While transfer of a protein encoded by a single nucleus to nearby nuclei in multinucleated cells has been known for almost 25 years, the biological consequences for gain-of-function diseases have not been considered. Here, we have investigated nuclear protein spreading and its potential consequences in two of the three most prevalent neuromuscular diseases. By performing co-cultures between diseased or control human myoblasts and murine C2C12 myoblasts, we demonstrate that in facioscapulohumeral dystrophy, although the transcription of the toxic protein DUX4 occurs in only a limited number of nuclei, the resulting protein diffuses into nearby nuclei within the myotubes, thus spreading aberrant gene expression. In myotonic dystrophy type 1, we observed that in human-mouse heterokaryons, the expression of a mutated DMPK from human nuclei titrates splicing factors produced by neighboring nuclei, inducing the mis-splicing of several pre-mRNAs in murine nuclei. In both cases, the spreading of the pathological phenotypes from one nucleus to another is observed, highlighting an additional mechanism that contributes to the dissemination and worsening of the muscle pathogenesis. These results indicate that nuclear protein spreading may be an important component of pathophysiology of gain of function muscular diseases which should be taken into consideration in the design of new therapeutic approaches.
Therapies for neuromuscular diseases are usually based on the rescue of a diseased multinucleated muscle fiber by healthy nuclei or on the correction of diseased nuclei by gene therapy. These approaches are successful for loss-of-function diseases and are mediated by the spreading of normal proteins within the cytoplasm (1). However, such approaches may not be sufficient for gain-of-function diseases since the muscle fiber is a syncytium in which a nuclear protein encoded by a single nucleus can be transferred to nearby nuclei within the same fiber (2,3), thus spreading a phenotype.
In this study, we examined the effects of nuclear spreading on two gain-of-function muscular diseases. Facioscapulohumeral dystrophy (FSHD) and myotonic dystrophy (DM1) are two of the three most prevalent muscular diseases with an estimated prevalence of 4/100 000 and 4.5/100 000, respectively (http://www.orpha.net). In FSHD, two loci have been described as being potentially involved in the pathological mechanism, leading to two forms of the disease, FSHD1 and FSHD2. The first one (FSHD1) is located on the subtelomeric part of chromosome 4 and corresponds to the contraction of a 3.3 kb repeated region called D4Z4, whereas control individuals carry between 11 and 150 D4Z4 repeats, FSHD1 patients carry between 1 and 10 D4Z4 repeats (4). The second one is located on chromosome 18 (FSHD2) where mutations in the SMCHD1 gene have been found, leading to epigenetic deregulations (5). These two loci are usually mutually exclusive but a few patients carrying both loci have been already characterized (6). Both loci induce chromatin relaxation allowing the expression of the transcription factor DUX4 [for review see (7)]. Indeed, each D4Z4 repeat contains the open reading frame (ORF) for the transcription factor DUX4 (8) and DUX4 thus appears to be the common point between FSHD1 and FSHD2 patients. However, there is some potential controversy regarding the expression of DUX4, whereas DUX4 mRNA is only expressed in 1/1000 nuclei (9), DUX4 protein can be found in 0.5–9% of nuclei (10,11), suggesting that although DUX4 is not transcribed in most nuclei, the resulting protein may spread further. The progressive decrease in the signal intensity in consecutive nuclei systematically observed after DUX4 antibody staining would support this hypothesis (11,12), but such a spreading has never been demonstrated in FSHD. In this study, using RNA-FISH, we demonstrate that Dux4 mRNA is produced in only a very limited number of nuclei within myotubes. However, in co-cultures of human FSHD cells with murine C2C12, we show that the resulting DUX4 protein is present in the murine nuclei (which do not produce endogenous DUX4). In addition the human DUX4 activates murine genes downstream of DUX4, thus provoking a spreading of the pathological phenotype from diseased nuclei to neighboring healthy nuclei by nuclear protein transfer.
