CS271011 was designed and synthesized at Chipscreen Biosciences (Shenzhen, Guangdong, China). The purity of CS271011 was over 99%. Resmetirom (MGL-3196) was purchased from MedChemExpress (Hycultec GmbH, Beutelsbach, Germany). Compounds were dissolved and suspended in sterile DMSO and 0.2% CMC-Na for in vitro and in vivo administration, respectively. The chow diet was supplied by Guangdong Medical Laboratory Animal Center (Foshan, Guangdong, China), and the high-fat diet (D12492i) was provided by SYSE Biotech Co., Ltd. (Changzhou, Jiangsu, China). Cholesterol (TC), alanine transaminase (ALT), aspartate transaminase (AST), alkaline phosphatase (ALP), and total bilirubin (TBIL) diagnostic kits were purchased from the Jiancheng Bioengineering Institute of Nanjing (Nanjing, Jiangsu, China). The Triglyceride (TG) Colorimetric Assay Kit was purchased from Elabscience (Wuhan, Hubei, China). All primers were synthesized by Genewiz (Suzhou, Jiangsu, China).

A luciferase reporter gene assay was applied to assess the capacity of drugs to bind and activate thyroid hormone receptor α or thyroid hormone receptor β. For each well, 5x103 CV-1 cells were seeded into 96-well plates and cultured in Dulbecco’s Modified Eagle Medium (DMEM; Hyclone, SH30022) containing 10% fetal bovine serum (FBS; Gibco, 16000–044) and 1% penicillin−streptomycin (P/S; Corning, 30-002-CI) for 24 hours at 37°C and 5% CO2. An expression plasmid mixture of THR-α or THR-β expression plasmids was transiently transfected into liposomes with a firefly luciferase reporter (2:2:1 thyroid hormone receptor α:TRE : GFP or 3:3:1 thyroid hormone receptor β:TRE : GFP) and mixed with the X-tremeGENE HP DNA Transfection Reagent (Roche, Basel, BS, Switzerland). The two different mixtures indicated above were added to the cells and incubated for 24 hours. Then, CS271011 (0.01-30 M), MGL-3196 (0.01-30 M), and T3 (0.000001-1 M) were diluted in DMSO and added to the transfected cells for 22 hours. Next, the medium was withdrawn, and 60 μl of 1X Cell Culture Lysis and 5X reagent (Promega, E153A) was added to each well for lysis. Subsequently, 50 μl of lysate was transferred into a 96-well white plate and measured on a Tecan microplate reader (Männedorf, Switzerland) to determine the GFP fluorescence value (Ex485/Em520). Finally, 30 μl of Luciferase Assay Buffer (Promega, E151A) was immediately added to each well to detect Luci values for 50% activation (AC50) calculation.

All animal experiments were performed at Chipscreen Biosciences (Shenzhen, China) according to the National Research Council’s Guide for the Care and Use of Laboratory Animals and the Regulations of Experimental Animal Administration issued by the State Committee of Science and Technology of China and approved by Shenzhen Peking University - The Hong Kong University of Science and Technology Medical Center’s Ethics Committee for the Welfare of Laboratory Animals (approval no. 2022-988). Forty C57BL/6J mice (4-5 weeks, male) were received and raised under a standard 12-hour light/dark cycle (lights off at 7:00 pm). Thirty-five mice were fed a fat-rich diet (SYSE Bio-tech, D12492i; 60% of kcal from fat; Changzhou, China), and the other five mice were fed a control diet (Guangdong Medical Laboratory Animal Center, Foshan, China) ad libitum for 12 weeks. Except for one mouse that died unexpectedly during the model construction, the other mice were randomized and divided into five groups at the beginning of the intervention: Group 1, chow diet control (n=5); Group 2, DIO control (n=8); Group 3, CS271011 1 mg/kg (n=9); Group 4, CS271011 3 mg/kg (n=9); and Group 5, MGL-3196 3 mg/kg (n=8). The mice from Group 2 to Group 5 were fed a high-fat diet. Treatments started on the thirteenth week and lasted 10 weeks, in which Groups 1 and 2 received solvent (0.2% CMC-Na), and the other groups were administered corresponding compounds orally once daily (the volume of solvent was 5 ml/kg). Body weight and food intake were measured once daily during treatment. All mice were sacrificed after treatment, and blood, hearts, and livers were collected for further validation. In addition, the dosage selection of CS271011 was based on the results of the luciferase reporter gene assay of THR-β and the experiment with MGL-3196 (3 mg/kg) applied in the DIO mouse model and DIO-NASH mouse model (15, 18).

