Elimination of fukutin reveals cellular and molecular pathomechanisms in muscular dystrophy-associated heart failure

Heart failure is the major cause of death for muscular dystrophy patients, however, the molecular pathomechanism remains unknown. Here, we show the detailed molecular pathogenesis of muscular dystrophy-associated cardiomyopathy in mice lacking the fukutin gene (Fktn), the causative gene for Fukuyama muscular dystrophy. Although cardiac Fktn elimination markedly reduced α-dystroglycan glycosylation and dystrophin-glycoprotein complex proteins in sarcolemma at all developmental stages, cardiac dysfunction was observed only in later adulthood, suggesting that membrane fragility is not the sole etiology of cardiac dysfunction. During young adulthood, Fktn-deficient mice were vulnerable to pathological hypertrophic stress with downregulation of Akt and the MEF2-histone deacetylase axis. Acute Fktn elimination caused severe cardiac dysfunction and accelerated mortality with myocyte contractile dysfunction and disordered Golgi-microtubule networks, which were ameliorated with colchicine treatment. These data reveal fukutin is crucial for maintaining myocyte physiology to prevent heart failure, and thus, the results may lead to strategies for therapeutic intervention.

T he dystrophin-glycoprotein complex (DGC) links the intracellular cytoskeleton to the extracellular basement membrane, thereby providing structural support for the sarcolemma 1,2 . Within this complex, dystroglycan (DG) links intracellular dystrophin to the extracellular matrix 3 . As heavily glycosylated α-DG binds tightly and with high affinity to laminin 4 , a glycosylation deficiency can dissolve the DGC-extracellular matrix connection. Mutations of dystrophin and the transmembrane sarcoglycans result in Duchenne/Becker and limb girdle muscular dystrophies, respectively [4][5][6][7] . Many patients with muscular dystrophies, particularly those with defects in the DGC, develop cardiomyopathies with chamber dilation and myocardial dysfunction [8][9][10][11][12] , which is the major cause of death in these patients.
Recently, the full functional glycan structure of α-DG required for its ligand binding activities was identified [13][14][15] , providing insight into the pathogenesis and therapeutic strategies for α-DGassociated diseases, including Fukuyama-type congenital muscular dystrophy (FCMD) 16 , muscle-eye-brain disease 17 , Walker-Warburg syndrome 18 , congenital muscular dystrophy types 1C and 1D 19 , and limb girdle muscular dystrophy type 2I 20 . Although the causative genes for these diseases differ, one biochemical feature that is common among them is abnormal glycosylation of α-DG. Thus, these conditions are considered α-DGpathies. FCMD is caused by mutation of the fukutin gene (FKTN) 16 , which encodes a Golgi-based ribitol phosphate transferase that catalyzes the biosynthesis of tandem ribitol phosphate structures on α-DG by sequential action with fukutin-related protein (FKRP) 13 . Mutations in FKTN and FKRP-associated α-DGpathy are both associated with cardiomyopathies 21,22 . However, the detailed molecular mechanisms for cardiac pathogenesis in these conditions remain unknown.
In this study, we show a crucial role for FKTN in the maintenance of myocyte structure and function using cardiac-specific Fktn knockout mice. The results from this study improve our understanding of the pathomolecular mechanism underlying muscular dystrophy-associated heart failure.

Results
Cardiac changes in MCK-Fktn-cKO (Fktn flox/flox ; MCK-Cre +/− ) mice. To examine the effect of FKTN deficiency, we crossed Fktn flox/flox (floxed) mice with a transgenic line expressing Cre recombinase under the control of the MCK promoter 23 . The reduced expression of FKTN and DGC proteins in the hearts of these MCK-Fktn-cKO mice was confirmed by immunoprecipitation and immunoblotting ( Supplementary Fig. 1a, b). The glycosylated form of α-DG and DGC immunoreactivity increased in the sarcolemma at 24 weeks after birth in normal control mice (Fig. 1a), suggesting that the physiological contribution of DGC proteins in the heart increases 6 months after birth. Notably, FKTN deficiency not only resulted in a reduction of α-DG glycosylation but also impaired the expression of DGC proteins in the sarcolemma at all developmental stages (Fig. 1a). We next examined global cardiac morphology and function in MCK-Fktn-cKO mice at different time points after birth. Cardiac gross morphology (Fig. 1b) and the heart-to-body weight ratio (Fig. 1c) were maintained in mice with Fktn deficiency. However, the cross-sectional areas of cardiomyocytes and fibrosis were increased in MCK-Fktn-cKO hearts 24-48 weeks after birth ( Fig. 1d-f). These mice also showed cardiac dysfunction and chamber dilation during diastole at this age (Fig. 1g, h). Only one or a few damaged areas were detected by anti-mouse IgG in 48-week-old MCK-Fktn-cKO mice ( Supplementary Fig. 2). The 48week-old heterozygous (Fktn flox/+ ; MCK-Cre +/− ) mice showed no abnormalities in overall cardiac morphology and function ( Supplementary Fig. 3). Thus, Fktn deficiency leads to pathological cardiac remodeling in 24-48-week-old mice.
