Desmin intermediate filaments and tubulin detyrosination stabilize growing microtubules in the cardiomyocyte

Dynamic microtubules are stabilized at the Z-disk and interact with desmin intermediate filaments

To study the dynamics of growing microtubules in mature cardiomyocytes, we treated adult rat cardiomyocytes with adenovirus containing GFP-labeled End-Binding Protein 3 (EB3-GFP) to directly visualize the plus-end of growing microtubules by time-lapse imaging (S. Movie 1). The dynamic properties of microtubules can be quantified as events that mark their transitions from growing (polymerization) to shrinking (depolymerization) states (Fig. 1a). These events consist of catastrophes (transitions from growth to shrinkage), rescues (transitions from shrinkage to growth), and pauses (neither growth nor shrinkage). Conveniently, EB3-GFP also provided a fainter, non-specific labeling of the protein-rich Z-disk region, enabling us to visualize where dynamic events occurred relative to a sarcomeric marker (Fig. 1b).

Fig. 1figure 1

Dynamic microtubules are stabilized at the Z-disk and preferentially interact with desmin intermediate filaments. a Schematic of the transition states of microtubule dynamics. b Representative kymograph from control cardiomyocytes transduced with AdV-EB3-GFP; black arrows denote Z-disk and colored arrows denote transition events. c Quantification of initiation, rescue, pause, and catastrophe events On and Off the Z-disk in control cardiomyocytes (N = 19 cells, n = 228 events). The bar represents mean ± 1SEM; statistical significance determined with Two Sample Kolmogorov–Smirnov Test. d Representative EM images from transverse sections of isolated cardiomyocytes. Microtubules are denoted by white arrows. In the right-hand panel, the area between the myofibrils is filled by membranous and filamentous structures consistent with intermediate filaments, which are bisected by microtubules. e Representative immunofluorescent images and f quantification of a-actinin-EB1 or desmin-EB1 PLA interactions in control cardiomyocytes (N = 3 rats, n = 10 cells per rat). The box represents the 25th and 75th percentiles ± 1SD, bolded line represents the mean; statistical significance was determined with two-sample Student’s T test

Under basal conditions, we observed a stark spatial bias in microtubule dynamic behavior, similar to that previously observed [13]. The initiation of microtubule growth, as well as pausing of growth, predominantly occurred at the level of the Z-disk (Fig. 1c). Conversely, catastrophes predominantly occurred off the Z-disk, while rescue from catastrophe again occurred more frequently at the Z-disk. As exemplified in S. Movies 1–2, myocyte microtubules tend to grow iteratively from one Z-disk to another, often pausing at each Z-disk region. If a microtubule undergoes catastrophe before reaching a Z-disk, it tends to shrink to a previous Z-disk, where rescue is more likely to occur. These data suggest factors at the Z-disk region strongly bias microtubule behavior and support the initialization and stabilization of growing microtubules.

Electron microscopy images of cardiomyocytes help illustrate the local environment surrounding microtubules at the nanoscale and suggest nearby elements that may stabilize microtubules. As seen in Fig. 1d, the microtubules running along the long-axis of the myocyte appear as 25 nm diameter tubes coming at the viewer in transverse sections, with a faint halo surrounding them where their C-terminal tails project. Microtubules most commonly run alongside, and not within, the sarcomere-containing myofibrils, squeezing in the gaps between myofibrils and the mitochondria or nucleus. Desmin intermediate filaments also occupy some of these gaps, wrapping around the myofibrils at the level of the Z-disk, and we observe microtubules bisecting through structures that resemble intermediate filaments and which surround the myofibrils at these locations (Fig. 1d, right). To orthogonally probe whether growing microtubules are more likely to interact with the intermediate filament vs. sarcomeric cytoskeleton, we utilized proximity ligation assay (PLA) to probe interactions between the endogenous microtubule plus-end tracking protein end-binding protein 1 (EB1) and either sarcomeric a-actinin or the intermediate filament desmin in adult rat cardiomyocytes. Although a-actinin is the most abundant protein in the Z-disk and expressed at substantially higher levels than desmin [8] (S. Fig. 1a), we observed ~ tenfold more abundant PLA puncta in the desmin-EB1 group compared to a-actinin-EB1, suggesting that the growing end of microtubules are frequently in close proximity to desmin intermediate filaments at the Z-disk (Fig. 1e, f).