Eighteen C57BL/6J mice (5-6 weeks, male) were randomly divided into two groups and treated with CS271011 3 mg/kg and MGL-3196 3 mg/kg once by gavage, and the volume of solvent was 5 ml/kg. Then, each group was subdivided into three groups for blood extraction: Group 1, sample collection after 0.5 hours and 2 hours of treatment (n=3); Group 2, sample collection after 1 hour and 4 hours of treatment (n=3); and Group 3, sample collection after 8 hours and 24 hours of treatment (n=3). Approximately 100 μl of blood was collected from the orbit and centrifuged (3500 g, 15 min, 4°C) to extract serum and placed on ice (0°C), and 20 μl of serum was added to 200 μl of extractant (containing 200 ng/ml methanol), which was assessed by liquid chromatography-tandem mass spectrometry (LC−MS/MS) procedures.

To detect the tissue distribution of drugs, the heart, liver, and kidney of mice in Group 1, Group 2, and Group 3 were collected at 2 hours, 4 hours, and 24 hours, respectively. Then, 1 ml of 1X PBS was added for tissue homogenization. 50 μl tissue homogenate was added to 500 μl of extractant (containing 200 ng/ml methanol). The protein of tissue samples was precipitated by vortex oscillation for 5 min and centrifuged at 13,000 rpm for 10 min, and 150 μL of supernatant was collected and determined by LC−MS/MS procedures. All the pharmacokinetic (PK) parameters of CS271011 and MGL-3196 were calculated by Excel 2019.

Blood samples were collected from the orbit and placed for 30 min at room temperature and then centrifuged (3500 g, 15 min, 4°C), and the serum was collected for TG, TC, AST, and ALT tests. For the liver TG and TC tests, the right lobe of the liver was collected and weighed accurately, and then isopropanol and absolute ethanol were added at a ratio of 1:9 (weight/volume) for tissue homogenization on ice. Next, the liver sample was centrifuged (10,000 g and 2500 g, 10 min, at 4°C for liver TG and TC extraction, respectively), and the supernatant was collected for liver TG and TC assays. Tests for TC, ALT and AST were performed following the manufacturer’s instructions.

After the animals were sacrificed, the left hepatic lobe was collected (approximately 100–200 mg) and fixed in 4% paraformaldehyde overnight. The liver tissue was sectioned at a thickness of 5 μm after being paraffin-embedded. Liver sections were stained with haematoxylin and eosin (H&E), and the NAFLD activity scores were determined based on the steatosis, lobular inflammation, and hepatocyte ballooning degeneration in accordance with the standard proposed by Kleiner (17).

To analyse the content of lipids in the liver, Oil Red O (ORO) staining of isobutyl alcohol-frozen liver sections was conducted by Wuhan Servicebio Technology Co., Ltd. (Wuhan, China). Each positive region of ORO staining was quantified for each sample using Imangepro plus 6.0 software (National Institute of Mental Health, USA).

The right hepatic lobe and heart tissues were randomly collected from three mice per group, stabilized overnight in RNAlater^® solution (Melunbio, Dalian, Liaoning, China), and stored at -80°C. RNA-seq was screened and analyzed by Tianjin Novogene Bioinformatic Technology Co., Ltd. (Tianjin, China). Volcano plot, GO, and KEGG enrichment analysis was performed on R software (version 4.2.1), and the UpSet plot was plotted by BioLadder (https://www.bioladder.cn/). Additionally, STRING (https://cn.string-db.org/) and Cytoscape (version v3.9.1) software were employed to study the protein−protein interaction (PPI) network, and the Gene Set Enrichment Analysis of DEGs was performed by using GSEA software (version 4.2.3).

