX) ( Figure 1H ). Pulsed GH administration restored Cyp2c11 mRNA levels whereas the effect of TP replacement alone was lower. However, the efficacy of GH to induce Cyp2c11 mRNA was reduced in the presence of TP ( Figure 1H ). TP ( Table S2 ), pulsed GH ( Table S3 ) or combinatory treatment ( Table S4 ) also increased expression of male-differentiated genes such as α-2u-globulin and major urinary proteins in TXOX rat liver. Paradoxically, TP replacement augmented the female-specific Cyp2c12 gene to higher level than age-matched intact and TXOX-vehicle treated rats ( Figure 1I ).

First, we performed a genome-wide analysis of gene expression to better understand the effects of T on liver transcriptome in TXOX rats. This analysis identified 611 DEGs (q<5%; Log₂>|0.32|) that were distinctly regulated by TP ( Table S2 ). Next, we identified genes involved in active biological processes (BPs) in the DEGs list by using GO functional clustering (39). GSEA of the 338 genes upregulated by TP ( Figure 2A ) revealed a significant over-representation of genes implicated in monocarboxylic acids metabolism, FA metabolism, biosynthesis of small molecules, generation of precursor metabolites and energy, response to nitrogen compounds, and glucose metabolism, among others. These BPs were also significantly upregulated by TP replacement in TXOX liver. In addition, TP administration to TXOX rats downregulated genes involved in the metabolism of organic and carboxylic acids, steroid metabolism, regulation of apoptosis, organic acid catabolism or lipid metabolism ( Figure 2B ). Accordingly, TP regulated several KEGG) intracellular signaling pathways linked to, among others, steroid metabolism (e.g., Cyp2c22; Cyp2d2; Ugt2b35; Hsd17b2), tricarboxylic acid cycle (TCA) cycle (e.g., Acly; Mdh1, Idh3B; Pck1), glycolysis/gluconeogenesis (e.g., Akr1a1; Gapdh; Pklr; Pck1), or chemical carcinogenesis (e.g., Cyp2c22; Ephx1; Cyp2c6) ( Table 1 ). Overall, these results revealed an extensive re-programming of hepatic transcriptome by TP replacement in TXOX rats, particularly of those genes that play a functional role in glucose and lipid metabolism.

Since the liver is a primary organ in which T and GH interact to enhance body growth and composition (13), we next performed a genome-wide analysis of gene expression to better understand the influence of TP on GH-regulated hepatic transcriptome in TXOX rats. First, we identified DEGs ( Table S3 ) and KEGG signaling pathways ( Table S4 ) induced by pulsed GH administration to TXOX rats. Second, the similarities in DEGs induced by TP or GH were analysed, and 125 significantly regulated transcripts (FDR<5%; q<5%; log₂>|0.32|) were identified in both TXOXTP and TXOXGH groups ( Figure 3A ). A Spearman rank correlation test comparing the effects of TP or GH on these DEGs yielded a strong positive correlation (r=0.6016; P<0.0001) between the effects of these hormones on liver transcriptome. Accordingly, 90% of these genes were regulated in the same way by TP and GH. Third, the transcription profiles of TXOX rats simultaneously treated with TP and GH were analyzed. A comparative analysis of DEGs in TXOXGH ( Table S3 ) or TXOXTPGH ( Table S5 ) groups revealed that T influenced a large fraction of GH-regulated genes ( Figures 3B–D ). Accordingly, in the absence of TP, the average expression changes (Log₂) across four independent hybridizations were 0.54 ± 0.01 and -0.52 ± 0.01 for GH-induced ( Figure 3C ) and GH-repressed ( Figure 3D ) genes, respectively, whereas the average fold regulation for the same set of DEGs in the presence of TP was 0.11 ± 0.01 and -0.05 ± 0.01, respectively. These significant differences (P<0.0001) showed that TP replacement inhibited several GH-regulated genes in the liver of TXOX rats. Finally, SAM multiclass analysis (37) ( Figures 3E, F and Table S6 ) identified genes that were regulated by GH whose mean expression values (SMD) were significantly different from those in TP- or TPGH-treated TXOX rats. DEGs by GH replacement in the presence of TP (i.e., TXOXTPGH) ( Table S5 ) were functionally linked to KEGG signaling pathways associated to TCA cycle, PPAR, chemical carcinogenesis, steroid hormone biosynthesis, FA metabolism, amino acids metabolism, glycolysis/gluconeogenesis and diseases associated to glucose and lipid metabolism (e.g., NAFLD) ( Table 2 ). Similarly, GH replacement whose SMD were significantly different from those in TP- or TPGH-treated TXOX rats ( Table S6 ) were functionally linked to KEGG pathways associated to ribosome regulation, TCA cycle, oxidative phosphorylation, FoxO (e.g., Igf1, Tgfbr1, Homer3, Kras), FA degradation (e.g., Gcdh, Echs1, Ehhadh), glycolysis/gluconeogenesis, arginine biosynthesis (e.g. Sds, Mat1a, Pah, Otc, Cps1, Cth), and to diseases such as prostate cancer and NAFLD. Extensive similarity was also detected between BPs that were upregulated by pulsed GH ( Figure 4 ) when compared with those regulated by GH in the presence of TP (i.e., TXOXTPGH) ( Figure 5 ).

