A novel mechanism of cardioprotection through TNF-α-induced ectopic expression of keratins K8 and K18

Abstract

The adult myocardium demonstrates a unique system of adaptation upon stress stimuli, in an effort to maintain its overall homeostasis. This compensatory mechanism remains a mystery¹. Tumor Necrosis Factor-α (TNF-α) is one of the major stress-induced pro-inflammatory cytokines that is up-regulated in heart failure^1,2 and its sustained expression is considered detrimental for the heart^1,3–9. Although previous studies have shown that lower levels of TNF-α confer cytoprotection in the myocardium following ischemic reperfusion injury¹⁰, such action in heart failure remains elusive. Here we propose a novel cardioprotective function for TNF-α overexpression in a genetic heart failure model, the desmin deficient mice, through NF-κB-mediated cardiomyocyte ectopic expression of keratin 8 (K8) and keratin 18 (K18)¹¹, two simple epithelia-specific Intermediate Filament (IF) proteins. The ectopically expressed K8 and K18 (K8/K18) form a cytoskeletal network that localizes mainly at the Intercalated Discs (IDs). This alternative K8/K18 cytoskeleton confers cardioprotection by a mechanism that maintains ID and mitochondrial integrity and function. Importantly, we demonstrated that K8/K18 ectopic induction takes place in other genetic and experimental models of heart failure and showed a cardioprotective function in mice subjected to transverse aortic constriction. Finally, we discovered that in cardiomyocytes of human failing myocardium, where TNF-α is induced², K8/K18 are also ectopically expressed and localize primarily at IDs, where desmin cannot be detected. This is the first report to propose a TNFα-mediated cardiac ectopic expression of K8/K18 IF proteins, which may act as stress-induced cardioprotective factors in the failing heart, a phenomenon of major clinical significance as it also extends to human heart failure.

The role of the pleiotropic intercellular cytokine TNF-α in the heart pathophysiology has been considerably debated in the scientific community¹². Although it is not constitutively expressed in the heart^13,14, it is rapidly and consistently expressed in response to various forms of myocardial injury as part of an intrinsic cardiac response system¹.

Accumulating data from various models established that TNF-α engenders deleterious effects in the myocardial structure and function, which mimic the characteristic phenotype of heart failure^1,3–9. On the other hand, little is known about the possible beneficial effects of TNF-α in the heart^10,15–17, while its potential cardioprotective mechanisms remain elusive. Uncovering the mechanisms of TNF-α-mediated cardioprotection, may explain the failure of anti-TNF-α clinical trials¹⁸ and allow the development of more efficient therapeutic approaches for human heart failure.

A critical step towards the delineation of TNF-α action in the myocardium, emerged from our recent study using the “MHCsTNFα” mice³ (myosin heavy chain promoter-driven cardiac-specific overexpression of TNF-α, hereafter named “TNF-α” mice), in which the TNF-α-mediated detrimental effects had been well demonstrated^3,5,19,20. We showed that desmin, the muscle-specific IF protein, is a major target in TNFα-induced cardiomyopathy^21,22. Specifically, desmin is cleaved by TNF-α-induced caspase-6, loses its proper ID localization and forms aggregates. In TNF-α mice expressing a caspase cleavage-resistant desmin mutant (D263E), cardiac myocyte apoptosis was attenuated, left ventricular (LV) wall thinning was prevented and cardiac function was improved, thus revealing an important role for desmin cleavage in the development of dilated cardiomyopathy and heart failure.

The clinical significance of desmin has been demonstrated by the discovery of a plethora of mutations in the desmin gene that have been associated with desmin-related myopathies^22,23. A milestone towards understanding the significance of the IF cytoskeleton in the myocardium was the generation of desmin-deficient (Des−/−) mice^24,25, a widely established model of Dilated Cardiomyopathy (DCM) and heart failure, characterized by myocardial necrosis, early inflammation and extensive fibrosis and calcification^24,26.

