FHL3 promotes the formation of fast glycolytic muscle fibers by interacting with YY1 and muscle glycolytic metabolism

The proportions of the various muscle fiber types are important in the regulation of skeletal muscle metabolism, as well as animal meat production. Four-and-a-half LIM domain protein 3 (FHL3) is highly expressed in fast glycolytic muscle fibers and differentially regulates the expression of myosin heavy chain (MyHC) isoforms at the cellular level. Whether FHL3 regulates the transformation of muscle fiber types in vivo and the regulatory mechanism is unclear. In this study, muscle-specific FHL3 transgenic mice were generated by random integration, and lentivirus-mediated gene knockdown or overexpression in muscles of mice or pigs was conducted. Functional analysis showed that overexpression of FHL3 in muscles significantly increased the proportion of fast-twitch myofibers and muscle mass but decreased muscle succinate dehydrogenase (SDH) activity and whole-body oxygen consumption. Lentivirus-mediated FHL3 knockdown in muscles significantly decreased muscle mass and the proportion of fast-twitch myofibers. Mechanistically, FHL3 directly interacted with the Yin yang 1 (YY1) DNA-binding domain, repressed the binding of YY1 to the fast glycolytic MyHC2b gene regulatory region, and thereby promoted MyHC2b expression. FHL3 also competed with EZH2 to bind the repression domain of YY1 and reduced H3K27me3 enrichment in the MyHC2b regulatory region. Moreover, FHL3 overexpression reduced glucose tolerance by affecting muscle glycolytic metabolism, and its mRNA expression in muscle was positively associated with hemoglobin A1c (HbA1c) in patients with type 2 diabetes. Therefore, FHL3 is a novel potential target gene for the treatment of muscle metabolism-related diseases and improvement of animal meat production.


Introduction
Skeletal muscle is the most abundant tissue in the mammalian body, accounting for approximately 40% of body mass [1]. Skeletal muscle is composed of distinct muscle fiber types with different metabolic activities, which can be defined by myosin heavy chain (MyHC) isoforms. In adult mammals, muscle fibers are classified as four major types: slow oxidative (type I), fast oxidative (type IIa), fast glycolytic (type IIb), and fast oxidative-glycolytic (type IIx) muscle fibers, which express MyHC1/slow, MyHC2a, MyHC2b, and MyHC2x, respectively [2,3]. Slow and fast oxidative muscle fibers are rich in mitochondria and myoglobin and have high oxidative capacity and good endurance. In contrast, fast glycolytic muscle fibers have greater strength, lower endurance, fewer mitochondria, and generate adenosine triphosphate (ATP) primarily by glycolysis [4]. The composition of muscle fiber types is associated with muscle metabolism and metabolic diseases, as well as meat growth and the quality of agricultural animals. Individuals with a low maximal aerobic capacity and a high proportion of fast muscle fibers are at higher risk of obesity and related metabolic diseases [5]. Patients with abdominal obesity and insulin resistance (IR) show identical abnormalities in muscle morphology, namely a low percentage of slow oxidative fibers and an elevated proportion of fast-twitch muscle (particularly fast glycolytic) fibers [6,7]. In animal production, a high proportion of fast glycolytic muscle fibers enhances pig muscle mass and muscle growth, but increases the occurrence of PSE (pale, soft, and exudative) meat [8]. Therefore, elucidation of the regulatory pathways that affect fiber-type conversion would improve both the treatment of human muscle metabolic diseases and animal meat-production traits.
Several regulatory factors are involved in muscle fibertype transformation, including myocyte enhancer factor 2 (MEF2) [9], peroxisome proliferator-activated receptor gamma (PPARγ) [10], peroxisome proliferator-activated receptor gamma coactivator-1 (PGC-1) [11,12], calcineurin [13], calcineurin-nuclear factor of activated T cells (CaN-NFAT) [3,14], and the six-protein family [15]. FHL3 belongs to a family of LIM-only proteins with four-and-a-half LIM domains [16], and regulates myocyte differentiation, cytoskeletal structure, and skeletal muscle formation by interacting with other proteins [17][18][19]. We previously reported that FHL3 differentially regulates the expression of MyHC isoforms at the cellular level [20]. However, whether FHL3 regulates the conversion of muscle fiber types in vivo and the underlying regulatory mechanism is unclear. In this study, we generated musclespecific FHL3 transgenic mice and conducted lentivirusmediated gene knockdown and overexpression in the muscles of mice and pigs. Overexpression of FHL3 in muscles significantly increased the proportion of fast-twitch (type II) fibers and muscle mass. Mechanistically, FHL3 promoted the expression of the fast glycolytic MyHC2b gene by interacting with YY1. Moreover, upregulation of FHL3 increased IR, and its mRNA expression in muscle was significantly associated with HbA1c in type 2 diabetic patients. Our findings suggest a potential target for the treatment of metabolic diseases and the improvement of animal meat production.

Production and identification of transgenic mice
The gel-purified pEGFP-N1L-MCK-FHL3 linearized transgene construct (Fig. S1A) was microinjected into the pronuclei of fertilized C57BL wild-type (WT) mice eggs by standard techniques to produce transgenic (TG) mice [21]. Transgenic founder mice (F0) and their offsprings were identified by PCR using primers shown in Fig. S1. Primer sequences were as follows: F 5′-AGG AGA CAG CGA GTA GCG AGC TCT -3′ and R 5′-TGT CAT AGC ACG GAA CGC AGT-3′. Transgenic lines were established by breeding F0 mice with wild-type mice to obtain heterozygous F1 TG mice. After that, the transgenic mice were established successively. All the subsequent studies were performed on F5 mice. Animal use and care for this study were approved by Institutional Animal Care and Use Committee of Huazhong Agricultural University.

Phenotype measurement
TG mice and WT littermates were weaned at 3-week-old, separated by gender and raised in each cage of 3 mice. Mice were given free access to a chow diet (10% kcal fat, ME3.85 kcal/g). Body weights were measured at 2, 3, 4, 5, 6, 7, and 8 weeks of age. Mice were euthanized by cervical dislocation, and hind limb, Gas, TA, and Qu muscles of TG and WT mice were collected and weighed, separately. Data were normalized to the body weight (mg/g). The strength test was performed using a grip strength meter (BIO-GS3; Bioseb, France). The activities of lactate dehydrogenase (LDH) and succinate dehydrogenase (SDH) in muscle were measured with commercial kits (Jincheng Bioengineering Institute, Nanjing, China) according to the manufacturer's instructions. Oxygen consumption measurements were performed using TSE lab master systems (TSE Systems, BadHomburg, Germany) [22]. All mice were acclimatized for 24 h prior to measurements, and then, the volume O 2 was measured over the course of the next 24 h. Mice were maintained at 25 °C under a 12 h light/12 h darkness cycle with free access to food and water. Hematoxylin-eosin (H&E) staining of muscle sections was performed according to a previous report [23]. Histological images were visualized and captured by a light microscope. Paraffin sections of muscle for immunofluorescence staining were prepared in 0.01 M sodium citrate solution (pH 6.0) for 30 min at 70 °C, and permeabilizing in 0.1% Triton X-100, then incubated in blocking buffer (P0100B, Beyotime Biotechnology) 37 °C 2 h, and incubated with rabbit anti-dystrophin, (Abcam, UK, ab275391, 1:200) anti-slow myosin skeletal heavy chain (slow-twitch) (Sigam, USA, M8421,1:1000), anti-fast myosin skeletal heavy chain (fast-twitch) (Servicebio, China, GB112130, 1:1000), anti-MyHC2b (Invitrogen, USA,14-6503-80, 1:1000), and anti-Myosin (Abclonal, China, A4963, 1:200) overnight at 4 °C. Sections were washed in PBS, incubated with secondary antibody (Beyotime Biotechnology, China, anti-mouse CY3, anti-rabbit CY3, anti-rabbit FITC). The images were visualized with a fluorescence microscope (IX51-A21PH, Olympus, Japan). For the glucose tolerance test (GTT) test, 8-week-old male mice were fasted for 16 h, and then intraperitoneally injected with glucose (1.5 g/kg). Blood was collected from the tail vein at 0, 15, 30, 60, 90, and 120 min after an intraperitoneally injection. For the insulin tolerance test (ITT) test, 8-week-old male mice were fasted for 4 h, and then intraperitoneally injected with insulin (1.0 U/kg). Blood was collected from the tail vein at 0, 15, 30, and 60 min after injection of insulin. Blood glucose levels were measured using a glucometer.

