Oxidative stress pathogenically remodels the cardiac myocyte cytoskeleton via structural alterations to the microtubule lattice

Abstract

In the failing heart, the cardiac myocyte microtubule network is remodeled, which contributes to cellular contractile failure and patient death. However, the origins of this deleterious cytoskeletal reorganization are unknown. We now find that oxidative stress, a condition characteristic of heart failure, leads to cysteine oxidation of microtubules. Our electron and fluorescence microscopy experiments revealed regions of structural damage within the microtubule lattice that occurred at locations of oxidized tubulin. The incorporation of GTP-tubulin into these damaged, oxidized regions led to stabilized “hot spots” within the microtubule lattice, which suppressed the shortening of dynamic microtubules. Thus, oxidative stress may act inside of cardiac myocytes to facilitate a pathogenic shift from a sparse microtubule network into a dense, aligned network. Our results demonstrate how a disease condition characterized by oxidative stress can trigger a molecular oxidation event, which likely contributes to a toxic cellular-scale transformation of the cardiac myocyte microtubule network.

Graphical Abstract

eTOC

Goldblum et al. demonstrate that oxidative stress leads to cysteine oxidation of tubulin, which is associated with damage to the microtubule lattice. In the presence of free tubulin, this damage is repaired with GTP-Tubulin, thus suppressing microtubule depolymerization. Thus, oxidative stress may facilitate densification of the microtubule network in cardiomyocytes.

Introduction

Oxidative stress is a hallmark of many cardiac pathologies, including the prolonged ischemia and reperfusion associated with ischemic heart disease, cardiac hypertrophy, and heart failure (Hori and Nishida, 2009; Keith et al., 1998; Mcmurray et al., 1993; Slezak et al., 1995; Thompson and Hess, 1986). Global ischemia in the heart leads to a ~250% increase in reactive oxygen species (ROS) (Slezak et al., 1995), suggesting that oxidative stress is an important driving force leading to changes in the cellular cytoskeleton that accompany ventricular remodeling in ischemic heart disease and heart failure (Hein et al., 2000; Iwai et al., 1990; Koide et al., 2000; Stones et al., 2013; Tsutsui et al., 1993; Wang et al., 2008; Zile et al., 2001). Similarly, oxidative-stress exposed cells isolated from healthy tissue undergo alterations to the microtubule network (Hinshaw et al., 1993; Kratzer et al., 2012; Lee et al., 2005; Valen et al., 1999; Wei-Guo and Qi-Ping, 2014; Yao et al., 2011). These alterations in the microtubule network suggest that oxidative stress can directly modulate microtubule dynamics or stability (Drum et al., 2016).

Importantly, pathological changes in the microtubule network organization have been shown to increase cellular stiffness and lead to contractile dysfunction of cardiac myocytes (Caporizzo et al., 2018; Cooper, 2006; Howarth et al., 1999; Nishimura et al., 2006; Swiatlowska et al., 2020; Tagawa et al., 1996; Tsutsui et al., 1993, 1994; Webster, 2002). Furthermore, microtubule network reorganization hinders intracellular transport of proteins to their targets at the cell surface (Drum et al., 2016; Prins et al., 2016). Many of these proteins, such as the potassium channels, Kv4.2, Kv4.3, and Kv1.5, are involved in cardiac myocyte electrical signaling. Failure of these channels to reach the plasma membrane may lead to cardiac arrhythmias (irregular heartbeat) (Drum et al., 2016; Prins et al., 2016; Zadeh et al., 2009). While it has been shown that oxidative stress leads to altered microtubule networks, the nature and mechanism of oxidative stress-mediated remodeling of the microtubule cytoskeleton is poorly understood.

Microtubules are cylindrical polymers of α/β-tubulin heterodimers that form a complex network throughout the cytoplasm (Desai and Mitchison, 1997; Kirschner and Mitchison, 1986, 2002; Mitchison and Kirschner, 1984; Ohi and Zanic, 2016). Microtubule plus-end dynamics are biphasic, characterized by periods of slow microtubule growth (polymerization) followed by intervals of abrupt shortening (depolymerization). Microtubule length and function depends on the tight regulation of microtubule growth and shortening rates, as well as the frequencies of “catastrophe” events – the transition from growth to shortening, and “rescue” events – the transition from shortening to growth.