Hypertrophic responses of hearts from MCK-Fktn-cKO mice. As mentioned above, young-adult MCK-Fktn-cKO mice did not exhibit significant cardiac dysfunction under physiological conditions (Fig. 1). Consistently, isolated cardiomyocytes from 10week-old MCK-Fktn-cKO mice showed normal contractility and Ca 2+ handling during E-C coupling ( Supplementary Fig. 4). To investigate the effect of FKTN deficiency under stress conditions, we examined cardiac structure, function, and intracellular signaling after inducing various hemodynamic stresses. Control mice showed incremental changes in heart weight-to-body weight ratios (Fig. 3a, b) and cross-sectional areas of cardiomyocytes ( Fig. 3c) without increased fibrosis (Fig. 3d) after administration of the adrenergic receptor agonists isoproterenol and phenylephrine as well as after prolonged pressure overload via 2-week thoracic aortic constriction. Conversely, these treatments caused chamber dilation and fibrosis with severe cardiac dysfunction in the hearts of MCK-Fktn-cKO mice (Fig. 3a, d, e). These observations suggest that Fktn-deficient hearts are vulnerable to hypertrophic stress and that FKTN in the heart is essential for compensatory hypertrophic responses against hemodynamic stress.
Next, in order to understand the molecular basis underlying vulnerability, we analyzed the changes in intracellular signaling in the hearts of Fktn-deficient mice treated with isoproterenol for 2 weeks. It is well known that Akt and Ca 2+ calmodulindependent protein kinase II (CamKII) phosphorylation mediate myocyte survival and apoptotic/necrosis pathways [25][26][27] , respectively. We found that isoproterenol increased the phosphorylation of Akt in control hearts but not in Fktn-deficient hearts, which instead showed increased phosphorylation of CamKII, suggesting a downregulation of survival and upregulation of apoptosis/ necrosis signaling. Thus, Fktn-deficient hearts showed an impaired hypertrophic response to stress.
It is also known that activation of protein kinase D (PKD) leads to the phosphorylation of histone deacetylases (HDACs) in myocyte pathological remodeling [28][29][30] . Whereas isoproterenol treatment slightly induced the phosphorylation of PKD in control hearts, basal levels of PKD phosphorylation in the absence of isoproterenol were already significantly higher in MCK-Fktn-cKO hearts than in controls (Fig. 3f). Similarly, the phosphorylation of HDAC9 in control hearts was increased by isoproterenol treatment, while the basal phosphorylation in MCK-Fktn-cKO hearts was already extremely high. These data indicate that Fktn elimination enhanced the PKD signaling pathways, provoking HDAC9 nucleocytoplasmic shuttling under physiological conditions. Therefore, Fktn elimination accelerates the progression from compensated cardiac hypertrophy to heart failure under hemodynamic stress conditions.  Anti-IIH6 (glycosylated form of αDG), βDG, dystrophin (Dys), αSG, βSG, and γSG antibodies (green) and DAPI (blue). Scale bar, 10 μm. b Cardiac morphology in floxed and MCK-Fktn-cKO mice. Scale bar, 5 mm. c Heart weight-to-body weight ratios (HW/BW) (n = 9 mice per group). d Masson's trichrome staining of the left ventricle. Scale bar, 50 μm. Cross-sectional areas (6w, n = 259 and 268 cells; 16w, n = 303 and 311 cells; 24w, n = 266 and 287 cells; 48w, n = 283 and 256 cells measured from 3 hearts per group.) (e) and fibrosis percentages (n = 3 mice per group) (f) from paraffin sections of left ventricles. Echocardiographic parameters of left ventricle diastolic dimension (g) and fractional shortening (h) in floxed and MCK-Fktn-cKO mice (n = 12 mice per group). Data are means ± s.e.m; #P < 0.05 between indicated groups based on Student's t-tests. Impaired hypertrophic response in MCK-Fktn-cKO cardiomyocytes. To more directly examine the hypertrophic responses of MCK-Fktn-cKO cardiomyocytes, we analyzed cell growth during culture with or without phenylephrine for 48 h. We confirmed successful Cre-lox recombination and expression of Cre recombinase in the nucleus in cultured neonatal cardiomyocytes from MCK-Fktn-cKO mice ( Supplementary Fig. 5). We also confirmed reduction of FKTN protein in 1-d-old MCK-Fktn-cKO mice ( Supplementary Fig. 5). In the presence of phenylephrine, cardiomyocytes from control mice showed enhanced sarcomere organization and upregulation of NCX1 expression (Fig. 4a), which are signs of myofibril maturation and intracellular Ca 2+ handling during E-C coupling 24 . By contrast, MCK-Fktn-cKO cardiomyocytes treated with phenylephrine had immature and tangled myofilaments (Fig. 4b) and showed no increase in cell area (Fig. 4c). We confirmed that all cKO myocytes treated with phenylephrine were alive using a LIVE/DEAD assay (Fig. 4d). Remarkably, cKO myocytes did not show spontaneous hypertrophy (at 48 h in the absence of phenylephrine; Fig. 4c). These myocytes exhibited no sarcomere formation, cell-cell interaction (Fig. 4a, upper right panels), or spontaneous beating, suggesting that Fktn deficiency also affects myocyte maturation in cKO cells. Ca 2+ content in the SRs of control cardiomyocytes increased during culture with phenylephrine ( Fig. 4e), whereas no such increase was observed in cardiomyocytes from MCK-Fktn-cKO mice. However, caffeine-induced release was still present, suggesting that these cells had functions characteristic of cardiomyocytes. As a control, mixed fibroblasts in the cultured myocyte preparation showed no response to caffeine (Fig. 4e). These observations suggest that Fktn elimination in myocytes impairs hypertrophic responses.
Using these cell models, we examined the subcellular localization of MEF2, which acts with class IIa HDACs, such as HDAC9, in myocyte differentiation and compensative hypertrophic responses 31 . MEF2 underwent nuclear translocation in control cardiomyocytes but not in MCK-Fktn-cKO cardiomyocytes cultured with 10% fetal calf serum (FCS) or 10 μM phenylephrine for 48 h (Fig. 5a). In control cardiomyocytes, serum (as well as adrenergic agonists) induced the nuclear export of HDAC9, which relieves transcriptional repression 28 . This phenomenon was not observed in MCK-Fktn-cKO cardiomyocytes (Fig. 5b). These observations suggest that MEF2 and HDAC9 translocation are impaired in FKTN-deficient cardiomyocytes.