The total RNA of fresh liver tissue was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA), and RNA quality was measured using a Nanodrop2000 by calculating the 260/280 ratio. Based on the manufacturer’s instructions, we reverse-transcribed 5 μg of isolated RNA with a Transcriptor First Strand cDNA Synthesis Kit (Roche Diagnostics, Mannheim, Germany). Real-time PCR (RT−PCR) was conducted to measure the relative expression of genes by using SYBR Green Master (ROX) dye (Roche Diagnostics, Mannheim, Germany) and the ABI Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA, USA). Changes in target mRNA were standardized to β-actin mRNA levels for each sample in triplicate. The 2-^△△Ct method was used to determine the relative mRNA expression levels. Supplementary Table 1 lists the primer sequences utilized in this investigation.

GraphPad Prism 8 (GraphPad Software, La Jolla, CA, USA) was applied to analyse all data in this study, and the data are represented as the mean ± standard deviation (SD). A 2-tailed Student’s t test compared significant differences between two groups, and one-way ANOVA was performed in three or multiple comparison analyses. P values were considered significant at p <0.05.

MGL-3196 was reported to have a 28-fold selectivity for THR-β over THR-α in a cell-free coactivator recruitment assay (13). To identify the potency and selectivity of CS271011, we conducted a cell-based luciferase reporter gene assay for both targets. The results showed that CS271011 reached the 50% activation (AC₅₀) of THR-α and THR-β with concentrations of 35.17 μM and 0.65 μM, respectively, whereas the result for MGL-3196 was 149.0 μM and 3.11 μM ( Table 1 ). Therefore, the activation ratios of THR-α and THR-β for CS271011 and MGL-3196 were 54.12 and 47.93, indicating that CS271011 showed higher THR-β activity over MGL-3196. Both CS271011 and MGL-3196 showed higher THR-β selectivity over T3, a negative control with robust potency in both isoforms but non-pronounced selectivity ( Figure 1A ). CS271011 also exhibited higher THR-α activity than MGL-3196, so we needed to calculate the safety window for both compounds. In addition, both drugs had a similar activity ascent interval, and CS271011 had stronger THRβ activity ( Figure 1B ).

We then studied the PK properties in normal C57BL/6J mice of a single administration of CS271011 and MGL-3196 at 3 mg/kg and 10 mg/kg by oral gavage ( Table 2 ). CS271011 exhibited favorable PK properties, including area under the curve (AUC), mean residence time (MRT), half-life (t_1/2), peak time (T_max), and peak concentration (C_max), in a dose-dependent fashion. The results showed that both doses of CS271011 highly accumulated in the liver than in other tissues, especially 2 and 4 hours after administration ( Figures 1C–E, H ). The liver/serum ratio of CS271011 was approximately 4:1 in both administrations, and the main ratio of liver/serum was approximately 26.2-fold and 5.96-fold higher than that of the heart and kidney, respectively ( Supplementary Table 2 ). MGL-3196 also demonstrated favourable PK properties and liver-targeted ( Figures 1C, F–H ) and presented higher bioavailability in vivo than CS271011 at the same dosage.

As illustrated in Figure 2A , mice were fed a chow diet or high-fat diet for 12 weeks and then received a 10-week consecutive treatment. The body weight and food intake of the animals were measured daily during treatment ( Figures 2B, C ). We observed a dramatic reduction in body weight in the MGL-3196-treated group but not in either CS271011 group ( Figure 2B ). However, neither CS271011 nor MGL-3196 impeded food intake in DIO mice ( Figure 2C ). The data suggest that CS271011 demonstrates a low impact on animal weight.