Luciferase reporter assay
C2C12 myoblasts were transfected with MyHC2b regulatory region constructs and pcDNA3.1-FHL3 vector or YY1 siRNA by Lipofectamine 2000 (Invitrogen, USA), and differentiated for 2 days in a 24-well plate, washed with PBS, lysed in 100 μL of lysis buffer. The harvested cells were assayed for regulatory region activity using a dual luciferase reporter assay system (Promega, USA). The enzymatic activity of luciferase was measured using a PerkinElmer 2030 Multilabel Reader (PerkinElmer). To normalize the transfection efficiency, the cells were transfected with 0.04 μg of the Renilla luciferase reporter plasmid (pRL-TK, Promega, USA).

Total RNA extraction, reverse transcription, and quantitative real-time PCR (qRT-PCR)
Total RNAs were extracted using the Trizol reagent (Invitrogen, USA). The concentration and quality of RNA were assessed with a NanoDrop 2000 (Thermo, USA) and agarose gel electrophoresis. One μg of total RNA was used for reverse transcription with the PrimeScript RTreagent kit with gDNA Eraser (Takara, Japan). qRT-PCR analysis was performed using performed in a LightCycler 480 II (Roche, Switzerland) system. All primers used in the study were presented in supplementary information Primers Excel 2-sheet3. The relative RNA expression levels were calculated using the Ct (2 -ΔΔCt ) method.

DNA electrophoretic mobility shift assay (EMSA)
DNA EMSA was performed using a Chemiluminescent EMSA Kit (Beyotime Biotechnology, China, GS009) according to the manufacturer's instruction. Biotin-labeled DNA probes were generated by in vitro and purchased from AuGCT (Wuhan, China). Briefly, recombinant GST-YY1 or GST-YY1-DD or His-FHL3, and 1 µM of biotin-labeled DNA probe were mixed, and then separated in 10% of native poly acrylamide gel. DNA-protein complexes were blotted with HRP-conjugated streptavidin and the results were visualized by electrochemiluminescence (ECL).

Co-immunoprecipitation (Co-IP) assays
Co-IP assays were performed as previously described [20]. C2C12 myoblasts were seeded in 10-cm dishes and differentiated for 2 days. Cells were harvested and lysed in 1 ml lysis buffer (Sangon, Shanghai, China) with protease inhibitor (Sangon, Shanghai, China). The lysates were centrifuged to remove insoluble components and incubated with anti-FHL3 monoclonal antibody (Santa Cruz, USA; sc-166917), YY1 (Santa Cruz, USA; sc-281), or IgG antibody (Beyotime, Jiangsu, China) overnight at 4 °C in the presence of Protein A + G Agarose beads (Beyotime, Jiangsu, China) after removing 25 μL lysates as the input control. The beads were washed four times using lysis buffer. The proteins were analyzed by western blotting as described above.

Protein prokaryotic expression and GST pulldown assays
The recombinant GST, GST-FHL3, GST-YY1, GST-FHL3 truncated proteins, GST-YY1 truncated proteins, and His-FHL3, His-YY1, and His-EZH2 (465-519) were produced by auto-induction in E. coli BL21. E. coli BL21 were grown to an OD 600 of 0.5 at 37 °C in LB supplemented with 60 μg/mL ampicillin (GST-tagged and FHL3) or 60 μg/mL kanamycin (His-tagged). Protein product was induced with 0.1 mM IPTG for 5 h at 37 °C. All GST-tagged proteins were purified using a GST spin purification kit (Beyotime Biotechnology, P2262) according to the manufacturer's instruction. In vitro translated FHL3 proteins were loaded to Superdex S200 SEC column (GE Healthcare). All Histagged proteins were purified using a His spin purification kit (Beyotime Biotechnology, China, P2226) according to the manufacturer's instruction. GST pulldown assays were conducted according to the GST pulldown Assay Kit (Boxin, China, Bes3012), and then analyzed by western blotting.

Chromatin immunoprecipitation (ChIP) assay
ChIP assays were performed after transfection of the pcDNA3.1-FHL3 plasmid or siRNA FHL3 into C2C12 myoblasts using the ChIP Assay Kit (Beyotime, Jiangsu, China). Each ChIP assay was performed using 1 μg of antibodies against YY1 (Cell Signaling Technology, USA; mAb#46395; 1:50), EZH2 (Abcam, UK; ab3748; 1:100) and H3K27me3 (Abcam, UK; ab6002; 1:100). IgG was used as the negative control. qPCR was conducted using the retrieved DNA and the primers for ChIP to check the Fig. 1 Targeted overexpression of FHL3 in skeletal muscles by random integration significantly increases muscle growth and the proportion of fast-twitch muscle fibers. A Representative photographs of 2-month-old mice showing that FHL3 transgenic mice (TG) have a larger body and whiter meat color than wild-type (WT) mice. B The growth curve of male WT and TG mice showed that the body weight of male TG mice was significantly higher than that of male WT mice at the same week of age (n = 10 for each group). C, D Representative photographs of whole hind limb, quadricep (Qu), gastrocnemius (Gas), and tibialis anterior (TA) muscles of 2-month-old male WT and TG male mice (C). Quantification analysis showed that the weights of Qu, Gas, and TA muscles of male FHL3 TG mice were significantly higher than those of WT mice (n = 10 for each group) (D). Data were normalized to the body weight (BW) (mg/g). E Representative images of dystrophin immunohistochemistry staining for Qu, Gas, and TA muscles from 2-month-old FHL3 TG and WT mice. Quantification in ten independent experiments indicated that FHL3 mice had higher mean cross-sectional areas of individual myofibers than WT mice. At least 150 myofibers were analyzed in an independent experiment. Scale bars, 50 μm. F The exhaustive swimming time of 2-month-old mice showed that the swimming time of TG mice was shorter than that of WT mice (n = 10 for each group). G Grip strength of 2-month-old mice showed that the muscle grip strength of TG mice was stronger than that of WT mice (n = 10 for each group). The strength test was performed using a grip strength meter (BIO-GS3; Bioseb, France). H Representative immunohistology images of fast-twitch and slow-twitch muscle fiber types for Gas muscles from 2-month-old FHL3 TG and WT mice. Quantification in five independent experiments indicated that FHL3 TG mice had a higher proportion of fast-twitch muscle fibers and a lower proportion of slowtwitch muscle fibers than WT mice. Fast-twitch muscle fibers were indicated by red, slow-twitch muscle fibers were indicated by green, and DAPI was indicated by blue, respectively. At least 150 myofibers were analyzed in an independent experiment. Scale bars, 50 μm. I Representative immunohistology images of MyHC2b (type IIb) muscle fiber types for Gas muscles from 2-month-old FHL3 TG and WT mice. Quantification in five independent experiments indicated that FHL3 TG mice had a higher percentage of MyHC2b muscle fibers than WT mice. MyHC2b muscle fibers were indicated by red, myosin muscle fibers were indicated by green, and DAPI was indicated by blue, respectively. At least 150 myofibers were analyzed in an independent experiment. Scale bars, 50 μm. J Western blotting results showed that the expression levels of FHL3, MyHC2a, and MyHC2b in Gas muscles of TG mice were significantly increased, while the expression level of MyHC1/ slow was significantly decreased compared with WT mice (n = 6 mice for each group). The relative protein levels were normalized to β-actin. K, L Muscle enzyme activity of Gas muscles from 2-month-old mice showed that the LDH enzyme activity (K) of TG mice was significantly higher than that of WT mice, but the SDH enzyme activity (L) of TG mice was significantly lower than that of WT mice (n = 9). M The O 2 consumption (VO 2 ) and quantification results showed the TG mice had lower oxygen consumption than WT mice during the dark cycle (n = 5 for each group). Mice were put individually into a metabolic cage. The VO 2 was measured by built-in detector. Dot plot reflected the data from independent experiment. The data were presented as mean ± SD of independent experiments; *P < 0.05, **P < 0.01, ***P < 0.001 ◂ enrichments of YY1, EZH2, and H3K27me3 at target gene regions. The YY1, EZH2, and H3K27me3 ChIP data were presented as % of input. The ChIP primers amplification with MyHC2b specific primers (F: GCC ATA AGC CTG ACG CAG TA; R: CCC AGT GGT CCC CTA TCA AA).