During microtubule growth, GTP-bound tubulin subunits are added to microtubule plus-ends. Once these GTP-tubulin subunits are buried within the microtubule lattice, they are hydrolyzed to GDP-tubulin (Carlier and Pantaloni, 1981; Carlier et al., 1984; Desai and Mitchison, 1997; Howard and Hyman, 2003; Mitchison and Kirschner, 1984; Weisenberg et al., 1976). This results in a GTP-tubulin ‘cap’ at the growing end of a predominantly GDP-tubulin microtubule filament. Importantly, GTP-tubulin is more stable within the microtubule lattice than GDP-tubulin, and thus the GTP-tubulin cap stabilizes microtubules and prevents rapid depolymerization. Recently, it has been shown that patches of GTP tubulin that are present along the length of the GDP-tubulin microtubule lattice may increase the likelihood of rescue events, thus promoting net microtubule elongation (Aher et al., 2020; Aumeier et al., 2016; Dimitrov et al., 2008; de Forges et al., 2016; Rai et al., 2019; Schaedel et al., 2015, 2019; Tropini et al., 2012; Vemu et al., 2018).

In this work, we found that oxidative stress contributes to a dramatic and maladaptive increase in microtubule density inside of cardiac myocytes. To dissect the mechanism behind this observation, we used cell-free reconstitution, mass spectrometry, and fluorescence and electron microscopy experiments. We found that microtubules subjected to oxidative stress undergo cysteine oxidation, and our electron and fluorescence microscopy experiments revealed that the locations of oxidized tubulin subunits within the microtubule lattice had structural damage, consisting of holes and lattice openings. For dynamic microtubules, the incorporation of stabilizing GTP-tubulin into these damaged lattice regions led to an increased frequency of rescue events (the transition from shortening to growth), and thus longer microtubules. Thus, oxidative stress led to a dramatic, pathogenic shift from a sparse microtubule network into a dense, longitudinally aligned microtubule network inside of cardiac myocytes, likely contributing to increased cellular stiffness and contractile dysfunction. Our results provide insight into myocardial changes in ischemic heart disease by describing a mechanism that can explain the dramatic remodeling of the microtubule cytoskeletal network within cardiac myocytes subjected to oxidative stress.

Results

Oxidative stress increases the density of the microtubule network in cardiac myocytes

Hydrogen peroxide (H₂O₂) is a reactive oxygen species that is upregulated in ischemic cardiac myocytes (Slezak et al., 1995). Thus, we explored H₂O₂ treatment conditions that could produce a moderate, physiological increase in cellular reactive oxygen species (ROS). We used stem-cell derived cardiac myocytes (H9c2 cells), and evaluated changes in cellular ROS with H₂O₂ treatment using dihydroethidium (DHE). DHE is a ROS indicator that, upon oxidation by cellular superoxide, intercalates into DNA within the cellular nucleus and becomes fluorescent (Devillard et al., 2008). Thus, H9c2 cells were loaded with DHE and incubated for 1 hour in the presence of H₂O₂ or media alone. Cells were then imaged using confocal microscopy (Fig. 1A, left), and the median nuclear DHE fluorescence, normalized to the background cytoplasmic DHE signal, was measured (see methods). We found that treatment with 100 μM H₂O₂ resulted in a ~40% increase in cellular ROS (Fig. 1A, right), while higher H₂O₂ concentrations led to cell blebbing and a substantial decrease in cell viability. Thus, we performed our cellular studies using 100 μM H₂O₂ treatment. This H₂O₂ treatment concentration is similar to other studies that evaluated the effects of oxidative stress inside of cells (Hinshaw et al., 1993; Kratzer et al., 2012; Lee et al., 2005; Livanos et al., 2014; Valen et al., 1999; Wei-Guo and Qi-Ping, 2014; Yao et al., 2011).

Figure 1: Oxidative stress leads to increased density of the microtubule cytoskeleton in cardiac myocytes.