Surprisingly, crossing TNF-α and Des−/− mice, two known heart failure models, results in a considerable rescue of the typical Des−/− extensive myocardial degeneration. Extensive fibrotic lesions (characterized as “replacement fibrosis”) and calcified areas, characteristic of Des−/− pathology^24,26, are totally absent in TNFαDes−/− myocardium (Fig. 1a). There is also a marked reduction of ventricular wall thinning and dilation (Fig. 1a), further verified by echocardiographic analysis (Supplementary Fig. 1b–c and Table 1). The typical myocardial necrosis and inflammation are significantly attenuated in TNFαDes−/− littermates (Fig. 1b and Supplementary Figs 2 and 3). Cardiomyocyte ultrastructural defects, particularly in mitochondrial morphology and distribution and secondarily in myofibril integrity, are hallmarks of the desmin-deficient cardiomyopathy²⁷. Transmission electron microscopy of both cardiac sections as well as isolated mitochondria demonstrated that the overall ultrastructure of TNFαDes−/− cardiomyocytes is significantly improved compared to Des−/− myocardium (Fig. 1c, 1e and Supplementary Fig. 4e–f). Specifically, the aberrant matrix fragmentation and abnormal cristae remodeling characteristic of Des−/− mitochondria²⁷ decrease extensively in TNFαDes−/− mitochondrial preparations. (Fig. 1e and Supplementary Fig. 5e). Furthermore, the overall cardiac mitochondrial function which is compromised in Des−/− heart is extensively rescued upon TNF-α overexpression. Specifically, the prominent decrease of Des−/− mitochondrial respiratory function, as estimated by measuring oxygen consumption (respiratory control ratio, State-III/State-IV) (Fig. 1d and Supplementary Fig. 5a) in isolated mitochondria of 3-mo-old hearts, was significantly improved in TNFαDes−/− hearts. This finding was in agreement with extensive rescue in TNFαDes−/− hearts of ATP production levels (Supplementary Fig. 5b) and redox activities, as measured by analysis of the reduction efficiency of GSSG and NADP⁺ to GSH and NADPH respectively (Supplementary Fig. 5c–d). Furthermore, this unexpected fundamental attenuation of heart abnormalities in TNFαDes−/− mice is further reflected to significant improvement of physiological cardiac functional properties, like fractional shortening (FS) (TNFαDes−/− vs Des−/−: 37.69% vs 34.63% at 3-mo-old and 35.73% vs 27.03% at 6-mo-old mice respectively; p-value<0.001) (Fig. 1f, Supplementary Fig. 1 and Table 1).

Figure 1. TNF-α overexpression rescues myocardial degeneration, ultrastructural abnormalities, mitochondrial defects and overall cardiac dysfunction caused by desmin deficiency.

(a) Representative pictures of whole hearts and the corresponding histology by Masson’s trichrome staining of cardiac sections of 3-mo-old mice (n=4 mice per genotype). Calcification (white regions) is depicted (asterisk) in Des−/− heart. In histological sections, fibrotic areas with collagen deposition detected in Des−/− hearts are stained in dark green. (b) Detection of necrotic areas by Evans Blue Dye (red fluorescence) staining of cryosections of TNFαDes−/− and Des−/− myocardium of 4-mo-old mice. Representative pictures of n=3 mice per genotype. Scale bars: 100 µm. (c) Electron microscopic analysis of the myocardium of 3-mo-old mice (n=3 mice per genotype).. Scale bars: 1 µm. (d) Oxygen consumption levels (d´) of isolated cardiac mitochondria from 3-mo-old mice, measured by Clark-type oxygen electrode and presented as respiratory control ratio (RCR; State-III/State-IV) levels. Representative cardiac mitochondrial respiration traces (d´´) of the indicated genotypes. 150 µg of mitochondria (arrow, “mito”) where added to 300 µl EB in presence of 5mM Glutamate - 2.5mM Malate. After 3 min, 400 µM ADP was added (“State-III”) and after 6 min, 1 µM oligomycin (“State-IV”), as indicated by arrows. For sample size, see Supplementary Fig. 5a. (d) Transmission electron microscopy of isolated cardiac mitochondria (n=4 mice per genotype). “Class-I” (arrows) and “class-IV” (arrowheads) mitochondria are depicted (as explained in Supplementary Fig. 5),. Scale bars: 1 µm. (e) Fractional shortening percentage (FS%) graphs from cardiac function analysis by 2D-directed M-mode echocardiography in 3-mo-old and 6-mo-old mice. For sample size, see Supplementary Table 1. All figure data are presented as mean ± SEM. Two-sided P-value: ** <0.01, *** <0.001, **** <0.0001 (one way ANOVA with Bonferroni/Dunn post-hoc test).

In an effort to further understand the mechanism by which TNF-α rescues the Des−/− myocardium, we analyzed the transcriptome changes in TNFαDes−/− relative to Des−/− hearts, by whole genome microarray analysis. Among 646 genes with significantly altered expression (FC≥±2, p-value≤0.05) in TNFαDes−/− heart, 206 are down-regulated and 440 up-regulated (Supplementary Fig. 3). Data meta-analysis revealed that the up-regulated genes (FC≥2) are mainly implicated in ‘cell death’ and ‘cell-to-cell signaling and interaction’ responses (Supplementary Fig. 3c). On the other hand, the most prominent down-regulation is observed in genes associated with inflammation/immunity and tissue remodelling related processes (Supplementary Fig. 4c, Table 2 and Data Set 1), reflecting the absence of leukocyte infiltration and inflammation in TNFαDes−/− hearts (Supplementary Fig. 2), in contrast to Des−/− hearts²⁶,.