Statistical analyses
All differences among groups were analyzed using unpaired or paired Student's t test. P < 0.05, P < 0.01, and P < 0.001 were statistical significance. All data are presented as mean ± standard deviation.

Targeted overexpression of FHL3 in skeletal muscles by random integration significantly increased muscle growth and the proportion of fast-twitch muscle fibers
To examine whether FHL3 regulates the conversion of muscle fibers in vivo, random integration technology was used to generate transgenic mice with muscle-specific FHL3 overexpression under the control of the muscle creatine kinase (MCK) promoter (Fig. S1A). Transgenic (TG) mice were identified by PCR analysis, and three F0 TG mice were obtained: two females (27♀, 51♀) and one male (46♂) (Fig. S1B). TG mice were mated with wild-type (WT) mice to produce offspring (Fig. S1C). In the FHL3 TG mice, the FHL3 copy number decreased in the generations from F1 to F5 mice, indicating an unstable copy number (Fig. S1D-F). F5-generation TG mice were identified by PCR analysis (Fig. S1G). To verify muscle-specific expression of exogenous genes, we analyzed FHL3 expression by western blotting in five tissues from 2-month-old TG and WT mice. Expression of FHL3 protein in skeletal muscle was significantly increased in TG mice compared with WT mice (Fig. S1H). Therefore, exogenous FHL3 was overexpressed in skeletal muscle of TG mice.
Morphological observations showed that 2-month-old TG mice had a larger body size than WT mice, and their skeletal muscle appeared whiter (Fig. 1A). The body weights of male and female TG mice were significantly higher than those of WT male and female mice, respectively ( Fig. 1B; S2A). FHL3 overexpression in muscles significantly increased the sizes and weights of the quadriceps (Qu), tibialis anterior (TA), and gastrocnemius (Gas) muscles in both male and female mice (Figs. 1C, D; S2B). H&E and immunofluorescence staining showed that TG mice exhibited a larger mean cross-sectional area of individual myofibers (Figs S2C; 1E). To investigate the effect of FHL3 on skeletal muscle strength, we evaluated exercise capacity. Interestingly, TG mice showed a shorter swimming time (Fig. 1F) and stronger muscle grip strength (Fig. 1G); therefore, FHL3 improves explosive exercise performance but decreases endurance exercise performance. Next, we investigated whether overexpression of FHL3 alters the composition of skeletal muscle fiber types. Immunofluorescence staining for muscle fiber types revealed that overexpression of FHL3 increased the percentage of fast-twitch myofibers in TG mice compared with WT mice (Fig. 1H), as well as the proportion of MyHC2b (type IIb) muscle fibers (Fig. 1I). Expression of MyHC1/slow, which encodes the slow-twitch type I myosin isoform, was significantly lower in TG mice. In contrast, the Lentivirus-mediated FHL3 knockdown in muscles significantly decreased muscle mass and the proportion of fast-twitch muscle fibers. A Injection diagram of the LV-shFHL3 and LV-shNC intramuscularly into the right and left legs of 1-month-old WT mice and representative photographs of hind limb, Qu, TA, and Gas muscles of 1-month-old mice showing the mice injected with LV-shFHL3 vector had redder meat color than mice injected with LV-shNC vector. The dose of lentivirus injection is shown in the schematic diagram. B Quantification of five independent experiments showed that lentivirus-mediated FHL3 knockdown significantly decreased the weights of hind limb, Qu, TA, and Gas muscles. P values are determined by paired t test. Data are normalized to the body weight (BW) (mg/g). C, D Representative images of dystrophin immunofluorescence (C) and H&E (D) staining for the Qua, TA, and Gas muscles of mice. Quantification in five independent experiments indicated that lentivirus-mediated FHL3 knockdown in muscles of mice significantly decreases the mean cross-sectional areas of individual myofibers. At least 150 myofibers were analyzed in an independent experiment. Scale bar, 50 μm. E Representative immunohistology images of fast-twitch and slow-twitch muscle fiber types for Gas muscles from 2-month-old mice injected with LV-shNC and LV-shFHL3 vectors. Quantification in five independent experiments indicated that lentivirus-mediated FHL3 knockdown in Gas muscles of mice significantly decreased the proportion of fast-twitch muscle fibers and increases the proportion of slow-twitch muscle fibers. Fast-twitch muscle fibers were indicated by red, slow-twitch muscle fibers were indicated by green, and DAPI was indicated by blue, respectively. At least 150 myofibers were analyzed in an independent experiment. Scale bars, 50 μm. F Representative immunohistology images of MyHC2b (type IIb) muscle fiber types for Gas muscles from 2-month-old mice injected with LV-shNC and LV-shFHL3 vectors. Quantification in five independent experiments indicated that lentivirus-mediated FHL3 knockdown in Gas muscles of mice significantly decreased the proportion of MyHC2b (type IIb) muscle fibers. MyHC2b muscle fibers were indicated by red, myosin muscle fibers were indicated by green, and DAPI was indicated by blue, respectively. At least 150 myofibers were analyzed in an independent experiment. Scale bars, 50 μm. G, H Western blotting (G) and qRT-PCR (H) results showed that lentivirus-mediated FHL3 knockdown in 2-month-old mice significantly decreased the expression levels of MyHC2a and MyHC2b, but significantly increased the expression level of MyHC1/slow. The relative protein and mRNA levels were normalized to β-actin. The data were presented as mean ± SD of independent experiments; P values were determined by paired t test. *P < 0.05, **P < 0.01, ***P < 0.001 ◂ 27 Page 10 of 24 expression of MyHC2a and MyHC2b, which encode fasttwitch myosin isoforms, was significantly greater in Gas muscles of 2-month-old TG mice than in WT mice (Fig. 1J).
Because slow oxidative fibers have more mitochondria and higher oxidative metabolic enzyme activities than glycolytic muscle fibers [4], we evaluated metabolism-related enzyme activity in WT and TG muscles. As expected, LDH activity was significantly increased and SDH activity was decreased in the Gas muscles of TG mice (Fig. 1K, L). TG mice also showed markedly lower oxygen consumption during the 12 h dark cycle than WT mice (Fig. 1M). Moreover, there was no significant difference in cardiac morphology between TG and WT mice (Fig. S2D, E). Therefore, FHL3 promotes the formation of fast-twitch myofibers and increases the activity of glycolytic metabolic enzymes in muscle.
To verify the effect of FHL3 on muscle mass and muscle fiber-type transformation, we injected LV-shNC and LV-shFHL3 intramuscularly into the left and right legs Qu, Gas, and TA muscles of 4-week-old WT mice ( Fig. 2A). Knockdown of FHL3 significantly decreased the size of the hind limb, Qu, Gas, and TA muscles ( Fig. 2A), as well as the weights of each of these muscle types (Fig. 2B). Immunofluorescence and H&E staining showed that the mean cross-sectional areas of individual myofibers of the Qu, Gas, and TA muscles were significantly decreased by FHL3 knockdown (Fig. 2C, D). Immunofluorescence staining for muscle fiber types showed that FHL3 knockdown significantly decreased the proportion of fast-twitch and MyHC2b (type IIb) muscle fibers in Gas muscle (Fig. 2E, F). Western blotting and qRT-PCR showed that the expression of MyHC2a and MyHC2b was significantly decreased by FHL3 knockdown, whereas that of MyHC1/slow was increased (Fig. 2G, H). These results show that FHL3 promotes the transformation from slow-to fast-twitch muscle fibers and muscle growth.