Phosphatidylinositol 4-phosphate (PI4P) in the Golgi apparatus is required for PKD activation in pathological hypertrophic remodeling 32,33 . Fktn-deficient myocytes showed abnormal accumulation of PI4P (Fig. 5c) and surprisingly, fragmentation of the Golgi apparatus under basal conditions, as revealed by staining with the cis-Golgi marker protein GM130 (Fig. 5d). These phenotypes may be indicative of PKD hyperactivation, which subsequently impairs the MEF2-HDAC axis following hypertrophic stimulation. Furthermore, PKD hyperactivation also may be associated with Golgi fragmentation. Treatment of HEK cells with the marine sponge metabolite ilimaquinone (a PKD activator) induced the vesiculation of the Golgi, as observed via GM130 staining (Fig. 5e, top middle). Alternatively, Fktn elimination per se may affect Golgi structure. Remarkably, we found that untreated FKTN-deficient HEK cells showed an elongated pattern of GM130 staining, suggesting that FKTN is involved in maintaining the structure of the Golgi apparatus (Fig. 5e, bottom left). In support of this notion, treatment with ilimaquinone or the PKD inhibitor CID755673 did not result in further disruption (Fig. 5e, bottom panels). In HAP cells, the structure of the Golgi apparatus was changed in an FKTNdependent manner (Fig. 5f). These observations suggest the possibility that FKTN contributes to the maintenance of the Golgi architecture in these cells.
Acute elimination of Fktn results in severe cardiac dysfunction.
Although FKTN protein is likely involved in the maintenance of myocytes, we did not observe structural and functional defects in cardiomyocytes from young-adult mice. To examine the direct   . Surprisingly,~50% of these mice died within 1 week after initiating tamoxifen treatments (Fig. 6c), suggesting an indispensable role for FKTN in the working heart. Fktn flox/+ ; αMHC-MCM +/− (hetero) mice showed no abnormalities in overall cardiac structure and function or mortality rate after tamoxifen treatment ( Supplementary Fig. 7). As previously mentioned (Fig. 1a), the glycosylation of α-DG (as detected with the IIH6 antibody) increases with age, such that levels are barely detectable in hearts from 10-week-old mice. The hearts of tamoxifen-treated αMHC-MCM-Fktn-cKO mice showed normal DGC protein localization and levels of glycosylated α-DG ( Supplementary  Fig. 8); this suggests that cardiac dysfunction at this stage (4 d after tamoxifen injection) is independent of α-DG glycosylation and DGC proteins in the sarcolemma. Cardiomyocytes isolated from tamoxifen-treated αMHC-MCM-Fktn-cKO mice had severely impaired contractility and Ca 2+ handling during E-C coupling with reduced amplitudes and delayed times to peak of electrically evoked Ca 2+ transients compared with those from vehicle-treated controls ( Fig. 6d-h).
Subcellular changes in αMHC-MCM-Fktn-cKO cardiomyocytes. We next investigated how Fktn elimination affects contractility and Ca 2+ handling during E-C coupling, particularly via the structure of T-tubules and myofilaments and expression of Ca 2+ regulatory proteins. We analyzed the T-tubule structure and found it was severely disordered in Fktn-deficient cardiomyocytes (Fig. 7a). The myocardia of tamoxifen-treated αMHC-MCM-Fktn-cKO mice also showed subdivided subcellular myofilaments and disorganized Z-line structures (Fig. 7b, black arrowheads). Thus, Fktn elimination directly leads to structural and functional cardiomyocyte defects, especially of the T-tubules that form the key structure of E-C coupling, independent of α-DG glycosylation. Despite the severe structural disorganization of T-tubules, the expression of Ca 2+ regulatory proteins, including NCX1 and LTCC, was not altered (Fig. 7c, d and Supplementary Fig. 9).
