E3-ligase knock down revealed differential titin degradation by autophagy and the ubiquitin proteasome system

The sarcomere protein titin is a major determinant of cardiomyocyte stiffness and ventricular distensibility. The constant mechanical stress on titin requires well-controlled protein quality control, the exact mechanisms of which have not yet been fully elucidated. Here, we analyzed E3-ligases potentially responsible for cardiac titin ubiquitination and specifically studied the involvement of the autophagosomal system in titin degradation. Pharmacological inhibition of autophagy and the proteasome in cultured primary rat cardiomyocytes significantly elevated titin ubiquitination and increased titin degradation. Using in-vitro pull down assays we identified binding of E3-ligases MuRF1-3, CHIP and Fbx32 to several titin domains. Immunofluorescence analysis showed sarcomeric localization of the E3-ligases. siRNA-mediated knock-down of the E3-ligases MuRF-1, -3 and a combination of CHIP/Fbx32 significantly reduced autophagy-related titin ubiquitination, whereas knock-down of MuRF-2 and -3 reduced proteasome-related titin ubiquitination. We demonstrated that the proteasomal and the autophagosomal-lysosomal system participate in degradation of the titin filament. We found that ubiquitination and degradation of titin are partially regulated by E3-ligases of the MuRF family. We further identified CHIP and Fbx32 as E3-ligases involved in titin ubiquitination.


Inhibition of autophagy increases titin ubiquitination in embryonic rat cardiomyocytes.
To identify a potential involvement of autophagy in sarcomeric titin degradation we inhibited autolysosomal activity by chloroquine (CQ, 10 µM) or bafilomycin (BAF, 100 nM) application to cultured embryonic rat cardiomyocytes (ERC) and analyzed the effect on the titin filament. To test the efficiency of inhibition by the two substances we performed immunofluorescent stainings and measured the protein levels of two autophagy markers, microtubule-associated protein light chain 3 (LC3 II) and Sequestosome-1 /p62 (SQSTM1, also known as ubiquitin-binding protein p62). LC3-II is widely used to monitor autophagy, as it becomes incorporated into the membranes of maturating autophagosomes 24 and contains a binding site for SQSTM1/p62, in the following termed p62. P62 is an autophagosome cargo protein that targets other proteins for selective autophagy. In response to autophagy inhibition, both proteins, LC3-II and p62, are expected to accumulate within the cell 25  www.nature.com/scientificreports/ In CQ and in BAF-treated ERCs we observed a strong increase in cytosolic LC3 and p62, most likely indicative for non-degraded autophagosomes (Fig. 1A). Under control conditions staining for p62 showed a mainly cytosolic but also some punctate sarcomeric A-band/M-line localization in about 20% of the cells, likely indicating p62 association to titin (Suppl. Fig. 1A) 22 . In addition, both LC3-II and p62 showed strongly increased protein levels in Western blot analysis after 24 h of inhibition by CQ and BAF (Fig. 1B). After proteasomal inhibition with MG132, p62 protein level was elevated and accumulation within cardiomyocytes could be observed, while LC3-II protein level remained unchanged and intracellular LC3 aggregates were not detected (Fig. 1B). We conclude from our observations that the combination of increased LC3 and elevated p62 clearly indicate inhibition of autophagy. To further confirm the effectiveness of CQ and BAF-mediated inhibition of autophagy we analyzed total ubiquitination and K63-dependent polyubiquitination of proteins in the molecular weight range of 70-250 kDa (Fig. 1C) and found markedly increased signal intensities in response to both treatments. Modification of the titin filament was tested using Western blot analyses of agarose-stabilized 2.2% SDS-PAGE, which allowed exclusive detection of titin in the molecular weight range of approx. 3 MDa (Suppl. Fig. 1B). Because no other proteins are known and detectable in this molecular weight range the ubiquitination signals on such blots can be considered to originate from titin. We observed that inhibition of autophagy with CQ for 2-24 h increased the K63-polyubiquitination of full-length titin (Fig. 1D). Incubation for 24 h with BAF resulted in a similar increase of ubiquitinated titin (Fig. 1E). To test if the increased titin ubiquitination is indicative for increased titin degradation, we analyzed the abundance of a specific titin degradation product called T2. This T2 degradation band has been recognized to be a product of the pre-digest and degradation intermediate of titin 6 . As a result of 24 h inhibition of autophagy the levels of the T2/T1 ratio, analyzed by 2.2% SDS PAGE, were significantly elevated (CQ: 33.5% ± 7,6%; BAF: 27% ± 3.3%) (Fig. 1F). A similar increase in the T2/T1 ratio (+ 50.9% ± 16.5%) was observed after blocking proteasomal activity with MG132 (Fig. 1F). After 3 days of cultivation, including 24 h inhibition of autolysosomal or proteasomal activity, composition of cardiac titin isoforms N2B and N2BA remained unchanged compared to control cells (Fig. 1F). To analyze if inhibition of the proteasome or autophagy also alters total titin levels we calculated the titin/actinin ratio. We detected significantly higher levels of titin after proteasomal (+ 56.1% ± 22.4%) or autophagy inhibition (+ 54.3% ± 20.5%; Fig. 1G), compared to controls suggesting accumulation of unprocessed titin.