Pig FHL3 has the conserved function of promoting the formation of fast-twitch muscle fibers
To assess the conserved function of FHL3 among species, we compared the FHL3 amino acid sequences of pig, mouse, human, bovine, ovis, and gallus. The FHL3 amino acid sequences had high homologies; for example, the amino acid sequence similarity of pig and mouse FHL3 was 93.77% (Fig. S3A, B). To explore the role of pig FHL3 in muscle development, lentivirus-mediated pig FHL3 overexpression and siRNA-mediated pig FHL3 interference experiments were carried out in pig skeletal muscle satellite cells (pSMSCs). Immunofluorescence staining showed that overexpression pig FHL3 lentivirus (LV-pFHL3) significantly reduced the number of MyHC1/ slow-positive nuclei muscle fibers and increased the number of MyHC2a-and MyHC2b-positive nuclei muscle fibers compared with control lentivirus (LV-Control) (Fig. 3A). Western blotting showed that the overexpression of pig FHL3 significantly promoted the expression of MyHC2a and MyHC2b but decreased the expression of MyHC1/slow (Fig. 3B). We designed three pairs of siRNAs against pig FHL3; the third siRNA resulted in the greatest interference of pSMSC (Fig. S3C) and was selected for further studies. siRNA-mediated pig FHL3 interference (si-pFHL3) in pSMSCs increased the number of MyHC1/ slow-positive nuclei muscle fibers and decreased the number of MyHC2a-and MyHC2b-positive nuclei muscle fibers compared with negative control (NC), as demonstrated by immunofluorescence staining (Fig. S3D) and western blotting (Fig. S3E). To verify the effect of pig FHL3 on muscle fiber-type transformation in vivo, we injected the Fig. 3 Pig FHL3 has the conserved function of promoting formation of fast-twitch muscle fibers. A Representative images of MyHC2a, MyHC2b, and MyHC1/slow immunofluorescence staining in pig skeletal muscle satellite cells (pSMSC) differentiated for 2 days. Quantification in three independent experiments indicated that lentivirus-mediated pig FHL3 overexpression (LV-pFHL3) significantly increased the proportion of MyHC2a-and MyHC2b-positive myotubes, and reduced the proportion of MyHC1/slow-positive myotubes compared with lentivirus-mediated pig empty vector (LV-Control). Nuclei were stained with DAPI. Scale bars, 50 μm. B Western blotting results in pSMSC differentiated for 2 days showed that lentivirus-mediated pFHL3 overexpression significantly increased the expression levels of MyHC2a and MyHC2b, but significantly decreased the expression levels of MyHC1/slow. The relative protein levels were normalized to β-actin. The data were presented as mean ± SD of three independent experiments. C Injection diagram of the lentivirus-mediated pFHL3 overexpression (LV-pFHL3) vector and empty control (LV-Control) vector intramuscularly into the right and left legs of 1-week-old pigs and representative images of dystrophin immunofluorescence staining for the Gas muscles of pigs. Quantification in three independent experiments indicated that lentivirus-mediated pFHL3 overexpression in Gas muscles significantly increased the mean cross-sectional areas of individual myofibers. At least 150 myofibers are analyzed in an independent experiment. P values were determined by paired t test. Scale bar, 50 μm. D Representative immunohistology images of fast-twitch and slow-twitch muscle fiber types for Gas muscles from 5-week-old pigs injected with LV-Control and LV-pFHL3 vectors. Quantification in three independent experiments indicated that lentivirus-mediated FHL3 overexpression in Gas muscles of pigs significantly increased the proportion of fast-twitch muscle fibers and decreased the proportion of slowtwitch muscle fibers. Fast-twitch muscle fibers were indicated by red, slow-twitch muscle fibers were indicated by green, and DAPI was indicated by blue, respectively. At least 150 myofibers were analyzed in an independent experiment. P values were determined by paired t test. Scale bars, 50 μm. E Representative immunohistology images of MyHC2b (type IIb) muscle fiber types for Gas muscles from 5-weekold pigs injected with LV-Control and LV-pFHL3 vectors. Quantification in three independent experiments indicated that lentivirusmediated FHL3 overexpression in Gas muscles of pigs significantly increased the proportion of MyHC2b muscle fibers. MyHC2b muscle fibers were indicated by red, myosin muscle fibers were indicated by green, and DAPI was indicated by blue, respectively. At least 150 myofibers were analyzed in an independent experiment. P values were determined by paired t test. Scale bars, 50 μm. F Western blotting results show that lentivirus-mediated pFHL3 overexpression (LV-pFHL3) in pig muscles significantly increased the expression levels of MyHC2a and MyHC2b, but significantly decreased the expression level of MyHC1/slow. The data were presented as mean ± SD of three independent experiments; P values were determined by paired t test. The relative protein levels were normalized to β-actin. *P < 0.05, **P < 0.01 ◂ LV-pFHL3 and LV-Control intramuscularly into the left and right leg Gas muscles of 1-week-old pigs (Fig. 3C). Immunofluorescence staining showed that the mean individual myofiber cross-sectional area of pig Gas muscle was significantly increased by pig FHL3 overexpression (Figs. 3C, S3F). Also, FHL3 overexpression in pig muscles significantly decreased the percentage of slow-twitch muscle fibers and increased that of fast-twitch and MyHC2b muscle fibers (Fig. 3D, E). Western blotting showed that the overexpression of pig FHL3 significantly increased the expression of MyHC2a and MyHC2b and inhibited that of MyHC1/slow (Fig. 3F). In conclusion, these results confirm that FHL3 has the conserved function of promoting the formation of fast-twitch glycolytic myofibers in pigs.