Interestingly, among the most pronounced and surprising changes was the high-level ectopic induction of krt8 and krt18 genes, encoding for the epithelial intermediate filament proteins, keratin 8 (K8) and keratin 18 (K18) (16.44 and 25.96 fold increase respectively in TNFαDes−/− vs Des−/− hearts; see Fig. 2a and Supplementary Table 2). K8 and K18 are normally expressed in the simple epithelium of various organs (such as liver, kidney, mammary gland etc.)¹¹ and have never been detected in the adult myocardium or in other muscle types, following the known strictly regulated tissue-specific pattern of IF proteins^11,28,29. However, very low expression of another keratin pair, K8/K19, has been reported in striated muscle³⁰. Real-Time PCR analysis confirmed the high levels of ectopic krt8 and krt18 RNA expression in TNFαDes−/− hearts (53.2 and 62.7 fold increase, respectively, in TNFαDes−/− vs WT) and further established that this induction takes place in TNF-α (Des+/+) hearts as well (Fig. 2b). The expression of the krt8 and krt18 genes was also confirmed at the protein level (Fig. 2c). Furthermore, we showed by immunofluorescence microscopy that K8/K18 are expressed throughout the myocardial area in TNFα-overexpressing mice (Fig. 2i). This expression pattern is evident in the hearts from neonatal (Supplementary Fig. 6) till aged TNFαDes−/− mice and persists at all age groups studied (data not shown). Control WT myocardium was always negative for K8/K18, both by western (Fig. 2c) and immunofluorescence microscopy analysis (Fig. 2i), as expected from previous studies^11,28,29. Intriguingly, in Des−/− myocardium K8/K18 expression is detectable, although less evident and in scattered islets of cardiomyocytes (Supplementary Fig. 7).

Figure 2. Keratin 8 (K8) and keratin 18 (K18) are ectopically expressed in the TNF-α-overexpressing cardiomyocytes.

(a) Microarray gene expression analysis comparing TNFαDes−/− versus Des−/− transcriptome (n=3 hearts per genotype, n=3 technical replicates). Heatmap view presenting the hybridization triplicates (a´), and MvA scatter plot (a´´) of all entities (FC≥±2, p-value<0.05), showing krt8 and krt18 among the top 3 of the mostly induced genes (a´: red boxes; a´´: red arrows) (Unpaired t-test and Benjamini-Hochberg correction). (b) Real-Time PCR analysis of total myocardial RNA for krt8 and krt18 expression relative to WT (n=6 per genotype, n=4 technical replicates, one way ANOVA with Bonferroni/Dunn post-hoc test). Beta-actin was used as reference gene. (c) Western blot analysis for K8 and K18 of total myocardial extracts from 3- and 6-mo-old mice (n=4 mice per genotype). Gapdh levels were used as loading control. (d) Chromatin-IP with anti-p65/RelA (n=3 mice per genotype, n=3 biological replicates) followed by RT-PCR analysis. “K18R1P3”, “K18R2P2”, “K8R2P1” and “K8K18R2P2” correspond to regulatory regions of krt8 and krt18 loci that showed significant enrichment in ChIPed DNA from TNFαDes−/− (TD) compared to WT hearts (for region analysis see Supplementary Fig. 6). (two way ANOVA with Bonferroni/Dunn post-hoc test). (e) Real Time-PCR for krt18 expression in the myocardium of IKK2^MyHC mice with cardiac-specific expression of constitutively active IKK2, 5-weeks and 12-weeks after doxycycline withdrawal relative to their appropriate controls (n=6 minimum per group, n=3 technical replicates, unpaired t-test). Rpl13 was used as reference gene. (f) Immunofluorescence staining for K8 (red) and desmin (green) of cardiac sections of IKK2^MyHC and their control mice, 12-weeks after doxycycline withdrawal (n=4 per group). Scale bars: 50 µm. (g) Real Time-PCR for krt8 and krt18 expression in the myocardium of 6-weeks-old TNFαM3 mice (cross of TNF-α and mice with cardiac-specific expression of IκBα suppersuppressor) relative to their TNF-α littermates (n=11 and n=6 per group, n=3 technical replicates, unpaired t-test). Beta-actin was used as reference gene. (h) Firefly luciferase reporter activity assays of WT (Krt8luc and Krt18luc) and mutant (µ2Krt8luc and µ2Krt18luc) 2-kb regulatory regions of krt8 and krt18, upon p65 induction. Luciferase activity is presented relative to the respective non-induced WT reporter. All experiments (n=8 minimum) were normalized to β-galactosidase activity (one way ANOVA with Bonferroni/Dunn post-hoc test). (i) Confocal laser-scanning immunofluorescence microscopy for K8 (green) expression in cardiac sections of 3-mo-old mice (n=5 mice per genotype). (K18 exhibits similar pattern (not shown).) Scale bars: 100 µm. (j) Co-staining for K8 and K18 in TNFαDes−/− (n=5 mice 3-mo-old) myocardium. Co-localization of K8 and K18 at the intercalated disc (arrowhead) and costamere-level (arrows) is presented. Scale bar: 20 µm. (k) Co-staining of K18 with desmoplakin (DSP) (k´) and β-catenin (β-cat) (k´´) of cardiac sections from 3-mo-old mice (n=4 mice). Scale bars: 20 µm. (l) Electron micrograph of immunogold labeling for keratin 8 (black dots) in TNFαDes−/− cardiac section (n=2 mice, 3-mo-old). ID plaques (arrows) and K8/K18 filaments (arrowhead) are shown. Scale bar: 300 nm. All figure data are presented as mean ± SEM. Two-sided P-values: * <0.05, ** <0.01, *** <0.001.