Desmin stabilizes growing and shrinking microtubules at the Z-disk

We next directly interrogated the role of desmin in regulating microtubule stability by adenoviral delivery of shRNA to acutely deplete desmin (desmin KD) in cardiomyocytes. Complementing our previous validation of this construct by western blotting [18], we measured a 40–50% reduction in desmin expression after 48 h of desmin KD (S. Fig. 1b). We first interrogated the effect on microtubule stability using a modified subcellular fractionation assay from Fasset et al. [15] that allowed us to separate free tubulin from polymerized tubulin in the dynamic (i.e. cold-sensitive) microtubule pool (Fig. 2a). Acute desmin depletion resulted in an increased free to polymerized ratio in the dynamic microtubule pool (Fig. 2b,c, S, Fig. 1c), suggesting that desmin coordinates the stability of dynamic microtubules. We next quantified microtubule acetylation and detyrosination, markers of long-lived microtubules, and found that both were decreased in desmin KD myocytes, without alterations in whole cell tubulin content (Fig. 2b, c), suggesting that desmin normally helps maintain microtubule stability.

Fig. 2figure 2

Desmin stabilizes dynamic microtubules at the Z-disk. a Overview of the cell fractionation assay adapted from Fassett et al. [15] that allows for the separation of free tubulin and polymerized microtubules within the dynamic tubulin pool. b Representative western blot and c quantification of α-tubulin in free and dynamic microtubule fractions (top) or of total dTyr-tubulin, α-tubulin, and acetylated tubulin in the whole-cell lysate (bottom) from control (Scram) or Desmin knock-down (Des KD) cardiomyocytes (N = 3 rats, n = 5 WB technical lanes for dtyr and 6 for acetyl and tubulin fractions). d Representative EB3-GFP kymograph from Scram (top) or Des KD (bottom) cardiomyocytes. e Quantification of catastrophe, pause, and rescue event frequencies and f event locations in Scram or Des KD cardiomyocytes (N = cells, n = events). The bar represents mean ± 1SEM; statistical significance for C was determined with two-sample Student’s T test, and for E and F was determined with two-sample Kolmogorov–Smirnov test

Next, we directly quantified plus-end microtubule dynamics by EB3-GFP upon desmin depletion. Blind quantification of global event frequency revealed that desmin depletion modestly increased the frequency of catastrophes while more robustly reducing both the frequency of rescues and pauses (Fig. 2d, e). As seen in S. Movies 3–4, upon desmin depletion (S. Movies 4) microtubule growth still initiated at the Z-disk, but the iterative, longitudinal growth from one Z-disk to another seen in control cells (S. Movies 3) was lost. Instead, microtubules often grew past Z-disk regions without pausing, and following catastrophe they were less likely to be rescued at the previous Z-disk (Fig. 2d, f). Interrogation of where dynamic events occurred in relation to the Z-disk revealed that desmin depletion specifically increased the number of catastrophes that occurred on the Z-disk while reducing the number of catastrophes that occurred off the Z-disk (Fig. 2f). More strikingly, desmin depletion markedly reduced the number of pauses and rescues that occur specifically on the Z-disk, while not affecting pause or rescue behavior elsewhere (Fig. 2f). Together, these results indicate that desmin spatially coordinates microtubule dynamics and stabilizes both the growing and shrinking microtubule at the Z-disk.

Cardiomyocytes from global, desmin germ-line knockout mice are characterized by misaligned and degenerated sarcomeres with a disorganized microtubule network [6, 31]. Gross restructuring of the myofilaments could affect microtubule dynamics due to a change in the physical environment that is permissive to microtubule growth, for example by increasing the spacing between Z-disks of adjacent myofilaments. To assess if our comparatively brief desmin depletion altered myofilament spacing or alignment, we performed quantitative measurements on electron micrographs from desmin KD cardiomyocytes. Blind analysis indicated that this relatively short-term desmin depletion did not detectably alter myofilament spacing or alignment (S. Figure 2), consistent instead with a direct stabilizing effect of desmin intermediate filaments on the microtubule network.

We next interrogated the functional consequences of this reduced microtubule stability driven by desmin depletion. As a reduction in detyrosinated microtubules and their association with the Z-disk is associated with reduced cardiomyocyte viscoelasticity [31], we hypothesized that desmin-depleted myocytes would be less stiff. To test this, we performed transverse nanoindentation of cardiomyocytes and quantified Young’s modulus of the myocyte over a range of indentation rates. Desmin depletion specifically reduced the rate-dependent viscoelastic stiffness of the myocyte without significantly altering rate-independent elastic stiffness (S. Fig. 3a, b). Reduced viscoelasticity is consistent with reduced transient interactions between dynamic cytoskeletal filaments.