The effects of MT depolymerization in αMHC-MCM-Fktn-cKO myocytes. Because MT densification can impede myocyte shortening [35][36][37][38] , we examined whether treatment with 10 μM colchicine, which inhibits MT polymerization, for 1 h resolved contractile dysfunction in tamoxifen-treated αMHC-MCM-Fktn-cKO cardiomyocytes. In control cardiomyocytes, α-tubulin immunostaining revealed a regulated meshwork pattern of MTs (Fig. 8a, top left), with some polymerized MTs perpendicular to the muscle fiber (Fig. 8a, top left inset). The meshwork pattern of MTs was lost after treatment with 10 μM colchicine (Fig. 8a, middle left), whereas cardiomyocyte shortening was not affected in vehicle-treated control myocytes (Fig. 8b). In tamoxifentreated αMHC-MCM-Fktn-cKO cardiomyocytes, we observed an accumulation of MTs in parallel with the muscle fiber (Fig. 8a, top right and inset), and shortening was impaired compared with that in the vehicle control (Fig. 8b). Notably, colchicine treatment reduced this accumulation of MTs and reversed the impaired shortening in tamoxifen-treated αMHC-MCM-Fktn-cKO cardiomyocytes (Fig. 8a, bottom right; Fig. 8b). Golgi elements in muscles are major sites for the nucleation of MTs 39   abnormal distribution of the Golgi apparatus, has recently been implicated in the progression of heart failure 34,35,40,41 . Interestingly, treatment with 10 μM colchicine disrupted the Golgi apparatus (Fig. 8a, middle left), consistent with the idea that MTs are crucial for proper localization of the Golgi in a wide range of cells 34 . Remarkably, this fragmentation of the Golgi did not affect cardiomyocyte shortening (Fig. 8b). Instead, the activation of PKD with ilimaquinone resulted in vesiculation in the Golgi apparatus, an accumulation of MTs (Fig. 8a, bottom left and inset), and contractile dysfunction (Fig. 8b) in control cardiomyocytes. In tamoxifen-treated αMHC-MCM-Fktn-cKO cardiomyocytes, we also observed abnormal fragmentation of the Golgi apparatus (Fig. 8a, top right, white arrowheads). These results suggest that contractile dysfunction in FKTN-deficient cardiomyocytes is caused by MT densification that may be associated with Golgi abnormalities. In tamoxifen-treated αMHC-MCM-Fktn-cKO hearts, the level of PKD phosphorylation was significantly higher than that in control hearts (Fig. 8c), suggesting that FKTN impacts the integrity of the Golgi structure by regulating PKD signaling pathways. We also observed abnormal distribution of the Golgi apparatus and the accumulation of MT structures in MCK-Fktn-cKO mice in late adulthood (Supplementary Fig. 10). Control myocytes from 10-month-old mice also exhibit the accumulation of MT structures. While membrane abnormalities were observed in cardiomyocytes isolated from 10month-old MCK-Fktn-cKO mice ( Supplementary Fig. 10), the severity was less than that observed in tamoxifen-treated αMHC-MCM-Fktn-cKO hearts (Fig. 7a). Additionally, phosphorylation of PKD was decreased in cardiomyocytes isolated from MCK-Fktn-cKO mice 48 weeks after birth ( Supplementary Fig. 11). On the other hand, reduction of FKTN protein immediately led to the abnormal fragmentation of the Golgi apparatus, MT densification, and PKD phosphorylation in tamoxifen-treated αMHC-MCM-Fktn-cKO hearts (Fig. 8). These data suggest that FKTN plays a crucial role in the maintenance of the Golgi-derived MT network and PKD signaling.
To elucidate how colchicine improves myocyte contractility, we assessed Ca 2+ handling during E-C coupling in cardiomyocytes isolated from tamoxifen-treated αMHC-MCM-Fktn-cKO hearts. We found that colchicine treatment enhanced intracellular Ca 2+ handling in tamoxifen-treated αMHC-MCM-Fktn-cKO myocytes (Fig. 8d). This was associated with improved myocyte contractile efficiency as evidenced by the ratio of cell shortening to the amplitude of Ca 2+ increase during E-C coupling (Fig. 8e). Thus, colchicine appears to improve myocyte contractility by improving Ca 2+ handling and reducing the stiffness of cardiomyocytes (Fig. 8e).
Finally, we showed that colchicine ameliorated the heart failure phenotype in tamoxifen-treated αMHC-MCM-Fktn-cKO mice. Chamber dilation, fibrosis, cardiac dysfunction, and survival rates after 10 d of tamoxifen treatment were all improved by colchicine administration (Fig. 9a-e). These observations suggest acute elimination of Fktn contributes to myocyte contractile dysfunction as a result of MT densification.
Mechanistic analysis for Golgi pathology in Fktn-deficient hearts.
To elucidate how FKTN regulates Golgi structure and function and thus leads to MT accumulation, we performed microarray analysis of tamoxifen-treated floxed or αMHC-MCM-Fktn-cKO hearts and investigated the putative pathway and biological processes induced by deletion of Fktn. A total of 31,762 probe sets obtained from an Agilent gene expression microarray platform (SurePrint G3 Mouse Gene Expression 8 × 60 K) were tested for differential expression. We found that 7408 genes were differentially expressed on the basis of a criterion of at least a twofold change and adjusted value of P < 0.05 when cKO was compared with floxed-control; 4536 transcripts were downregulated and 2872 transcripts were upregulated in αMHC-MCM-Fktn-cKO hearts ( Table I in the online-only Data Supplement). Hierarchical clustering revealed a high level of homogeneity within and a clear separation between groups ( Supplementary Fig. 12). A heatmap displaying expression levels of 1,611 differentially expressed Golgi apparatus-related genes annotated from a global cross-database NCBI search showed excellent consistency of changes within groups (Fig. 10a). Thus, the elimination of Fktn impacts the expression profile of Golgi-related genes.
Integrity of the Golgi apparatus requires the MT network 42 . We focused on genes known to be relevant to nucleation and dynamics of Golgi-derived MTs: Tpx2 (coding for a targeting protein of Xklp2), Rab3a (coding for RAS oncogene family member 3a), Pde4dip (coding for myomegalin), Pcnt (coding for pericentrin), and Clasp1 (coding for CLIP-associating protein 1) 42 . The positions of these genes in the volcano plot are depicted in Fig. 10b. Although the MT nucleation factor encoded by Tpx2 was increased, the anchoring proteins of the γ-tubulin ring complex (γ-TuRC) of the cis-Golgi encoded by Pde4dip and Pcnt were reduced in cKO hearts. The MT-stabilizing protein encoded by Clasp1 was also reduced in cKO hearts. These observations suggest that Fktn elimination leads to a decline in the stability and dynamics of the Golgi-derived  Pathway enrichment analysis of the downregulated genes in cKO hearts (FDR < 0.001, Log 2 FC < −1.5) revealed enrichment of pathways involved in cardiac physiology (i.e., muscle contraction and cardiac conduction; Supplementary Fig. 12). On the other hand, upregulated genes in cKO hearts (FDR < 0.001, Log 2 FC < −1.5) revealed enrichment of pathways involved in the cell cycle, Rho GTPase signaling, extracellular matrix organization or degradation, integrin cell surface interaction, and extracellular matrix proteoglycan as analyzed by ReactomePA (https:// reactome.org/; Fig. 10c). These pathways were influenced by the stability and dynamics of Golgi-derived MT, protein synthesis at the Golgi apparatus, and membrane transport via MT in cells. Thus, these observations emphasize the possibility that disintegration of the Golgi apparatus and Golgi-derived MT network is one etiology of muscular dystrophy-associated heart failure in Fktn-deficient mice.