E3-ligase interaction with the titin filament.
Mono-and poly-ubiquitination reactions are performed in 3 steps and are finally catalyzed by E3-ligases. For the ubiquitination reaction, E3-ligases need to be in contact with their target proteins. We therefore tested the interaction of human striated muscle specific E3-ligases, or E3-ligases which are highly expressed in striated muscle, with 4 different human titin fragments by in vitro pull-down assays ( Fig. 2A/B). In addition to the previously described titin interaction partners MuRF-1 and 2, we analyzed MuRF-3, CHIP and Fbx32. E3-ligases were immobilized on sepharose beads and incubated with purified titin fragments. All used titin fragments showed a high purity grade (Suppl. Fig. 2A). The affinity of the E3-ligases to the different titin fragments were analyzed by comparison of the input signal to the bound signal (Suppl. Fig. 2B). As the A168-170 fragment showed low binding to the GST-tag control, the GST-bound signal was subtracted for all E3-ligases. MuRF-1 and -3 showed binding to the N2B domain and A168-170. For MuRF-2 we observed similar interaction with the PEVK domain and the A168-170 fragment. CHIP associated to the N2B domain and A168-170 with a trend for higher binding to the N2B domain. Fbx32 interacted with the N2B, A168-170 and the N2A fragment, with the strongest binding to A168-170 (Suppl. Fig. 2B). We further studied intracellular localization of the E3-ligases in cultured ERCs and adult rat cardiomyocytes (ARC). Cardiomyocyte-specific localization was confirmed by co-staining with α-actinin, which is a specific marker for sarcomeric Z-discs in cardiomyocytes. At culture day 3 a subset of ERCs showed sarcomeric localization of CHIP and Fbx32 in the Z-disc/I-band region and A-band, respectively. MuRF-1, -2 and -3 did not show sarcomeric localization after 3 days. However, after 8 days in culture, we detected partial localization of all MuRF isoforms at the A-band (MuRF-1) and the Z-disc/I-band in a small number of cells (< 10%) (MuRF-2 and -3) ( Fig. 2C and Suppl. Fig. 3A). In ARCs, all investigated E3-ligases showed sarcomeric localization in almost all cells. In ARC Fbx32 localized to the Z-disc region instead of the M-line, whereas all other E3-ligases had the same localization as seen in ERCs ( Fig. 2D and Suppl. Fig. 3B). Magnifications and intensity profiles of the antibody signals further demonstrate localization of the E3-ligases within the Z-disc/I-band region or the A-band, respectively (Suppl. Fig. 3B). As many E3-ligases need proteases like calpains for ubiquitination of their substrates we further checked intracellular localization of calpain-1 in cultured ERCs, using immunofluorescence. In the majority of ERCs (> 90%) we found calpain-1 localization in the cytosol and in the I-band of the sarcomere (Suppl. Fig. 3C).

Downregulation of MuRF1-3 alters titin ubiquitination after autophagy and proteasomal inhibition.