To directly test if the reduction in desmin alters microtubule buckling between sarcomeres, we performed a semi-automated, blind analysis of microtubule buckling, as in our previous work [31]. In control cells, most microtubules buckle in a clear sinusoidal pattern with a wavelength corresponding to the distance of a contracted sarcomere (~ 1.5–1.9 µm) (S. Fig. 3c, d) (S. Movie 5). Upon desmin depletion, fewer polymerized microtubules were observed in general, with more chaotic deformations and organization upon contraction (S. Movie 6). For microtubules that did buckle, we observed reductions in the amplitude of buckles (S. Fig. 3d) and the proportion of microtubules that buckled at wavelengths corresponding to the distance between 1 and 2 sarcomeres (1.5–1.9 or 3.0–3.8 µm, respectively) (S. Fig. 3e, f). Combined, these results are consistent with desmin coordinating the physical tethering and lateral reinforcement of detyrosinated microtubules at the cardiomyocyte Z-disk to regulate myocyte viscoelasticity.

Tyrosination alters the dynamics of the microtubule network

Next, we sought to determine the effect of detyrosination on the dynamics of the cardiomyocyte microtubule network. To reduce detyrosination, we utilized adenoviral delivery of TTL into isolated adult rat cardiomyocytes [31]. TTL binds and tyrosinates tubulin in a 1:1 complex, and this binding leads to tubulin sequestration. Hence, to separate the effects of tubulin tyrosination from tubulin sequestration, we utilized adenoviral delivery of TTL-E331Q (E331Q), a verified catalytically dead mutant of TTL that binds and sequesters tubulin but does not tyrosinate [9]. We have previously confirmed that TTL overexpression under identical conditions reduces detyrosination below 25% of initial levels, while TTL-E331Q does not significantly affect detyrosination levels with similar overexpression [9]. To specifically quantify the effects of reducing detyrosination on the dynamic microtubule population, we fractionated free and polymerized tubulin as outlined above (Fig. 2a). Expression of TTL, but not E331Q, resulted in significantly less detyrosinated tubulin in the dynamic microtubule pool (Fig. 3a). Further, only TTL expression shifted tubulin away from the polymerized fraction towards the free tubulin fraction, resulting in an increased ratio of free:polymerized tubulin (Fig. 3a, S. Fig. 4a). This suggests that tyrosination affects the cycling of tubulin within the dynamic microtubule pool. If indeed tyrosinated microtubules are more dynamic, then levels of acetylation, a canonical marker of long-lived microtubules [40], should also be decreased by TTL. Consistent with this, TTL, but not E331Q, led to a robust reduction in levels of microtubule acetylation, suggesting that tyrosination reduces microtubule lifetime in the cardiomyocyte (Fig. 3b).

Fig. 3figure 3

TTL reduces microtubule stability through its tyrosinase activity. a Representative western blot (top) and quantification (bottom) of α-tubulin and detyrosinated (dTyr) tubulin in free and cold-sensitive dynamic microtubule fractions from adult rat cardiomyocytes treated with null, TTL, or TTL-E331Q adenoviruses; detyrosinated tubulin values are normalized to α-tubulin in cold-sensitive fraction (N = 4 rats, n = 8 WB technical lanes). b Representative western blot (top) and quantification (bottom) of α-tubulin and acetylated tubulin in whole-cell lysate from null, TTL, or E331Q expressing cardiomyocytes (N = 3 rats, n = 6 WB technical lanes). c Validation of HDAC6 and αTAT1 constructs and Tubastatin A (TubA) treatment. Representative western blot (top) and quantification (bottom) of a-tubulin and acetylated tubulin in whole-cell lysate from adult rat cardiomyocytes treated with null, HDAC6, or αTAT1 adenoviruses, or DMSO or 1 mM TubA treatment overnight (N = 3 rats, n = 6 WB technical lanes). d Representative western blot (top) and quantification (bottom) of α-tubulin and acetylated tubulin, in free and polymerized dynamic fractions. Lysates from cardiomyocytes were infected with null, HDAC6, or αTAT1 adenoviruses, or DMSO or 1 mM TubA overnight (N = 3 rats, n = 6 WB technical lanes). e Representative western blot (top) and quantification (bottom) of α-tubulin and detyrosinated tubulin in whole-cell lysate from adult rat cardiomyocytes treated with null, HDAC6, or αTAT1 adenoviruses, or DMSO or 1 mM TubA treatment overnight (N = 4 rats, n = 8 WB technical lanes). The bar represents mean ± 1SEM; statistical significance for (a) and (b) was determined with one-way ANOVA with post hoc test, and for (c) to (e) was determined with Two-sample Student’s T test

