Desmin enters the nucleus of cardiac stem cells and modulates Nkx2.5 expression by participating in transcription factor complexes that interact with the nkx2.5 gene Open Access

The heart is the first organ developing in mammalian embryos and has to be functional throughout life (Vincent and Buckingham, 2010). Cardiomyogenesis is guided by a hierarchically structured network of transcription factor (TF) genes receiving versatile extrinsic and intrinsic signals during commitment of primitive mesoderm to the cardiogenic lineage and differentiation of cardiac progenitor cells (CPCs) (Taubenschmid and Weitzer, 2012). The transition from primitive mesoderm to early stages of CPCs is facilitated by the TFs brachyury (King et al., 1998), eomesodermin (Costello et al., 2011), and Mesp1 (Bondue and Blanpain, 2010; Bondue et al., 2008; David et al., 2008,, 2011) and certainly by many others that have not been well studied in this respect yet. Nkx2.5 and GATA4 are expressed downstream of the above mentioned TFs, and contribute together with Isl1 to fate decisions along the myocardial lineage (Den Hartogh et al., 2015; Dorn et al., 2015), finally leading to the formation of the four-chambered mammalian heart (Srivastava, 2006).

The homeobox TF Csx/Nkx2.5, a homolog of the invertebrate's TF tinman (Akazawa and Komuro, 2005; Chen and Schwartz, 1995; Harvey et al., 2002; Schwartz and Olson, 1999), plays a critical role in ensuring spatial and temporal discrimination between progenitor and differentiated states during cardiomyogenesis (Dorn et al., 2015; Prall et al., 2007), and contributes to homeostasis in the adult heart (Reamon-Buettner and Borlak, 2010). Nkx2.5 expression is regulated by a vast number of TFs, such as Mesp1, Tbx5, and GATA4 (Taubenschmid and Weitzer, 2012). On the other hand, Nkx2.5 interacts with TFs such as GATA4, Tbx5, and Mef2C in regulating downstream target genes. Cardiomyogenesis is severely hampered when Nkx2.5 is mutated or absent in vertebrates (Bodmer, 1993; Bruneau et al., 2000; Chen and Fishman, 1996; Kasahara et al., 2000), and mutant Nkx2.5 cause severe congenital heart diseases in humans (Reamon-Buettner and Borlak, 2010). The consequences of nkx2.5 haploinsufficiency in mice (Biben et al., 2000; Jay et al., 2005) and a negative auto-regulatory Nkx2.5 feedback loop (Prall et al., 2007; Tanaka et al., 1999) suggest that fine-tuning of the Nkx2.5 expression level is critical for proper CPC specification, cardiomyogenesis, and homeostasis of the adult heart (Akazawa and Komuro, 2003). This hypothesis is strengthened by the observation that over-expression of the muscle-specific intermediate filament (IF) protein desmin causes an up-regulation of brachyury and Nkx2.5 expression in CPCs followed by a significantly improved cardiomyogenic differentiation in embryoid bodies (EBs) (Hofner et al., 2007).

Desmin is one of the earliest cardiac muscle specific proteins expressed in mesodermal cells committed to the myocardial lineage (Capetanaki et al., 2015; Kuisk et al., 1996; Li and Capetanaki, 1994), in satellite cells (Allen et al., 1991), and in cardiac muscle side population stem cells (CSPCs) (Pfister et al., 2005). Desmin is a type III IF protein and a member of a large family of more than 70 proteins (Oshima, 2007). These proteins were originally believed to provide a static framework supporting the cytoarchitecture of all metazoan cells but there is accumulating evidence demonstrating that IFs are highly dynamic structures and that their subunits seem to contribute to a plethora of regulatory processes involved in differentiation, homeostasis, aging, and disease (Hyder et al., 2011). Knock out of the desmin gene, although without any obvious phenotypic consequences during murine embryogenesis (Li et al., 1996; Milner et al., 1996), causes severe cardiac defects during adulthood (Mavroidis et al., 2015; Milner et al., 2000,, 1996; Psarras et al., 2012; Thornell et al., 1997). In line with these data, a variety of mutations in the desmin gene have been linked to human skeletal and cardiac myopathies (Capetanaki et al., 2015). Absence of desmin in muscle cells leads to structural and functional mitochondrial defects (Milner et al., 2000; Papathanasiou et al., 2015; Weisleder et al., 2004), however, the consequences of its deficiency to nuclear functions is the most intriguing one. Skeletal muscle specific myogenic TFs MyoD and myogenin are down-regulated in the absence of desmin in C2C12 myoblasts (Li et al., 1994) and during embryonic stem cell (ESC) differentiation in desmin knockout EBs (Weitzer et al., 1995). Further, overexpression of desmin in differentiating ESCs increases the expression of the TFs brachyury and Nkx2.5 in developing CPCs (Hofner et al., 2007), and accelerates the commitment and differentiation of primitive mesodermal cells to rhythmically contracting cardiomyocytes (CMCs). Deletion of desmin's amino-terminal domain or mutation of serine residues 6, 7, and 8 or 31 and 32 to alanine, causes a significant down regulation of early Nkx2.5 expression in EBs and severely hampers cardiomyogenesis (Höllrigl et al., 2007,, 2002). Finally, in vivo, desmin and Nkx2.5 co-localize and are first detectable in the precardiac mesoderm between embryonic day 7.5 and 7.8 (Kasahara et al., 1998; Kuisk et al., 1996).

All these results suggest that desmin may also influence transcription of genes important for early cardiomyogenesis, in particular Nkx2.5, directly. This notion is supported by data demonstrating that desmin and vimentin interact with DNA in vitro (Tolstonog et al., 2005; Traub and Shoeman, 1994b). Desmin has been detected in nuclei of BHK21 cells (Kamei, 1986) and nestin in nuclei of brain tumor cells (Krupkova et al., 2011). It has been suggested that vimentin enters the nucleus via a piggyback mechanism (Hartig et al., 1998) and binds to DNA via its amino-terminal domain particularly at the nuclear matrix attachment regions (Tolstonog et al., 2001). Strong nuclear vimentin signals have also been found in lymph node metastasis from nasopharyngeal carcinoma (Luo et al., 2012). These independent lines of evidence suggest that type III IF proteins may enter the nucleus under certain circumstances, and influence transcriptional processes by interacting with DNA and proteinaceous components of the chromatin directly.