(A) Left: DHE-loaded live H9c2 cells treated with 0 mM or 0.1 mM H₂O₂. Right: Median nuclear DHE fluorescence intensity normalized to median cytoplasmic background signal for each cell; p<<0.0001, t-test (see methods). Sample sizes in parenthesis indicate number of images analyzed, with multiple cells per image. Dashed indicates baseline value, as all values were normalized to the overall median value of the control data set. (B) Left: Live H9c2 cells that express GFP-tubulin treated with 0 mM or 0.1 mM H₂O₂. Right: Microtubule density as assessed by dividing total microtubule area by total cell area (see methods); p<<0.0001, t-test (see methods). Sample sizes in parenthesis indicate number of cells. (C) Left: Primary cardiac myocytes exposed to 0 mM or 0.1 mM H₂O₂ and immunostained for α-tubulin (green). Right: Microtubule density as assessed by dividing total microtubule area by total cell area (see methods); p<<0.00037, t-test (see methods). Sample sizes in parenthesis indicate number of cells. (D) Tip displacement of representative microtubules from live H9c2 cells that express GFP-tubulin, treated with 0mM or 1mM H₂O₂. (E) Average fraction of time microtubules spent in growth/pausing (left; p<0.00001), or shortening (right; p<0.00001). Sample sizes as shown in parenthesis represent number of microtubules analyzed. P-values determined from calculated Z-statistics for proportions. All box and whisker plots: Crosses: Mean; Bar: Median; Box: first-third quartiles.

We then examined the effect of oxidative stress on the microtubule network in cells. H9c2 cells were transformed to express Tubulin-GFP (CellLight^® BacMam 2.0 (ThermoFisher)), and live-cell confocal microscopy was performed. We found that, compared to untreated cells, the H₂O₂ treated cells appeared to exhibit a denser microtubule cytoskeleton, with bright microtubule bundles (Fig. 1B, left). To quantify this observation, we used MATLAB to measure cellular microtubule density, which we defined as the visible microtubule area within each cell divided by the total cell area (see methods). We found that the microtubule density was increased by ~67% in H9c2 cells after H₂O₂ treatment (Fig. 1B, right (p<<0.0001, t-test)).

To determine whether the phenotype that we observed using H9c2 cells would also translate to primary cardiac myocytes, we performed similar experiments using primary rat ventricular cardiac myocytes. Thus, isolated primary cardiac myocytes were treated for 1 hour in media alone, or with 100 μM H₂O₂.

To characterize the microtubule cytoskeletal network, cardiac myocytes were then fixed and stained with anti-α-tubulin antibody, and imaged with confocal microscopy. Control cardiac myocytes displayed a sparse, complex network of microtubules, similar to previous reports (Fig. 1C, left, top) (Robison et al., 2016). In contrast, cardiac myocytes treated with 100 μM H₂O₂ had a dramatic cytoskeletal reorganization: we observed bright microtubule bundles that were longitudinally aligned along the long axis of the cell (Fig. 1C, left, bottom). This result is similar to the pathological microtubule remodeling reported in failing cardiac myocytes (Tagawa et al., 1998; Tsutsui et al., 1993, 1994). We then quantified microtubule density in the primary cardiac myocytes and, similar to the H9c2 cells, we found that the microtubule density was increased by ~88% in H₂O₂-treated cells relative to untreated controls (Fig. 1C, right; p=0.00037, t-test).

Microtubule shortening is suppressed in cardiac myocytes with oxidative stress

The increase in microtubule density after H₂O₂-treatment in two different cell types suggests that oxidative stress leads to net microtubule polymerization, a pathological condition that could disrupt cardiac myocyte contractility. To directly examine whether oxidative stress favors microtubule growth, and/or the suppression of microtubule shortening, we used our Tubulin-GFP expressing H9c2 cells, and collected movies to observe growing and shortening microtubule ends (Fig. 1D). In untreated H9c2 cells, microtubules displayed slow elongation followed by extended periods of rapid shortening (Fig. 1D, top). However, in H₂O₂ treated cells, extended shortening events were less common, and long “pause” periods were observed, with rapid switching between growth and shortening events (Fig. 1D, bottom). We summed up the total time spent in a growth/pause state, as well as in a shortening state, for microtubules in the control and H₂O₂ treated cells. We found that there was a ~13% increase in the total fraction of time spent in a growth/pause state in the H₂O₂ treated cells as compared to the controls (Fig. 1E, right, p<0.00001, Z=−6.53), with a larger ~66% decrease in the total fraction of time that microtubules spent in a shortening state in the H₂O₂ treated cells (Fig. 1E, right, p<0.00001, Z=18.04). Thus, H₂O₂ treatment suppresses microtubule shortening in H9c2 cells, which could explain the increased microtubule density that was observed in both H9c2 and primary cardiac myocytes (Fig. 1B,C).