The analysis of protein-protein interactions on TPGH (or GH)-regulated transcriptome in TXOX liver predicted interactions among proteins playing a crucial role in the regulation of biosynthesis of FA and steroid metabolism ( Figure 6 ; labelled by black circle). However, several genes involved in FA biosynthesis were not modulated by pulsed GH but upregulated by TP (e.g., SCD, FASD, APOE, ME1, ENO1, DGAT2, PCK) ( Figure 6 ; gray nodes). Interestingly, most of these genes were not upregulated by GH as a single treatment but were induced when GH was combined with TP (e.g., SRD5A1). This supports the hypothesis that the cooperation between T and GH is needed to induce the transcription of several genes and to stimulate the interaction between proteins (e.g., PKLR, BAAT, ELOVL5, APOC2, LDHA) ( Figure 6 ; black nodes). TP also cooperates with GH to increase expression of genes involved in long chain FA (LCFA) synthesis (e.g., Acly) which suggests that this mechanism can promote LCFA synthesis. To validate the cooperative interaction between T and pulsed GH, several lipid biomarkers were explored by RTPCR ( Figure 7 ). Castration of hypothyroid male rats (i.e., TXOX) reduced Srebp2, Srebp1, Acc1, Scd1, Fas or Cyp4f1, to a greater extent than non-orchiectomized but hypothyroid rats (i.e., TXSO). Interestingly, these changes were restored upon TP replacement and were further increased when TP was combined with pulsed GH which supports a positive cooperative interaction between these hormones to regulate lipid metabolism. This was not observed with the Fas gene which encodes the enzyme that catalyze de novo synthesis of FAs. In this case, the stimulative effect induced by TP on Fas gene was inhibited when TP was combined with GH ( Figure 7F ). Interestingly, only when TP was present, GH increased Srd5a1 gene expression ( Table S5 ), which supports the hypothesis that an enhanced conversion of T to DHT could contribute to potentiate AR-dependent signaling in the liver. Overall, these results reveal an extensive overlapping and cooperation between T and pulsed GH in liver transcriptome linked to lipid metabolism.

The transcriptome data suggested that T and GH cooperate to maintain lipid composition in the liver. Therefore, the effects of T on serum and hepatic lipid composition were first evaluated by quantitative analysis of lipid classes in TXOX rats ( Table 3 ). The development of a hypothyroid state in male rats was accompanied by altered levels of circulating lipids (i.e., 2-fold increase of CHO and 3-fold decrease of TG). However, neither TP alone nor its combination with GH were able to normalize these variations. In contrast, TP replacement caused complex changes of hepatic lipids ( Table 3 ). Given the complexity of these data, we initially performed Principal Component Analysis of lipid classes to discriminate experimental groups based on their lipid profiles ( Figure 8A ). The outcomes revealed that the two principal components explained 64% of overall variance. Principal Component 1 was negatively related to TG, cholesteryl esters and total neutral lipids; and positively to phosphatidylcholine, phosphatidylethanolamine, sphingomyelin, and total polar lipids. Principal Component 2 was mainly related to phosphatidylinositol and free FA (FFA) ( Figure 8A , left panel). When we computed factor scores 1 and 2 from PC1 to pursue group simplification ( Figure 8A , right panel), we obtained that INTACTSO group were represented by a discrete cluster with a small degree of intersection with TXOX, TXOXGH and TX groups. Best distinguishable groups were TXOXTP, TXOXT3GH, TXOXT3 and TXOXTPGH (these later two groups displaying nearly complete overlapping). Overall, it may be concluded that TP, GH and T3 differentially modify hepatic lipid metabolism, but their effects are clearly modulated by their combinations. This was particularly obvious for the interactions of the effects of TP or T3, on one side, and the presence of GH, on the other ( Figure 8A , right panel).