Identification of genes acting upstream of nkx2.5 is crucial for understanding the transcriptional network and the interwoven paracrine and autocrine signals that regulate the development of primitive mesoderm to functional CMCs. Thus, we decided to test if desmin promotes cardiogenic commitment and myocardial differentiation directly by influencing the transcriptional network in CPCs. Herein, we demonstrate that desmin is indeed involved in a temporary, cell autonomous molecular process influencing Nkx2.5 transcription in CPCs during early cardiomyogenesis and in adult CSPCs. Desmin activates transcription of nkx2.5 reporter genes and rescues the haploinsufficient phenotype in heterozygous nkx2.5 knockout EBs by increasing Nkx2.5 expression. Moreover, absence of desmin in CSPCs from adult desmin-knockout mice causes a reduced Nkx2.5 expression. These effects are mediated by the temporary presence of desmin in the nuclei of differentiating CPCs and the physical interaction of desmin with enhancer and promoter elements of the nkx2.5 gene. Thus, desmin contributes to the transcriptional regulation of the nkx2.5 gene in CPCs during a short period of time at the beginning of cardiomyogenesis and in CSPCs in the hearts of adult mice.

Previous results have demonstrated that early and increased expression of desmin in differentiating embryonic stem cells (ESCs) promotes cardiomyogenesis (Hofner et al., 2007). This becomes evident by temporarily increased expression of the cardiac transcription factor (TF) gene nkx2.5, followed by an increase in the number of rhythmically contracting cardiomyocytes (CMCs) in embryoid bodies (EBs). To test whether desmin influences Nkx2.5 expression in a cell-autonomous fashion or not, we chose two cell lines with different myogenic potentials and the ability to activate the nkx2.5 gene. The first cell line, 10T1/2 fibroblasts, expresses small amounts of desmin, if any at all, but have a latent myogenic (Weintraub et al., 1989) and cardiogenic (Zhou et al., 2012) potential, whereas the second one, C2C12 myoblasts, expresses significant amounts of desmin and can be induced to differentiate into multinucleated myotubes (Yaffe and Saxel, 1977). The influence of desmin on nkx2.5 gene expression was monitored first with the luciferase (LUC) reporter plasmid pNKE24 (Searcy et al., 1998), containing the proximal enhancer and promoter region (PEPR) of the nkx2.5 gene and then with the LUC reporter plasmid pMCE, containing the minimal cardiac specific enhancer region (MCE) (Lien et al., 1999) in addition to the PEPR (Fig. 1A; for precise localization along the nkx2.5 gene see Fig. 4A).

Co-transfection of 10T1/2 fibroblasts with pNKE24 and a desmin-expressing plasmid (desmin-ect.) resulted in a significant decrease of the LUC activity (Fig. 1B). This inhibitory effect of desmin could be reverted by co-transfection with a desmin-specific short hairpin RNA (shRNA) plasmid (des-shRNA) resulting in desmin expression reduced to basal levels (Fig. S1). In sharp contrast, co-transfection with pMCE conferred a desmin-dependent increase in the expression of the nkx2.5 LUC reporter.

Transfection of C2C12 myoblasts with pNKE24 resulted in a much larger basal LUC activity compared to 10T1/2 fibroblasts, suggesting that C2C12 myoblasts express a more appropriate set of transcriptional co-activators of the nkx2.5 gene than 10T1/2 cells (Fig. 1C). Nonetheless, again co-expression of desmin resulted in an attenuation of the nkx2.5 LUC reporter expression when only the PEPR was present in pNKE24, while des-shRNA expression partially compensated the negative effect caused by desmin. Addition of the MCE to the PEPR resulted in significantly increased pMCE-mediated LUC activity and desmin co-expression caused an additional 40% increase of the LUC activity which was attenuated by des-shRNA expression. These two sets of concordant data suggest that desmin influences Nkx2.5 expression in fibroblasts and myoblasts through the MCE and the PEPR of the nkx2.5 gene.

To test if the effect of desmin on Nkx2.5 expression has some physiological relevance in maturing CMCs, which already express substantial amounts of desmin, we co-transfected primary heart cells from newborn mice with the LUC reporter plasmids pNKE24 and pMCE and the des-shRNA plasmid (Fig. 1D). Transfection with the PEPR containing reporter plasmid pNKE24 led to a two-fold increase in LUC activity as compared to 10T1/2 fibroblasts, and co-transfection with the des-shRNA plasmid reduced the LUC activity to 60% of the basal activity. Likewise, transfection of heart cells with the MCE+PEPR containing reporter plasmid pMCE resulted in a five-fold increase in LUC activity, which again was reduced to 72% by desmin-specific des-shRNA. The positive effect of desmin on PERP activity in primary cardiac cells, in contrast to the negative effect in 10T1/2 and C2C12 cells, suggests that desmin requires a cardiac cell specific set of co-factors to positively influence the PEPR region of the nkx2.5 gene, whereas presence of the MCE conferred activation in all tested cell types. These results together indicate that desmin contributes to the modulation of Nkx2.5 expression in developing muscle cells and substantiate previous findings that desmin over-expression in ESCs undergoing cardiomyogenesis increases Nkx2.5 expression.

To demonstrate that desmin promotes expression of the endogenous nkx2.5 locus in ESC-derived cardiac progenitor cells (CPCs) and differentiating CMCs, we generated heterozygous nkx2.5^+/EGFP ESCs with one allele of the nkx2.5 gene converted to an EGFP reporter allele by homologous recombination with a nkx2.5::EGFP knock-in vector (Hidaka et al., 2003). Heterozygosity at the nkx2.5 locus causes haploinsufficiency (Biben et al., 2000; Kasahara et al., 2000) with evident functional consequences at the very beginning of cardiomyogenesis in EBs (Fig. 2A). In EBs derived from nkx2.5^+/EGFP des^+/+ ESCs the onset of cardiomyogenic differentiation was significantly delayed and reduced when compared to wild type nkx2.5^+/+des^+/+ EBs (Weitzer et al., 1995) and desmin over-expressing nkx2.5^+/+des^+/+des^ect EBs (Hofner et al., 2007). By contrast, desmin over-expression in nkx2.5^+/EGFP des^+/+des^ect EBs reversed the nkx2.5-related haploinsufficency phenotype and cardiomyogenesis commenced as in nkx2.5^+/+des^+/+des^ect EBs (Fig. 2A). Fluorescence microscopy revealed a strongly increased EGFP fluorescence signal in the presence of desmin in differentiating nkx2.5^+/EGFP des^+/+des^ect CMCs (Fig. 2B). Accordingly, quantification of EGFP intensity in single contracting CMCs by image analysis of fluorescence microscopy data demonstrated increased fluorescence intensity, and therefore increased expression of the nkx2.5^EGFP allele in CMCs (Fig. 2C). The effect of desmin on the nkx2.5⁺ allele was also demonstrated by RT-PCR using desmin- and nkx2.5-specific primer pairs (Fig. 2D). Likewise, quantification of EGFP fluorescence in CPCs and early CMCs from EBs at day 7.5 using flow cytometry provided evidence for an increased number of cells with a strongly increased EGFP fluorescence in the presence of desmin (Fig. 2D), suggesting that desmin enhances even mono-allelic Nkx2.5 expression in differentiating CMCs, and thus rescues this Nkx2.5-related haploinsufficient phenotype.

