All studies were approved by the Columbia University Institutional Animal Care and Use Committee (IACUC). Male wild-type (WT) C57BL/6 and Ldlr^−/− mice of 3–4 months of age, purchased from the Jackson Laboratory, were used. WT C57BL/6 and Ldlr^−/− mice (n = 5–6/group) were fed a Western-type diet (WTD) containing: 42% fat, 42.7% carbohydrate, 15.2% protein, 0.15% cholesterol; total 4.5 Kcal/g (Harlan Teklad) for 1 week. After 1 week, C57BL/6 and Ldlr^−/− mice were continued on the WTD and were divided into groups receiving vehicle (58.5% saline + 40% DMSO + 1.5% 1M NaOH) or iodothyronines—T3 (0.75 mg/kg) (EMD Chemicals/Calbiochem) or T2 (12.5 mg/kg) (Santa Cruz)—via daily gavage for another week and then sacrificed. The doses of T3 and T2 were chosen to obtain the same cholesterol reducing effect (Goldberg et al., 2012). Mass spectrometry and NMR profiles showed no T3 or T4 contamination in the T2. All blood samples obtained during the non-terminal portion of the study were taken after a 4-h fast.

Protein extraction and sample preparation for 2D-E were performed as reported in Silvestri et al. (2006). Liver tissue was suspended in sample buffer [20 mM Tris, 7 M urea, 2 M thiourea, 4% CHAPS, 10 mM 1,4-dithioerythritol (DTE), 1 mM EDTA, and a mixture of protease and phosphatase inhibitors]. The suspensions were homogenized using a Polytron homogenizer, sonicated for 30 s, and centrifuged at 150,000 g for 45 min. The obtained supernatants contained the total liver proteins solubilized in the isoelectrofocusing (IEF)-compatible agents. The protein content of each sample was determined by Bio-Rad's DC method (Bio-Rad Laboratories, Hercules, CA). Total protein extracts were prepared for each animal, and each individual was assessed separately.

Samples of 650 μg of protein were applied to immobilized pH gradient (IPG) strips (pH 4–7, 17 cm) (Bio-Rad). Samples of 1 mg of protein were utilized for preparative gels (IPG strips, pH 4–7, 17 cm). Focusing started at 200 V, with the voltage being gradually increased to 3500 V and kept constant for a further 66,500 V/h (PROTEAN IEF System, Bio-Rad). Prior to SDS-PAGE, the IPG strips were incubated for 15 min with a solution of Tris-HCl buffer (pH 8.8), urea (6 M), glycerol (30%, v/v), SDS (2%, w/v), and DTT (2%, w/v). Strips were then equilibrated for another 15 min in the same buffer containing iodoacetamide (2.5%, w/v) instead of DTT. The second-dimensional separation was performed in 12% SDS-polyacrylamide gels. After protein fixation, the gels were stained with colloidal Coomassie Blue, according to the manufacturer's instructions. Molecular masses were determined by running standard protein markers, covering the range 10–200 kDa. The pI values used were those given by the supplier of the IPG strips.

Digital images of the gels were recorded using a calibrated densitometer (GS-800, Bio-Rad) and analyzed using PDQuest software (Bio-Rad) (Silvestri et al., 2006). For each matchset analysis, maps corresponding to protein extracts from animals of the same experimental group were organized into “Replicate Groups” (each containing 4 maps), named Ldlr^−/−, Ldlr^−/−+T2, and Ldlr^−/−+T3. Statistical analysis was performed using a Student's t-test. Spots for which the p-value was <0.05 were considered to display significant changes.