As acetylation itself is linked to microtubule stability [14, 42], the TTL-dependent change in the dynamic microtubule pool (Fig. 3b) could be directly related to tyrosination, or it could be a secondary effect due to the reduction in acetylation. To discriminate between these two hypotheses, we directly modulated acetylation. To this end, we developed adenoviral constructs encoding histone deacetylase 6 (HDAC6) and α tubulin acetyltransferase 1 (αTAT1). HDAC6 expression reduced total microtubule acetylation to 25% of initial levels (Fig. 3c) and αTAT1 expression increased acetylation 12-fold (Fig. 3c). Because αTAT1 has been shown to modulate microtubule dynamics independent of enzymatic activity [19], we also used a pharmacological inhibitor of HDAC6, Tubastatin A (TubA) to increase acetylation through an orthogonal approach (Fig. 3c). Having validated robust tools to modulate acetylation, we next determined the effect of acetylation on the dynamic microtubule pool utilizing the same fractionation assay. Neither increasing nor decreasing acetylation altered the free:polymerized tubulin ratio (Fig. 3d, S. Fig. 4b). Given that modulating tyrosination altered levels of acetylation (Fig. 3c), we also asked whether this relationship was reciprocal. However, whole-cell levels of detyrosination were largely unaffected by modulating acetylation (Fig. 3e), except for a modest increase with HDAC6 expression that may be related to HDAC6 association with microtubules increasing their stability and availability for detyrosination [2]. Together, these results suggest tyrosination directly alters cardiomyocyte microtubule stability, independent of corresponding changes in acetylation.

Tyrosination promotes catastrophe of growing microtubules

Next, to precisely quantify the effects of tyrosination on the dynamics of individual microtubules, we overexpressed either Null, TTL, or E331Q viruses in conjunction with EB3-GFP in adult rat cardiomyocytes. Although EB interaction is thought to be unaffected by microtubule detyrosination [27], we first wanted to validate that EB3 labeling of microtubules did not systematically differ with TTL expression. EB3 fluorescence intensity along the length and at the tip of the microtubule was unchanged in control, TTL, or E331Q expressing cells (S. Fig. 4c), indicating that EB3 expression or labeling of microtubules was not altered by our experimental interventions.

As seen in S. Movie 7, microtubules in TTL-expressing cells still initiated growth at the Z-disk, but often had shorter runs and underwent catastrophe before reaching a subsequent Z-disk. Consistently, TTL overexpression significantly increased the frequency of catastrophes, while reducing the frequency of pausing (Fig. 4a, b). E331Q expression did not alter event frequency compared to control cells (S. Movie 8), suggesting a tyrosination-specific effect on microtubule dynamics (S. Fig. 4d). Further examination of spatial dynamics revealed that the effect of TTL on microtubule breakdown was agnostic to subcellular location; TTL similarly increased the number of catastrophes both on and off the Z-disk. In contrast, TTL reduced the number of pauses specifically on the Z-disk (Fig. 4c). As a readout of inefficient growth, TTL increased the tortuosity of microtubule trajectories, defined as the ratio of growth distance to net growth (Fig. 4d). Combined, the lack of stabilization at the Z-disk and more frequent catastrophes resulted in tyrosinated microtubules depolymerizing ~ fivefold as often before successfully crossing a Z-disk when compared to either null or E331Q expressing cells (Fig. 4d). In sum, this data indicates that tyrosination increases the stochastic transition to microtubule breakdown irrespective of subcellular location and that tyrosinated microtubules inefficiently navigate successive sarcomeres with fewer stabilizing interactions at the Z-disk.

Fig. 4figure 4

Tyrosinated microtubules are more dynamic. a Representative kymographs from cardiomyocytes treated with EB3-GFP plus null, TTL, or E331Q adenoviruses. b Quantification of catastrophe and pause event frequencies and c event locations in cardiomyocytes treated with EB3-GFP plus null, TTL, or E331Q adenoviruses (N = cells, n = events). d Gross measurements of microtubule dynamics. (Left) Tortuosity, the distance a microtubule grows divided by its displacement, & (right) number of catastrophes in relation to the number of successful Z-disk crossing in cardiomyocytes treated with EB3-GFP plus null, TTL, or E331Q adenoviruses. e Z-disc bias score (log2 transformation of the ratio of events that occurred On vs. Off the Z-disk) for all experimental conditions. The bar represents mean ± 1SEM; statistical significance was determined with Kruskal–Wallis ANOVA with post hoc test

To summarize how our different interventions (tyrosination, desmin depletion) affected the spatial organization of microtubule behavior, we took the ratio of events that occurred on vs. off the Z-disk and performed a log2 transform, calculating a “Z-disk bias” for each type of dynamic event (Fig. 4e). Of note, this metric only reflects the spatial bias of events, not their frequencies. TTL reduced the preference for microtubule pausing at the Z-disk but did not affect the spatial preference of rescues, catastrophes, or initiations. Desmin depletion, on the other hand, virtually eliminated the typical Z-disk bias for pauses, rescues, or fewer catastrophes. Initiations had a strong Z-disk bias regardless of intervention, which likely reflects nucleating events from microtubule organizing centers at Golgi outposts proximal to the Z-disk that are not affected by these manipulations [26].