To demonstrate the in vivo relevance of desmin in sustaining proper Nkx2.5 expression levels in adult cardiac stem cells, we isolated cardiac side population stem cells (CSPCs) from 90 day old desmin^−/− mice (Fig. 3A) and compared the expression levels of nkx2.5 by qRT-PCR to desmin-expressing CSPCs isolated from wild type mice of the same age. First we verified desmin expression in CSPCs (Fig. 3B), as previously described (Yamahara et al., 2008). In the absence of any desmin expression, nkx2.5 levels were reduced by one third in CSPCs from adult mouse hearts (Fig. 3C), indicating that desmin partially contributes to the maintenance of Nkx2.5 expression in adult CSPCs in vivo as well.

To further delineate the cell-autonomous effect of desmin on Nkx2.5 expression in CPCs, we used chromatin immuno-precipitation (ChIP) to screen the nkx2.5 gene for the presence of desmin in TF complexes bound to DNA. Primer pairs for PCR were chosen to amplify the minimal cardiac specific enhancer (MCE; also called AR1), located between base pairs −9432 and −8922 of the murine nkx2.5 gene (Brown et al., 2004; Lien et al., 1999) and the essential proximal enhancer (PE; also called AR2) (Brown et al., 2004; Searcy et al., 1998), located between base pairs −3059 and −2554, and adjacent 5′- and 3′-flanking sequences (Fig. 4A). As a source of CPCs we used EBs expressing wild-type and mutant desmin alleles. ChIP with desmin-specific antibodies was performed at the time when first rhythmically contracting CMCs are observed in EBs (Fig. 4B). At that time desmin is exclusively expressed in CPCs and early CMCs (Hofner et al., 2007; Weitzer et al., 1995). Desmin interacted with the PE but not with adjacent DNA sequences in CPCs from wild type (des^+/+) EBs between day 6 and 8 (Fig. 4C). In desmin over-expressing EBs with one additional RSV-driven desmin allele (des^+/+des^ect), association of desmin with the nkx2.5 gene was already observed from day 5 on but did not last longer than in des^+/+ EBs. In EBs, homozygously expressing a mutant desmin lacking the amino-terminus (des^{Δ1-48/Δ1-48}), as in desmin knock-out EBs (des^−/−), no signals could be detected, suggesting that these 48 amino acids of the amino-terminus potentially mediate this interaction of desmin with DNA or TFs, or might be necessary for nuclear transport. Likewise, desmin was present in a TF complex interacting with the MCE of the nkx2.5 gene in des^+/+ and des^+/+des^ect CPCs between days 5 and 7, but not in des^{Δ1-48/Δ1-48} and des^−/− CPCs (Fig. 4D). Data from five independent biological replicates demonstrated a desmin concentration dependent increase in the number of desmin containing TF complexes bound to the PE and MCE of the nkx2.5 gene in ESC-derived CPCs (Fig. 4E). As a positive control, we repeated these experiments with a Mesp1 specific antibody and confirmed the already reported interaction of Mesp1 with the MCE and PE of the nkx2.5 gene (Bondue et al., 2008). Desmin was also present in a TF complex bound to the PE in differentiating cardiovascular progenitor cells (CVPCs), isolated as phenotypically stable cell lines from hearts of newborn mice (Hoebaus et al., 2013), and in primary heart cells from newborn mice (Fig. 4F). Additionally, desmin could be identified as a component of TF complexes interacting with the nanog, brachyury, mesp1, and desmin genes in ESC- and CVPC-derived CPCs (Table S2), suggesting that nuclear localization of desmin during early cardiomyogenesis might even contribute to the transcriptional regulation of several genes.

Identification of desmin-containing TF complexes bound to the nkx2.5, nanog, brachyury, mesp1, and desmin genes and the premature but not permanent up-regulation of Nkx2.5 expression in des^+/+des^ect ESC-derived CPCs (Hofner et al., 2007) suggest that desmin might be present in nuclei of CPCs only during a relative short period of time, and therefore had been overlooked in former studies. Hence we screened different developmental states of CPCs derived from ESC and CVPC lines stained with desmin specific antibodies, by optical sectioning of nuclei with a confocal immunofluorescence microscope, and determined the percentage of cells with an unambiguous presence of desmin in the nucleus (Table 1). In developing ESC-derived CPCs, desmin protein could be detected in many cells at day 5.5 after aggregation, which is two days after desmin mRNA has become detectable by northern blotting (Höllrigl et al., 2002) and RT-PCR (Hofner et al., 2007). Indeed, in about 7% of these cells desmin was found in the nucleus (Fig. 5A,B; Table 1), partially co-localizing with vimentin. As expected, desmin was not detected in undifferentiated des^+/+ ESCs (Fig. S2A) which, however, express low amounts of vimentin, consistent with previous data in human ESCs (Van Hoof et al., 2008). Ectopic expression of desmin in des^+/+des^ect ESCs resulted in the cytoplasmic localization of desmin in all undifferentiated ESCs (Fig. 5C) and in 0.2% of these cells desmin was detected in the nucleus (Fig. 5F). These few cells are most likely already prematurely differentiating along the myocardial lineage, because over-expression of desmin favours myocardial commitment of ESCs (Hofner et al., 2007). In contrast to wild type desmin, mutant desmin protein lacking the amino-terminus was never localized in the nuclei of des^{Δ1-48/Δ1-48} ESCs (Fig. 5D). Any potentially misleading cross-reactivity of the desmin antibodies were excluded by the absence of a detectable signal in differentiating des^−/− ESCs (Fig. 5E). In undifferentiated CVPCs nearly no desmin was detected by immunofluorescence microscopy (Table 1 and Fig. S2C), whereas at day 7 after aggregation and initiation of myocardial differentiation desmin localized to the nucleus in 10% of the CVPC-derived CPCs (Table 1; Fig. S2D,E). This percentage decreased to 4% at day 13 when CMCs, already contracting from day 11, became even more mature. However, a small number of the emerging cardiac troponin T-positive CMCs still had some desmin localized in the nucleus (Fig. S2F,G).