Spots from 2D-E were manually excised from gels, triturated, and washed with water. Proteins were in-gel reduced, S-alkylated, and digested with trypsin, as previously reported (D'Ambrosio et al., 2008). Protein digests were subjected to a desalting/concentration step on μZipTipC18 pipette tips (Millipore Corp., Bedford, MA, USA) and then analyzed by nano-liquid chromatography (nLC)-electrospray ionization (ESI)-linear ion trap (LIT)-tandem (MS/MS) mass spectrometry, using a LTQ XL mass spectrometer (Thermo Fischer Scientific, USA) equipped with a Proxeon nanospray source connected to an Easy-nanoLC (Proxeon, Odense, Denmark). Peptide mixtures were separated on an Easy C18 column (100 × 0.075 mm, 3 μm) (Thermo, USA) using a gradient of acetonitrile containing 0.1% formic acid in aqueous 0.1% formic acid; acetonitrile was ramped from 5 to 35% over 10 min, from 35 to 95% over 2 min, and remained at 95% for 12 min, at a flow rate of 300 nL/min. Spectra were acquired in the range m/z 400–2000. Acquisition was controlled by a data-dependent product ion-scanning procedure over the 3 most abundant ions, enabling dynamic exclusion (repeat count 2; exclusion duration 1 min). The mass isolation window and collision energy were set to m/z 3 and 35%, respectively.

MASCOT software package version 2.2.06 (Matrix Science, UK) (Cottrell, 2011) was used to identify spots unambiguously from a Mus Musculus protein sequence database retrieved from UniProt repository (76,058 sequences, 10/2011). Raw data from nanoLC-ESI-LIT-MS/MS were searched using a mass tolerance value of 2 Da for precursor ion and 0.8 Da for MS/MS fragments, trypsin as proteolytic enzyme, a missed-cleavages maximum value of 2, and Cys carbamidomethylation and Met oxidation as fixed and variable modifications, respectively. Protein candidates with more than 2 assigned peptide sequences, with MS/MS ion score >30 and a peptide expectation value <0.05, were further evaluated by comparison with their calculated mass and pI values, using the experimental values obtained from 2D-E.

The lists of differentially expressed proteins were input into the IPA platform (Ingenuity Systems, http://www.ingenuity.com) for the identification of canonical pathways and functions differing between the treatments. The cutoff used was 1.5 for the fold change and 0.05 for the p-value. Fisher's exact test was used to calculate a p-value indicating the probability that each biological function and/or disease assigned to that dataset might be so assigned due to chance alone. The results of Fisher's exact test were corrected for multiple testing using the false discovery rate (FDR). Comparison analysis was also performed to compare the roles of the proteins among the generated lists. In order to generate the reported networks, a list of differentially expressed proteins in the experimental conditions was overlaid onto a global molecular network developed from information contained in the Ingenuity Pathways Knowledge Base (IPKB). Networks of these focus-gene products were then algorithmically generated on the basis of their connectivity. The IPA platform uses a curated database to construct functional regulatory networks from a list of individual proteins. To build networks, the program utilizes the IPKB containing large numbers of individually modeled relationships between proteins (obtained from the upgraded literature contained in IPA platform). The algorithm then determines a statistical score for each network. This is done by comparing the number of focus proteins that contribute to a given network relative to the total number of occurrences of those proteins in all networks or pathways stored in the IPKB. Then a score is assigned to each network. The score is the negative log of P, and it denotes the likelihood that the focus proteins in the network might be found together by chance. Therefore, scores of 2 have at least 99% confidence of not being generated by chance alone. In addition, the biological functions assigned to each network are ranked according to the significance of that biological function to the network.

As already described in Cioffi et al. (2010), livers were dissected and minced in ice-cold isolation buffer (consisting of 220 mM mannitol, 70 mM sucrose, 20 mM Tris-HCl, 1 mM EDTA, and 5 mM EGTA, pH 7.4) and then were homogenized in a Potter-Elvehjem homogenizer. To isolate nuclei, the liver homogenate was centrifuged at 500 g for 10 min at 4°C. The obtained pellet was subsequently resuspended and spun through a sucrose cushion (30% sucrose, 10 mM Tris-HCl, pH 7.5, 10 mM NaCl, and 3 mM MgCl₂) at 1300 g for 10 min at 4°C. The nuclear pellet was washed with cold 10 mM Tris-HCl, pH 7.5, and 10 mM NaCl.