Bioinformatics analysis of differentially expressed genes in WT and FHL3 knockout C2C12 cells
FHL3 regulates the expression of MyHC2a and MyHC1/slow by interacting with CREB and MyoD, respectively [20]. However, how FHL3 promotes the formation of fast glycolytic muscle fibers is unclear. To investigate the mechanisms by which FHL3 promotes the formation of fast glycolytic muscle fibers, we designed two sgRNAs to knock out FHL3 in C2C12 cells by CRISPR/Cas9 technology (Fig. S4A). FHL3 knockout (KO) cells were obtained by using flow cytometry and DNA sequencing. Two sgRNAs (sgRNA-1 and sgRNA-2) efficiently knocked out FHL3 (Fig. S4A, B). sgRNA-1 knockout cell line was used for subsequent studies. Compared with WT cells, the number of MyHC1/ slow-positive muscle fibers in KO cells was significantly increased, whereas that of MyHC2a-and MyHC2b-positive muscle fibers was significantly decreased (Fig. S4C); these findings were confirmed by the results of western blotting (Fig. 4A). RNA-seq analysis of WT and FHL3 knockout cells differentiated for 4 days identified a total of 3436 differentially expressed genes (DEGs) (adjusted P value < 0.05, |FC|> 1.5), comprising 1389 upregulated genes and 2047 downregulated genes in KO cells (Fig. 4B). Gene ontology (GO) analysis showed enrichment of transition of fast and slow fibers, skeletal muscle cell differentiation, regulation of striated muscle contraction, ATP metabolic process, and glycoprotein metabolic process (Figs. 4C, S4D). Gene set enrichment analysis (GSEA) indicated that genes related to the transition from the fast to slow fiber pathway were significantly upregulated by FHL3 KO (Fig. 4D). The DEGs heatmap showed that the expression of genes related to fast-twitch fibers was decreased, whereas that of genes related to slow-twitch fibers was increased significantly (Fig. 4E). qRT-PCR results confirmed reduced expression of DEGs associated with fast-twitch fibers, such as Myh1, Myh2, Myh4, Bdnf, Igfn1, and Sox6, as well as increased expression of DEGs associated with slow-twitch fibers, such as Myh7, Tnni1, Tnnc1, Myl3, Myoz2, and Actc1 (Fig. 4F). Interestingly, FHL3 knockout did not affect the expression of some genes that regulate the expression of fast-twitch muscle genes, such as MyoD, IGF1, MEF2C, and CaN-NFAT pathway-related genes (Nfatc1, Nfatc2, and Ppp3r1) (Fig. S4E). Because FHL3 does not directly bind DNA and regulates target gene expression by interacting with other proteins [16], we identified the protein partners of FHL3 that regulate MyHC2b expression. We used BART online software (http:// bartw eb. org/) to analyze DEGs and found 87 predicted regulatory factors (Supplementary Excel 1: sheet1). Next, we identified the potential binding motifs of transcription factors in the MyHC1/slow, MyHC2a, and MyHC2b regulatory region using online software (PROMO, http:// alggen. lsi. upc. es/ cgi-bin/ promo_ v3/ promo/ promo init. cgi? dirDB= TF_8.3) (Supplementary Excel 2: sheet 2-4), and identified 12 potential specific transcription factors in MyHC2b regulatory region (Fig. S4F), including NF-1, CUTL1, Sox2, YY1, AP-1, and ERR1 (Supplementary Excel 2: sheet5). Overlapping analysis of 12 transcription factors and 87 regulatory factors showed that YY1 was the common factor, suggesting that FHL3 may promote MyHC2b expression via YY1 (Fig. 4G).

YY1 directly binds to the MyHC2b regulatory region and specifically represses its expression
To determine whether the transcriptional activity of MyHC2b is affected by YY1, we generated eight truncated fragments driving the transcription of the luciferase reporter gene in the regulatory region between − 2000 bp and + 100 bp relative to the translation start site. Then, we detected the effects of si-YY1 on the luciferase activity of eight truncated fragments. The results showed that the transcription activity of the region containing nucleotides from − 1110 bp to − 910 bp (D5-D7) was significantly improved by si-YY1 (Fig. S5A), implying that there may be the binding motifs of YY1 in this region. To assess whether FHL3 regulates MyHC2b expression by YY1, we further examined the effect of FHL3 overexpression on the luciferase activity of eight truncated fragments, and the results showed that the transcription activity of the D5-D7 region was also significantly improved by FHL3 overexpression (Fig. 5A). Online software analysis (PROMO) of the nucleotides sequence from − 1110 bp to − 910 bp revealed two potential YY1 binding motifs. We mutated the two YY1 binding motifs and conducted luciferase reporter assays. The results showed that Fig. 4 Bioinformatics analysis of differentially expressed genes between WT and FHL3 knockout C2C12 cells revealed that YY1 may participate in the regulation of the formation of fast glycolytic muscle fibers by FHL3. A Western blotting results showed that FHL3 knockout significantly decreased the expression levels of MyHC2a and MyHC2b, but significantly increased the expression levels of MyHC1/slow in C2C12 cells differentiated for 4 days. B Volcano plot of differentially expressed genes (DEGs) between WT and FHL3 KO C2C12 cells. Blue: downregulated DEGs, red: upregulated DEGs. C GO enrichment dot plot of the differentially upregulated genes after FHL3 knockout. D Gene set enrichment analysis (GSEA) of DEGs related to 'transition between fast and slow fiber.' E Heatmap of differentially expressed fast-twitch and slow-twitch related genes. F qRT-PCR results showed that the expression levels of fast-switch related genes were significantly decreased, and the expression levels of slow-switch genes were significantly increased in FHL3 knockout cells compared with WT cells. G Veen diagram showing that YY1 was the common gene between 12 potential transcription factors that specifically bind to MyHC2b regulatory region by PROMO analysis and 87 regulatory factors predicted by RNA-seq DEGs BART analysis. The relative protein and mRNA levels were normalized to β-actin. The data were presented as mean ± SD of three independent experiments; P values were determined by paired t test. *P < 0.05, **P < 0.01, ***P < 0.001 ◂ the overexpression of FHL3 did not increase fluorescence activity when binding motif A was mutated (Fig. 5B). ChIP results showed that YY1 bound to the regulatory region of MyHC2b in C2C12 cells differentiated for 2 days (Fig. 5C). EMSA was conducted to determine whether purified GST-YY1 directly binds to the MyHC2b regulatory region in vitro. A DNA-protein complex was detected after the addition of GST-YY1 to MyHC2b probe, but not for GST or free H 2 O with MyHC2b motif A probe (Fig. 5D). These results suggested that the GST-YY1 fusion protein directly bound to the MyHC2b regulatory region (Fig. 5D). To confirm the role of YY1 in MyHC2b expression, we performed overexpression and knockdown experiments in C2C12 cells. Immunofluorescence staining showed that overexpression of YY1 significantly reduced the number of MyHC2bpositive nuclei muscle fibers but did not affect the number of MyHC2a-or MyHC1/slow-positive nuclei muscle fibers (Fig. 5E). Western blotting showed that overexpression of YY1 significantly downregulates MyHC2b expression but had no significant effect on the expression of MyHC2a and MyHC1/slow (Fig. 5F). These results were confirmed by YY1 knockdown experiments (Fig. S5B, C). Therefore, YY1 represses the expression fast glycolytic muscle fiber gene MyHC2b by binding to its regulatory region.