Due to its gigantic size we hypothesized that titin is likely ubiquitinated by more than just one E3-ligase. We therefore aimed to identify E3 ligases that ubiquitinate the titin filament and treated ERCs at cultivation day 5 for 24 h with siRNAs targeting MuRF-1, -2 and -3, CHIP and Fbx32. After 72 h (cultivation day 8), we additionally inhibited the proteasome or autophagy for 24 h using MG132 and BAF, respectively. If the listed E3 ligases were responsible for proteasome-or autophagy-targeting titin ubiquitination, siRNA-mediated knock-down should prevent increased titin ubiquitination after inhibition by MG132 or BAF. To exclude off target effects of the siRNAs we additionally used two independent siRNA negative controls. Titin ubiquitination after inhibition of autophagy or the proteasome was not altered by the control siRNAs (Suppl. Fig. 4A). Since members of the MuRF family can functionally compensate the loss of each other 18 , we started with knocking down all three MuRFs as well as CHIP and Fbx32 in common to narrow down the candidates. Efficiency of the siRNA treatment was analyzed by Western blot analysis and demonstrated that protein levels of the E3-ligases were markedly decreased by the siRNA treatment (Fig. 3A). Total titin protein levels were not significantly www.nature.com/scientificreports/ www.nature.com/scientificreports/ affected by MuRF-1, -2 and -3 or Fbx32/CHIP siRNA treatment without additional inhibition of proteasome or autophagy, but we observed a strong variability in total titin: actinin ratio within groups (Fig. 3B). We hypothesized that at this time of cultivation ERCs have already differentiated to a more adult state. This maturation involves the loss of proliferative ability leading to a substantial reduction in the degradation and synthesis of sarcomeric proteins. In support of this hypothesis we stained for the cell cycle marker Ki67 (Suppl. Fig. 4B) and found an almost complete loss of Ki67 staining after 8 days. Additional inhibition of autophagy or the proteasome did not reduce ubiquitination of cardiomyocyte protein within the molecular weight range of 70-250 kDa (Fig. 3C). In contrast, levels of ubiquitinated titin were significantly decreased in cells with a knock down of MuRF-1, -2 and -3 under basal conditions (− 64.4% ± 9.43%), after inhibition of autophagy (− 69.3% ± 2.91%) and proteasomal activity (− 66.6% ± 8.37%) ( Fig. 3D/E). In cells with a knock-down of CHIP and Fbx32 we determined significantly lower titin ubiquitination levels after inhibition of autophagy (-33.9% ± 4.56%), but not under basal conditions or after inhibition of the proteasome (Fig. 3D/E).

Discussion
In a beating heart, sarcomeres are exposed to permanent mechanical stress. In order to maintain proper cardiac function, turnover and renewal of sarcomeric components as well as adaptation to altered mechanical strain have to be precisely managed by protein quality control mechanisms. In our present study, we present first evidence of K63-dependent poly-ubiquitination of the titin filament, which is strongly increased after inhibition of autophagy. We conclude from these data that degradation of titin is not exclusively performed by the proteasome, but also by autophagy. www.nature.com/scientificreports/ To test the efficiency of autolysosomal inhibition the expression level of autophagy marker proteins was tested using immunofluorescence and Western blot analyses. In response to inhibited autophagy, we detected increased protein levels of LC3-II. However, using only LC3-II as marker is not sufficient to draw a conclusion concerning activation or inhibition of autophagy because increased LC3-II protein levels could indicate both higher autophagic activity by elevated LC3-I lipidation and formation of autophagosomes, or impaired lysosomal degradation of unprocessed autophagosomes. In addition, we therefore analyzed p62 protein levels, which binds LC3 at the inner membrane of the autophagosome and becomes degraded after fusion with a lysosome 25 . Protein levels of p62 are increased by inhibition and decreased by activation of autophagy 26 .We also found p62 accumulation after inhibition of the proteasome (Fig. 1B) suggesting that p62 protein levels can increase independent of the formation of autophagosomes. These findings are hints for proteasomal degradation of p62 and its shuttled substrates. Proteasomal inhibition did not affect the protein level of LC3-II (Fig. 1B). The combination of increased LC3-II and p62 expression levels observed in our study can be considered as an indicator for effective autophagy inhibition.