Consequences of nuclear spreading were also investigated in DM1, characterized by a repeat expansion in the DMPK gene localized on chromosome 19. The 3′ untranslated region of this gene normally contains 5–37 copies of a CTG trinucleotide repeat while in DM1 patients, expansion of this unstable motif to 50–5000 copies have been reported leading to the expression of mutant CUG exp-RNAs that remain trapped in the nucleus as aggregates and sequester splicing factors such as MBLN1 [for review see (13)]. MBNL1 loss-of-function results in the misregulation of the alternative splicing of several pre-mRNAs including CLCN1 and BIN1, associated, respectively, with myotonia and muscle weakness (14,15). Using heterokaryons generated between human and murine myoblasts, we demonstrate that mutated DMPK mRNAs transcribed in the diseased nuclei titrate splicing factors in murine nuclei (which do not express the mutated RNA), thus inducing mis-splicing of several murine pre-mRNAs including MBLN1, BIN1 and SERCA1 in neighboring nuclei.
In both FSHD and DM1, we demonstrated the diffusion of the molecular pathological phenotype through nuclear protein spreading, highlighting an additional mechanism that contributes to dissemination and worsening of muscle pathogenesis. Nuclear protein spreading may be an important component of pathophysiology of gain-of-function muscular diseases which should be taken in consideration in therapeutic approaches.
DUX4-FL mRNA is expressed in only few nuclei within myotubes
To determine whether DUX4 transcription occurs in only a small percentage of myonuclei in our cellular model, RNA-FISH directed against DUX4 was performed on FSHD myotubes at Day 5 of differentiation. A probe generated by DIG-PCR allowing the recognition of both DUX4-FL1 (spliced intron 1) and -FL2 (non-spliced intron 1) isoforms was used (Fig. 1A). A very discrete number of myonuclei, with often only one nucleus per myotube, were found to express DUX4 (Fig. 1B and Supplementary Material, Fig. S1). In contrast using the DUX4 E5.5 antibody (16), we always observed several DUX4 protein-positive nuclei within each myotube. This demonstrates that DUX4 is transcribed in one or only very few nuclei within myotubes in vitro, but confirms that the resulting DUX4 protein spreads to the neighboring nuclei.
DUX4 spreads within myotubes
To verify if DUX4 is able to spread to several nuclei within a myotube, co-cultures between murine C2C12 cells (which do not express DUX4) and either FSHD or control human muscle cell lines were performed. At Day 5 of differentiation, cells were fixed and immunostainings were performed. Human nuclei were characterized by a bright human-specific LAMIN A/C labeling, whereas murine nuclei were identified by the presence of heterochromatin revealed after Hoechst treatment (Fig. 2A and B). Figure 2A confirms that murine and human myoblasts can fuse together in vitro, giving rise to myotubes containing both murine and human myonuclei. The presence of DUX4 was assessed by specific immunostaining (E5.5 antibody) as a control, no DUX4-positive nucleus was ever detected in co-cultures of murine C2C12 and human control myoblasts, which is in agreement with the fact that DUX4 is not expressed by either control human cells or murine cells (data not shown). In contrast, DUX4 protein was detected in heterokaryons formed in co-cultures of FSHD and C2C12 cells. DUX4 was seen in both human FSHD and murine nuclei and the consecutive gradient of DUX4 intensity often observed (Fig. 2B) indicates that DUX4 protein is able to spread from human into murine nuclei. The longest distance measured between human and murine DUX4-positive nuclei was 51 µm (Supplementary Material, Fig. S2).