Kyoto Encyclopedia of Gene and Genomes (KEGG) functional network analysis of the 7408 genes showed large fluctuations with high significance for Parkinson's, Huntington's, and Alzheimer's diseases (Fig. 10d). Structural and functional alteration of the Golgi apparatus is recognized as a constant pathological hallmark of neurodegenerative disease, including Parkinson's, Huntington's, and Alzheimer's diseases 43 . A pathway related to hypertrophic cardiomyopathy and dilated cardiomyopathy was also prominent. These results suggest that the pathological mechanism of cardiac dysfunction induced by Fktn elimination partly overlaps with that of neurodegenerative disease.
To investigate the cause of the Golgi pathology observed in Fktn elimination, we examined localization of the α-DG protein using antibody against the core region of α-DG (core α-DG) in cultured neonatal cardiomyocytes isolated from MCK-Fktn-cKO hearts (Fig. 10e). In control myocytes, core α-DG co-localized with the post-Golgi marker Rab3a beneath the membrane (Fig. 10e, upper panels). On the other hand, although core  c Representative images of immunoblots. Phosphorylation level of PKD is shown relative to total PKD (n = 6 mice per group). *P < 0.05 vs. vehicle-treated floxed mice; # P < 0.05 between indicated groups based on Tukey-Kramer tests. d Normalized peak amplitudes (vehicle-intact, n = 32 cells from 4 mice; vehicle-colchicine, n = 39 cells from 3 mice; vehicle-ilimaquinone, n = 26 cells from 3 mice, tamoxifen-intact, n = 14 cells from 4 mice; tamoxifencolchicine, n = 11 cells from 4 mice). e Contractile efficiency (vehicle-intact, n = 8 cells; vehicle-colchicine, n = 6 cells; vehicle-ilimaquinone, n = 6 cells, tamoxifen-intact, n = 5 cells; tamoxifen-colchicine, n = 6 cells from 3 mice per group). # P < 0.05 between indicated groups based on Tukey-Kramer tests (b, d, e). α-DG also co-localized with Rab3a in cKO myocytes, its signal accumulated intracellularly on the vesicle structure (Fig. 10e, lower panels). Next, we examined whether accumulation of core α-DG is a critical factor in disorganization of the Golgi apparatus (Fig. 10f). We observed faint signals of anti-core α-DG antibody beneath the membrane in HEK cells (Fig. 10f, upper panels). On the other hand, FKTN-KO HEK cells showed internally accumulated core α-DG, which mainly co-localized with Rab3a, and fragmentation of the Golgi apparatus (Fig. 10f, middle panels). Forced expression of full-length DG induced accumulation of core α-DG in the post-Golgi area, which co-localized with Rab3a (Fig. 10f, left lower panels). However, we did not observe fragmentation of the Golgi apparatus by GM130 labeling (Fig. 10f,  right lower panels). These observations suggest that accumulation of core α-DG does not trigger the Golgi pathology observed in Fktn-deficient cells and thus that FKTN may be critical for the maintenance of Golgi structure and function.

Discussion
As DG glycosylation and the DGC play essential roles in the connection between the basement membrane and sarcolemma in skeletal muscle 23,44 , the loss of membrane integrity has been thought to be a cause of muscle damage in cardiac tissues 45 . In support of this hypothesis, it was shown that DG limits cardiac myocyte membrane damage 46 . In addition, our present study demonstrates, for the first time, the cellular and molecular pathomechanisms of heart failure associated with α-DGpathy.
Studies of skeletal muscle-specific Fktn-cKO mice indicate that the loss of functional α-DG protein likely contributes to the pathogenesis of muscular dystrophy 23,44 . Surprisingly, our study shows a lack of cardiac dysfunction in young-adult MCK-Fktn-cKO mice despite the marked reduction in DGC proteins in the cardiac sarcolemma, suggesting that membrane fragility is not the sole etiology of cardiac dysfunction. Cardiac dysfunction with impaired myocyte shortening in old age (Figs. 1 and 2) or vulnerability to hemodynamic stress at a young age (Fig. 3) in MCK promoter-driven cKO hearts suggests that the FKTN protein is critical for maintaining cardiac structure and function under conditions of aging and hemodynamic stress. Therefore, MCK promoter-driven cKO hearts may have an unknown compensatory mechanism that prevents mice from manifesting any obvious phenotypes of FKTN deficiency during the young-adult age under normal physiological conditions. Therefore, it is possible that the acute Fktn deficiency results in severe cardiac dysfunction and reduced survival in the young mice. Notably, DGC localization was not affected in these young mice, suggesting that reduced α-DG glycosylation and DGC expression may not be the sole reasons for cardiac dysfunction.