Designing cell-free experiments to replicate oxidative stress conditions in cells

To investigate the intrinsic mechanism for suppressed microtubule shortening under conditions of cellular oxidative stress, we turned to cell-free reconstitution experiments. Because it is likely that growing microtubule ends interact with cell edges or cardiac myocyte intermediate filaments to alter their dynamics (Robison et al., 2016; Smyth et al., 2010), we reasoned that cell-free reconstitution experiments would allow us to evaluate the direct effect of oxidative stress on microtubule dynamics, at single-microtubule, nanoscale resolution, and independently of cardiac myocyte health or the interactions of growing microtubule ends with cardiac myocyte sarcomeric structures. However, to directly compare the cell-free reconstitution experiments to the experiments in cells, we first explored conditions for a cell-free assay that would be comparable to the ~40% increase in ROS that was observed in the H9c2 cells with H₂O₂ treatment (Fig. 1A).

Typically, cell-free microscopy experiments are performed in a buffer that contains oxygen scavenging enzymes, to eliminate ROS generated from the imaging process itself. Thus, we first performed an experiment to determine the level of ROS generating from the fluorescence imaging process itself, in the absence of oxygen scavenging enzymes and without any H₂O₂ treatment. Using an Amplex Red ROS indicator, we found that, in the absence of oxygen scavenging enzymes, light-induced ROS from the imaging process itself generated a ~120% increase in ROS relative to experiments without oxygen scavenging enzymes (Fig. 2A), which was well above the ROS increase in cells, and which led to rapid microtubule disintegration (Fig. S1A). Therefore, we then used buffers with oxygen scavenging enzymes, and we titrated H₂O₂ into these buffers, allowing us to increase ROS more gradually. In each case, we measured the ROS increase using Amplex Red. We searched for an H₂O₂ concentration range that would replicate the experimentally measured cellular ROS increase (Fig. 2A) and found that, in the presence of oxygen scavenging enzymes, an H₂O₂ concentration range of 0–1 mM produced a 0–50% Increase in ROS (Fig. 2B; 0–30 μM residual ROS (Fig. S1B)), thus encompassing the cellular ROS increase of 40% (Fig. 1A).

Figure 2: Oxidative stress increases rescue frequency in dynamic microtubules.

(A) Representative time courses during fluorescence imaging of Amplex Red fluorescence intensity as a readout for residual ROS (see methods). All intensity values matched at 6 minutes imaging time to compensate for differing initial conditions. (B) Normalized Amplex Red intensity at 40 min. Amplex Red intensity measurements were normalized to the mean intensity of the control condition (blue). Green dashed line represents the fold increase in ROS for H₂O₂-treated H9c2 cells, shown in Fig. 1A. (C) Top: Schematic of TIRF experiments to measure microtubule length dynamics. Bottom: Representative kymographs of dynamic microtubules in the presence of 0 mM H₂O₂ (left), 0.5 mM H₂O₂ (middle), and 1 mM H₂O₂ (right) (red, GMPCPP-stabilized microtubule seeds; green, Alexa-Fluor-488 dynamic microtubule extensions; yellow arrows: rescue events). (D) Microtubule growth rate as a function of H₂O₂ concentration (n=280, 265, 149, 123 from 0 to 1 mM on graph). (E) Microtubule shortening rate as a function of H₂O₂ concentration (n=41, 39, 41, 39 from 0 to 1 mM on graph). (F) Microtubule catastrophe frequency as a function of H₂O₂ concentration (n=280, 265, 149, 123 from 0 to 1 mM on graph). (G) Microtubule rescue frequency, calculated as the frequency of a rescue event per catastrophe, and normalized to mean microtubule length at catastrophe, n=280, 265, 149, 123 catastrophe events from 0 to 1 mM on graph, absolute number of rescue events observed, from 0 to 1 mM on graph: n=20, 45, 26, 90. All error bars, in all panels, mean ± S.E.M.

