The intragenic microRNA miR199A1 in the dynamin 2 gene contributes to the pathology of X-linked centronuclear myopathy

Mutations in the myotubularin 1 (MTM1) gene can cause the fatal disease X-linked centronuclear myopathy (XLCNM), but the underlying mechanism is incompletely understood. In this report, using an Mtm1−/y disease model, we found that expression of the intragenic microRNA miR-199a-1 is up-regulated along with that of its host gene, dynamin 2 (Dnm2), in XLCNM skeletal muscle. To assess the role of miR-199a-1 in XLCNM, we crossed miR-199a-1−/− with Mtm1−/y mice and found that the resultant miR-199a-1-Mtm1 double-knockout mice display markers of improved health, as evidenced by lifespans prolonged by 30% and improved muscle strength and histology. Mechanistic analyses showed that miR-199a-1 directly targets nonmuscle myosin IIA (NM IIA) expression and, hence, inhibits muscle postnatal development as well as muscle maturation. Further analysis revealed that increased expression and phosphorylation of signal transducer and activator of transcription 3 (STAT3) up-regulates Dnm2/miR-199a-1 expression in XLCNM muscle. Our results suggest that miR-199a-1 has a critical role in XLCNM pathology and imply that this microRNA could be targeted in therapies to manage XLCNM.

Mutations in the myotubularin 1 (MTM1) gene can cause the fatal disease X-linked centronuclear myopathy (XLCNM), but the underlying mechanism is incompletely understood. In this report, using an Mtm1 ؊/y disease model, we found that expression of the intragenic microRNA miR-199a-1 is up-regulated along with that of its host gene, dynamin 2 (Dnm2), in XLCNM skeletal muscle. To assess the role of miR-199a-1 in XLCNM, we crossed miR-199a-1 ؊/؊ with Mtm1 ؊/y mice and found that the resultant miR-199a-1-Mtm1 double-knockout mice display markers of improved health, as evidenced by lifespans prolonged by 30% and improved muscle strength and histology. Mechanistic analyses showed that miR-199a-1 directly targets nonmuscle myosin IIA (NM IIA) expression and, hence, inhibits muscle postnatal development as well as muscle maturation. Further analysis revealed that increased expression and phosphorylation of signal transducer and activator of transcription 3 (STAT3) up-regulates Dnm2/miR-199a-1 expression in XLCNM muscle. Our results suggest that miR-199a-1 has a critical role in XLCNM pathology and imply that this microRNA could be targeted in therapies to manage XLCNM.
Centronuclear myopathies (CNMs) are a series of fatal muscle diseases with muscular pathology characterized by atrophic muscle fibers, disordered sarcomere organization, and a large number of centronuclear myofibers (1). Among these diseases, X-linked CNM is associated with the most severe form of myopathies, and most patients die in the first year after birth due to respiratory failure of muscle weakness (2). Although these diseases have long been characterized, no clinical treatment is available as yet (1).
The DNM2 protein is encoded by the 114-kb DNM2 gene, which contains 22 exons, and four intragenic miRNAs are located in different introns (17). DNM2 is ubiquitously expressed in different tissues under physiological conditions (18). In CNM, normal DNM2 protein is up-regulated in diseased muscles (19,20). Overexpression of WT DNM2 causes CNM phenotypes (8,21), such as centralized nuclei, muscle atrophy, and deformed T-tubules, whereas down-regulation of DNM2 expression inhibits CNM (19,20). There are reports that downregulated GTPase of DNM2 could protect against CNMs (22,23). Alternatively, as the intragenic miRNAs of DNM2 are transcribed along with its host gene and regulated related biology process (24,25), we proposed that miR-199a, a uniquely conserved intragenic miRNA of Dnm2 among species, is central to the myopathy process.
To address the model that miR-199a is a contributor to XLCNM, we used a mouse model with deletion of the Mtm1 gene (Mtm1 Ϫ/y ) as an XLCNM model and assessed the role of the intragenic miRNA miR-199a-1 in myopathy. We found that Dnm2 mRNA and miR-199a-1 were simultaneously elevated in This article contains supporting information. ‡ These authors contributed equally to this work. * For correspondence: Zhu Min-Sheng, zhums@nju.edu.cn; Chen Hua-Qun, chenhuaqun@njnu.edu.cn; Zhang Xue-Na, zhangxn@nicemice.cn.