A main finding is that FKTN is crucial for maintaining contractile function and Ca 2+ handling during E-C coupling in individual cardiomyocytes. By contrast, previous studies of dystrophin-deficient mdx mice found changes in Ca 2+ signaling with only slightly elevated resting intracellular concentrations and potentially arrhythmogenic E-C coupling in ventricular myocytes 47,48 . Homeostasis in young-adult Fktn-deficient mice may have been achieved by reduced Ca 2+ influx through LTCC and increased efflux by NCX1 or SERCA ( Supplementary Fig. 13 Akt and upregulated CamKII, suggesting altered Ca 2+ -dependent signaling. Thus, the inability to precisely regulate intracellular Ca 2 + after Fktn elimination likely impacted cell survival in the long run. Indeed, we also observed an upregulation of the hypertrophic mediator calcineurin 51 and a downregulation of its endogenous inhibitor MCIP 52 , along with strong induction of the hypertrophic marker ANP, in MCK-Fktn-cKO mice at 16 weeks of age ( Supplementary Fig. 13). Thus, at this age, hypertrophic remodeling, mediated by Ca 2+ -calcineurin signaling, is already underway before abnormalities in morphology and contractility emerge. In our study, Fktn-deficient cardiomyocytes showed hyperphosphorylation and altered localization of HDAC9, which is a direct transcriptional target of MEF2 and acts as signalresponsive repressor of cardiac hypertrophy 31 . HDAC9-knockout mice are ultrasensitive to hypertrophic stimuli and spontaneously develop cardiac hypertrophy with advanced age 31 . Thus, HDAC9 alterations can be associated with the vulnerability to stress in Fktn-deficient myocytes. The constitutive phosphorylation of PKD, which transduces signals for diacylglycerol biogenesis 53 , and hyperphosphorylation of HDAC9 under basal conditions in the hearts of Fktn-deficient mice suggest there may be excessive transcriptional activation in response to stress.
Another key finding is that FKTN is essential for maintaining the Golgi-derived MT network in cardiomyocytes. Constitutively phosphorylated PKD in myocytes from young MCK-Fktn-cKO mice or tamoxifen-treated αMHC-MCM-Fktn-cKO myocytes may remain associated with the cytoplasmic surfaces of Golgi membranes and regulate the fission of vesicles that carry protein and lipid cargo to the plasma membrane 54 . We also observed the vesiculation of Golgi and accumulation of MT in the myocytes from old MCK-Fktn-cKO mice ( Supplementary Fig. 10), supporting the notion that FKTN is involved in the maintenance of the Golgi-MT network. Alternatively, PKD phosphorylation in myocytes from old MCK-Fktn-cKO mice was lower than that in control myocytes (Supplementary Fig. 11). This contradiction may be explained by differences in timing of Fktn deletion, disease progression course (period), and hemodynamic stress conditions, thus reflecting pathogenesis and/or pathological consequences. Nonetheless, our data show that FKTN protein plays a role in the maintenance of Golgi-MT network in cardiomyocytes. Considering the results that old MCK-Fktn-cKO mice showed very mild pathology compared to isoproterenol-treated young MCK-Fktn-cKO mice or tamoxifen-treated αMHC-MCM-Fktn-cKO mice, our data also suggest the presence of a compensatory mechanism during long-term progression in the presence of Fktn deficiency.
The altered distribution of GM130 in FKTN-deficient HAP and HEK cells indicates that the Golgi morphology was also altered independently of PKD activity (Fig. 5e, f). Impairment of glycosylation by Fktn deficiency may accelerate fragmentation of the Golgi apparatus by enhancing the budding of vesicles from the Golgi membrane or disrupting protein trafficking. We observed the accumulation of core α-DG in the post-Golgi area and fragmentation of the Golgi apparatus in Fktn-deficient cardiomyocytes and HEK cells (Fig. 10e, f). HEK cells expressing full-length α-DG also showed accumulation of core α-DG in the post-Golgi area (Fig. 10f). However, we did not observe abnormal fragmentation of the Golgi apparatus in cells, suggesting a pivotal role of FKTN in the maintenance of Golgi integrity (Fig. 10f).
FKTN is targeted to the medial Golgi apparatus through its Ntermini and transmembrane domain 55 . Therefore, the lack of FKTN per se might affect the integrity of the Golgi ribbon. Alternatively, FKTN also forms a complex with other Golgiresident glycosyltransferases 56,57 , and thus, the lack of FKTN may change cellular glycosylation, affecting Golgi structure and function. The results of our microarray analysis also suggest the impact of FKTN on expression of Golgi-related genes. Elimination of Fktn led to the downregulation of genes encoding the anchoring protein of γ-TuRC at cis-Golgi and MT-stabilizing protein (Fig. 10b). Golgi-derived MT networks serve important roles in post-Golgi trafficking, maintenance of Golgi integrity, cell polarity, and cellular function 42 . The disruption of MT dynamics in Fktn-deficient myocytes may impact Golgi vesiculation, with the accumulation and densification of MTs affecting both the central cellular localization of Golgi and ER-Golgi-plasma membrane trafficking, which indirectly impacts the size and morphology of the Golgi 58 . Recent studies show that disintegration of Golgi structure and function is one of the main causes of Alzheimer's disease 42 . In this disease, precipitation of hyperphosphorylated tau protein is also related to Golgi structure and function 42 . Remarkably, KEGG analysis suggested that heart failure induced by acute Fktn elimination has a common etiology with neuromuscular disease (Fig. 10d). Our results further suggest a mechanistic connection between the Golgi apparatus and myocyte T-tubule organization, which is crucial for Ca 2+ homeostasis in cardiomyocytes. Further investigations will clarify the FKTN-dependent maintenance of Golgi-derived MT networks.