Oxidative stress directly increases microtubule rescue frequency in cell-free experiments

Thus, we used our optimized H₂O₂ with imaging buffer assay to perform cell-free experiments with dynamic microtubules, allowing us to evaluate the direct effects of oxidative stress on microtubule growth and shortening behavior. Tubulin purified from swine brain was labeled with Alexa-488, and the green microtubules were imaged growing from stabilized, coverslip-attached rhodamine-labeled (red) microtubule “seeds” using Total Internal Reflection Fluorescence (TIRF) microscopy (Gell et al., 2010) (Fig. 2C, Supplementary Movies S1, S2; see methods).

The growth and shortening rates of dynamic microtubules were estimated, along with the frequency of switching between these two states. To determine whether oxidative stress would directly alter microtubule dynamics, these parameters were then quantified for microtubules that were exposed to increasing concentrations of H₂O₂ in the presence of imaging buffer (Fig. 2D–G). We found that microtubule growth and shortening rates were not significantly altered by increasing H₂O₂ concentrations (growth: Fig. 2D, p=0.43, R²=0.213; shortening: Fig. 2E, p=0.57, R²=0.18). Thus, the kinetics of microtubule assembly and disassembly were not altered.

We then measured the growth lifetime of each microtubule prior to shortening, and inverted these lifetime values to determine the “catastrophe frequency” (min⁻¹). We found that the catastrophe frequency of microtubules was unchanged with increasing H₂O₂ concentrations (Fig. 2F, p=0.89, R²=0.65), indicating that the average microtubule growth lifetimes were not altered.

Finally, we measured the frequency of “rescue” events, in which a post-catastrophe microtubule would stop shortening and begin to grow again, prior to complete depolymerization to the red stabilized “seed” (Fig. 2C, bottom, yellow arrows)). The rescue frequency per catastrophe was normalized to the maximum microtubule length prior to catastrophe, and then plotted as a function of increasing H₂O₂ concentration (Fig. 2G). Strikingly, we found that rescue frequency increased exponentially with increasing H₂O₂ concentration (Fig. 2G, Chi-squared goodness of fit test to exponential curve, p=1, χ² test statistic=0.005), with a ~7-fold increase in rescue frequency at 1 mM H₂O₂. An increase in rescue frequency in our cell-free assays is consistent with the suppression of microtubule shortening that was observed in the H9c2 cells under conditions of oxidative stress (Fig. 1E).

Interestingly, while oxidative stress led to a dramatically increased rescue frequency in our cell-free reconstitution assay, we found that H₂O₂ did not slow the hydrolysis rate of GTP-tubulin (Fig. S1C), or increase the stability of GTP-tubulin subunits in the lattice, as assessed via the microtubule nucleation rate (Fig. S1D). Thus, we next explored how H₂O₂ treatments could potentially modify tubulin itself to increase rescue frequency.

Mass Spectrometry: Oxidative stress leads to cysteine oxidation of polymerized microtubules

Tubulin dimers contain 20 cysteine residues and the oxidation of several has been previously reported (Clark et al., 2014; Landino et al., 2002, 2011). We therefore performed mass spectrometry experiments to determine whether H₂O₂ exposure oxidizes the tubulin that has polymerized into microtubules. Purified microtubules were grown in the presence or absence of H₂O₂, centrifuged, and the pellet containing polymerized microtubules was digested for mass spectrometry (Fig. 3A). We note that in order to resolve post-translational tubulin modifications, 1 mM H₂O₂ was added to the growing microtubules in the absence of oxygen scavenging buffers, thus leading to a ~40-fold higher ROS concentration than in the cell-free dynamic microtubule assays.

Figure 3: H₂O₂ exposure leads to tubulin cysteine oxidation.