The hypothesis that titin degradation products are in part processed by autophagosomes is further supported by the presence of PEST motifs in the so-called PEVK-domain of the filament protein. Similar to K48-dependent www.nature.com/scientificreports/ ubiquitination, K63-dependent polyubiquitination preferably occurs at these PEST motifs, which have been described as potent protein degradation-targeting signals 27 . These regions are rich in proline (P), glutamic acid (E), serine (S), and threonine (T). Sequence analysis of titin revealed such PEST motifs in the PEVK domain titin, which is rich in the amino acids proline (P), glutamic acid (E), valine (V) and lysine (K). Previous reports have demonstrated titin proteolysis by calpains potentially defining calpains as central components of titin turnover 9,28,29 . Matrix-metallo-protease 2 has also been shown to bind within the sarcomere and degrade titin resulting in a higher abundance of the titin degradation product T2 and an increased T2/ T1 ratio 21 . It has therefore been widely accepted that a molecule of the size of titin requires pre-digestion for being released from the pseudocrystalline sarcomeric structure. In our present study we report that inhibition of autophagy and proteasomal function resulted in ubiquitination not only of the T2-band but to a large extent of non-processed titin. This finding indicate that ubiquitination occurred at the full-length titin suggesting that the ubiquitination process is independent of and happens before the pre-digest. Nevertheless, calpain and MMP2-mediated proteolysis may be required for E3 ligases to reach some of their target sequences along the titin filament, particularly in the tightly packed A-band region that is closely associated with the myosin filament. The specific titin degradation band T2 was increased after inhibition of autophagy or the proteasome suggesting that under these conditions the titin degradation is stalled after proteolytic pre-digest, resulting in accumulation of the pre-digest product T2. At the same time, protein levels of full-length titin were increased indicating that inhibition of autophagy or the proteasome additionally results in significantly lower amounts of pre-digested full-length titin. This could in part be explained by previous findings reporting that peptide aldehyde inhibitors like MG132 can reduce calpain and MMP2 activity 30,31 . Of note, titin T2 band has a similar molecular weight than the recently identified cronos titin isoform, which is highly expressed early in fetal cardiomyocyte development and has been suggested to be important for normal sarcomere maturation and function 32 . We currently cannot exclude the possibility that in the ERC experiments part of the observed increase in the T2 band may derive from higher expression of the titin cronos isoform.
We further aimed to identify E3-ligases involved in titin degradation via the proteasome and autophagy and performed in vitro pull-down experiments with selected E3 ligases and recombinant titin fragments. Immunfluorescent stainings of E3 ligases in ERC and ARC confirmed the observations from pull down experiments: MuRF-1 was detected within the A-band/M-line and showed binding to the titin fragment A168-170. We also observed binding of MuRF-1 to the titin N2B-domain, which may provide an explanation for the previous observation that MuRF-1 localizes at the Z-disc/I-band region 33 . We found MuRF-2 localization in the Z-disc/I-band region of cultured embryonic and adult rat cardiomyocytes, which is line with its in vitro binding to the titin PEVK-domain. Unlike Lange et al. 22 we did not observe MuRF-2 localization within the A-band/M-line. We further identified MuRF-3, Fbx32 and CHIP as novel interaction partners of titin. Our data suggest interaction of MuRF-3 and Fbx32 with the titin A168/170 domain, which is supported by their sarcomeric localization at the M-line of cultured cardiomyocytes. Fbx32 further showed binding affinity to the N2B and N2A titin fragments and intracellular localization in Z-disk proximity in adult cardiomyocytes. This is in line with a previous report showing localization of Fbx32 at the Z-disk 28 . Of note, in the fluorescence microscope used here, it was not possible to clearly distinguish between Z-disk and I-band localization. Therefore, a signal in Z-disk proximity is quite plausible for Fbx32 binding to the N2A or N2B segment. Considering the gigantic size of titin, binding of E3-ligases to other titin domains than the ones tested here is of course possible. Also unspecific or non-functional binding of E3-ligases to titin in the immunofluorescence stainings cannot be entirely excluded. Figure 5 summarizes the main findings of the interaction and localization studies and illustrates possible regulation of E3-ligase based titin ubiquitination and degradation by autophagy or the proteasome.