cro ARTICLE
Mtm1 Ϫ/y skeletal muscle. Deletion of miR-199a-1 is suggested to attenuate the severity of myopathy in Mtm1 Ϫ/y mice, as evidenced by increased muscle mass, improved muscle strength, and prolonged animal lifespan. Our analyses suggest that miR-199a-1 targeted nonmuscle myosin IIA (NM IIA) and thereby inhibited muscular postnatal development as well as muscle maturation. In addition, we found that the overexpression of DNM2/miR-199a-1 was regulated by ectopic Stat3 activation in XLCNM muscle. Our study revealed an important role for an intragenic microRNA in XLCNM pathological progression.

Expression of Dnm2 and intragenic miRNAs are elevated in Mtm1 ؊/y skeletal muscle
Using CRISPR technology, we established two mouse lines with a 5-bp (Mtm1 ⌬5/y ) and 7-bp (Mtm1 ⌬7/y ) deletion within the Mtm1 gene (Fig. 1A). Note that Mtm1 is encoded in the X chromosome and the Y chromosome does not encode a Mtm1 allele, and the mutant male mice were referred to as Mtm1 Ϫ/y . These mutant mice showed a retarded body growth and smaller muscle sizes (Fig. 1, B and C), which were comparable with those reported previously (26). We used these lines as a XLCNM disease model.
The Western blotting assay showed a higher level of Dnm2 in Mtm1 Ϫ/y tibialis anterior (TA) muscle in contrast to the control (Fig. 1D). As miR-199a is the intragenic microRNA of the Dnm2 gene and evolutionarily conserved across species (Fig. 1E), we measured miR-199a-1 expression level in the mutant TA muscles. Interestingly, the miR-199a-1 level was also significantly elevated ( Fig. 1, F and G). These results indicated a simultaneous elevation of Dnm2 and miR-199a-1 expression in the myopathy muscle. Overexpression of Dnm2/miR-199a-1 is present in XLCNM mice. A, truncated genomic sequence and protein of Mtm1 surrounding the targeted exon 3. The red marker reflects the mutant region, and the blue letters show the premature stop codon. aa, amino acid. B, comparisons of body and muscle size between 5-week-old Mtm1 ⌬5/y and WT mice. C, comparisons of body and muscle size between 6-week-old Mtm1 ⌬7/y and WT mice. D, Western blotting for Mtm1 and Dnm2 in lysates prepared from 3-week-old WT or Mtm1 ⌬5/y or Mtm1 ⌬7/y tibialis anterior (n ϭ 4). E, the presence of intragenic miR-199a in the Dnm2 gene was conserved among different species. Hsa, human; Mml, Rhesus monkey; Ptr, chimpanzee; Ssc, pig; Bta, cattle; Cfa, dog; Mmu, mouse. F and G, quantitative PCR to detect the expression of Dnm2 mRNA and miR-199a-1 in D (n ϭ 3-4). Graphs represent mean Ϯ S.D. (error bars). *, p Ͻ 0.05; **, p Ͻ 0.01 (two-tailed Student's t test).

miR-199a-1 regulates X-linked centronuclear myopathy Ablation of miR-199a-1 improves the myopathy of Mtm1 ؊/y mice
To assess the role of miR-199a-1 in XLCNM pathogenesis, we first established miR-199a-1 deletion lines of mice ( Fig. 2A). The birth of mutant pups followed an expected Mendelian ratio. The mice were fertile and reached adulthood without any obvious abnormalities (Fig. 2, B-D). Western blotting assays showed no apparent alteration of Dnm2 protein expression after miR-199a-1 deletion (Fig. 2, E and F).
We next measured the contractile response of the mutant EDL to electric stimulation (27). Mtm1 ⌬5/y muscle displayed a typical contraction action, but the maximal force was significantly less than that of WT muscle (3.44 Ϯ 0.47 g versus 0.45 Ϯ 0.06 g, p Ͻ 0.01). The Homo-1 muscle, in contrast to the Mtm1 ⌬5/y muscle, showed a significant increase in maximal force tension (0.99 Ϯ 0.16 g versus 0.45 Ϯ 0.07 g, p Ͻ 0.01) (Fig.  3G). Because the muscle masses were different among these muscles, we normalized the values and found that the maximal force of Homo-1 muscle was significantly higher than that of the Mtm1 ⌬5/y muscle (1.95 Ϯ 0.25 N/cm 2 versus 1.20 Ϯ 0.14 N/cm 2 , p Ͻ 0.05), but still much lower than that of the WT control (1.95 Ϯ 0.25 N/cm 2 versus 3.73 Ϯ 0.33 N/cm 2 , p Ͻ 0.01) (Fig. 3H). The phenotype was similar in Mtm1 ⌬7/y and Homo-2 muscle (Fig. S1, G and H). This result showed a functional improvement of the muscle after the deletion of miR-199a-1 in Mtm1 Ϫ/y mice.