(A) Schematic of mass spectrometry sample preparation for microtubules reconstituted in vitro. (B) Relative abundance of tubulin post-translational modifications measured via mass spectrometry as described in panel A. Parentheses indicate residues associated with each modification. Results for Trypsin digestion shown, results for Chymotrypsin digestion in Fig. S3A. Inset: number of oxidized Cysteine residues (C) Schematic of DCP-Rho1 western blotting sample preparation for microtubules reconstituted in vitro, to detect the concentration of oxidized cysteine residues. (D) Representative western blot for microtubules treated with increasing H₂O₂ concentrations. Anti-Rhodamine antibody was used to detect DCP-Rho1-bound tubulin. (E) Quantification of DCP-Rho1 blot intensities, normalized to total α-tubulin in each case. Circle, square, and asterisks: 3 independent Western blots. (F) Schematic of DCP-Rho1 western blotting sample preparation for H9c2 cells, to detect the concentration of oxidized cysteine residues. (G) Representative western blot for H9c2 cells treated with increasing H₂O₂ concentrations. Anti-Rhodamine antibody was used to detect DCP-Rho1-bound tubulin. The intensity of bands between 45–66 kD were summed for quantification in each lane. (H) Quantification of DCP-Rho1 blot intensities for H9c2 cells, normalized to total α-tubulin in each case. Circle, square, and asterisks: 3 independent Western blots.

Trypsin-digested and chymotrypsin-digested microtubule samples were analyzed in mass spectrometry scans for post-translational modifications of the polymerized tubulin, and their relative abundance was quantified using PEAKS software (Fig. 3B, Fig. S3). We found that H₂O₂-treated microtubules contained similar abundances for all identified post-translational modifications other than cysteine oxidation, which was present in the H₂O₂-treated microtubules, but absent in the controls (Fig. 3B, Fig. S3A). Further analysis of individual residues revealed that four tubulin cysteine residues were oxidized in the H₂O₂ treated, purified microtubule samples: cysteines 213, 295, and 305 of alpha-tubulin and cysteine 129 in beta-tubulin (Fig. 3B inset; Fig. S3B). Thus, our mass spectrometry data suggests that cysteine oxidation is the only substantial tubulin modification in H₂O₂-treated microtubules. We note that, because polymerized tubulin was exclusively analyzed in our mass spectrometry experiments, our analysis does not include modifications to tubulin that may have prevented its polymerization into microtubules.

Western blots reveal a linear increase in tubulin cysteine oxidation with increasing H₂O₂ concentrations in purified samples and in cells

To directly probe the stoichiometry between H₂O₂ treatment and cysteine oxidation, western blotting of purified microtubules was performed. Purified microtubules were grown in the presence of 0–1 mM H₂O₂ and DCP-Rho1, which selectively binds sulfenic acid, and is thus a positive indicator of cysteine oxidation (Poole et al., 2007) (Fig. 3C). Microtubules were grown and centrifuged, and western blotting was performed on the resulting pellet to detect for DCP-Rho1 and total tubulin (Fig. 3D). The anti-rhodamine (DCP-Rho1) band intensity was then normalized to the total tubulin band intensity for all H₂O₂ treatment conditions (Fig 3E). We found that tubulin cysteine oxidation increased linearly with H₂O₂ concentration (Fig. 3E, R²=0.94; p<<0.0001 for null hypothesis of a zero slope).

We then used western blotting to probe for tubulin cysteine oxidation in cells. H9c2 cells were treated with increasing concentrations of H₂O₂ for 1h, and then loaded with DCP-Rho1 for 15 minutes (method adapted from (Klomsiri et al., 2014)) (Fig. 3F). Western blotting was performed on the cell lysates to detect for DCP-Rho1 and total tubulin (Fig. 3G). The anti-rhodamine (DCP-Rho1) band intensity was then normalized to the total tubulin band intensity for all H₂O₂ treatment conditions. Similar to the purified, polymerized tubulin, we found that tubulin cysteine oxidation in cells increased linearly with H₂O₂ treatment concentration (Fig. 3H, R²=0.92; p<<0.0001 for null hypothesis of a zero slope).