We next aimed to establish the functional importance of the identified E3 ligases by targeted deletion using specific siRNAs and determined the effects on titin ubiquitination. Ubiquitination of titin has previously been shown in skeletal muscle by Western blot analysis and electron microscopic images using ubiquitin antibodies 18 . However, the study did not determine whether titin was targeted for degradation by autophagy or the proteasome. Here we provide direct evidence for ubiquitination of titin by five E3-ligases. In addition, our study discriminates for the first time between autophagy and proteasome dependent ubiquitination of titin. The members of the MuRF family are able to build hetero-dimers and thereby target different domains within the titin filament for ubiquitination. Varying hetero-dimers could also explain autophagy-or proteasome-related ubiquitination. Other reports assumed a redundancy of MuRF members concerning degradation of several target proteins 18 . Here, knocking down single MuRFs in cardiomyocyte cultures differentially altered titin ubiquitination, indicating that MuRF 1, -2 or -3 are not able to compensate the loss of each other completely. Simultaneous knock-down of CHIP and Fbx32 also significantly lowered titin ubiquitination after autophagy inhibition. Why single knock down of Fbx32 and CHIP in our model did not significantly alter titin ubiquitination still remains elusive at this point. CHIP has been reported to play an important role in ischemia-reperfusion injury as knock out of CHIP worsens the outcome in mouse models 19 . Our group previously demonstrated that titin ubiquitination is strongly increased in the non-ischemic area during the first days after cardiac ischemia, possibly because of increased turnover in response to elevated mechanical stress 23 . We now hypothesize that this titin ubiquitination is in part performed by CHIP and the other E3-ligases described here. Decreased titin ubiquitination and impaired titin degradation could therefore contribute to the detrimental effects of CHIP knockdown for myocardial performance after ischemia 19 .
Somewhat unexpectedly, downregulation of E3-ligases did not result in differences in total titin protein levels compared to untreated cells, suggesting that titin degradation was not altered by the siRNA treatment. Another unexpected result was the unchanged protein ubiquitination in the molecular weight section from 70 to 250 kDa after siRNA treatment and inhibition of autophagy or the proteasome. A possible explanation for both observations could be the fact that E3-ligases were strongly reduced by the siRNA treatment, but not completely absent so sufficient levels of ubiquitination and turnover may still occur. Besides, it is conceivable that MuRFs  34 may also compensate for the loss of others.
The binding assays showed two main clusters of E3-ligase interaction with titin: one in the I-band region, the other at the A-band/M-band junction (summarized in Fig. 5). The Titin T2 fragment, as a result of the predigest, has been reported to contain mainly the A-band part of titin 6 . From our data we conclude that particularly MuRF2 and MuRF3, which bind to titin in the I-band region, could be responsible for I-band titin ubiquitination and its subsequent processing by the UPS. The A-band part of titin is tightly associated to the myosin-filaments and therefore not easily accessible, even after proteolytic removal of the proximal I-band part. We therefore hypothesize that the entire A-band part of titin could be degraded via autophagy. Along the same line, p62, which associates to the titin kinase domain in close M-line proximity 22 possibly serves as an anchor protein for the formation of autophagosomes.
Unfolded domain oxidation (UnDOx) within the titin filament was recently reported to alter titin-based properties 35 . Many pathologies including ischemia are linked to alterations of the oxidative state. Titin oxidation has been shown to occur in the ischemic region of hearts from mice that underwent permanent ligature of the LAD 35 . Unfolding events may also increase the vulnerability of the titin filament and subsequently ubiquitination and degradation.
Revealing the mechanisms involved in titin turnover is important also from a clinical point of view, as disturbances of protein degradation are hallmarks of several diseases including cardiovascular diseases 36 . Insufficient titin turnover and modification of myocardial elasticity could be a part of this dysregulation. In skeletal muscle M-line titin has been reported to be a binding partner as well as a substrate for calpain-3 and impaired titin-calpain interaction causes titinopathies with increased titin turnover 9,29 . Titin-truncating variants are the most common genetic cause of dilated cardiomyopathy 37 , which is often associated with changes in the cellular metabolism and decreased autophagic degradation of myocardial proteins. In turn, impairment of autophagy or the proteasome could lead to accumulation of dysfunctional titin filaments or even titin aggregates. Pathological hyper-activation of the degradation systems could potentially result in mislead turnover of functional filaments. Both situations likely affect passive stiffness and diastolic properties of the myocardium.