miR-199a-1 inhibits myoblast fusion by targeting nonmuscle myosin IIA
Whereas XLCNM muscle has no apparent defect of myogenesis during embryonic development (26), other reports suggested developmental defects for this muscle (28). We isolated single myofiber from 3-week-old Mtm1 Ϫ/y mice and observed fewer nuclei in the muscle also (  . B, whole-body weights of the indicated genotype mice were monitored from 3 to 8 weeks of age (n ϭ 13-23). Results of WT mice are also presented in Fig. 2B. C, the relative muscle weight (muscle mass/body weight) of these mice was also measured (n ϭ 5-8). The data of WT mice are also presented in Fig. 2C. D, muscle pathological sections of the indicated mice. Scale bar, 20 nm. E, the frequency of myofibers with central nuclei (n ϭ 5). F, the myofiber ratio of the indicated genotype mice was grouped according to fiber area (n ϭ 5). G, typical EDL contraction curve induced by electric stimulus. H, quantification of maximal forces in G (n ϭ 4 -7). All mice were sacrificed at 6 weeks of age. SOL, soleus. Graphs represent mean Ϯ S.D. (error bars). NS, no significant difference; *, p Ͻ 0.05; **, p Ͻ 0.01 (two-tailed Student's t test).
To identify the target gene of miR-199a-1, we used the bioinformatics tools TargetScan (RRID:SCR_010845), PicTar (RRID: SCR_003343), and DIANA TOOLS (RRID:SCR_018425) to predict candidate target genes. Given that deletion of miR-199a-1 led to improvement of myopathic muscle, the function of its target gene was expected to promote myoblast fusion or similar processes. Following this rationale, we identified four candidate target genes, MYH9, SULF1, RAD23B, and PPARGC1A, for miR-199a-5p. We transfected miR-199a-5p into C2C12 cells and measured the target proteins by Western blotting. Among these proteins, only NM IIA expression, the product of Myh9 translation, was inhibited by overexpression of miR-199a-5p (Fig. 6 (A and B) and Fig. S3A).
To assess whether NM IIA was involved in the XLCNM, we first detected its expression in Mtm1 Ϫ/y muscle. As we expected, a striking reduction in NM IIA was measured in these mice ( Fig. 6L and Fig. S3B). As NM IIA participated in myofibrillogenesis that might reflect myotube fusion and muscle maturation (31), we examined the muscle fibers by staining for ␣-actinin. The WT fibers showed clear and well-organized sarcomeres, whereas Mtm1 ⌬5/y sarcomeres showed misalignment of Z-lines, as reported previously (19). NM IIA was abundantly co-localized with ␣-actinin in WT muscle, but its abundance was greatly decreased in the Mtm1 ⌬5/y sarcomeres. However, in Homo-1 sarcomeres, there appeared to be an improvement of the Z-line arrangement, together with an increase in the amount of NM IIA (Fig. 6M). These results indicated that the impaired maturation of the mutant muscles was attributable to the down-regulation of NM IIA by miR-199a-1.

Stat3 signaling contributes to the ectopic expression of the Dnm2 gene and miR-199a-1
To investigate the regulatory mechanism of overexpression of Dnm2/miR-199a-1 in XLCNM muscle, we analyzed the regulatory region of the Dnm2 gene with JASPAR (RRID: SCR_003030). We found conserved signal transducer and activator of transcription 3 (Stat3) binding sites upstream of the Dnm2 gene. Given the important role of Stat3 during myoblast fusion (32), we hypothesized that the abundant Stat3 cis-elements located within the region might contribute to the up-regulation of Dnm2/miR-199a-1 (Fig. 7A). To address the potential Stat3 regulation of the Dnm2/miR-199a-1 locus, we first measured Stat3 signaling in XLCNM muscle and found a significant elevation of the levels and phosphorylation of Stat3 (Pho-Stat3) compared with that of WT littermates (Fig. 7B). We then made a series of luciferase reporters containing Dnm2