To address the mechanism of the K8/K18 ectopic induction in TNFαDes−/− myocardium, we explored the potential involvement of Nuclear Factor-κB (NF-κB), known to be activated by TNF-α, on the transcriptional induction of both krt8 and krt18 genes. Due to very limited information on krt8 and krt18 gene regulation, a bioinformatics analysis (including ENCODE Consortium data and transcription factor binding sites prediction tool, Supplementary Tables 3 and 4) was performed, which revealed several potential regulatory regions in krt8/krt18 loci (Supplementary Fig. 8). Following in vivo cardiac chromatin immunoprecipitation experiments for the p65/RelA subunit of NF-κB, these regions were used as targets for RT-PCR enrichment of the precipitated chromatin fragments. Our data imply that NF-κB is involved in the ectopic activation of both krt8 and krt18 gene transcription in TNFαDes−/− heart, by binding to at least one genomic regulatory locus of each gene (K8R2P1 region for krt8, K18R1P3 and K18R2P2 regions for krt18) and to an intergenic region of potential regulation (K8K18R2P2) (Fig 2d and Supplementary Fig. 8).

The NF-κB involvement in the ectopic induction of K8/K18 in the heart was further verified using an in vivo model of cardiac-specific transgenic expression of constitutively active IKK2 (IKK^MyHC mice), the kinase responsible for canonical NF-κB activation³¹. Indeed, in contrast to control group, IKK^MyHC mice showed ectopic expression of K8/K18 in the myocardium (Fig. 2e–f and Supplementary Fig. 9a–b). Induction was evident both before (5-weeks of doxycycline withdrawal) and after (12-weeks of doxycycline withdrawal) manifestation of an overt heart failure phenotype, which in these mice usually starts to develop 6- to 8-weeks post doxycycline discontinuation³¹. Furthermore, complementary to the IKK^MyHC gain-of-function in vivo model, a ‘loss-of-function’ in vivo approach was also utilized to verify the NF-κB involvement in K8/K18 ectopic induction in the heart. Crossing the TNF-α mice with a model of NF-κB signaling blockade, conferred by cardiac-specific expression of an IκBa superreppressor (IκBα-3M mice)³², resulted in a significant decrease of both krt8 and krt18 ectopic expression in the heart (29% and 41% respectively, p-value<0.05)(Fig 2g).

Reporter assays also verified the in vivo data on the involvement of NF-κB in the transcriptional regulation of krt8 and krt18. By using a realistic system of a 2-kb-long genomic region upstream of their translation start sites (see Supplementary Fig. 9c), thus including all the regulatory loci, as described in Supplementary Figure 8, we showed that both krt8 and krt18 are significantly responsive to p65 induction (1.91 and 2.08 respectively, fold change relative to non-induced, p-value<0.001)(Fig. 2h). Furthermore, mutations on the specific κB-sites found to bind p65 in vivo (Fig. 2d and Supplementary Fig.8), were adequate to significantly reduce the p65-dependent krt8 and krt18 induction (23% and 40% respectively, relative to WT-induced, p-value<0.05)(Fig. 2h and Supplementary Fig. 9e). The effect of the specific κB-site mutations on the krt8 and krt18 regulation was also observed under baseline conditions (Supplementary Fig. 9d–e).