FHL3 directly interacts with YY1 and decreases its binding to the MyHC2b gene regulatory region
To evaluate how FHL3 regulates MyHC2b expression via YY1, we co-transfected FHL3 and YY1 into C2C12 myoblasts differentiated for 4 days. siRNA-mediated mouse YY1 interference (si-YY1) significantly inhibited the promotion of MyHC2b expression by FHL3 overexpression (Fig. 6A), whereas siRNA-mediated mouse FHL3 interference (si-FHL3) significantly increased the inhibitory effect of YY1 overexpression on MyHC2b expression (Fig. 6B), suggesting that FHL3 affects the transcriptional activity of YY1 on MyHC2b. However, there was no significant change in the expression of YY1 after FHL3 knockout ( Fig. S4E; S5D), and western blotting results showed that si-FHL3 did not affect the protein expression of YY1 (Fig. S5E). Therefore, FHL3 may regulate the expression of MyHC2b by interacting with YY1. Indeed, immunofluorescence staining in C2C12 cells differentiated for 2 days showed that FHL3 (red) protein colocated with YY1 (green) in the nucleus (Fig. 6C). Co-IP also indicated that FHL3 interacted with YY1 in C2C12 cells differentiated for 2 days in vivo (Fig. 6D, E). To investigate whether FHL3 interacts with YY1 directly, we expressed and purified fusion proteins of GST with different truncated LIM domains of FHL3 according to the FHL3 LIM structure in vitro. GST pulldown assays showed that in vitro translated His-YY1 protein was immunoprecipitated by GST-FHL3 fusion protein, but not by GST, indicating that FHL3 directly interacts with YY1 (Fig. 6F). Also, each LIM domain of FHL3 directly interacted with His-YY1 in vitro (Fig. 6F) [27,28] were generated and purified in vitro (Fig. 6G). We then performed GST pulldown assays using the YY1 truncated domains and in vitro translated His-FHL3 protein and His-EZH2 (465-519). EZH2 (465-519), which binds to the YY1-RD region, was used as the positive control (Fig.  S6A). YY1-RD (170-225), YY1-DD (294-414), and fulllength YY1 (1-414) interacted with FHL3 (Fig. 6H). GST pulldown of YY1-DD truncated domains confirmed that FHL3 directly interacts with the zinc finger domain ZN2 Fig. 5 YY1 directly binds to MyHC2b gene regulatory region and specifically represses its gene expression. A Luciferase reporter assay in C2C12 cells differentiated for 2 days indicated that the FHL3 overexpression significantly increased the transcription activities of the D5, D6, and D7 truncated fragments containing two putative YY1 binding motifs. Left panel, schematic diagram of eight truncated fragments of MyHC2b gene regulatory region linked to luciferase gene in pGL3 vector. The nucleotides were numbered relative to the translation start site that was assigned as + 1. Schematic diagram of two potential YY1 binding motifs (A, B) in truncated fragments of MyHC2b gene regulatory region were displayed. Right panel, the relative activities of a series of truncated fragments of the pGL3-MyHC2b construct determined by luciferase assay after transfection of FHL3 overexpression vector (pcDNA3.1-FHL3). B Schematic diagram of two YY1 binding sites (A, B) in MyHC2b gene regulatory region between nucleotides − 1100 bp and − 910 bp which were predicted by PROMO and mutated motifs. The dual luciferase assays showed that overexpression of FHL3 did not significantly enhance the fluorescence activity of MyHC2b D5 truncated fragment only when motif A was mutated. C ChIP-qPCR results showed that YY1 could bind to MyHC2b gene regulatory region in C2C12 cells differentiated for 2 days. IgG was performed as controls and precipitated DNA was amplified by PCR with primers for the MyHC2b gene regulatory region. D EMSA assays were used to analyze the binding of GST, free H 2 O, and GST-YY1 fusion protein to MyHC2b regulatory regions in vitro. GST, and free H 2 O were used for negative control and blank control, respectively. The specific DNA-protein complex band was observed in the WT probes incubated with GST-YY1. E Representative images of MyHC1/slow, MyHC2a, and MyHC2b immunofluorescence staining and quantification in three independent experiments indicated that YY1 overexpression (pcDNA3.1-YY1) significantly decreased the proportion of MyHC2b-positive myotubes, and did not significantly affect the proportion of MyHC1/slow-and MyHC2a-positive myotubes in C2C12 cells differentiated for 4 days. The myotubes were stained with anti-MyHC1/slow antibodies (red), anti-MyHC2a antibodies (red), anti-MyHC2b antibodies (red), and DAPI (blue). Scale bars, 50 μm. F Western blotting results showed that YY1 overexpression significantly decreased the expression levels of MyHC2b, and had no significant effects on the expression levels of MyHC1/slow and MyHC2a in C2C12 cells differentiated for 4 days. The relative protein levels were normalized to β-actin. The data were presented as mean ± SD of three independent experiments; *P < 0.05, **P < 0.01, N.S. indicates statistical non-significance ◂ of YY1-DD (Fig. 6I). EMSA showed that the ZN1, ZN2, and ZN3 domains directly bound to the MyHC2b regulatory region motif A probe, and the binding abilities of the ZN2 and ZN3 domains to the probe were significantly stronger than that of the ZN1 domain (Fig. 6J). Therefore, the ZN2 domain of YY1 mediates its binding to FHL3 and the MyHC2b regulatory region, implying that FHL3 prevents the binding of YY1 to its targets. To verify this, we performed ChIP experiments in C2C12 cells differentiated for 2 days after FHL3 overexpression or knockdown. FHL3 overexpression significantly decreased the enrichment of YY1 at the MyHC2b regulatory region (Fig. 6K), whereas FHL3 knockdown significantly increased the enrichment of YY1 at the MyHC2b regulatory region (Fig. S6B). Next, we performed EMSA experiments using the His-FHL3 fusion protein to verify the influence of FHL3 on the direct binding of YY1-DD to the MyHC2b regulatory region. GST-YY1-DD formed a DNA-protein complex with MyHC2b probe, and the addition of His-FHL3 reduced a complex and super-shift band (Fig. 6L). Therefore, FHL3 decreases YY1 binding to the MyHC2b regulatory region by competitively interacting with the DNA-binding domain of YY1.