miR-199a-1 regulates X-linked centronuclear myopathy
promotor regions with either intact or mutant Stat3-binding sites (Fig. 7C). Upon transfecting HEK 293T cells with the reporters, we found that luciferase activity was significantly reduced in the Stat3-binding site mutant groups compared with the intact group (ϳ50%; Fig. 7C). Furthermore, we treated C2C12 myoblasts with Stat3 siRNA, measured the expression of Dnm2/miR-199a-1, and found that miR-199a-1 level was decreased by ϳ50% with reduced Dnm2 expression (Fig. 7, D-F). Considering that phosphorylation is essential for Stat3 activation, we applied the specific inhibitor (5,15-DPP) of Pho-Stat3 to mice injured by BaCl 2 and found a predicted reduction in Dnm2/miR-199a-1 level and an increase in NM IIA expression (Fig. 7, G-I). These results indicate that Stat3 activation drives Dnm2/miR-199a-1 expression in muscle.
Collectively, aberrant activation of Stat3 in XLCNM muscle induced expression of the Dnm2 gene and intragenic miR-199a-1, which inhibited myoblast fusion and muscle maturation during early postnatal development via direct down-regulation of NM IIA (Fig. 8).

Discussion
Mutation of Mtm1 causes multiple skeletal muscle pathologies, including growth retardation, skeletal muscle wasting, and other severe complications (2). As observed in previous studies, dysfunction of the Mtm1 gene is the leading cause of XLCNM (33). In this report, we found that, during the pathological process, the DNM2 gene, along with its intragenic miR-199a-1, was secondarily up-regulated in XLCNM muscle. Importantly, inhibition of miR-199a-1 expression attenuated the severity of the myopathy and, hence, prolonged lifespan significantly. As miR-199a-1 was able to apparently inhibit the process of myoblast fusion both in vitro and in vivo, we suggested that, during the process of myopathy, the elevated miR-199a-1 functioned as an inhibitor of myoblast fusion and myofiber maturation. In other words, the myopathy phenotype of XLCNM is partially attributable to the ectopic expression of miR-199a-1. Based on the collective results, we suggest a working model for XLCNM pathological progression, in which the Mtm1 mutation causes activation of Stat3 signaling and hence promotes Dnm2/miR-199a-1 expression, which inhibits myoblast fusion and muscle maturation. It was reported that up-regulation of miR-199a was present in muscle of eight major muscular disorders (34). The pathogenesis of miR-199a may therefore extend to other myopathies or muscle injuries. It is noted that even in XLCNM, the severer form of myopathy mice, inhibition of miR-199a-1 is still able to prolong lifespan by ϳ30% as reported here. This implies that targeting miR-199a-1 might shows more efficacy in the case of other myopathy with relative mild phenotypes.
miR-199a has been demonstrated to be involved in several biological processes (35)(36)(37) in which several genes are directly targeted. Based on our observations, nonmuscle myosin IIA serves as a new target of miR-199a-1 in the process of muscular

miR-199a-1 regulates X-linked centronuclear myopathy
development. During myofibrillogenesis, NM IIA is predominantly expressed at nascent premyofibrils and then disappears in mature myofibrils, showing a featured turnover of muscle and nonmuscle myosin molecules (31). As the reduction in NM IIA impaired sarcomere organization in XLCNM muscle, we suggest that the miR-199a-1-regulated NM IIA might function in muscle myosin turnover during myofibril maturation also. In addition, as the nonmuscle tissues are able to express NM IIA along with DNM2, it is reasonable to predict that the downregulation of NM IIA by miR-199a-1 might also contribute to the pathology of nonmuscle tissues in XLCNM patients, who showed a high percentage of nonmuscular comorbidity with ear and kidney (2). The similar phenotypes observed in patients with NM IIA mutations also support this speculation (38,39).