Using immunofluorescence microscopy, we demonstrated that the ectopic expression of both K8/K18 proteins in TNF-overexpressing hearts is restricted to cardiomyocytes, where they fully co-localize (Fig. 2j), while they cannot be detected in cardiac fibroblasts (Supplementary Fig. 10). The ‘area composita’ (fusion of desmosomes and fasciae adherente)³³ of the intercalated discs (IDs) is the major localization site of the K8/K18 pair in TNFαDes−/− cardiomyocytes (Figs 2j–k and Supplementary Fig. 4a). Similarly to desmin in normal myocardium, K8/K18 co-localize with major components of desmosomes and adherens junctions, such as desmoplakin (DSP) and β-catenin (β-cat) respectively, at the IDs of TNFαDes−/− myocardium (Fig. 2k and Supplementary Fig. 4a). Z-disc localization was not detected, however, a sarcoplasmic striated pattern of K8/K18 at the M-line level of the myofibrils extending to the costameres could be observed (Fig. 2j and Supplementary Fig. 4b). Immunoelectron microscopy further confirmed the keratin 8 specificity for the cardiomyocyte ID plaques and allowed the detection of the filamentous organization of the ectopic K8/K18 network (Fig. 2l and Supplementary Fig. 4c).

Keratins have been linked to several mechanical and non-mechanical cytoprotective functions^34–40. Notably, K8/K18 have been involved in several diseases, from liver pathophysiology to cancer, in modulation of signal transduction pathways, protection from cell death, protein targeting and organelle regulation and transport^{34–37,39–43}. Therefore, we asked whether the highly expressed and specifically localized K8/K18 in TNFαDes−/− cardiomyocytes could be responsible for the TNF-α-mediated rescue of the myocardial degeneration and dilated phenotype observed in the Des−/− mice. To address this issue, we generated TNFαDes−/−Krt18−/− mice, which provide a K8/K18 deficient model, given that keratin 8 is degraded in the absence of its partner⁴¹.

While gross morphology of TNFαDes−/−Krt18−/− hearts appears similar to those from TNFαDes−/− mice, histological sections reveal a more prominent dilated phenotype (Fig. 3a), and this is further verified by echocardiographic analysis (Supplementary Fig. 1 and Table 1). Additionally, interstitial fibrosis appears to be increased in TNFαDes−/−Krt18−/− hearts and areas of replacement fibrosis are occasionally evident (Fig. 3a´), in contrast to TNFαDes−/− myocardium. Echocardiographic data prove that the TNF-α-mediated cytoprotective action is lost in the absence of K8/K18, manifested by a significant reduction in the overall cardiac function in TNFαDes−/−Krt18−/− compared to TNFαDes−/− mice (FS: 31.64% vs 37.69% at 3-mo-old and 27.68% vs 35.73% at 6-mo-old respectively; p-value<0.0001) (Fig. 3b, Supplementary Fig. 1 and Table 1). While the worse pathophysiology of TNFαDes−/−Krt18−/− hearts clearly demonstrates the K8/K18 cardioprotective effect, in the TNFαKrt18−/− myocardium such an effect is not profound. No significant changes in fractional shortening (Supplementary Table 1) and histology (data not shown) compared to TNF-α mice, are observed, consistent with the known dominant effect of the defective desmin network in this model²¹. However, lack of K8/K18 leads to a more dilated phenotype in TNFαKrt18−/− mice (Supplementary Fig. 1 and Table 1), demonstrating a cardioprotective effect of K8/K18 in TNF-α model as well. Krt18−/− control mice appear similar to WT by morphological, histological and echocardiographic analysis (Figs 3a–b), while the Des−/−Krt18−/− myocardium demonstrates similar to Des−/− pathologic characteristics, including extensive replacement fibrosis, calcification and myocardial degeneration (data not shown).

Figure 3. Ectopic expression of K8/K18 confers cardioprotection through maintenance of mitochondrial function and intercalated disc integrity in TNFαDes−/− mice.

(a) Representative pictures from morphological and histological analysis by Masson’s trichrome staining of myocardium from 3-mo-old mice, comparing TNFαDes−/−Krt18−/− with TNFαDes−/− models (n=4 mice per genotype). The myocardium of Krt18−/− control mice presents normal phenotype. Replacement fibrosis of TNFαDes−/−Krt18−/− is depicted in dark green in a´ (enlargement). (b) Fractional shortening percentage (FS%) by 2D-directed M-Mode echocardiographic analysis of 3-mo-old and 6-mo-old mice from all groups. For sample size see Supplementary Table 1. The data are presented as mean ± SEM. Two-sided P-values: *** <0.001, **** <0.0001 (one way ANOVA with Bonferroni/Dunn post-hoc test). (c) Electron microscopic analysis of TNFαDes−/−Krt18−/− myocardium of 3-mo-old mice (n=2). Severe mitochondrial abnormalities, including fragmentation, disorganization and degeneration (arrows) are shown, together with abnormal organization of myofibrils compared to TNFαDes−/− cardiomyocytes (see Fig. 1c and Supplementary Fig. 3e–f). Scale bar: 2 µm. (d) Oxygen consumption levels (d´) of isolated cardiac mitochondria from 3-mo-old mice, measured by Clark-type oxygen electrode and presented as respiratory control ratio (RCR; State-III/State-IV) levels. Representative cardiac mitochondrial respiration traces (d´´) of the indicated genotypes. 150 µg of mitochondria (arrow, “mito”) where added to 300 µl EB in presence of 5mM Glutamate - 2.5mM Malate. After 3 min, 400 µM ADP was added (“State-III”) and after 6 min, 1 µM oligomycin (“State-IV”), as indicated by arrows. For sample size, see Supplementary Fig. 5a. (e) Immunofluorescence microscopy for the ID proteins desmoplakin (DSP), β-catenin (β-cat) plakoglobin (PG) and with the Z-disc protein α-actinin in cardiac sections of 3-mo-old mice (n=3 mice per group) reveals the mis-localization of the ID proteins and their disrupted pattern in TNFαDes−/−Krt18−/− compared to TNFαDes−/− and WT control cardiomyocytes. Scale bars: 20 µm. (f) Electron microscopic analysis comparing the highly abnormal intercalated disc structures (arrowheads) of TNFαDes−/−Krt18−/− compared to that of TNFαDes−/− and control WT cardiomyocytes (n=2 mice of 3-mo-old per genotype). Scale bars: 500 nm.