In hearts from neonatal mice, desmin was present in 4% of the nuclei of primary immature CMCs (Table 1, Fig. 5G,H; Fig. S3A). By contrast, desmin was neither found in nuclei of desmin-positive cardiac fibroblast-like cells (Fig. S3B), nor in nuclei of fully differentiated CMCs (Fig. S3C), wherein desmin was predominantly localized at the Z-discs of myofibrils. Taken together, these data demonstrate that desmin was exclusively found in nuclei of CPCs which were committed to the myocardial lineage but had not fully completed differentiation to CMCs.

To further confirm the nuclear localization of desmin without using desmin-specific antibodies, we generated ESC and CVPC lines, expressing either a desmin-mCherry or a mCherry-desmin fusion protein, and as controls, expressing merely mCherry. In fibroblasts these fusion proteins integrated well in the pre-existing vimentin IF network but displayed a diffuse and spotty staining in the cytoplasm of undifferentiated ESCs and CVPCs (Fig. S4A) which express vimentin and cytokeratins but most likely do not contain IFs (de Souza Martins et al., 2011; Gao et al., 1994). Although expressed under the control of the CMV promoter in all cells, mCherry-tagged desmin was found in nuclei of less than 10% of the differentiating CVPCs (Fig. S4B). Nuclear localization of the mCherry-tagged desmin proteins was demonstrated in nuclei isolated and purified from differentiating CPCs (Fig. 6A; Fig. S4C). Again, only 10.7% (n=440) of the purified nuclei of CVPC-derived CPCs contained the mCherry-tagged desmin proteins. Quantification of the nuclear fluorescence signals in differentiating CPCs demonstrated that, similar to endogenous desmin, both desmin-mCherry and mCherry-desmin fusion proteins, but not mCherry alone, can enter the nucleus (Fig. 6B).

Analysis of the cytoplasmic, nucleoplasmic, and chromatin fractions of mCherry-tagged desmin expressing CPCs by western blot analysis with mCherry specific antibodies corroborated the association of the mCherry-tagged desmin proteins with chromatin (Fig. 6C; Fig. S4D). mCherry expressed in control cell lines was exclusively found in the cytoplasmic fraction. As controls, the same cellular fractions were tested with desmin antibodies and the nuclear localization of endogenous desmin was confirmed. Furthermore, use of gp130 and histone H3 specific antibodies, respectively, demonstrated the efficient separation of cytoplasm and membranes from the chromatin. Finally, fluorescence microscopy of isolated nuclei confirmed the absence of any IFs at the nuclear periphery.

Applying the same isolation procedure to murine heart cells, isolated three days post-partum, also demonstrated the nuclear localization of desmin in early postnatal heart cells (Fig. 6D). The significant larger amounts of desmin in the cytoplasmic fraction of CMCs, as compared to stem cells, originates from a substantial increase in desmin expression in maturing CMCs and from the concomitantly occurring enormous increase in the cytoplasmic-to-nuclear volume ratio.

To validate the interaction of desmin with the regulatory elements of the nkx2.5 gene without using desmin-specific antibodies, ChIP analysis was performed in differentiating desmin-mCherry and mCherry-desmin-expressing ESC- and CVPCs-derived CPCs with a monoclonal mCherry specific antibody (Fig. 6E; Fig. S4E). Both mCherry-tagged desmin proteins bound to the MCE, the PE and to the promoter of the nkx2.5 gene in a manner comparable to that of endogenous desmin. ChIP analysis in cell lines expressing only mCherry did not give any positive signal.

In order to better understand the transcriptional regulation of cardiomyogenesis in cardiac progenitor cells (CPCs), we investigated the influence of the very early expressed muscle cell specific IF protein desmin on the transcription of the nkx2.5 gene, one of the key regulators of cardiomyogenesis. This endeavor was based on the observation that desmin expression significantly enhanced Nkx2.5 expression and the development of beating cardiomyocytes (CMCs) in differentiating embryonic stem cells (ESCs). Using reporter-gene assays in different cell types and Nkx2.5 expression analysis in cardiac side population stem cells (CSPCs), combined with nuclear localization and ChIP assays, we demonstrate herein that desmin contributes to the transcriptional regulation of the nkx2.5 gene in CPCs during a short period of cardiomyogenesis and in CSPCs of the adult mouse.

The present study strengthens the evidence of a functional interaction between desmin and the nkx2.5 gene in CPCs, suggested by previous investigations (Hofner et al., 2007; Höllrigl et al., 2007) and provides evidence for a physical and functional interaction of desmin with the nkx2.5 gene in CPCs for the first time. Furthermore, it suggests a physiologically relevant role of desmin for the regenerative capacity of the adult heart. Finally, it introduces Nkx2.5 as an additional factor potentially contributing to the worsening of the desmin knockout heart defects with increasing age and of human desmin related cardiomyopathies (Capetanaki et al., 2015; McLendon and Robbins, 2011; Milner et al., 1999).

Activation of nkx2.5-reporter transgenes by desmin in fibroblasts, myoblasts, and ESC-derived CPCs, in combination with its inhibition, mediated by desmin-siRNA in primary heart cells from newborn mice, as well as the reduced expression of Nkx2.5 in adult desmin null CSPCs (Figs 1–3), provide solid evidence for the involvement of desmin in the transcriptional regulation of the nkx2.5 gene. Rescue of cardiomyogenesis in nkx2.5^+/− haploinsufficient CPCs by desmin expression substantiates in vivo the direct positive influence of desmin on Nkx2.5 expression and on cardiomyogenesis in general. This subtle contribution of desmin to the transcriptional regulation in CPCs was not directly evident in vivo because desmin knockout mice did not show an obvious embryonic phenotype, most likely due to unknown redundant mechanisms safe-guarding cardiomyogenesis. However, reduced expression of Nkx2.5 in adult CSPCs lacking desmin might contribute to the severe defects observed in aging human and mouse hearts with mutant desmin alleles (Capetanaki et al., 2015). These very diverse defects might be co-founded by the inability of adult CSPCs lacking desmin or expressing dominant negative mutant desmin proteins to properly reactivate or up-regulate Nkx2.5 expression. The inadequate upregulation of Nkx2.5 expression in desmin null stem cells after injury or stress might contribute to a compromised regenerative potential in desmin-related heart diseases. The role of desmin as a co-factor acting in a dose-dependent manner upstream of nkx2.5 is also supported by the previous finding that knock-down of desmin in C2C12 myoblasts and knockout in ESCs resulted in blocking myotube formation and the partial down-regulation of the myoD and myogenin genes (Li et al., 1994; Weitzer et al., 1995), which are key players of myogenic differentiation.