To explore the effect of CS271011 on ameliorating serum lipid content in DIO mice, serum TG and TC were measured ( Figures 3A, B ). DIO mice displayed 1.56-fold and 1.96-fold higher serum TG and TC concentrations compared with chow diet controls, indicating that the model was successfully established for further evaluation. After 10 weeks of treatment, CS271011 and MGL-3196 showed a significant reduction in serum TG and TC compared to the DIO control, whereas CS271011 dose-dependently decreased serum TC. Furthermore, we observed similar efficacy of CS271011 and MGL-3196 at the same dosage for both serum indices. We also observed that serum ALT and AST were not significantly elevated in DIO mice and the treatment group, suggesting that this model exhibited no severe hepatic injury, and no influence of CS271011 and MGL-3196 was observed on ALT and AST levels, suggesting that neither drug may cause liver damage ( Figures 3C, D ). Besides, serum ALP and TBIL level was tested for further validation of drug hepatotoxicity ( Supplementary Figure 1 ), no significant change was observed in different concentrations of CS271011 group compared to control group, suggesting that CS271011 exhibited hepatic safety.

After sacrifice, the mice were weighed, and the livers and hearts were removed and weighed to calculate the ratio of liver weight to body weight (LW/BW). As expected, DIO mice exhibited a larger liver volume ( Figure 4A ) and higher liver weight ( Figure 5A ) than those of chow diet controls. The LW/BW between control and model mice was not significantly different due to the higher increase in body weight of DIO mice ( Figure 5B ). After a 10-week drug intervention, both CS271011 and MGL-3196 dramatically reduced liver weight and LW/BW, in which CS271011 showed a dose-dependent and comparative effect with MGL-3196 at the same dose ( Figure 5A ). Moreover, CS271011 and MGL-3196 did not influence heart weight or the ratio of heart weight to body weight (HW/BW), suggesting minor adverse effects on the heart for both drugs ( Supplementary Figure S2) .

As the liver is responsible for lipogenesis and is mainly targeted by CS271011, the pathological changes in DIO mice, whether they were reverted by the agonist, are the gold standard for dyslipidaemia or NAFLD; thus, we qualified liver indices such as steatosis and ballooning by H&E and ORO staining ( Figure 4A ). The results showed that both CS271011 and MGL-3196 at the same dose (3 mg/kg) decreased the ORO staining area ( Figure 4B ) and steatosis score ( Figure 4C ), while 1 mg/kg CS271011 showed no significant effect in either assay. Both drugs at a dose of 3 mg/kg also mitigated hepatic ballooning with a considerable effect but not lobular inflammation ( Figure 4C ), while CS271011 1 mg/kg displayed no significant mitigation for either index. The average NAS score for the DIO model was approximately 5, which meets the criteria for mild NASH. CS271011 at 3 mg/kg but not 1 mg/kg dramatically reduced the NAS score in DIO mice with equivalent efficacy to that of the MGL-3196-treated group.

To confirm the role of CS271011 in hepatic lipid metabolism, we also performed assays to determine liver TG and TC levels ( Figures 5C, D ). As expected, CS271011 3 mg/kg along with MGL-3196 markedly reduced both hepatic TG and TC concentrations, while CS271011 1 mg/kg only significantly lowered liver TC. Together, at the same dose, CS271011 and MGL-3196 share comparative pathological effects on the liver in DIO mice.

To further explore the mechanism of CS271011 in lipid metabolism, we assessed the differentially expressed genes (DEGs) of the liver in different drug-treated groups by RNA-sequence analysis, and the distribution of DEGs in five groups is shown in the UpSet plot ( Figure 6A ). The standard for significance differences of DEG screening included an adjusted P value (padj) < 0.05 and log2FoldChange > 1. Compared with chow diet controls, DIO mice showed 55 upregulated and 35 downregulated genes, most of which were intimately associated with lipid localization, lipid transport, and metabolism pathways ( Supplementary Figures S3A, C and Figure S4A ). Compared with the DIO control group, mice treated with CS271011 1 mg/kg ( Supplementary Figures S3B, D and Figure S4B ), CS271011 3 mg/kg ( Figure 6C ), and MGL-3196 3 mg/kg ( Figure 6D ) showed significant changes in 142, 220 and 456 genes, respectively. GO and KEGG analyses showed that DEGs in the CS271011 3 mg/kg group were primarily enriched in steroid metabolic process, bile secretion, and cholesterol metabolism ( Figures 7A, C ). In comparison, the enrichment of significant DEGs in the MGL-3196 treatment group was mainly in the fatty acid metabolic process, steroid metabolism, and PPAR signaling pathway ( Figures 7B, D ). Gene Set Enrichment Analysis (GSEA) revealed the enriched DEGs in both groups. DEGs in the CS271011 3 mg/kg group were primarily enriched in steroid biosynthesis, the phosphatidylinositol signaling system, and cholesterol metabolism ( Figure 8A ), while the results in the MGL-3196 treatment group were primarily enriched in steroid biosynthesis, oxidative phosphorylation, and autoimmune thyroid disease ( Figure 8B ). In conclusion, the results of the difference in pathways between CS271011 and MGL-3196 group showed that CS271011 might prefer impact the metabolism of steroid and cholesterol, while MGL-3196 also impacted the fatty acid metabolism except for influencing the same pathway of CS271011. The results of GSEA in the DIO model and CS271011 1 mg/kg group are shown in Supplementary Figures S4C, D .