DUX4 induces a diffusion of the pathological phenotype in myotubes
We next investigated whether the human DUX4 proteins were functional and able to activate genes downstream of DUX4 not only in human but also in murine nuclei. RT–PCRs were performed on mRNA extracted from the different co-cultures at Day 5 of differentiation. DUX4-FL2 was only detected in the presence of FSHD cells co-cultures (Fig. 3A) as expected. The expression of human genes downstream of DUX4 was analyzed by RT–qPCR. A 12.3- and 6-fold increase was observed for both human ZSCAN4 and DEFB103 [two genes previously shown to be activated by DUX4 in FSHD myotubes (17)] in co-cultures with immortalized FSHD cells when compared with co-cultures with control cells (Fig. 3B), showing that DUX4 is active in human nuclei. Interestingly, when the expression levels of ZSCAN4 and DEFB103 were analyzed in control or FSHD myotubes in the absence of murine cells, a stronger activation of the two genes was observed when compared with what was obtained in the presence of murine cells, thus suggesting that the diffusion of the pathological molecular phenotype is correlated to the number of nuclei being exposed to DUX4 (Supplementary Material, Fig. S3A). We also analyzed the expression level of two murine genes previously described to be DUX4 targets in a transgenic mouse over-expressing DUX4 (18). We confirmed after transfection of a DUX4-expression plasmid that Zscan4c and Wfdc3 are activated in the presence of DUX4 (Supplementary Material, Fig. S3). Using murine-specific primers (Fig. 3C), we detected a fold increase of 5 and 10.2 for murine Wfdc3 and Zscan4c, respectively, in C2C12-FSHD cell co-cultures when compared with C2C12-control cell co-cultures (Fig. 3D). This demonstrates that DUX4 is able to activate target genes in murine nuclei initially devoid of any DUX4 protein. This co-culture system allows us to conclude that after transcription of DUX4 in a very small number of human nuclei within myotubes, DUX4 mRNA is translated in the cytoplasm and the resulting proteins spread into both human and murine nuclei where they activate genes downstream of DUX4. Thus, the molecular signature of FSHD is spread throughout myonuclei within myotubes, which can explain how a very rare toxic mRNA, present in only 1 in 1000 nucleus (9) can participate efficiently to the FSHD pathophysiology.
DM1 nuclei induced splicing defects in murine nuclei in heterokaryons
Consequences of nuclear protein spreading were also analyzed in DM1. We investigated by using co-cultures whether splicing factors produced by murine nuclei, can be titrated by DM1 CUGexp-RNAs trapped in neighboring human nuclei within the same myotube and consequently, induce splicing misregulation events in the murine nuclei. Co-cultures between DM1, control or FSHD human cell lines and C2C12 cells were performed and harvested at Day 5 of differentiation. After reverse transcription (RT), splicing of Serca1 exon 22 and Bin1 exon 11, that were shown to be mis-spliced in DM1 muscle cells (15,19), were analyzed. Using human (SERCA1) or murine (Serca1) specific primers, we observed that in heterokaryons formed between C2C12 and DM1 human cells, splicing of both SERCA1 and Serca1 exon 22 are impaired in the presence of DM1 cells when compared with co-cultures with control or FSHD cells (Fig. 4A). Indeed, the percentage of inclusion of human SERCA1 exon 22 is 53% in heterokaryons formed between C2C12 and control cells and 20% when the heterokaryons are formed between C2C12 cells and DM1 cells. Similarly, in the co-cultures performed between C2C12 and non-DM1 cells, the percentage of inclusion of mouse Serca1 is 49% but only 30% in the C2C12-DM1 co-cultures (Fig. 4B). Splicing of Serca1 exon 22 is thus impaired in both human and murine nuclei by the presence of the mutated DMPK mRNA. Splicing of BIN1/Bin1 is also altered in both human and murine nuclei in the co-cultures with DM1 cells. In the absence of the mutated CUGexp-RNAs, the percentage of inclusion of murine Bin1 exon 11 represents 72%, whereas in the presence of the mutated mRNA, this percentage drops to 47%. Finally, we observed that Mbln1 exon 7 splicing is also impaired in the DM1-C2C12 co-cultures when compared with FSHD or control co-cultures (Fig. 4C).
These results show that within a syncytium, pre-mRNAs transcribed in murine nuclei surrounded by DM1 nuclei develop splicing defects, suggesting a trans-dominant effect of human CUGexp-RNA on murine splicing and the spreading of the DM1 phenotype to other nuclei within the myotubes.
CUGexp-RNA aggregates are localized only in human nuclei
Finally, we investigated whether the splicing defects we observed were related to the presence of the CUGexp-RNAs within the murine nuclei. A RNA-FISH directed against CUG expansion was performed and revealed that CUGexp-RNA aggregates are only localized in human nuclei, confirming that the mis-splicing events observed in the murine nuclei are due to the trans-dominant depletion of murine splicing factors by the CUGexp-RNAs present in human nuclei and not to the presence of the human CUGexp-RNAs in aggregates within the murine nuclei (Fig. 4D). As for FSHD, the molecular signature of DM1 is spread from human to murine nuclei throughout myotubes.