miR-199a-1 regulates X-linked centronuclear myopathy
ously (46). Briefly, a sgRNA (5Ј-GTAACTCCCCTGGGAGCC-GA-3Ј) targeting the third exon of Mtm1 was cloned into the pUC57-U6-sgRNA vector with the BbsI cleavage site. Then the sgRNA was transcribed and purified in vitro, whereas the Cas9 mRNA was purchased from the Nanjing Biomedical Research Institute of Nanjing University. Next, the sgRNA and Cas9 mRNA were co-injected into the pronucleus and cytoplasm of C57BL/6 zygotes, respectively. Then the manipulated embryos were implanted into the oviducts of pseudopregnant female mice to generate chimeras. Mice with a 5-/7-bp deletion of Mtm1 in the germline were backcrossed for at least 6 generations into the C57BL/6 background for subsequent research. The generation of Mir199a-1 KO mice was similar to that of Mtm1-deficient mice with an sgRNA for miR-199a-5p (GAA-CAGGTAGTCTGAACATC).

Intramuscular injection operation
The acute muscle injury model was generated by injection of BaCl 2 (50 l of 1.2% (w/v) in saline) into the TA of one hind limb as described previously (47); another TA with saline (50 l) treatment was used as a control. For the miR-199a-5p oligonucleotide operation, the acute injury model was first induced in two hind limbs of adult C57BL/6 mice by BaCl 2 . On the second day, the regenerative muscle of one hind limb was injected with agomir or antagomir (1 nmol, RiboBio Ltd. Co., Guangzhou, China). Meanwhile, the introduction of Scrambled (1 nmol; RiboBio) into another regenerative TA was used as a control. The injections were consecutively performed at intervals of 1 day until the TA muscle was isolated and subjected to experimental measurements, within a week. In addition, 5,15-DPP (2.5 nmol; D4071, Sigma-Aldrich), a specific Stat3 inhibitor, was introduced into acutely injured TA in a similar manner as the miR-199a-5p oligonucleotides, and an equal volume of DMSO was used as a control.

Cell culture
All cell lines were purchased from MuCyte Ltd. Co. (Nanjing, China). C2C12 myoblasts and HEK 293T cells were maintained in Dulbecco's modified Eagle's medium (12100046, Gibco) supplemented with 10% fetal bovine serum (growth medium (GM)) at 37°C and 5% CO 2 . Myoblasts were induced to differentiate after the cells reached ϳ90% confluence by replacing the GM with Dulbecco's modified Eagle's medium containing 2% horse serum (differentiation medium or DM).

Plasmid construction and luciferase assays
To construct miR-199a-5p expression vectors, a region including miR-199a-1 was amplified from C57BL/6 mouse genomic DNA by PCR. The PCR fragment was digested with BglII and SalI and then inserted into pIRES2-EGFP (BD Biosciences Clontech) using the same cutting sites. For the pGL3-Myh9-3Ј-UTR vector, a 1.2-kb fragment of the Myh9 3Ј-UTR including putative miR-199a-5p-binding sites was amplified from C57BL/6 mouse cDNA by PCR. The PCR product was digested with SpeI and cloned into the firefly luciferase gene downstream of the pGL3-promoter vector (Promega), which was digested with XbaI. The mutation vector of the miR-199a-5p-binding site of the Myh9 3Ј-UTR (pGL3-Myh9-3Ј-UTRmut) was constructed using a MutanBEST Kit (D401, TaKaRa) with PCR. Similarly, 2 kb upstream of the Dnm2 gene, including the WT or B1/B2 mutant Stat3-binding site, was inserted into the multiple-cloning site of the pGL3-Basic vector (Promega). The primers are listed in Table S1.
HEK 293T cells were seeded in plates and cultured overnight. The transfection procedures were performed using LipoMax or Lipofectamine 2000 according to the manufacturer's protocol (32012, Sudgen; 11668019, Invitrogen). After 24 h, the cells were collected and analyzed using the Dual-Luciferase reporter system (E1910, Promega). The Renilla luciferase reporter plasmid (pRL-TK, E2241, Promega) was used as an internal control. Relative luciferase activity was calculated as the ratio of firefly luciferase to Renilla luciferase.

Oligonucleotide transfection
C2C12 myoblasts were transfected with 100 nM negative control oligonucleotides (199-C), miR-199a-5p inhibitor (199-I), or miR-199a-5p mimic (199-M) (RiboBio) in GM, respectively. After 24 h, the myoblasts were switched to DM, and immunoblots and immunofluorescence analysis were performed at 3 days after culture in DM. The siRNA oligonucleotides for Myh9 and Stat3 were obtained from Invitrogen and RiboBio, respectively. In total, 20 nM oligonucleotides were transfected into C2C12 myoblasts using LipoMax according to the manufacturer's instructions. Next, the myoblasts were cultured in DM until immunoblot or immunofluorescence analysis on the third day.