One of the important functional roles of K8/K18 stems from their interaction with desmosomes, the intercellular adhesive junctions of epithelial cells. This critical interaction confers stabilization and maintenance of the desmosome structures⁴⁴, proper targeting of desmosomal proteins⁴³, subsequent maintenance of surface membrane integrity⁴⁵ and facilitation of intercellular mechanical and ionic communication^45,46. Complementary to protein targeting function, another important role of IFs is organelle modulation, specifically that of mitochondria^{27,37,47–49}. Evidence for the importance of IF cytoskeleton in mitochondrial structure and function was provided by previous studies in Des−/− mice^27,50, supporting the data of Figure 1. Additionally, recent data suggest a similar cytoprotective mechanism of K8/K18 (or K8/K19) at the mitochondrial level^42,49,51,52. Therefore, we hypothesized that the mechanism of protection by K8/K18 in TNFαDes−/− mice could be through maintenance of mitochondrial, ID and myofibril integrity.

Electron microscopy studies revealed that the TNFαDes−/−Krt18−/− myocardium presents severe ultrastructural defects. Mitochondria accumulate and aggregate atypically and are disarranged (Fig. 3c and Supplementary Fig. 11a), while myofibril organization is also defective, compared to TNFαDes−/− myocardium (Fig. 3c and Supplementary Fig. 4e–f). Mitochondrial abnormalities resemble the characteristic defects in desmin-deficient cardiomyopathy⁴⁷, reflecting a central role of IFs in mitochondrial homeostasis and suggesting that K8/K18 can compensate for proper mitochondria regulation in the absence of desmin. Indeed, partially or totally swollen mitochondria (“phase-III” and “phase-IV”, see Supplementary Fig. 5 legend) are prominent in TNFαDes−/−Krt18−/− hearts (Fig. 3c and Supplementary Fig. 5e). Importantly, the ultrastructural abnormalities also reflect a significant impairment of mitochondrial functional status in the absence of K8/K18 network, as their respiration is severely defective (Fig. 3d and Supplementary Fig. 5a). Furthermore, ATP levels (Supplementary Fig. 5b) and redox activities (Supplementary Fig. 5c–d) are abnormal compared to TNFαDes−/− cardiac mitochondria, similarly to Des−/−. These findings support a mitoprotective effect of K8/K18 network in TNFαDes−/− cardiomyocytes, as a mechanism of TNF-mediated cardioprotection in Des−/− myocardium.

Furthermore, the IDs of TNFαDes−/−Krt18−/− cardiomyocytes are severely abnormal, being discontinuous and amorphous compared to typical IDs of WT and TNFαDes−/− myocardium (arrowheads in Fig. 3f). Consistent with this finding is the mis-targeting of critical components of the ‘area composita’, including Desmoplakin, Plakoglobin and β-Catenin in TNFαDes−/−Krt18−/− myocardium (Fig. 3e) and Supplementary Fig. 11b). This finding suggested that the proper targeting or maintenance of these proteins to IDs is part of the mechanism by which the K8/K18 network compensates for the desmin deficiency in TNFαDes−/− hearts.