FHL3 reduces EZH2 recruitment to the regulatory region of MyHC2b by competitively interacting with the RD2 domain of YY1
YY1 recruits EZH2 to induce trimethylation of lysine 27 on histone H3 (H3K27me3) at target regions [29]. In this study, GST pulldown showed that YY1-RD mediates the binding of FHL3 and EZH2 (Fig. 6H, S6A), indicating that FHL3 and EZH2 competitively bind to YY1-RD. YY1-RD has a HAT/HDAC interaction domain (YY1-RD1) and a REPO domain (YY1-RD2) (Fig. 6G), and YY1-RD2 is necessary for PcG-dependent transcriptional Fig. 6 FHL3 directly interacts with ZN2 domain of YY1 and decreases its binding capacities to MyHC2b gene regulatory region. A C2C12 myoblasts were co-transfected with pcDNA3.1-FHL3 and YY1 siRNA (indicated at the bottom) and differentiated for 4 days. The cell lysates were subject to western blotting with anti-FHL3, anti-YY1 and anti-MyHC2b (indicated at the left). Western blotting results showed that knockdown of YY1 could reduce the promotion effect of FHL3 overexpression on MyHC2b expression. The relative protein levels were normalized to β-actin. The data were presented as mean ± SD of three independent experiments; *P < 0.05, **P < 0.01. B C2C12 myoblasts were co-transfected with FHL3 siRNA and pcDNA3.1-YY1, and then differentiated for 4 days. The cell lysates were subject to western blotting with anti-FHL3, anti-YY1, and anti-MyHC2b. Western blotting results showed that FHL3 knockdown could increase the inhibitory effects of YY1 overexpression on MyHC2b expression. The relative protein levels were normalized to β-actin. The data were presented as mean ± SD of three independent experiments; *P < 0.05, ***P < 0.001. C Immunofluorescence staining showed that FHL3 protein co-located with YY1 in cell nucleus of C2C12 myoblasts differentiated for 2 days. C2C12 myoblasts were stained with anti-FHL3 (red), anti-YY1 (green), and DAPI (blue), and were imaged by confocal laser scanning microscopy. Scale bars, 5 μm. D, E The co-immunoprecipitation (Co-IP) results showing the interaction of FHL3 and YY1 in vivo. C2C12 myoblasts were differentiated for 2 days, harvested and co-immunoprecipitated with anti-FHL3 or anti-IgG (D) and with anti-IgG or anti-YY1 (E). IgG was used as a negative control. The total cell lysates before immunoprecipitation were used as input to verify expression of FHL3 and YY1. at the MyHC2b regulatory regions were significantly decreased after FHL3 overexpression in C2C12 myoblasts differentiated for 2 days. IgG was performed as controls, and the precipitated DNA was amplified by PCR with primers for the MyHC2b gene regulatory region. *P < 0.05. L EMSA results showing the effect of the FHL3 on the binding capacity of YY1-DD to the MyHC 2b gene regulatory region in vitro. GST, and free H 2 O were used as negative and blank control, respectively. A specific DNA-protein complex band was observed in the wild-type probes incubated with GST-YY1-DD, while the DNAprotein complex band became weaker and a super band formed when His-FHL3 was added repression [27]. Therefore, we performed GST pulldown assays of YY1-RD1 and YY1-RD2. Both FHL3 and EZH2 interacted with YY1-RD2 (Figs. 7A, S6C). Co-IP in C2C12 cells indicated that the binding of YY1 to EZH2 was inhibited by FHL3 overexpression (Fig. 7B) but increased by FHL3 knockdown (Fig. 7C). GST pulldown showed that the addition of purified FHL3 fusion protein decreased the interaction of His-EZH2 fusion protein with GST-YY1 in vitro (Fig. 7D). Therefore, FHL3 decreases EZH2 recruitment at YY1 binding sites by competitively interacting with YY1-RD2, thereby increasing target gene expression. To investigate whether FHL3 affects EZH2-mediated histone modification, we used an antibody against H3K27me3 for ChIP-qPCR in C2C12 cells transfected with pcDNA3.1-FHL3 or FHL3 siRNAs. As expected, FHL3 overexpression in differentiated C2C12 cells significantly decreased the binding of EZH2 and H3K27me3 to the MyHC2b regulatory regions (Fig. 7E,  F), while the binding of EZH2 and H3K27me3 to the Fig. 7 FHL3 reduces EZH2 recruitment to regulatory regions of MyHC2b by competitively interacting with YY1-RD2 domain. A GST pulldown results indicated that GST-YY1-RD1 and GST-YY1-RD2 could interact with His-FHL3 in vitro. In vitro translated proteins of GST, GST-YY1-RD1, GST-YY1-RD2, GST-YY1-RD, and His-FHL3 were used for GST pulldown assays. The interaction of GST-YY1 truncated proteins with His-FHL3 was detected by western blotting with anti-His (up). The GST-tagged proteins were detected by western blotting with anti-GST (down). B, C The Co-IP results showed that FHL3 overexpression (B) and knockdown (C) significantly decreased and increased the interaction of EZH2 with YY1 in vivo, respectively. C2C12 myoblasts differentiated for 2 days were co-immunoprecipitated with anti-YY1 or anti-IgG. The immunoprecipitated proteins were subject to western blotting with anti-FHL3, anti-YY1, and anti-EZH2 (indicated at the left). IgG was used as a negative control. The total cell lysates before immunoprecipitation were used as input to verify expression of FHL3, EZH2, and YY1. D The GST pulldown results indicated that addition of FHL3 fusion protein reduced the interaction His-EZH2 (465-519) with YY1 in vitro. In vitro translated proteins of GST, GST-YY1, His-EZH2 (465-519), and FHL3 were used for GST pulldown assays. The FHL3 fusion protein, GST, GST-YY1, and His-EZH2 (465-519) were detected by western blotting with anti-FHL3, anti-GST, and anti-His, respectively. E, F ChIP-qPCR results showed the EZH2 (E) and H3K27me3 (F) enrichments at the MyHC2b regulatory regions were significantly decreased after FHL3 overexpression in C2C12 myoblasts differentiated for 2 days. IgG was used as negative control, and the precipitated DNA was amplified by PCR with primers for the MyHC2b regulatory regions. G, H ChIP-qPCR results showed the EZH2 (G) and H3K27me3 (H) enrichments at the MyHC2b regulatory regions were significantly increased after FHL3 knockdown in C2C12 myoblasts differentiated for 2 days. IgG was used as negative control, and the precipitated DNA was amplified by PCR with primers for the MyHC2b regulatory regions. The data were presented as mean ± SD of three independent experiments; **P < 0.01, ***P < 0.001 ◂ MyHC2b regulatory region was significantly increased after FHL3 knockdown (Fig. 7G, H). In conclusion, FHL3 competes with EZH2 to interact with YY1-RD2 and reduces H3K27me3 enrichment at regulatory regions of MyHC2b gene.

FHL3 reduces glucose tolerance and its expression is positively correlated with HbA1c in patients with type 2 diabetes
The transformation of the skeletal muscle fiber type is accompanied by a change in metabolic function. Oxidative fibers have high mitochondrial content and strong aerobic oxidative metabolism, and require more oxygen. In contrast, glycolytic fibers mainly engage in anaerobic fermentation. RNA-seq of FHL3 KO cells showed that the expression of pkfkb4, kbfkb2, and pdk4 (related to glycolysis) was decreased, whereas that of genes associated with the TCA cycle, and oxidative phosphorylation (OXPHOS) (e.g., Cox5a, ATP5c1, EGR1, Acaa2, Gcdh) was increased (Fig. 8A). This result was confirmed by qRT-PCR (Fig. 8B). Therefore, knockdown of FHL3 enhanced muscle aerobic metabolism. Interestingly, KEGG analysis of DEGs showed that the pathways related to diabetic cardiomyopathy, glucagon signaling, PI3K-Akt signaling, and MAPK signaling were enriched, implicating FHL3 in diabetes (Fig. 8C). Next, we investigated the effect of FHL3 on glucose homeostasis. At 2 months of age, WT and TG mice were fed standard chow and subjected to the intraperitoneal GTT and ITT. FHL3 TG mice exhibited reduced glucose tolerance and higher IR than WT mice (Fig. 8D, E). Skeletal IR is important in metabolic diseases such as obesity, type 2 diabetes, and hypertension [30,31]. Therefore, we analyzed microarray data of 12 healthy controls and 12 diabetic patients (GSE 29221) to assess the correlation between FHL3 and diabetes. FHL3 mRNA expression in muscle was significantly upregulated by 1.9-fold in diabetic patients compared with healthy controls (P = 0.0467) (Fig. 8F). In addition, we analyzed the relationship between the expression of FHL3 and HbA1c in 42 diabetic patients (GSE 202295). The mRNA expression of FHL3 in muscle was significantly and positively correlated with HbA1c in diabetic patients (r = 0.43, P = 0.0041) (Fig. 8G), implicating FHL3 in diabetic IR.