Activation of gene expression by desmin in a cell-autonomous manner requires either autocrine or mechanochemical activation of signal-transduction, sequestration of transcriptional inhibitors, or direct interaction with euchromatin in the nuclei of certain types of CPCs. Focusing on the latter, we were able to demonstrate the physical interaction of desmin with the minimal cardiac specific enhancer (MCE), the proximal enhancer (PE) and promoter region of the nkx2.5 gene in CPCs by ChIP analysis (Figs 4 and 6; Table S2). Simultaneous binding of desmin to the distant MCE, the PE, and the promoter region of the nkx2.5 gene with different affinity, and its ability to form dimers and tetramers (Quinlan et al., 1986) suggest a model wherein desmin might be part of different TF complexes bringing the distal MCE into proximity to the PE and the promoter region. By binding to three regulatory elements of the nkx2.5 gene with different affinity, desmin may variably contribute to DNA bending and scaffolding TF complexes in front of the DNA-RNA polymerase II complex, and may by this way influence transcription. The significance of the desmin-nkx2.5 gene interactions is also supported by the fact that it is only prominent at the beginning of cardiomyogenesis and is at/or below the detection limit in undifferentiated CVPCs, ESCs and fully developed CMCs. Furthermore, desmin was also present in TF complexes composed of nanog, brachyury, Mesp1, and Nkx2.5 in variable configuration. These TF complexes interact with the nanog, brachyury, mesp1, nkx2.5, and desmin genes in differentiating CPCs and thus allow the speculation on a potential broader modulating role of desmin in cardiomyogenesis.

The present data, however, do not discriminate between indirect and direct binding of desmin to the MCE and PEPR DNA. The former possibility is supported by the fact that the enhancer and promoter elements of the nanog, brachyury, mesp1, nkx2.5, and desmin genes, where desmin could be localized by ChIP, are also occupied by TFs nanog, brachyury, Mesp1, and Nkx2.5 (Table S2) (Bondue et al., 2008; David et al., 2011; Doppler et al., 2014). The latter possibility is supported by the fact that desmin also binds to a few DNA regions where none of the studied TFs were present, and by previous work demonstrating that desmin and other type III IFs directly interact with purified and synthetic DNA (Li et al., 2002,, 2003; Tolstonog et al., 2005; Traub and Shoeman, 1994a,,b). It will be a challenging task for future research to investigate the interaction of desmin with DNA at the molecular level in more detail.

Independently of the mode of interaction between desmin and DNA, ChIP data strongly suggest that low amounts of desmin must be present at least in some of the CPC nuclei at the beginning of cardiomyogenesis. Indeed, in differentiating ESCs and CVPC lines desmin was detected by confocal immunofluorescence microscopy in 7% and 10% of the cells, respectively. Notably, desmin was never detected in nuclei of undifferentiated ESC lines or in histological specimen from adult hearts. Absence of desmin in nuclei of mature CMCs is most likely due to the fact that desmin subunits are simply sequestered by abundantly emerging desmin-rich sub-cellular structures like Z-disks, costameres, and intercalated discs in addition to the IFs closely associated with mitochondria (Capetanaki et al., 2015). Preparation of nuclei from CPCs expressing mCherry-desmin fusion proteins allowed exclusion of the presence of any remnants of perinuclear localized desmin IFs by fluorescence microscopy and further demonstrated the nuclear localization of mCherry-tagged desmin proteins and desmin by western blot analysis with mCherry- and desmin-specific antibodies (Figs 5 and 6; Table 1).

Nuclear transport and presence of IF proteins has been demonstrated for tail-less cytokeratin 8, 18, and 19 in 3T3 fibroblasts (Bader et al., 1991), for vimentin in nasopharyngeal carcinoma lymph node metastasis (Luo et al., 2012), for nestin in various brain tumors (Krupkova et al., 2011), and for desmin in BHK21 cells (Kamei, 1986).

Desmin has no perfect nuclear localization signal in its carboxy-terminal domain like the lamins. However, it has a putative and suboptimal nuclear localization signal between amino acids R141 and R163 and a putative weak nuclear export signal located between amino acids L182 and L199 (NetNES 1.1 prediciton server, CBS) in its linker L1 and helix 1B domains. Alternatively, desmin may be transported to the nucleus by an unknown shuttle mechanism. A single study has suggested that vimentin may be transported to the nucleus by a cytoplasmic DNA-mediated transport mechanism (Hartig et al., 1998) and another, that desmin might be transported piggyback by filaggrin (Mack et al., 1993; Pearton et al., 2002). Finally, desmin may serve as a cytoplasmic docking site for nuclear proteins during the M-phase and in turn some of these proteins might facilitate the nuclear transport of desmin in a piggyback mechanism under certain circumstances in a small set of CPCs. Limitation of the nuclear presence of desmin and the desmin-chromatin interactions to a small subpopulation of CPCs during a short period of time at the beginning of cardiomyogenesis supports its specificity. Furthermore, these findings, together with the observation that even ubiquitous CMV-driven expression of mCherry-tagged desmin resulted in a nuclear localization in only 10% of nuclei of differentiating CVPCs suggest the existence of obligatory co-factors facilitating the nuclear transport of desmin and perhaps also its recruitment to the enhancer and promoter regions of the nkx2.5 gene. Hence the major challenge for future investigations will be to unravel the mechanism of the nuclear transport of desmin in differentiating CPCs and to identify either the consensus DNA sequence or the chromatin components to which desmin binds in the nucleus of CPCs.

In summary, desmin, though not a typical TF may function as an auxiliary factor or scaffold with or without lamins in the nucleus, selectively modulating euchromatin, and thus permitting transcription of cardiac specific genes to proceed expeditiously and more efficiently.

10T1/2 fibroblasts and C2C12 myoblasts were maintained in DMEM supplemented with 2 mM L-glutamine, 0.05 mg/ml streptomycin, 0.03 mg/ml penicillin, and 10% (v/v) fetal bovine serum (FBS) from Gibco (M10). Murine des^+/+ (AB2.2 and W4) (Lauss et al., 2005), des^+/+des^ect (DC6) (Hofner et al., 2007), des^−/− (Weitzer et al., 1995), des^{Δ1-48/Δ1-48} (Höllrigl et al., 2002) ESCs and A5 cardiovascular progenitor cells (CVPCs) (Hoebaus et al., 2013) were maintained on SNL76/7 feeder cells in the same medium but supplemented with 15% (v/v) FBS (Hyclone, SH3007001) (M15) instead. Differentiation of ESCs and CVPCs was achieved by aggregation to embryoid bodies (EBs) and cardiac bodies as previously described (Hoebaus et al., 2013; Hofner et al., 2007).