Cardiomyocytes isolated from acute Fktn deficiency in young mice had dramatically disrupted T-tubule membranes and MT densification, as seen in failing hearts of human patients and animal models 35,59,60 . Studies show that MT-mediated defects in JP2 trafficking contribute to myocyte T-tubule remodeling in heart failure 34,38 . Here, we show that Fktn elimination also led to a reduction and redistribution of JP2. Thus, disorganization of the T-tubule membrane in Fktn-deficient mice might be caused by MT-mediated defects in JP2 trafficking. These changes contributed to reduced myocyte contractility and cardiac dysfunction, which were ameliorated by colchicine treatment. These data suggest that the accumulation of MTs after Fktn elimination is a major contributor to myocyte dysfunction and that an effective therapeutic strategy for heart failure involves MT depolymerization. MT densification increases the stiffness of cells, thereby mechanically impeding sarcomere motion 40 , and alters T-tubules, which are important for E-C coupling 34 . We show that treatment with colchicine not only restored myocyte shortening, Ca 2+ handling during E-C coupling, and contractile efficiency but also ameliorated severe cardiac dysfunction and improved the survival of mice with FKTN deficiency. One question is whether colchicine treatment can be applied to other forms of muscular dystrophy-associated cardiomyopathies. Answering this question requires a system that can tightly control the timing of eliminating the expression of the responsible gene, as we did in this study. However, a previous report suggested that colchicine treatment improves Ca 2+ handling in mdx mice 38 , a model for DMD; although, the exact mechanism remains to be elucidated. Importantly, as colchicine is already an FDA-approved drug for the treatment of familial Mediterranean fever and acute gout flares, our findings suggest that this drug may also be useful for treating muscular dystrophy-associated heart failure.

Methods
Thoracic aortic constriction surgery. Left ventricular pressure overload was induced by thoracic aortic constriction. The 10-week-old mice (weighing 20-22 g) were anesthetized by i.p. injections of a mixture of ketamine (100 mg/kg) and xylazine (5 mg/kg). After orotracheal intubation, cannulae were connected to a volume-cycled ventilator (SN-480-7; Shimano, Japan) with room air at a tidal volume of 0.2 ml and a respiratory rate of 110 breaths per minute. The chest cavities were entered through small incisions to the second intercostal space, and the transverse aortas were constricted with 7-0 nylon strings by ligating with a blunted 27-gauge needle, which was later removed.
Neonatal cardiomyocyte culture. Primary cardiomyocyte cultures were prepared from ventricles of 1-d-old mice by very gentle trypsinization at room temperature using a modified method for rat neonatal hearts 46 . Briefly, hearts were rapidly removed from neonatal MCK-Fktn-cKO or control (floxed) mice anesthetized with an overdose of pentobarbital (30 mg/kg, i.p.). The ventricles were excised, cut into several pieces, and washed three times with 10 ml ice-cold PBS for 1 min by gentle shaking. The tissue pieces were digested three times with 0.06% trypsin in DMEM (8 ml) for 8 min at 37°C by gentle agitation. The cells were resuspended in DMEM with 10% FCS to stop trypsinization and centrifuged at 1400 × g for 3 min. The cell pellets were resuspended in fresh DMEM containing 10% FCS, plated on collagencoated 24-well dishes at a density of 4 × 10 4 cells per well, and maintained in DMEM containing 10% FCS. After 24 h, the cells were divided into three groups and then maintained for up to 72 h in DMEM alone or DMEM with 10% FCS or with 10 μM phenylephrine. The formation of myocytes clusters and spontaneous synchronized beating were confirmed by observation under an inverted microscope (CKX41; Olympus). Cellular viability was investigated at 48 h of culture by using a The LIVE/DEA Fixable Green Dead Cell Stain Kit (Invitrogen). Cells viability were validated on three independent experiments for cell type.
Isolation of adult mouse ventricular myocytes. Ventricular myocytes were obtained from 10-month-old male MCK-Fknt-cKO, αMHC-MCM-Fknt-cKO, or floxed-control mice. Hearts were rapidly removed from adult mice anesthetized with an overdose of pentobarbital (300 mg/kg, i.p.) and Langendorff-perfused at a constant hydrostatic pressure of 70 cm H 2 O at 37°C using cell-isolation buffer (CIB; 130 mM NaCl, 5.4 mM KCl, 0.5 mM MgCl 2 , 0.33 mM NaH 2 PO 4 , 22 mM glucose, 50 nM/ml bovine insulin, and 25 HEPES-NaOH (pH 7.4)) supplemented with 0.4 mM EGTA, which chelates calcium within the heart. Insulin was used from a 1 U/ml stock solution in 0.1 mM HCl (pH 4.0). EGTA was from a 400 mM stock in 1 M NaOH (pH 7.8). The perfusate was then switched to the enzyme solution (15 ml) of CIB supplemented with 0.3 mM CaCl 2 , 1 mg/ml collagenase, 0.06 mg/ml trypsin, and 0.06 mg/ml protease. Once the tissue had undergone complete digestion, the ventricles were excised, cut into several pieces, and further digested in fresh enzyme solution (15 ml) for 15-20 min at 37°C until they were mostly dissociated. The CaCl 2 level in this enzyme solution was increased to 0.7 mM and supplemented with 2 mg/ml BSA. The cell suspension was centrifuged at 1400 × g for 3 min. The cell pellet (~0.1 ml) was resuspended in CIB supplemented with 1.2 mM CaCl 2 and 2 mg/ml BSA, then incubated at 37°C for 10 min, centrifuged (1400 × g for 3 min), and resuspended in 10 ml Tyrode's solution (140 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl 2 , 0.5 mM MgCl 2 , 0.33 mM NaH 2 PO 4 , 11 mM glucose, and 5 mM HEPES-NaOH (pH 7.4)) supplemented with 2 mg/ml BSA. Isolated cardiomyocytes were incubated at 37°C in Tyrode's solution containing 10 μM colchicine for 1 h to depolymerize MTs or 25 μM ilimaquinone for 15 min to induce vesiculation of Golgi membranes.