Discussion
FHL3 differentially regulates the expression of MyHC isoforms in C2C12 myoblasts [20], but its function in skeletal muscle fiber-type transformation in vivo and muscle metabolism-related diseases is unclear. FHL3 is highly and specifically expressed within fast-twitch muscle fibers [32]. In this study, we generated TG mice with muscle-specific FHL3 overexpression. The growth rate, muscle mass, and the proportion of fast-twitch muscle fibers of TG mice were significantly higher than those of WT mice. Composition of muscle fiber types is significantly associated with muscle growth and meat quality of agricultural animals [33]. Muscle mass is mainly determined by the number of muscle fibers, the cross-sectional area of muscle fibers, and the composition of muscle fiber type. An increase in fast glycolytic muscle fiber numbers leads to muscle fiber thickening, muscle hypertrophy, and greater muscle mass [34,35], but results in whiter meat color and the decline of meat pH value after slaughter [36]. TG mice have a higher proportion of fasttwitch muscle fibers and a larger muscle fiber cross-sectional area than WT mice. Lentivirus-mediated knockdown of FHL3 in mouse muscle and overexpression of pFHL3 in pig muscle confirmed that FHL3 increases the proportion of fast-twitch muscle fibers and the cross-sectional area of muscle fibers in vivo. Therefore, upregulation of FHL3 is an important approach to improve animal meat production. However, FHL3 overexpression may also have a negative effect on meat quality such as meat color and pH value, and further experiments are needed to verify this effect. Also, pathways related to diabetic cardiomyopathy, glucagon signaling, PI3K-Akt signaling, and MAPK signaling pathway were enriched in DEGs, observed by RNA-seq. FHL3 TG mice displayed the increased number of glycolytic fibers and LDH activity, and the decreased number of slow oxidative fibers and SDH activity; FHL3 TG mice also exhibited the reduced glucose tolerance and the higher IR than WT mice. Skeletal muscle is the largest insulin-sensitive tissue, and about 80% of glucose is used by skeletal muscle [37]. Muscle fiber type affects whole-body physiology and metabolism [38,39]. Individuals with type 2 diabetes mellitus or obesity have more glycolytic fibers and fewer oxidative fibers than healthy individuals [40,41]. Fiber-type distribution is closely related to glucose uptake and IR in human [6]. The increase of fast-twitch muscle has an important impact on IR in skeletal muscle, because the glycolytic fibers have lower oxygen consumption and lower glucose uptake capacity than the oxidative fibers [42], while slow oxidative fibers have a higher activity of oxidative metabolic enzymes and oxidative capacity than glycolytic muscle fibers [43]. Interestingly, published microarray and RNA-seq data showed that FHL3 was significantly upregulated in diabetic patients and positively correlated with HbA1c in type 2 diabetics, respectively. HbA1c, which is used to monitor the effects of Fig. 8 FHL3 reduces glucose tolerance and its expression is positively correlated with blood glucose values in patients with type 2 diabetes. A Heatmap of DEGs associated with glycolysis, TCA cycle, and OXPHOS. B qRT-PCR results showed that the expression levels of glycolysis-related genes were significantly decreased, and the expression levels of TCA cycle and OXPHOS-related genes were significantly increased in FHL3 knockout cells. C KEGG pathway of DEGs between WT and FHL3 KO C2C12 cells. D GTT results of WT and TG male mice showed the clearance of glucose from the circulation during GTT was significantly lower in TG mice compared with WT mice (n = 8 for each group). E ITT results of WT and TG male mice showed the clearance of glucose from the circulation during ITT was significantly lower in TG mice compared with WT mice (n = 8 for each group). F Volcano map of DEGs between normal people and diabetic patients (GSE: 29221) showed that the mRNA expression of . FHL3 directly interacts with YY1-ZN2, weakening their binding to the MyHC2b regulatory region and thus upregulating MyHC2b gene expression. Moreover, FHL3 competes with EZH2 to bind YY1-RD2, and decreases H3K27me3 enrichment at the MyHC2b regulatory region, thereby increasing target gene expression. The relative mRNA levels were normalized to β-actin. The data were presented as mean ± SD of three independent experiments; *P < 0.05, **P < 0.01 ◂ diabetes treatments, reflects the mean blood glucose level in the past 2-3 months. Clinically, it is an index to evaluate the long-term control of blood glucose, and its expression accurately and indirectly reflects the blood glucose level of diabetics [44][45][46][47]. We speculated that FHL3 reduced muscle aerobic oxidative metabolism by upregulating the proportion of fast glycolytic muscle fibers, and increased IR and the HbA1c level. However, it should be pointed out that IR is a complex physiological process, and is also affected by obesity, exercise, impaired insulin signaling pathway, glucose metabolism dysfunction, etc. [48]. For example, exercise can promote glucose intake and increase insulin sensitivity, thereby reducing the occurrence of insulin resistance [49]. Therefore, individuals with large muscles (body builders) do not necessarily suffer from insulin resistance.
Skeletal muscle fiber-type remodeling is regulated by multiple signaling pathways, including NFATc1 [14] and PGC1a [11]. We identified two novel pathways by which FHL3 promoted the formation of fast glycolytic muscle fibers. FHL3 regulates fast glycolytic muscle fiber gene expression by interacting with YY1 and decreasing its repressive activity. YY1 is a myogenic transcriptional repressor that recognizes a core 5′-CCATNTT-3′ sequence flanked by flexible nucleotides [50]. YY1 as an activator or repressor of gene expression binds DNA via four C-terminal zinc finger domains [27,51]. YY1 represses the synthesis of late-stage differentiation genes by directly binding to the regulatory region or enhancer elements of MyHC2b [29]. However, the effect of YY1 on MyHC2b expression and its molecular mechanism is unclear. We found that YY1 specifically inhibits MyHC2b expression but does not affect the expression of MyHC2a and MyHC1/ slow. The ZN1, ZN2, and ZN3 regions of YY1 mediate its binding to the MyHC2b regulatory region. FHL3 and YY1 colocalized in the nucleus and each LIM domain of FHL3 protein could bind to the YY1 domain. In addition, FHL3 bound to the YY1-RD and YY1-ZN2 regions of YY1 simultaneously. FHL3 reduced the binding of YY1 to the MyHC2b regulatory region by interacting with the ZN2 domain of YY1, attenuating the inhibitory effect of YY1 on MyHC2b expression. FHL3 also decreased EZH2 recruitment at the MyHC2b regulatory region by competitively interacting with the RD2 domain of YY1. EZH2 epigenetic regulation increased the H3K27me3 level, a hallmark of gene repression [52], and the binding of EZH2 to chromatin is dependent on YY1 [28,53]. We propose that FHL3 disrupts the YY1-EZH2 interaction and reduces the H3K27me3 level at the MyHC2b regulatory region, thereby promoting target gene expression.
In summary, FHL3 is a determinant of glycolytic muscle fiber type and muscle metabolism, as well as IR and glucose tolerance. Mechanistically, FHL3 directly interacts with YY1-DD, weakening its binding to the MyHC2b regulatory region and thus upregulating fast glycolytic fiber genes. Moreover, FHL3 competes with EZH2 to bind YY1-RD, and decreases H3K27me3 enrichment at the MyHC2b regulatory region, thereby increasing target gene expression (Fig. 8H). Finally, FHL3 reduces glucose tolerance, and its expression is positively associated with HbA1c in patients with type 2 diabetes.
Author contribution BZ and ZX contributed to the study conception and design; WB and YZ are the main authors of the experiment; Bioinformatics analyses were performed by JM, YH, XZ,WJ and SL; MD, HX, JW, LZ, WL, XL, XZ, XL, QL, YM, MS, YY, and SZ participated in the experiments; YP and JL designed sgRNAs in CRISPR/Cas9 genome-editing. The manuscript was written by WB and YZ, BZ, ZX, and MD. All the authors read and approved the final manuscript.

Data availability
The RNA-sequencing data in this study have been deposited in the Gene Expression Omnibus database (accession number: GSE213996).

Conflict of interest
The authors declare no competing interests.
Ethics approval All animal studies were conducted in strict accordance with the Guide for Care and Use of Laboratory Animals of Laboratory Animal Centre, Huazhong Agriculture University, and all experiments conform to the relevant regulatory standards. The experiments and protocols were approved by the Animal Management and Ethics Committee of Huazhong Agriculture University (Assurance number HZAUMO-2020-0053).