Then, the PPI network was applied to verify the top 10 cluster genes upon treatments, most of which in the CS271011 3 mg/kg group were involved in lipid and steroid metabolism, such as 3-hydroxy-3-methylglutaryl-coenzyme A reductase (Hmgcr), cholesterol 7α hydroxylase (Cyp7a1) and squalene epoxidase (Sqle), in which the first two genes encode rate-limiting enzymes for cholesterol and bile acid synthesis, and the latter one functions in steroid biosynthesis ( Figure 8C ). Interestingly, MGL-3196 diversely regulated these pathways but through different genes, such as cytochrome P450, family 2, subfamily e, polypeptide 1 (Cyp2e1) and cytochrome P450, family 3, subfamily a, polypeptide 11 (Cyp3a11) in lipid metabolism and hydroxy-delta-5-steroid dehydrogenase, 3 beta- and steroid delta-isomerase 2 (Hsd3b2) in steroid function ( Figure 8D ). The hub genes of the DIO and CS271011 1 mg/kg groups analyzed by the PPI network are shown in Supplementary Figures S4E, F .

Subsequently, we validated these dye-regulated genes by RT−PCR ( Figure 9 ). Genes such as cysteine sulfinic acid decarboxylase (Csad), solute carrier family 25 member 30 (Slc25a30), and trefoil factor family peptide 3 (Tff3), which were downregulated in DIO mice, were rescued by both CS271011 and MGL-3196. In contrast, lipogenesis genes such as sterol regulatory element binding transcription factor 1 (Srebf1) was strongly downregulated by both agonists, and cytochrome P450, family 4, subfamily a, polypeptide 10 (Cyp4a10) was obviously downregulated by MGL-3196, limiting hepatic lipid accumulation. Furthermore, both CS271011 and MGL-3196 positively regulated the expression of genes involved in cholesterol-related metabolism, such as low-density lipoprotein receptor (LDLR), Cyp7a1, cytochrome P450 family 17 subfamily A member 1 (Cyp17a1), sterol regulatory element binding transcription factor 2 (Srebp2), and Sqle, and negatively regulated ATP binding cassette subfamily C member 3 (Abcc3). In addition, in the regulation of steroid metabolism, both agonists upregulated the expression of cytochrome P450 family 39 subfamily A member 1 (Cyp39a1) and diminished the expression of hydroxysteroid 17-beta dehydrogenase 6 (Hsd17b6). Furthermore, the THR-related genes, such as thyroid hormone responsive (Thrsp), iodothyronine deiodinase 1 (Dio1), and malic enzyme 1 (Me1), were also tested for the validation of THR activation.

The RNA-seq of the heart indicated that both dosages of CS271011 caused no significant DEGs of the heart compared to the DIO model ( Supplementary Figure S5 ), and the expression of genes associated with myocardial energy metabolism, like mitochondrial glycerol 3-phosphate dehydrogenase (mGPDH) was validated by RT-PCR, showing no significant changes among all groups ( Figure 9E ). Furthermore, in the kidney cell line (HMC and NRK49F), both CS271011 and MGL-3196 exhibited no effect on cell viability, demonstrating that both compounds might be safe in kidney cells ( Supplementary Figure S6 ).