The fact that muscle is a multinucleated organ has often been seen as a protection against the effect of some mutations. For instance and despite the random X chromosome inactivation, females heterozygous for mutations in the dystrophin gene will not develop Duchenne Muscular Dystrophy thanks to the synthesis of DYSTROPHIN by the healthy allele in 50% nuclei and the diffusion of the functional protein throughout the myotube at the sarcolemma. Another example is the accumulation of mutations in the mitochondrial genome of fibers can be somewhat limited by the fusion of satellite cells, which do not accumulate these mutations at the same rate (20,21). This phenomenon is not true for all muscle proteins and depends on the size of the nuclear domain that might impair intracellular movement of proteins (22). The spreading of nuclear proteins has been described several years ago using heterokaryons (3,23). However, the consequences on pathophysiology and therapies for muscular diseases had not been investigated.
In this study, we investigated whether nuclear protein spreading could be evidenced in myotubes. Concerning FSHD, one of the key questions about the pathophysiology of this disease is how such a rare event the transcription of DUX4 in only a few nuclei can trigger the worsening of entire muscle fibers? Even though the hypothesis was raised that within myotubes, DUX4 may be transcribed in few nuclei but the resulting protein may spread to other nuclei inducing a cascade of abnormal gene activations, it was never demonstrated (11,24). Here, we show the spreading of the DUX4 protein from one FSHD nucleus to neighboring nuclei, thus demonstrating that spreading of nuclear proteins can exist and in additional are functional since genes downstream of DUX4 are activated. This spreading may explain that, when investigating the presence of the protein DUX4, a decrease in signal intensity is observed in consecutive nuclei within myotubes. This consecutive gradient of DUX4 intensity is observed in both human and murine nuclei. Indeed, as all proteins, DUX4 is translated in the cytoplasm and is redirected to the nuclei by three monopartite nuclear localization signals (25). The probability to reach one specific nucleus may depend on the distance and on the proteins involved in the nuclear import. The longest distance we measured between one human DUX4-positive nucleus and one murine DUX4-positive nucleus was 51 µm. However, the nucleus of a myoblast moves rapidly after fusion towards the central myotube nuclei. It is thus difficult to conclude that DUX4 can migrate as far as 51 µm because two nuclei can be adjacent at a given time and then rapidly migrate at the two extremities of the myotubes (26). However, it is possible that in vivo, during fiber repair, the nucleus newly incorporated in muscle fiber is briefly close enough to a DUX4 expressing nucleus to be ‘contaminated’ and thus spread DUX4-associated molecular phenotype.
Another interesting point raised about DUX4 propagation within myotube is whether or not the transactivation of either human or murine gene is due to a strong activation from a very limited number of nuclei or a weak activation from many nuclei (24). It has been very recently published that the number of DUX4-positive nuclei is 6–8-fold increased when myoblasts are differentiated using 20% knock out serum replacer (KSOR) instead of classical conditions (serum depletion+ITS) (10). In parallel, the activation of the genes downstream of DUX4 is increased in the KSOR cultures, showing that the amplification of the expression of downstream targets of DUX4 is correlated to the number of DUX4-positive nuclei. Consequently, the effect of the spreading is more likely linked to a weak activation from most of the DUX4-positive nuclei rather than a strong activation from one adjacent nucleus.
Concerning DM1, we show in DM1-C2C12 heterokaryons that nuclear CUGexp-RNAs transcribed by DM1 nuclei sequester splicing factors produced by C2C12 nuclei, inducing murine pre-mRNA mis-splicing. In both FSHD and DM1 heterokaryons, a spreading of the pathological phenotype from one nucleus to others is thus confirmed. Such diffusion may be detrimental only for gain-of-function diseases, and similar mechanisms may occur in other muscular diseases involving nuclear protein dysfunctions such as oculopharyngeal muscular dystrophy (OPMD) in which mutated PABPN1 is aggregated with numerous nuclear proteins (27). The physiological consequences of this spreading may depend on the size of the D4Z4 repeats for FSHD, the level of expansion of the CTG trinucleotide repeat for DM1 or on the expansion of the trinucleotide (GCN) repeat in the first exon of PABPN1 for OPMD. Moreover, several other parameters such as the half-life of the mRNA or proteins involved, their expression levels and their functions may also be important.