For Western immunoblotting analysis, the nuclear pellets were homogenized in lysis buffer containing 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 2.5 mM Na₂H₂P₂O₇, 1 mM b-CH₃H₇O₆PNa₂, 1 mM Na₃VO₄, 1 mM PMSF, 1 mg/ml leupeptin, and 1% (w/v) Triton X-100 by using an Optima TLX Ultraturrax (Beckman Coulter, Milan, Italy), then centrifuged at 13,400 g for 10 min at 4°C. Protein concentration was determined by using the Bio Rad's DC method (Bio-Rad Laboratories, Hercules, CA). The following primary antibodies were used: anti-HNF4α (ab41,898, Abcam, mouse monoclonal); anti-PPARα (ab8934, Abcam, rabbit polyclonal); anti-LXRα (ab41,902, Abcam, mouse monoclonal); anti-TRβ (GTX113278, GeneTex, rabbit polyclonal); anti-PGC-1α (AB3242, Merck Millipore, rabbit polyclonal); anti-α tubulin (ab4074, Abcam, rabbit polyclonal). Proteins were detected by a chemiluminescence protein-detection method based on the protocol supplied with a commercially available kit (Millipore) and by using the appropriate secondary antibodies. Signals were quantified by means of a Bio-Rad ChemiDoc™ XRS, using dedicated software (QuantityOne, Bio-Rad Laboratories).

The serum levels of FT3 and FT4 were measured by commercially available kits by means of immunoassay [materials and protocols supplied by Byk-Sangtec Diagnostica (Dietzenbach, Germany)].

Data are expressed as the mean ± SD as indicated in the figures. Comparisons between two groups were performed using student's t-test. Comparisons amongst three groups were performed using one-way ANOVA and followed by the Newman-Keuls test. For all analyses, p < 0.05 was considered statistically significant.

Through a 2D-E-based proteomic approach, the hepatic pathways and the molecular mediators involved in the T2- and T3- induced reductions in circulating cholesterol levels in Ldlr^−/− mice were investigated. At the detection-limits set, the image software counted 450 matched proteins among the various maps in the liver (Figure 1A). Pair-wise comparisons were performed to analyze the differential expression pattern associated with the T2 and T3 treatment (Ldlr^−/−+T2 and Ldlr^−/−+T3 vs. Ldlr^−/−). When the interest was limited to a differential expression of at least 2-fold and a statistical significance of at least 95% (p < 0.05), 57 (about 12.4% of total entries), and 59 spots (about 12.8% of total entries) showed significant quantitative changes in liver following T2- and T3-treatment, respectively. Importantly, the differential expression produced by T2 and T3 overlapped on 33 protein products (Figure 1B) corresponding to 40% of the total amount of differentially expressed proteins (Figure 1E). The remaining, specifically affected either by T2 (Figure 1C) or by T3 (Figure 1D), represented 29 and 31% of the total amount of differentially expressed proteins (Figures 1F,G, respectively).

Considering that in a previous study (Goldberg et al., 2012) we demonstrated a modulation exerted by T3 and T2 on intrahepatic content of apolipoproteins and that ApoE and ApoA1 are well resolved on a standard 2D map of total soluble liver proteome, ApoE and ApoA1 were localized by means of map comparison (Fountoulakis et al., 2001) and subjected to quantitative analysis. ApoE, corresponding to spot 43 (Figure 1C), was significantly affected only by T2 with a differential expression of −40% vs. Ldlr^−/− control levels (p < 0.1) (Figure 2). ApoA1, corresponding to spot 69 (Figure 1D), tended to be reduced in Ldlr^−/−+T3 mice by −40% vs. Ldlr^−/− control levels, although without statistical significance (Figure 2). These two proteins together with other 67 spots among those automatically detected by the matching software, were manually excised, trypsinyzed, and subjected to nanoLC-ESI-LIT-MS/MS analysis. Fifty-two protein spots were unambiguously identified (Supplementary Material 1). Among these, 25 protein products were affected by both T2 and T3, 9 only by T2, 18 only by T3.