Conclusion
Taken together, we demonstrated that protein quality control of the sarcomeric titin filament not only involves degradation by the proteasome but also by the autolysosomal system. Our data further suggest that ubiquitination of full-length titin does not necessarily require proteolytic pre-digestion. Finally, we propose that different muscle specific E3-ligases specifically ubiquitinate titin domains for targeted degradation by either autophagy or the proteasome. We conclude that more detailed knowledge of the fine-tuned ubiquitination and turnover process of titin is needed to understand how pathological modification of these processes may influence sarcomere function and affect the properties of the myocardium.

Materials and methods
Isolation of embryonic rat cardiomyocytes (E18). All animal procedures were conducted in accordance with the local animal care and use committee and the responsible authorities (Landesamt für Natur, Umwelt-und Verbraucherschutz Nordrhein-Westfalen) that reviewed and approved the experimental protocols www.nature.com/scientificreports/ (Az. 84-02.04.2017.A145). Hearts were obtained from embryos (gestational day 18) of pregnant adult Wistar rats (Janvier Labs). Adult animals were anesthetized with ketamine/xylazin (100 and 10 mg/kg body weight) embryos were removed by C-section and embryos as well as the adult animal were euthanized by decapitation. Embryonic rat cardiomyocytes (ERC) were isolated by enzymatic digestion of the embryonic heart and plated in 6-well dishes (day 0). Cells were cultured in DMEM + 20% FBS (culture medium) or serum-starved (DMEM + 1% charcoal filtered serum). ERCs were plated on gelatine-coated (1% in PBS) cover slips. Cells were then fixed with 2% paraformaldehyde. Primary antibodies were incubated overnight at 4 °C. Rabbit or mouse AlexaFluor488 and AlexaFluor555-coupled secondary antibodies were incubated for 2 h at room temperature. For inhibition of autophagy or proteasomal activity medium was supplemented with chloroquine (CQ, 10 µM, Sigma), bafilomycin (BAF, 100 nM, Cayman Chemical Company) or MG132 (1 µM, Sigma) for up to 24 h. The study was carried out in compliance with the ARRIVE guidelines.
SiRNA treatment in embryonic rat cardiomyocytes (E18). Rat  Isolation of adult rat cardiomyocytes. Adult rat CMs were isolated using a Langendorff perfusion system, as previously described 38,39 . Briefly, male Wistar rats were terminally anaesthetized with isoflurane and killed by cervical dislocation. The heart was dissected and the aorta was cannulated and connected to a perfusion apparatus. After perfusion with gassed digestion buffer single cardiomyocytes were isolated (for detailed information see online supplement). For immunofluorescent stainings cells were plated on laminin/gelatine coated cover glasses. Cells were then fixed with 2% paraformaldehyde. Primary antibodies were incubated overnight at 4 °C. Rabbit or mouse AlexaFluor488 and AlexaFluor555-coupled secondary antibodies were incubated for 2 h at room temperature.
Pull down interaction assays. All human E3-ligases and titin fragments were recombinant expressed and purified with the GST-Fusion System. With the GST-tag E3-ligases were immobilized on GSH-sepharose beads and incubated with supernatant containing purified titin fragments for 90 min at 4 °C. All titin fragments were tested for an affinity to the GST-tag. The beads were washed four times to eliminate unbound protein, and the interaction was detected by Western blotting with overnight incubations of anti-titin antibodies (α-PEVK; α-N2B; α-N2A and α-A186) and 1 h incubations of horseradish peroxidase-conjugated anti-rabbit secondary antibody (dilution 1:10.000). Bands were visualized using a Fusion FX imaging system (Vilber & Lourmat, France). Binding affinity was analyzed by calculating the ratio of input signals to bound signals.
Statistical analysis. Data are presented as mean ± standard error of the mean (SEM). Statistical significance was analyzed by using unpaired Student t test. Multiple comparisons were performed using one-way ANOVA Bonferroni t-test or Dunn`s multiple comparison. P values < 0.05 were considered significant and are indicated in the figures by asterisks. Analysis was performed using Sigma Plot (version 13.0).