Another aspect of the consequences of such diffusion is a potential impact on therapeutic approaches. Questions such as how many nuclei will have to be repaired by gene transfer or how many normal cells have to be grafted to counteract the pathological spreading needs to be investigated, and the answers may be patient specific. However, several studies have now started targeting the nuclear machinery to avoid viral replication in the nucleus or improve drug delivery (28) and similar approaches could be developed for FSHD to avoid DUX4 dissemination.
Finally, multinucleated cells are present in several human organs among them the liver (hepatocytes), the brain (Purkinje cells) and bones (osteoclasts) where similar spreading may occur and may impact therapy efficiency. For instance, somatic cell fusion events between bone marrow derived-cells and Purkinje cells have been proposed to counteract spinocerebellar ataxia type 1, but since the pathological mechanism of this disease involves expansion of the polyglutamine tract in ATXN1 which induces a loss of RORα from Purkinje cell nuclei (29), such strategies may not be successful.
In conclusion, we show that spreading of nuclear proteins or mRNA participates to the dissemination of a pathological molecular phenotype within multinucleated fibers, and may represent a novel mechanism to explain pathophysiology. This should be taken into account in the design of new therapeutic strategies.
MATERIALS AND METHODS
Immortalized FSHD myoblast clones were generated from one mosaic patient as previously described (30) and the immortalized DM1 myoblasts used in this study carries 2500 CTG repeats in the 3′UTR of the DMPK gene. Human clones were cultivated in proliferation medium [4 vols of DMEM, 1 vol of 199 medium, FBS 20%, gentamycin 50 µg/ml (Life technologies, Saint Aubin, France)] supplemented with insulin 5 µg/ml dexametasone 0.2 µg/ml, β-FGF 0.5 ng/ml, hEGF 5 ng/ml and fetuine 25 µg/ml. Murine C2C12 were cultivated in DMEM, FBS 10%, gentamycin 50 µg/ml. Differentiation medium for both human and murine cells was composed of DMEM supplemented with insulin (10 µg/ml). For FSHD experiments, heterokaryon formation was obtained by mixing clones and C2C12 at ratio 4 : 1 (1.5 × 104 and 3.75 × 103 cells per cm2, respectively) on gelatin-coated plates or slides in proliferation medium. Two days later, proliferation medium was replaced by differentiation medium. The fusion efficacy was improved by treating cells with PEG 1000 at a concentration of 50% (W/V) in DMEM pH 7.4 for 60 s at 37°C according to the method as previously described (23). Cells were carefully rinsed and cytosine arabinoside (Ara-C, 10e-5 M, Sigma-Aldrich) was added to reduce the proportion of unfused cells. Five days after the induction of differentiation, the cells were trypsinized and filtered using pre-separation filters of 30 µm (Miltenyi Biotech, Paris, France) to selectively retain myotubes. For DM1 heterokaryon experiments, a ratio 4 : 1 between DM1/C2C12 cells has been used. The myoblasts were plated on gelatin-coated plates and differentiation was induced 2 days later. Five days after triggering differentiation, cells were harvested and filtered to retain myotubes.
RNA extraction, PCR and real-time PCR
Trizol was directly added on filters and RNA extraction was performed according to the manufacturer protocol (Life technologies, Saint Aubin, France). RNA concentration was determined using a nanodrop ND-1000 spectrophotometer (Thermo Scientific, Wilmington, USA). The RT was carried out on 1 µg of total RNA with Roche Transcriptor First Strand cDNA Synthesis Kit (Roche, Meylan, France) at 60°C for 50 min with 1 µl of oligo dT and 2 µl of random hexameres in a final volume of 20 µl.