A critical question raised by our findings was whether the K8/K18 expression in TNF-overexpressing mice reflects a common response in heart failure. To address this important issue we introduced additional, well established, genetic and experimental models of heart failure and investigated the krt8 and krt18 expression status in the heart. Importantly, we found that both these genes are ectopically induced in transgenic hearts with cardiac-specific overexpression of constitutively active Calcineurin (MHC-Cn)⁵³, a calcium-dependent phosphatase, and in muscle–specific LIM-only protein knockout (MLP−/−)⁵⁴ hearts (Fig. 4a´). Notably, they present the same protein localization properties described for TNF-overexpessing cardiomyocytes (Fig. 2j–k), in a significantly lower extent though, forming a network localized in the cardiomyocyte IDs and costameric striations (Fig. 4a´´–a´´´). Moreover, we showed that K8/K18 ectopic network is also evident in experimentally induced HF (Fig. 4b). Cardiomyocytes of WT mice subjected to either transverse aortic constriction (TAC) (Fig. 4b´–b´´ and Supplementary Fig. 12a–d) or myocardial infarction (MI) by permanent ligation of the left anterior descending (LAD) coronary artery, exhibit the characteristic pattern as discussed for the genetic models of HF. All appropriate WT control groups studied were never found positive for K8/K18 expression. These data provide strong evidence for a global nature of K8/K18 ectopic induction in stressed or failing cardiomyocytes.

Figure 4. Cardiomyocyte ectopic induction of K8 and K18 is a common response in several genetic and experimental cardiomyopathy models and expands in human heart failure.

(a) Real-Time PCR (a´) for krt8 and krt18 RNA (n=5 at least per group, n=3 technical replicates, unpaired t-test) and immunofluorescence staining for K8 and a-actinin (a´´) (n=2 per group) of the myocardium of 12-weeks-old transgenic mice overexpressing α-MHC promoter-driven active Calcineurin (“MHC-Cn”) and muscle LIM-protein knockout (“MLP−/−“) mice. Expression is presented relative to WT groups. Rpl13 was used as reference gene. Scale bars: 20 µm. (b) Real-Time PCR (b´) for krt8 and krt18 RNA (n=5 at least per group, n=3 technical replicates, unpaired t-test) and immunofluorescence staining for K8 and a-actinin (b´´) (n=2 per group) of the myocardium of 8-weeks-old WT C57/BL6 mice, 2-weeks and 12-weeks after transverse aortic constriction (“2-w pTAC” and “12-w pTAC”, respectively) and myocardial infarction by permanent ligation of the left anterior descending (LAD) coronary artery for 4-weeks (“MI”). Expression is presented relative to WT groups. Rpl13 was used as reference gene. Scale bars: 20 µm. (c) 2D-directed echocardiographic analysis of fractional shortening (FS) (c´) and relative wall thickness (RWT) (c´´) of 4-months-old WT and krt18−/− mice, before subjected to transverse aortic constriction (TAC) surgery (baseline, “BSL”) and 2-weeks and 4-weeks post surgery (“2-w pTAC” and 4-w pTAC” respectively) (for number of animals per group see Supplementary Table 5). Two-sided P-values: * <0.05, ** <0.01, *** <0.001 relative to baseline values, two-sided P-values: # <0.05, ## <0.01 relative to WT (c´) and to Krt18−/− (c´´) group (two way ANOVA with Bonferroni/Dunn post-hoc test). (d) Immunofluorescence co-staining for K8, a-actinin and β-catenin of cardiac sections of WT post TAC mice used in Figure 4c, 8-weeks post surgery (n=3). Scale bars: 20 µm. (e) Western blot analysis for K8, K18 and desmin of human end-stage heart failure cardiac samples (n=10 different patients), obtained after transplantation. Representative blots of total cardiac protein extracts from patients #1 – #6 (additional n=4 patient samples are shown in Supplementary Fig. 13a) and control human myocardial sample (n=1 sample of healthy human myocardium) are presented. Arrows at the lower panel show caspase-6-mediated cleavage products of desmin at 24 and 29 kDa. Gapdh levels were used as loading control. (f) ELISA for human TNF-α in the serum of the same patients used in the study (n=8). Data are presented as fold change (FC) relative to a normal group (n=10) (unpaired t-test). (g) Immunofluorescence microspopic analysis by co-staining for keratin 18 (K18) and desmoplakin (DSP) of human end-stage heart failure cardiac sections (n=3 different patient samples), reveals their co-localization at the IDs. Scale bar: 30 µm. (h) Co-staining for K8 with K18 (h´) and K8 with Desmin (h´´) in cardiomyocytes of human patients of end-stage HF (n=3 patients). Scale bars h´: 30 µm; h´´: 20 µm. (i) Co-staining of Desmin (DES) and Desmoplakin (DSP) in human myocardial sections of control sample (n=2) (i´). Desmin and Desmoglein-2 (DSG2) co-staining in cardiac sections of patient samples (n=3) (i´´). Notice the loss of desmin from patient IDs (h´´, i´´). Scale bars: 20 µm. Projection images from confocal stacks used at (g–i). All figure data are presented as mean ± SEM. Two-sided P-values: * <0.05, ** <0.01.