Cell shortening and Ca 2+ transients in isolated myocytes. Isolated cardiomyocytes were loaded with 10 µM Indo-1 AM and electrically stimulated at 1 Hz using a two-platinum electrode insert connected to a bipolar stimulator (SEN-3301; Nihon Kohden) on the stage of an inverted microscope (IX71; Olympus) with a 20× water immersion objective lens (UApo N340; Olympus). Calcium transients were measured as the ratio of fluorescence emitted at 405/480 nm, after excitation at 340 nm, using a high-performance Evolve EMCCD camera (Photometrics). Cardiomyocytes were maintained under continuous flow in standard Tyrode's solution, exchanged using a microperfusion system. The experiments were recorded and analyzed using MetaMorph software (version 7.7.1.0; Molecular Devices). To estimate the effect of colchicine on cell contraction, we calculated the efficiency of converting elevated Ca 2+ levels to cell contraction by dividing the percentage of cell shortening by the Ca 2+ amplitude.
Generation of knockout cell line with CRISPR/Cas9. The CRISPR/Cas9 targeting sequence for FKTN was GAGTAGAATCAATAAGAACG. The oligonucleotides for this sequence were inserted into the Cas9 Smart nuclease all-in-one vector (System Biosciences, Mountain View, CA) with an additional 8-base sequence at the 5′ terminus. The vector was transfected into HEK293 (American Type Culture Collection, Manassas, VA, USA; CRL-1573) or HAP1 cells (Haplogen, Vienna, Austria; C631). IIH6-negative cells were sorted by fluorescenceactivated cell sorting (MoFlo, Beckman Coulter, Brea, CA). HEK293 cells were cultured in DMEM supplemented with 10% FCS and penicillin/streptomycin. HAP1 cells were cultured in Iscove's modified Dulbecco's medium (Wako Pure Chemical Industries) supplemented with 10% FCS and penicillin/streptomycin. Each cell clone was verified for IIH6 reactivity by Western blot analysis and DNA sequencing. The mutations in each clone were as follows: (HEK293 cell is a triploid) HEK FKTN1o-2 (27-bp deletion, 1-bp deletion, and 1-bp insertion in exon 2, causing frameshifts) and HAP FKTN5 (34-bp deletion in exon 2, causing a frameshift).
Histology. Hearts were excised and immediately fixed in buffered 4% paraformaldehyde, embedded in paraffin, and sectioned at a thickness of 4 µm.
Immunoblotting. Mice hearts were homogenized using a Hiscotron homogenizer (NITI-ON) in lysis buffer containing 20 mM HEPES (pH 7.4), 150 mM NaCl, 1% sodium deoxycholate, 1% SDS, 2 µg/ml leupeptin, 1 µg/ml aprotinin, 200 µM phenylmethylsulfonyl fluoride, and 200 µM benzamidine hydrochloride. The lysates were centrifuged at 100,000 × g for 20 min and the supernatants were used for immunoblot analysis. DG from solubilized myocardium was enriched with wheat germ agglutininagarose beads (Vector Laboratories). To detect the expression of FKTN protein, cardiac muscles were solubilized with 1% Triton X-100 in TBS. The solubilized samples were assayed for total protein concentration and then 4 mg of proteins were subjected to FKTN antibody-conjugated beads to enrich FKTN proteins as previously reported 65 . The bound materials were eluted with 0.1 M glycine-HCl (pH 2.5) and then analyzed by Western blotting using FKTN antibody (RY213). Immunoreactive bands were visualized using a chemiluminescence detection system (Perkin Elmer or Amersham Biosciences Corp.) and an LAS3000 luminescent image analyzer (Fuji Film). Uncropped and unprocessed full scan images are shown in Supplementary  information (Supplementary Fig. 14).
Microarray analysis. Microarray analysis was performed with an Agilent gene expression microarray platform (SurePrint G3 Mouse Gene Expression 8 × 60 K). We visualized genes with the largest variance using a hierarchical clustering heatmap. For each gene, we compared expression levels between floxed and Fktn-cKO RNAs. Gene expression differences were evaluated using the eBayes function in the limma package 66 , microarray data were fit to a linear model, and differential expression was estimated by empirical Bayes moderation. The resulting P-values were corrected via the Benjamini and Hochberg method using the topTable function. Differentially expressed genes were defined as those with Log 2 changes of at least 1.5-fold between a pair of samples at an FDR of 0.001 for genes. Supplementary Table 1 provides the complete dataset. For differentially expressed genes, we carried out functional annotation analysis using DAVID. Differentially expressed genes were used as the input gene list, and all mouse genes that were expressed in the heart were used as the background. We looked for enrichment of genetic associations within KEGG pathways. Enrichment pathways were analyzed using ReactomePA.
Data analysis. Data were analyzed by individuals who were blinded to the genotype, drug treatment, and operation. Data presented here were reproducible in at least three independent experiments. Results are shown as the means ± s.e. m. Paired data were evaluated using a Student's t-test. For multiple comparisons, analyses of variance with Tukey-Kramer tests were performed as appropriate using GraphPad Prism. The Kaplan-Meier method, with a log-rank test, was used for survival analysis. A P value of < 0.05 was considered statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request. The source data underlying Figs. 1e, 3c, 4c, 4e, and 8b, d are provided as Source Data file. The gene expression data were deposited at NCBI's Gene Expression Omnibus (GEO). It is accessible through GEO series accession number of GSE138280.