To provide a quantitative assessment of hormonal interactions, we next performed two-way ANOVA on factor scores as well as on individual lipid classes. We found that there was a dependent relationship between all the hormones studied (particularly for GH), and interactions between them were identified for the first two components. For factor score 1, main dependency was obtained for GH (F=19.191, P=0.000). Further, a clear interaction was observed for TP and T3 (F=30.11, P=0001), and between GH and TP (F=19.32, P=0.000). For factor score 2, the main determinant of variance was the state of TP (F=8.36, P=0.007) followed by T3 (F=3.10, P=0.088). There were also significant interactions between T3 and TP (F=13.49, P=0.001) as well as between GH and T3 (F=24.69, P=0.000). Provided these multivariate results, we next compared individual lipid classes in the liver focused on the main parameters determining factor score 1. Thus, TXOX animals exhibited highest hepatic CHO content and decreased levels of TG and DG compared to the INTACTSO group. The highest levels of hepatic CHO were found in TXOX rats whereas the lowest values were observed in TXOXGH rats [ Table 3 and (33)]. In general, changes in cholesteryl esters levels were found to be negatively related to variations in CHO levels across groups (r=0.5901, P=0.0002). Cholesteryl esters were highly increased by GH treatment and reached lowest levels after TP (that together exhibited a certain degree of antagonism). Compared to INTACTSO rats, TG levels were not affected in vehicle-treated TXOX but were considerably increased by GH (TXOXGH) and reached minimal levels after TP (TXOXTP) treatment ( Table 3 ).

The effects of TP were counteracted by the presence of GH, indicating an antagonistic effect of GH on T activity. In parallel, GH caused a dramatic reduction in hepatic levels of FFA, while TP induced the opposite effect. Noteworthy, changes in FFA negatively correlated with changes in DG (r=0.5296; P=0.0006). Interestingly, TP treatment of TXOX animals resulted in further FFA accumulation in the liver. The highest reduction of DG and increase of FFA was observed in TXOX rats treated with T3 plus GH (i.e., TXOXT3GH) (data not shown). However, the opposite response was observed in TXOX rats treated with TP plus GH (i.e., TXOXTPGH). These results indicate that, while GH effects are modulated by TP, the effects of this interaction are permissive. Furthermore, the combination of GH and TP significantly increased total neutral lipids due to changes in hepatic contents of CHO, cholesteryl esters, TG and DG. On the other hand, total polar lipids showed a reciprocal behavior to total neutral lipids. These data indicate a strong remodeling effect of hormonal treatments on hepatic phospholipids. Combination of TP and GH (but not GH alone) caused a significant reduction of total polar lipids, due to decreased phosphatidylcholine and phosphatidylethanolamine. Overall, these results suggest that hormonal replacement with GH, TP or TPGH can induce different hepatic lipid profiles in hypothyroid subjects. The main difference between GH and TP treatments was that TG levels were significantly increased by GH and reduced by TP. These data pinpoint to an anabolic effect leading to esterification of FFA into DG and TG triggered by GH, which was turned into catabolic by TP.