An important finding supporting the cardioprotective effect of K8/K18 ectopic induction in the heart was provided using the TAC experimental model of HF. We demonstrated that mice deficient in K8/K18 network (Krt18−/−) exhibit very early, by two weeks post-TAC, a heart failure response, as their hearts show a significant decrease of 25.1% (Fig. 4c´ and Supplementary Table 5) in the fractional shortening compared to WT baseline levels (baseline FS: 52.21%; 2-weeks post-TAC FS: 39.09%) (Fig. 4c and Supplementary Fig. 12e and Table 5). Most possibly, this is due to failure of Krt18−/− hearts to compensate for the pressure overload stress in contrast to WT mice, given that they do not undergo the normal hypertrophic response observed in WT mice justified by a significant increase of relative wall thickness (RWT)(Fig. 4c´´). Instead, Krt18−/− hearts develop dilation as measured by a significant decrease in left ventricular dimensions in systole (LVIDs) (Supplementary Fig. 12e´), in contrast to WT. Consistent to the findings of Figure 4b, the WT TACed mice used in this study were found positive to K8/K18 network formation presenting the typical ID and striated pattern in cardiomyocytes (Fig. 4d and Supplementary Fig. 12a–d).

Finally, we set to answer the next important question about the relevance of our findings in human heart failure as well, where TNF-α has been shown to be up-regulated². Using cardiac samples from patients that underwent heart transplantation, we demonstrated that, indeed, both keratin 8 and keratin 18 proteins are expressed in human failing heart, showing variable levels of K8 and K18 ectopic expression between the patients (Fig. 4e and Supplementary Fig. 13a). Serum TNF-α levels in the group of these patients were found significantly increased compared to a normal group (Fig. 4f), as expected². Similarly to our findings in murine models of HF, we found that K8 and K18 co-localize in the cardiomyocytes of the human failing myocardium (Fig. 4h´ and Supplementary Fig. 13c), supporting the clinical importance of our observations. Similarly to our studies with mice, K8/K18 expression occurs specifically in cardiomyocytes and not in cardiac fibroblasts (Fig. 4g and Supplementary Fig. 10b). Moreover, K8/K18 are clearly detected at the IDs where they co-localize with desmosomal (Fig. 4g) and adherens junction proteins (data not shown). No K8/K18 formation was detected in cardiomyocytes of normal human myocardium (Supplementary Fig. 13b), in agreement with the literature^11,28,29. Thus, the significantly lower levels of K8/K18 detected by WB in control myocardium (Fig. 4e and Supplementary Fig. 13a) may correspond to non-cardiomyocyte cells of the epicardium or endothelial cells of the heart microvasculature⁵⁵.

Based on our previous data in mice²¹, we also investigated the characteristics of desmin network in human heart failure. Importantly, we found that desmin is significantly accumulated in human failing myocardium in all cases tested (Fig. 4e). Desmin protein levels are elevated in the patient samples compared to healthy myocardium, most possibly due to lower degradation rate of aggregated desmin²¹. Furthermore, immunofluorescence microscopy revealed that, in agreement with our previous findings in mice²¹, desmin loses extensively its proper ID localization (Fig. 4i). Notably, K8/K18 show more efficient ID localization compared to desmin (Fig. 4h´´ and Supplementary Fig. 13c´´), consistent with our data in TNF-α mice (Supplementary Fig. 14). This could be possibly related to the potential modification of desmin by caspase-6-mediated cleavage, as we had previously demonstrated in TNF-α failing myocardium²¹. Indeed, desmin-cleavage products are evident in some of the patient samples and not in control human myocardium (arrows in Fig. 4e). This is the first report of desmin cleavage and mis-localization in human end-stage heart failure, the latter abnormality reported before only in Carvajal Syndrome⁵⁶.

Collectively our results demonstrate a novel cardioprotective role for TNF-α through de novo expression of epithelial IF proteins in the heart. This is the first study to report a reprogramming mechanism by which this cytokine may contribute to overcoming restrictions required to maintain tissue-specific gene expression in adult cardiomyocytes, thus leading to keratin 8 and keratin 18 expression. These proteins form obligatory heteropolymers, a prerequisite for the generation of a properly functioning intermediate filament network³⁴. Therefore, the potential of TNF-α to ectopically activate both genes through NF-κB plays a major role in the achieved cardioprotection. Furthermore, the ability of K8/K18 to maintain ID integrity and mitochondrial function strongly support the notion that this keratin network might be an ideal alternative IF cytoskeletal system to compensate for desmin deficiency.

These novel discoveries suggest the reconsideration of the present knowledge of myocardial biology and denote the need for re-evaluation of our current interpretation of mechanisms involved in heart pathophysiology, which could potentiate more effective therapeutic approaches in human heart disease. Furthermore, the potential diagnostic and even prognostic value of these findings for different types and stages of heart pathology is substantial and should be thoroughly investigated and clinically evaluated.