Proteins detected as a mixture of components were excluded from further analyses and thus were not discussed further in this report. When proteins were identified as multiple spots on the same map, putatively reflecting the occurrence of post-translational modifications, the pattern of changes was fairly similar among the various species.

Both T2 and T3 reduced expression of a number of hepatic proteins compared to the expression levels obtained for Ldlr^−/− control mice (91 and 79% of differentially expressed proteins, respectively). A significant increase produced by both iodothyronines was observed for cellular retinol-binding protein 1 also named CRBP1 (spot 1) and major urinary protein 1 (MUP1) (spot 2), with the effect of T3 being more pronounced than that of T2 (Figure 3). Key metabolic enzymes were among the proteins decreased by both iodothyronines (Figure 4). Importantly, these enzymes are involved in major hepatic processes such as amino acid metabolism [catechol O-methyltransferase (spot 4), ornithine aminotransferase, mitochondrial (spot 21) and histidine ammonia-lyase (spot 29)] (Figure 4A), substrate metabolism [aldose 1-epimerase (spot 15), malate dehydrogenase (spot 17), and fructose-1,6-bisphosphatase 1 (spot 19)] (Figure 4B), and cellular stress [lactoylglutathione lyase (spot 3), regucalcin (spot 13), mitochondrial aldehyde dehydrogenase (spot 25)] (Figure 4C). Of note, these proteins are regulated to the same extent by either T3 or T2, highlighting a common effect of the two iodothyronines.

T2 specifically decreased proteins such as fatty acid-binding protein (spot 31), succinyl-CoA ligase (spot 38), and glycerol kinase (spot 42), which are involved in lipid, oxidative, and carbohydrate metabolism, respectively (Figure 5A). In contrast, glycerol-3-phosphate dehydrogenase (spot 68) and isocitrate dehydrogenase [NAD] (spot 58), involved in oxidative and substrate metabolism, were significantly increased only by T3 (Figure 5B). As a whole, these results suggest that, at the doses used, T2 and T3, although modulating overlapping metabolic events (e.g., amino acid and intra-mitochondrial energy metabolism), also have their own specific molecular targets.

To further characterize the effects elicited by either T2- or T3-treatment in liver of Ldlr^−/− mice, proteomic data were analyzed by using the IPA platform that, based on known interactions between affected proteins, defines common functional and canonical pathways as well as protein networks, thereby offering additional information about the complex interactive links between modulated proteins following the treatments under study. The in silico analysis confirmed that the most significant T2/T3- dependent changes altered lipid-, amino acid-, carbohydrate-, and energy- metabolism (Supplementary Material 2). These changes are mediated by effects on pathways such as glycolysis/gluconeogenesis, citrate cycle, pentose phosphate, glutathione, and amino acid metabolism (Figures 6A,B). Overall, in terms of modulated functions and pathways, T2 and T3 exerted a similar effect with a few exceptions. Of particular metabolic relevance the peroxisome proliferator-activated receptor α (PPARα)/retinoid X receptor α (RXRα) pathway was affected only by T3 (Figure 6B).

The protein network analysis for T2 produced the highest scored node (the value being 28) corresponding to the hepatocyte nuclear factor 4α (HNF4α), a nuclear receptor well known to act as a master regulator of liver-specific gene expression orchestrating lipid and cholesterol metabolism (Figure 7A). HNF4α is directly interconnected with some focus proteins acquired in 2D-E analysis such as aldehyde dehydrogenase (ALDH2, ALDH1), malate dehydrogenase (MDH1), and fatty acid-binding protein (FABP), which are involved in substrate and lipid metabolism (Figure 7A). Strictly in line with the pathways analysis, as far as T3 effects are concerned, network tool revealed PPARα as the main hub of the highest scored network (IPA score 30), a nuclear receptor that regulates diverse aspects of lipid metabolism, including fatty acid oxidation, and lipoprotein metabolism (Figure 7B). Within the network, PPARα directly interacts with some focus proteins acquired in 2D-E analysis such as aldehyde dehydrogenase (ALDH2), mitochondrial glycerol-3-phosphate dehydrogenase (GPD2), maleylacetoacetate isomerase (GSTZ1), major urinary protein 1 (MUP1), and peroxiredoxin-6 (PRDX6), all of which are involved in substrate and energy metabolism and in cellular stress (Figure 7B).