Primary cardiomyocytes (CMCs) were prepared from newborn Balb/C and C57BL/J6 mice (Mus musculus Linnaeus, 1758), respectively as previously described (Stary et al., 2005). Animal studies were conducted according to approved European, Austrian, and Greek Institutional animal guidelines (Directive 2010/63/EU). Heart tissue fragments were triturated three times in PBS containing 4% (v/v) pancreatin (Sigma, P3292) and 0.5 mg/ml collagenase II (Worthington, CLS-2) for 5 min, debris was removed by centrifugation, and lastly cardiac fibroblasts were removed by adsorption to a gelatin coated tissue culture plate for 45 min at 37°C. CMCs were maintained in DMEM supplemented with 2 mM L-glutamine, 0.05 mg/ml streptomycin, 0.03 mg/ml penicillin, and 4% (v/v) FBS (Sigma, F2442).

Adult cardiac side population stem cells (CSPCs) were isolated from 8- to 12-weeks old 129SV wild type and des^−/− mice (Mus musculus Linnaeus, 1758) as previously described (Pfister et al., 2010). Briefly, minced cardiac tissue (pool of six mouse hearts per genotype) was digested with 0.1% collagenase B (Roche, 11088815001), 2.4 U/ml Dispase II (Roche, 04942078001) and 2.5 mM CaCl₂ in Hanks' balanced salt solution (HBSS) buffer supplemented with 2% FBS and 10 mM HEPES at 37°C for 30 min. Digestion was terminated by addition of HBSS buffer, digested tissue was filtered through a 70 μm filter, washed with HBSS buffer supplemented with 2% FBS and 10 mM HEPES and passed once more through a 40 μm filter. Total mono-nucleated cell number was determined and cells were re-suspended in high glucose DMEM (GIBCO, 41966-029), supplemented with 2% FBS and 10 mM HEPES at a concentration of 10⁶ cells/ml. Cells were stained with 2.5 μg/ml Hoechst 33342 (Sigma, B2261) at 37°C for 90 min with occasional shaking during incubation to ensure even staining. A negative control sample was also prepared by including 50 μM verapamil during the entire Hoechst staining procedure, in order to distinguish CSPCs. After staining, cells were washed twice with HBSS buffer supplemented with 2% FBS and 10 mM HEPES, resuspended in PBS w/o Mg⁺²/Ca⁺², pH 7.4 (GIBCO, 70011-036) supplemented with 2% FBS and 2 mM EDTA and kept on ice until FACS analysis.

Freshly isolated CSPCs were cultured at a density of 2×10⁵ cells on 0.1% gelatine-coated 35×10 mm culture dishes in DMEM-F12 (GIBCO, 31330-038) supplemented with 10% FBS (GIBCO, 12484-028), 1% Penicillin/Streptomycin (GIBCO, 15070-063), 10 ng/ml basic fibroblast growth factor (Peprotech, 450-33), 20 ng/ml epidermal growth factor (Peprotech, 315-09), 1000 U/ml leukemia inhibitory factor (Sigma, L5158) and 1× insulin-transferrin-selenium stock (Invitrogen, 41400045), at 37°C, 5% CO₂. Medium was renewed every three days and the resulting adherent cultures were passaged at 80% confluency using standard trypsinization techniques.

Fluorescence activated cell sorting of CSPCs was performed using a FACS AriaII-Upgraded (BD Biosciences) instrument equipped with triple lasers. The Hoechst dye was excited using a near UV laser (375 nm) and fluorescence emission was collected with a 450/20 nm filter (Hoechst Blue) and a 575/26 nm filter (Hoechst Red). A 502 nm long pass dichroic filter was used to separate the emission wavelengths. 7-amino-actinomycin D (7-AAD) was added to cells at a concentration of 0.25 μg/10⁶ cells immediately before sorting to exclude dead cells. Acquired data were analyzed using FlowJo software (FlowJo, LLC).

Total RNA was prepared from wild type and des^−/− CSPCs, either freshly isolated or cultured for 10 days, with Tri Reagent (Sigma-Aldrich, T9424) according to manufacturer's instructions, with an additional DNase I treatment (Roche, 04716728001). cDNA was synthesized from 2 μg RNA using M-MLV reverse transcriptase (Invitrogen, 28025-013) and random primers (Invitrogen, 48190-011), according to manufacturers' instructions. The quantitative RT-PCR analysis (qRT-PCR) was performed from 5 ng cDNA with the LightCycler^® 480 Instrument (Roche) using the 480 SYBR Green I Master kit (Roche, 04707516001). The expression of nkx2.5 RNA in wild type and des^−/− CSPC was assayed in duplicate in five independent qRT-PCR experiments and was normalized to the expression of ribosomal S26 RNA (RPS26). The primers used for qRT-PCR amplification were as follows: nkx2.5: fwd: 5′-ACATTTTACCCGGGAGCCTA-3′, rev: 5′-GGCTTTGTCCAGCTCCACT-3′; RPS26: fwd: 5′-GCCATCCATAGCAAGGTTGT-3′, rev: 5′-GCCTCTTTACATGGGCTTTG-3′. The obtained data were expressed as mean±standard deviation (s.d.) of the fold expression of nkx2.5 in des^−/− versus wild type CSPCs.

The plasmid pCSX-EGFP-PPDT (Hidaka et al., 2003) containing a nkx2.5 transgene in which the exons have been replaced by the EGFP cDNA and a puromycin selection cassette, was introduced by homologous recombination into AB2.2 ESCs exactly as previously described (Höllrigl et al., 2001). The cells were selected with 1 µg/ml puromycin (Life Technologies) for four days, subcloned, tested for correct knock-in by PCR as described (Hidaka et al., 2003), and nkx2.5^+/EGFP des^+/+ ESC clones were tested for EGFP expression during differentiation of EBs by fluorescence microscopy.

To introduce an ectopic constitutive desmin allele under the control of the RSV promoter into nkx2.5^+/EGFP des^+/+puro^R ESCs, these cells were fused with nkx2.5^+/+des^+/+des^ect neo^R DC6 ESCs (Hofner et al., 2007) using the polyethylene method. Cells selected with 1 µg/ml puromycin and 180 µg/ml active G418 (Gibco, 11811031) were sub-cloned and tested for EGFP and Desmin expression. Isolated clones not expressing desmin were used as controls for in vitro differentiation experiments.

Cells from EBs were separated by incubation with a mixture of trypsin solution (Invitrogen, 27250018), with 4% (v/v) pancreatin (Sigma, P3292) and 0.5 mg/ml collagenase (Whortington, CLS-2) for 20 min at 37°C two times and re-suspended in serum containing medium. Cells were harvested, resuspended in PBS, and EGFP fluorescence intensity was analyzed with a Becton & Dickinson LSR1 flow cytometer after elimination of dead cells positively stained for propidium iodide.