Quantitative PCRs were performed in a final volume of 9 µl with 0.4 µl of RT product, 0.18 µl of each forward and reverse primers 20 pmol/µl (Table 1), and 4.5 µl of SYBR®Green mastermix 2× (Roche, Meylan, France). The qPCR was run in triplicates on a LightCycler 480 Real-Time PCR System (Roche, Meylan, France). The qPCR cycling conditions were 94°C for 5 min, followed by 50 cycles at 95°C for 30 s and 60°C for 15 s and 72°C for 15 s. The PCR for BIN1/Bin1 and SERCA1/Serca1 were performed as previously described for BIN1 (31). PSMA2 and Psma2 were used, respectively, as human and murine normalizer as determined by the geNorm software.
For co-immunostaining of DUX4 and LAMIN A/C, cells were fixed in 2% paraformaldehyde for 10 min at room temperature and washed twice with PBS 1×. Cells were permeabilized with 1% Triton X-100 (Sigma) in PBS for 10 min at room temperature with gentle rocking. After blocking non-specific sites with PBS-FBS 2%, E5.5 DUX4 antibody directed against the C-terminal region of DUX4 (dilution 1:10) and antibody against human LAMIN A/C (clone JOL2, mouse IgG1; 1:100; Abcam, Cambridge, MA, USA) were diluted in PBS-FBS 2% and incubated with the cells overnight at 4°C. After three washes with PBS-0.5% Triton X-100, the cells were incubated with the secondary antibodies [goat anti rabbit IgGH+L–Alexa 488 (1/200) and goat anti mouse IgG1-Alexa 555 (1/200)] for 1 h in the dark. The cells were washed three times with PBS-0.5% Triton X-100 and incubated 5 min with Hoescht 33342 (1/1000, Thermo scientific, Courtaboeuf, France) before being mounted on microscope slides using Dako fluorescent mounting medium (Dako, Glostrup, Denmark).
The DUX4 probe was synthetized using the PCR DIG Probe Synthesis Kit (Roche, Meylan, France) with a 1 : 6 ratio of DIG-dUTP. The primers are DUX4_FL2_fw and DUX4_FL2_rev allowing a DUX4-specific amplificon of 164 bp. The PCR was performed in a final volume of 50 µl using the taq platinium and 1 µl of cDNA, according to a manufacturer protocol. The probe was purified using the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, Hoerd, France). The RNA-FISH was performed on immortalized FSHD myotubes at Day 5 of differentiation. The cells were fixed for 10 min in PFA 4%, rinsed with PBS and incubated with hybridization buffer (formamide 40%, 2×SSC, 0.2% BSA)+20 ng of DUX4 probe for 90 min at 50°C. The cells were then washed and incubated with PBS-tween 0.1% for 30 min at 55°C, and non-specific binding was blocked with PBS-FBS 2% for 30 min at room temperature. Cells were then incubated with anti-digoxigenin-Rhodamine Fab Fragments (dilution 1/200, Roche diagnostics) for 1 h in the dark. After three washes with PBS, the cells were incubated 5 min with Hoescht 33342 and mounted on microscope slides using Dako fluorescent mounting medium.
The probe used to detect DM1 foci was a 2′OMe oligo containing the sequence (CAG)7-Cy3. The cells were fixed in PFA 4% for 10 min, washed in PBS and incubated in hybridization buffer+probe (dilution 1/1000) for 90 min at 37°C. The cells were then washed and incubated in PBS-tween 0.1% for 30 min at 42°C, then rinsed in PBS before being incubated with Hoescht 33342 and mounted on microscope slides using Dako fluorescent mounting.
Conflict of Interest statement. None declared.
This work was supported by the Association Française contre les Myopathies (AFM-Téléthon, France), UPMC-Emergence 2010 (to J.D.), FSHD Global research foundation Ltd (to M.F.), the FSH society (FSHS-22012-03 to V.M.), and the Agence Nationale de la Recherche (FSHDecipher, ANR-13-BSV1-0004 to J.D.). The authors acknowledge the platform of the Institut de Myologie and in particular Kamel Mamchaoui for immortalization of human cells. The E5.5 DUX4 antibody was a kind gift from Dr S. Tapscott and pci-neo and pci-DUX4 a gift from Dr A. Belayew.