Considering that IPA network analysis highlighted HNF4α and PPARα as the highest scored nodes grouping the major number of interactions with differentially regulated proteins by T2 and T3, respectively, Western blot analysis was performed to further investigate the putative involvement of such nuclear factors into the hypocholesterolemic effects of both iodothyronines. Both T2 and T3 administration to Ldlr^−/− mice produced a significant reduction of the nuclear content of HNF4α, with the effect being slightly stronger in Ldlr^−/−+T2 than in Ldlr^−/−+T3 mice (Figures 8A,B). By contrast, only T3 administration produced a pronounced and significant decrease of nuclear PPARα levels vs. those in control Ldlr^−/− mice. According to what is suggested by the in silico analysis, this result is consistent with the hypothesis that PPARα has a specific role in the metabolic effects exerted by T3 in Ldlr^−/− mice.

In view of the fact that a partnership of HNF4α with its coactivator, peroxisome proliferator-activated receptor γ coactivator-1α (PGC-1α) in the regulation of lipoprotein metabolism has been characterized (Rhee et al., 2006), the effect of T2 and T3 on the hepatic nuclear content of PGC-1α was analyzed. As shown in Figures 8A,B, only T2 significantly reduced expression levels of PGC-1α, hence suggesting an impaired PGC-1α/HNF4α-dependent signaling in the plasma cholesterol-lowering effects of T2.

Nuclear receptors other than HNF4α and PPARα also participate in the transcriptional regulation of key factors involved in the intrahepatic cholesterol metabolism. Specifically, it has been reported that T3 exerts its hypocholesterolemic effect mainly by binding to liver TRβ but, in the absence of this, also acting through the liver X receptor α (LXRα) (Gullberg et al., 2000). Nuclear receptor expression levels may be predictive for the activity of their ligands. In view of this and considering that LXRα is a central hepatic cholesterol sensor, Western blot analysis was performed to verify whether the hypocholesterolemic effect of T2 and T3 in Ldlr^−/− mice correlates with LXRα and TRβ hepatic expression levels. In parallel with the lower plasma cholesterol levels detected in Ldlr^−/−+T2 and Ldlr^−/−+T3 mice (vs. Ldlr^−/−), both iodothyronines reduced LXRα nuclear content (Figures 8A,B). Of note, LXRα nuclear content tended to increase when Ldlr^−/− mice were consuming the WTD (Figures 8C,D). A significant reduction of TRβ levels (about −60% vs. Ldlr^−/− control mice) was observed only in Ldlr^−/−+T3 mice (Figures 8C,D).

In order to evaluate whether the absence/presence of LDLr might be a major determinant of the effects of T2 and T3 on the expression level of the aforementioned transcription factors, Western blot analyses were performed on liver tissue from WT mice fed either chow diet or WTD. The effects elicited by T2 on HNF4α, PGC-1α, and LXRα levels in livers from Ldlr^−/− mice were not observed in WT mice fed chow or WTD while T3 treatment, without significantly altering PPARα nuclear content, increased both HFN4α and PGC-1α hepatic levels, in WTD WT mice (Supplementary Material 3).

Finally, due to the large dose of T2 used, we sought to exclude displacement of T3 from thyroid binding globulin as a reason for the efficacy of the T2-treatment. To do this, we assessed serum FT3 and FT4 levels in T2 treated mice; neither was increased [the actual values being for FT3 (pg/ml): 8.55 ± 1.8; 3.2 ± 1.9^*; for FT4 (ng/dl): 1.24 ± 0.8; 0.15 ± 0.06^*, in WTD WT and WTD-WT+T2 animals, respectively; ^*P < 0.05 vs. WTD WT]. Therefore, T2 did not displace T3 from the thyroid binding globulin. Thus, increased endogenous T3 does not account for the effects of T2.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.