To generate red fluorescent mCherry-tagged desmin fusion proteins the murine desmin cDNA (NCBI Ref. Se.NP_034173.1) was amplified by PCR with primer pairs facilitating the in-frame insertion of the desmin cDNA either in front of the mCherry protein coding sequence into pmCherry-N1 or right after the mCherry coding region into pmCherry-C1 vectors (Clontech, 632523 and 632524). Synthesis and correct ligation of the CMV-promoter-driven mCherry-desmin and desmin-mCherry transgenes was verified by sequencing of both DNA strands. DraIII-linearized vector DNA was transfected into W4 ESCs and A5 CVPCs by electroporation as previously described (Höllrigl et al., 2001). After selection with 180 µg/ml G418 (GIBCO, 11811031) monoclonal cell lines were tested for mCherry-desmin and desmin-mCherry expression by fluorescence microscopy and western blotting.

ESCs, CVPCs, and primary CMCs were processed and stained as previously described (Hofner et al., 2007). Briefly, cells were fixed in 4% paraformaldehyde in PBS for 20 min at room temperature, permeabilized with 0.15% saponin in PBS, and stained with antibodies against desmin (Santa Cruz, sc7559, 1:500, Sigma Aldrich, D1033, 1:50, and Abcam ab8592, 1:80), vimentin (Sigma, V6630, 1:200), cardiac troponin T (Thermo Scientific, MS-295, 1:200, and Santa Cruz, 20025, 1:200), emerin (Novocastra, NCL-Emerin; 1:50), and lamin A/C (Roblek et al., 2010; 1:500) for 60 min, washed three times with PBS, and stained with FITC- and TR-conjugated secondary antibodies (Dianova, 711-095-152, 1:200; 711-095-151, 1:200; 711-075-152, 1:200), and Alexa Fluor 488-conjugated secondary antibodies (Molecular Probes, A-11001, 1:1000) for 60 min. Alternatively, cells on cover slips were fixed in 96% ethanol at −20°C for 20 min, dried and blocked with 1% BSA in PBS for 10 min. Nuclei were counterstained with DAPI (Sigma, D9542). Desmin was scored to be intra-nuclear if the signal appeared and disappeared within the emerin-, lamin A/C-, or DAPI stained volume while scanning vertically through the nucleus. Spots at the periphery of nuclei were always excluded from the analysis (see Fig. 5H). For visualization of the mCherry-desmin, desmin-mCherry, and mCherry proteins, cells and nuclei, respectively, were fixed with 100% ethanol at −20°C for 5 min. Photomicrographs were taken on a Zeiss LSM 510 confocal microscope. 12, 16 or 25 Z-stacks of optically sectioned cells were obtained with a step size of 0.5 and 0.37 µm, respectively. Routine analysis was performed with an Axiovert 135TV microscope.

Nkx2.5-driven EGFP fluorescence in single living nkx2.5^+/EGFP des^+/+puro^R and nkx2.5^+/EGFP des^+/+des^ect neo^R cardiomyocytes were quantified by image analysis. Photomicrographs were taken on a Zeiss LSM 510 confocal microscope. Fluorescence images of single contracting cardiomyocytes were quantified with Adobe Photoshop software. To quantify EGFP fluorescence, images obtained with a 530 nm filter were imported into Photoshop CS2™ as tiff files and pixel intensity values from several areas of a cell were extracted by the built-in histogram tool. Pixel intensities were subtracted by the mean background pixel intensity obtained from several cytoplasmic areas of non-cardiomyocytes of the same image.

mCherry fluorescence images of nuclei were obtained from isolated nuclei taken from the fraction prepared for western blot experiments and nuclear fluorescence was quantified with Adobe Photoshop software. To quantify mCherry fluorescence, images obtained with a 615 nm filter were imported into Photoshop CS2™ as tiff files and pixel intensity values within several areas of cells and nuclei, respectively, were extracted by the built-in histogram tool. Pixel intensities were subtracted by background pixel intensity obtained from several cell- or nucleus-free areas of the images.

Preparation of soluble chromatin and ChIP assays were carried out as previously described (Hauser et al., 2002). Shortly, ESCs, CVPCs and primary CMCs were fixed using formaldehyde and lysed with a buffer containing 10% SDS, Roche complete protease inhibitor cocktail, and PMSF. Equal amounts of chromatin were sonicated, and DNA-protein complexes were immuno-precipitated over night with antibodies against desmin (Abcam, ab8592, Sigma, D8281 and D1030), brachyury (Santa Cruz, sc17743 and sc17745), and mCherry (St. John's Laboratory, STJ34373), respectively. As positive and negative controls, antibodies against the C-terminus of histone H3 (Abcam, ab7228), c-Myc (AbCam, 1791), vimentin (Sigma, V4630), and IgG (Santa Cruz, sc2027) were used. Using pre-immune IgG as a negative control did not result in any signals above the limit of detection. ChIP DNA and control input DNA were analyzed by PCR, using 35 to 40 cycles and primer pairs (for sequences see Table S1) situated in the 5′-region of the murine nkx2.5 locus (NCBI: Gene ID: 18091, updated on 14 July 2015) as indicated in Fig. 4A and Fig. 6E. For location of primer pairs used to analyze the interaction of nanog (antibodies: Santa Cruz, sc33760; Cell Signaling, 8822), brachyury (antibody: Santa Cruz, sc17745), mesp1 (antibody: Santa Cruz, sc1763078), nkx2.5 (antibodies: Santa Cruz, sc8697 and 14033), and desmin (antibodies: Abcam, ab8592; Sigma, D8281 and D1030), with the nanog, brachyury, mesp1, nkx2.5, and desmin gene, see Fig. S5. Quantification was performed by measuring the luminosity of ethidium bromide stained PCR products from at least two independent experiments by Adobe Photoshop CS2™ tools.

10T1/2 and C2C12 cells were transiently transfected by the classical calcium-phosphate method with a DMSO shock, and primary CMCs with the Lipofectamine 2000 (Invitrogen, 11668-019) method. DNA for transfection was isolated with the Endo Free Plasmid Maxi Kit (Qiagen, 12263) from the following plasmids: pGL3b, containing the Photinus pyralis luciferase (LUC) cDNA (Promega, E1751) and phRL-TK (Promega, E2241), containing the Renilla reniformis luciferase cDNA as reporter and transfection efficiency control; pNKE24 (Searcy et al., 1998), a pGL3b backbone, containing the nkx2.5 proximal enhancer and promoter region (PEPR), located between base pairs −3059 to −2554 of the nkx2.5 5′-region; pMCE, containing the minimal cardiac specific enhancer (MCE) located between base pairs −9432 to −8922 of the nkx2.5 5′-region (Lien et al., 1999), in addition to the PEPR in the LUC reporter plasmid; pRSV-desmin (‘desmin-ect.’), containing the desmin cDNA under the control of the RSV promoter (Hofner et al., 2007); pshRNA desmin knock-down plasmids (‘des-shRNA’), which performed equally well in reducing desmin mRNA levels; clone 1: 5′-TCCTACACCTGCGAGATTGAT-3′; clone 2: 5′-GATCGCGTTCCTTAAGAAAGT-3′; clone 3: 5′-CAAGGGCTCCTCGAGTTCAAT-3′; clone 4: 5′-AGACCATCGCGGCTAAGAACAT-3′; pshRNA-neg (‘nc-shRNA’): 5′-GGAATCTCATTCGATGCATAC-3′, as negative control (SA Biosciences); and pUC19, for leveling the concentration of transfected DNA.