Tyrosination increases EB1 and CLIP170 association on microtubules

Next, we wanted to determine why tyrosinated microtubules exhibit increased catastrophe frequencies. Several pieces of evidence suggest that the tyrosinated or detyrosinated status of the microtubule alone is likely insufficient to alter microtubule dynamics [21, 41], but instead the PTM exerts its effect by governing the interaction of stabilizing/destabilizing MAPs with the microtubule [10, 28]. There are two prominent examples of tyrosination altering interactions with depolymerizing effector proteins in the literature. First, mitotic centromere-associated kinesin (MCAK/Kif2C) is a depolymerizing MAP that preferentially binds and depolymerizes tyrosinated microtubules [28]. Second, a recent in vitro reconstitution study indicates that tyrosination promotes the binding of CLIP170 on microtubule plus ends, which synergizes with EB1 to increase the frequency of catastrophes [10]. This mechanism has not been examined in cells. Due to its low abundance in the post-mitotic cardiomyocyte, our attempts to detect and knock down MCAK levels were unreliable; we thus hypothesized that tyrosination may promote the interaction of EB1 and CLIP170 on microtubules to promote their destabilization and catastrophe.

To test this hypothesis, we utilized a PLA to test whether EB1 and CLIP170 interactions on cardiac microtubules were guided by tyrosination. We first performed control assays to ensure the specificity of this PLA assay and ask whether EB1-CLIP170 interactions are observed on intact microtubules. No PLA puncta were observed when primary antibodies against EB1 or CLIP170 were excluded from the PLA assay (S. Fig. 5a). Further, the majority of EB1-CLIP170 interactions co-localized directly on super-resolved microtubules (Fig. 5a) indicating that interactions occur primarily on the polymerized microtubule. We next evaluated whether this interaction was sensitive to tyrosination. First, we ensured that global levels of EB1 or CLIP170 were not changing due to TTL or E331Q expression (Fig. 5b, c). We then quantified specific interactions of EB1-CLIP170 that were occurring on microtubules by thresholding the microtubule and PLA images, quantifying the fractional area covered by their overlap, and normalizing that area to the microtubule coverage in the same image plane (S. Fig. 5b). As shown in Fig. 5d, TTL increased the number of EB1-CLIP170 interactions per microtubule area by ~ fourfold relative to control or E331Q transduced cardiomyocytes (Fig. 5d), despite unchanging levels of EB1 or CLIP170. As this interaction has been demonstrated to be sufficient to robustly increase the catastrophe frequency of dynamic microtubules [10], we conclude that tyrosination destabilizes cardiac microtubules at least in part by promoting increased association with the destabilizing effector complex of EB1 and CLIP170.

Fig. 5figure 5

Tyrosination promotes EB1 and CLIP170 interactions on cardiomyocyte microtubules. a Representative AiryScan Joint Deconvoluted immunofluorescent images of EB1-CLIP170 PLA interactions in adult rat cardiomyocytes treated with null, TTL, or TTL-E331Q adenoviruses. b Representative western blot (top) and quantification (bottom) of EB1 in whole-cell lysate from adult rat cardiomyocytes treated with null, TTL, or E331Q adenoviruses for 48 h (N = 3 rat, n = 3 WB technical lanes). c Representative immunofluorescent images (left) and quantification (right) of CLIP170 in adult rat cardiomyocytes treated with null, TTL, or E331Q adenoviruses for 48 h (N = 3 rats, n = 10 cells per rat). d Quantification of EB1-CLIP170 PLA interactions in adult rat cardiomyocytes treated with null, TTL, or TTL-E331Q adenoviruses (N = 3 rats, n = 10 cells per rat). The bar represents mean ± 1SEM, and the middle line in the box graph represents mean ± 1SEM; statistical significance for (b) was determined with one-way ANOVA with post hoc test, and for (c) and (d) was determined with Kruskal–Wallis ANOVA with post hoc test

留言 (0)

沒有登入
gif