In order to better understand the influence of hormonal replacement on hepatic FA profiles ( Table 4 ), we performed multivariate analyses of the whole dataset. Principal Component Analysis results showed that two Principal Components may explain 52% of total variance ( Figure 8B , left panel). Principal Component 1 was negatively related to 22:5n6, 18:1n9, and 18:1n7, and positively to saturated FA (SFA) (23:0, 24:0), n-3 long chain polyunsaturated FA (LCPUFA) (22:5n3, 20:5n3), and n-6 LCPUFA (20:3n6). Principal Component 2 was mainly related to 20:4n6 (negatively) and 16:1n7 (positively). Computed factor scores 1 and 2 allowed the discrimination between different hormonal treatments into four major clusters. The INTACTSO and TXSO groups were represented by two discrete clusters whereas the T3 and TP groups appeared in differentiated clouds with certain degree of intersection ( Figure 8B , right panel). Untreated TXOX group exhibited marginal intersections with TXOXGH and TXOXTP, but none with TP in the presence of GH ( Figure 8B , right panel). Analyses of dependencies and interactions for the effects of different hormones on FA from total lipids were performed by multiple analyses of variance. The results for factor score 1 indicated that T3 was the main discriminating hormone (F=158.68, P=0.000), which also displayed some interaction with GH (F=3.47, P=0.072). However, for factor score 2, main dependences were observed for the three hormones individually (highly significant for T3: F=8.58, P=0.006), but only significant interactions with GH were detected for TP (F=8.87, P=0.006). Given these results, we next focused on main dependencies displayed by hormonal treatments on the composition of major FAs. Thus, compared to TXOX, TXOXTP animals displayed slightly lower levels of monoenoic FA (MUFA), while SFA and polyunsaturated FA (PUFA) remained unaffected ( Table 4 ).

Hypothyroidism resulted in the accumulation of SFA in the liver, which was maintained upon castration. Comparison of TXOX data with those of TXSO suggested that many of the effects on FA contents were due to deficiency of gonadal hormones. Further, in the presence of GH (i.e., TXOXTPGH), the most notable changes were those introduced by T ( Table 4 ). GH treatment was able to restore SFA to the levels of INTACTSO animals, an effect further enhanced by co-treatment with T. However, the comparison of TXOX group with TXSO revealed a discrete effect of gonadal hormones on hepatic FAs. This apparent absence of T effects was significantly changed when combined with GH. In the presence of T, (i.e., TXOXTPGH group), SFA levels dropped, while MUFA increased. These effects could be explained by the significant reduction of stearic acid (18:0) and the elevation of oleic acid (18:1n-9). For MUFA group, and particularly for 18:1n-9, a permissive action of GH on the effects of TP was evident, since levels of this major FA were highest in TXOXTPGH group, and significantly higher than in INTACTSO, TXOX, TXSO, TXOXGH and TXOXTP groups. Within n-3 or n-6 PUFAs series, no significant changes were observed by TP or TP plus GH treatments. However, linoleic acid (18:2n-6) was slightly increased and arachidonic acid (20:4n-6) reduced (as well as its elongation/desaturation intermediates 20:2n-6 and 20:3n-6) in the TXOXTPGH group, indicating that GH decreased the biosynthesis of essential n-6 LCPUFA in the presence of T. Within the n-3 LCPUFA series, TP alone or combined with GH (i.e., TXOXTPGH group) failed to modify levels of docosahexaenoic acid (DHA; 22:6n-3) and eicosapentaenoic acid (EPA; 20:5n-3) when compared to TXOX group. In this sense, the interaction between T and GH is particularly illustrative since DHA levels increased when GH was administered alone ( Table 4 ). In agreement, comparison of TXOX and TXOXTPGH groups with TXSO rats, suggested an important role of gonadal hormones on DHA and arachidonic acid levels since they were higher and lower in TXSO rats, respectively. This effect may be caused by orchiectomy. It might be concluded a main role of T plus GH in the configuration of hepatic FA profiles, but these effects are modulated by complex interactions between them.

Finally, the potential androgenic, estrogenic or both effects of TP replacement on liver transcriptome and lipid composition were assessed. First, T-regulated transcriptome ( Table S2 ) was compared to those genes upregulated by E2 (33) but no significant correlation between the DEG datasets was found, and GSEA comparison of the BPs upregulated by TP and E2 showed limited overlap ( Figure 9A ). E2 treatment induced expression of genes involved in FA oxidation (e.g., PPAR-dependent signaling) while T upregulated those genes linked to FA biosynthesis which agrees with an increased FFA accumulation in TXOXTP liver ( Table 4 ). Second, TPGH-regulated transcriptome ( Table S3 ) was compared to those genes upregulated by E2GH (33) but a limited (6% of the genes) and significant correlation (q<0.0001) was found ( Figure 9B ). Strikingly, the overlap between TPGH and E2 administered alone was 16% (q<0.0001) ( Figure 9C ). However, when TP- (or TPGH-) upregulated genes were compared to those modulated by E2 (or E2GH), there were global different effects, despite some similarities were also found. In summary, these results suggest that T causes an extensive re-programing of liver transcriptome which is mainly an androgenic effect.