The pMCE LUC reporter plasmid was generated by cloning the MCE into the SalI restriction site of pNKE24. This resulted in the location of the MCE upstream of base pair −2641, 5′ to the PEPR in the pNKE24 plasmid. PCR primers were designed for the production of the MCE insert from C3H mouse genomic DNA, with flanking SalI restriction sites: SalI fwd: 5′-TAATGTCGACTCTGGGTCCTAATGC-3′and SalI rev: 5′-TATTGTCGACTCAAGGTGCACATGA-3′. The PCR product was generated using Pyrococcus furiosus polymerase (Thermo Scientific, EP0501) and both strands were sequenced to verify accurate synthesis.

Transfected cells were incubated in M10 medium for 48 h and then lysed and firefly and renilla LUC activity was measured with the Dual Luciferase^® Reporter System (Promega, E1910) using a Centro LB 960 luminometer (Berthold Technologies). Renilla reniformis LUC activity from co-transfected phRL-TK plasmids was used to correct firefly LUC activity for varying transfection efficiency. LUC activity was always normalized to the basic activity of the promoter-less pGL3b plasmid. Expression of desmin in cells transfected with pshRNA-anti-Desmin clones 1-4 and pshRNA-neg was quantified by RT-PCR.

mRNA was isolated from transfected 10T1/2 and C2C12 cells with the Qiagen RNeasy kit. cDNA was synthesized with RevertAid M-MuLV reverse transcriptase (Thermo Scientific, EP0441). RT-PCR was performed with Taq polymerase (Thermo Scientific, EP0402) using primers for desmin as follows: fwd: 5′-TGATGAGGCAGATGAGGGAG-3′, rev: 5′-TGAGAGCAGAGAAGGTCTGG-3′. Amplification conditions: Tm: 52°C, 35 cycles. Quantification was performed by measuring the luminosity of ethidium bromide stained RT-PCR products by Adobe Photoshop CS2™ tools of two independent experiments and four RT-PCR analyses.

To obtain cytoplasm-free nuclei from cells undergoing cardiomyogenesis a high-quality biochemical fractionation protocol (Rosner et al., 2013) was employed with some modifications allowing the separation of the nuclear membranes from the chromatin (Aaronson and Blobel, 1975). Cells from 6.7-day old EBs and 9.3-day old cardiac bodies, respectively, expressing either mCherry, the desmin-mCherry, or the mCherry-desmin fusion protein, were dissociated by incubation with a mixture of trypsin solution (Invitrogen, 27250018), with 0.1% (m/v) pancreatin (Sigma, P3292) and 0.05% (m/v) collagenase type 2 (Whortington, CLS-2) at 37°C for 25 min. Cells were suspended in M15, cell aggregates removed by low-speed centrifugation, and single cells were recovered by centrifugation. Cells were washed once with ice cold PBS and then carefully lysed in 5× the pellet volume of 20 mM Tris-HCl pH 7.6, 50 mM β-mercaptoethanol, 0.1 mM EDTA, 2 mM MgCl₂, 174 µg/ml PMSF (Sigma, P7626), 2 µg/ml leupetin (Sigma, L2884), 2 µg/ml aprotinin (Sigma, A6279), and 0.3 µg/ml benzamidine chloride (Sigma, B6506) (solution 1) at room temperature for 2 min and then at 4°C for 10 min. After adding 10% (v/v) of a 10% Nonidet-P40 (Sigma 74385) solution to the lysates, cell fragments were carefully suspended 3× with a 200 µl Gilson tip and nuclei were pelleted at 600×g and 4°C for 5 min. The supernatant containing the cytoplasmic fraction of the cells was removed and nuclei were washed to remove all cytoplasmic components from their surface. To maintain the integrity of the nuclei, they were carefully suspended 3× with a 200 µl Gilson tip in 0.5 ml of solution 1 containing 1% Nonidet-P40 and then pelleted at 600×g and 4°C for 5 min. This procedure was repeated twice and purity of nuclei was assessed by fluorescence microscopy. After removal of the last wash solution, nuclei were lysed in 15× the pellet volume of 6 mM NaH₂PO₄ pH 7.1, 603 mM KCl, 171 mM NaCl, 1% Triton X-100, 8 mM β-mercaptoethanol, and 348 µg/ml PMSF in a micro douncer at 4°C for 1 h. Insoluble chromatin was pelleted at 17,000×g at 4°C for 25 min, the supernatant containing the karyoplasm was removed, and the chromatin solubilized in 1× SDS sample buffer (Laemmli, 1970) containing 1 mg/ml DNase I (Boehringer, 104159) by repeated pipetting at room temperature and final incubation at 96°C for 5 min. Samples of the cytoplasmic, karyoplasmic, and chromatin fraction were stored in 1× SDS sample buffer at −20°C.

Western blot analysis of the samples was performed by standard procedures with antibodies against mCherry (St. John's Laboratory, STJ34373), desmin (Abcam, ab8592 and Sigma, D8281), histone H3 (Abcam, ab7228), and gp130 (Santa Cruz, sc656), respectively, and secondary-alkaline phosphatase-conjugated antibodies (Promega S372B, S373B, and V1151).

All data are presented as the arithmetic mean±s.d. σ_x(n−1). Statistical significance was evaluated using Student's t-test and values of P≤0.05 were considered to indicate statistical significance.

We would like to thank Brigitte Gundacker, Matthias Scheinast, Sabrina Bail, Thomas Sauer and Wolfgang Weber for technical assistance; Dr Katherine Yutzey, Cincinnati, OH, USA, for the pNKE24 reporter construct; Dr Takayuki Morisaki, Suita, Japan, for the pCSX-EGFP-PPDT knock-in vector; Dr Sasha Martens, Vienna, Austria, for the pmCherry vectors; and Dr Elisabeth Simböck for crucial experimental details and help with ChIP experiments.