RT-PCR
Total RNA was extracted from cells or tissues using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. Mature miR-199a-5p expression was detected using previously described procedures with PCR (48). Briefly, the mRNA miR-199a-1 regulates X-linked centronuclear myopathy was extracted and linked to poly(A) tail with Escherichia coli poly(A) polymerase (New England BioLabs Inc., M0276). Then the reverse transcription was performed with specific RT primer (5Ј-GCTGTCAACGATACGCTACGTAACGGCAT-GACAGTGTTTTTTTTTTTTTTTTTTTTTTA-3Ј), which recognizes poly(A) to generate an mRNA cDNA library. Next, expression of miR-199a-5p was tested by real-time PCR with specific primer (forward, 5Ј-CCCAGTGTTCAGACTACC-TGTTC-3Ј; reverse, 5Ј-GCTGTCAACGATACGCTACGTA-ACG-3Ј). The 5S rRNA was detected as an internal reference. Reverse transcription was performed using a PrimeScript RT reagent kit (DRR037A, TaKaRa). Real-time PCR was performed using SYBR Premix Ex Taq (TaKaRa) and the ABI Prism Step One system. The primers are listed in Table S1.

Histology test
Fresh TA or gastrocnemius (GAS) muscle was isolated and immediately frozen in precooled isopentane. Sections of 10-m thickness were made and stained with hematoxylin and eosin. Next, the stained sections were analyzed with Image-Pro Plus software. Cross-sectional area was calculated from 3-5 mice/ group with over 200 fibers for each mouse. The percentage of myofibers with paracentral or central nuclei was measured from 3-5 mice/genotype with over 200 fibers for each mouse.

Immunofluorescence analysis
Cells for immunofluorescence analysis were plated on 0.1% gelatin-coated glass coverslips and cultured until the indicated time. Then the cells were washed with D-Hanks', fixed in 4% paraformaldehyde in PBS, permeabilized with 0.5% Triton X-100 in PBS, and stained with the appropriate antibodies. The slides were costained with 4Ј,6-diamidino-2-phenylindole (Sigma) to mark nuclei. For NM 2A location analysis, the gastrocnemius of the indicated mice was isolated and fixed in 4% paraformaldehyde overnight. Then a single muscle fiber was isolated and manipulated using a protocol similar to that described for the cells above. The primary antibody dilutions were as follows: MYOG, 1:100; MyHC, 1:100; NM 2A, 1:100; ACTN2, 1:100. Alexa Fluor-conjugated secondary antibodies were used at a dilution of 1:250. For myonucleus number analysis in myofiber, the EDL muscles of the indicated 3-week-old mice were isolated and underwent a procedure similar to that above.

Muscle force measurement
The EDL muscles from the indicated mice were isolated and then mounted on force-displacement transducers (MLT0202, ADInstruments), which were connected to a recording device (PowerLab, ADInstruments). The muscles were incubated in 37°C Krebs-Ringer buffer (NaCl, 118.07 mM; KCl, 4.69 mM; CaCl 2 , 2.52 mM; MgSO 4 , 1.16 mM; NaH 2 PO 4 1.01 mM; NaHCO 3 , 25 mM; glucose, 11.1 mM). The maximal force was generated by a 10-V electric stimulus with a frequency of 100 Hz for 250 ms. Meanwhile, the physiological muscle optimal length (L0) and muscle weight (W) were measured. Assuming that the density of muscle was 1.06 g⅐cm Ϫ3 (), the cross-sectional area (S) was equal to the ratio of W and (L0 ϫ ), and the maximal force was calculated as the maximal specific force/S.

Statistics
All of the data are presented as the mean Ϯ S.D. (n Ն 3). The differences between two groups were determined using twotailed Student's t test analysis with GraphPad Prism version 7. For survival experiments, the log-rank (Mantel-Cox) test was performed. p Ͻ 0.05 was considered significantly different.

Study approval
All animal experiments were approved by the Animal Care and Use Committee of the Model Animal Research Center at Nanjing University, which is a member of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).

Data availability
All data are contained within the article.