Electron microscopy reveals that oxidative stress causes structural damage to the GDP-Tubulin microtubule lattice

Because H₂O₂ exposure oxidizes cysteine residues within tubulin subunits, and tubulin cysteine oxidation has been previously implicated in weakened tubulin-tubulin interactions (Clark et al., 2014; Landino et al., 2002, 2011) and damage (Guo et al., 2006), we then tested whether oxidative stress could alter microtubule structure, perhaps by altering the stability of GDP-Tubulin subunits within the microtubule lattice. Thus, we used negative-stain transmission electron microscopy (TEM) to examine the effect of H₂O₂ treatment on microtubule structure at nanoscale resolution. In these experiments, microtubules were grown and centrifuged, and then the pellet re-suspended in buffer without free tubulin, but containing 20 μM Taxol, in order to stabilize the microtubules and allow for high-resolution imaging in the absence of free tubulin. The Taxol-stabilized microtubules were then treated with 0 – 500 μM of H₂O₂ for 1 hour, before placing on a pre-warmed grid for imaging.

We note that, because no oxygen scavenging buffers were used during the H₂O₂ treatment, the ROS concentration is up to 20-fold higher in the experiments with Taxol-stabilized microtubules than for the dynamic microtubule experiments in Fig. 2. This is because we found that Taxol-stabilized microtubules were relatively resistant to H₂O₂ treatment, and thus required higher concentrations to explore its effect on microtubule structure. However, to ensure that our high resolution imaging results in Taxol would be consistent with lower H₂O₂ concentration results in dynamic microtubules, we tested the predictions from our Taxol experiments using dynamic microtubules at lower H₂O₂ concentrations (see below, Fig. 6).

Figure 6: Oxidative stress leads to GTP-tubulin repair islands in dynamic microtubules.

(A) Schematic of Mal3-GFP binding assay for dynamic microtubules in the presence of oxygen scavenging enzymes. Left: In a dynamic microtubule assay, Mal3-GFP typically binds to growing microtubule ends. Right: Mal3-GFP may bind to damaged areas on the lattice into which GTP-tubulin has incorporated. (B) Left: Representative TIRF images of Mal3-GFP binding assay, in which Mal3-GFP (green) binds the dynamic microtubule extension (red) polymerized from stabilized seeds (white). Mal3-GFP binds to growing microtubule tips, as expected (cyan arrows), but can also be observed within the microtubule lattice (white arrow), especially in the presence of H₂O₂ (bottom). Right: Quantification of Mal3-GFP coverage fraction on microtubules. Crosses: Mean; Bar: Median; Box: first-third quartiles; sample sizes represent number of images, many microtubules per image. p<<0.0001, calculated from single-factor ANOVA. Individual comparisons calculated via t-test: 0 and 0.25mM H₂O₂ (p<0.0001), 0 and 0.5mM H₂O₂ (p<0.0001), 0.25 and 0.5mM H₂O₂ (p<0.001) (C) Incorporation of GTP-tubulin (red) into the GDP-tubulin lattice (grey) may lead to stabilized “hot spot” areas within the GDP-tubulin lattice. (D) Left: kymograph with sequential rescue events, and measurement of Δr, which is the distance between two sequential rescue events along the length of the microtubule. Yellow arrows indicate rescue events. Double magenta arrow indicates photobleaching of stabilized microtubule lattice beyond the rescue “hot spot”. Right: Two sequential rescue locations are nearer to each other than if a random rescue location is selected on the second microtubule (p<0.0001, t-test). Crosses: Mean; Bar: Median; Box: first-third quartiles; sample sizes indicated number of sequential events analyzed. (E) Model for oxidative stress mediated remodeling of the cardiac myocyte microtubule cytoskeleton: the switch from a sparse microtubule network (left) into a dense network (right) suggests that lengthened dynamic microtubules (red) may be stabilized by interaction and binding of the growing microtubule plus-ends with intermediate filaments at the Z-disks (blue) (Robison et al., 2016).

We found that Taxol-stabilized microtubules treated with H₂O₂ appeared to develop small openings and defects within the microtubule lattice (Fig. 4A, red arrows). To quantify these small structural defects, we used our previously published semi-automated analysis code, which accounts for changes in microtubule width and curvature due to structural defects within the microtubule lattice (Fig. 4B, see methods) (Reid et al., 2017). This analysis produced a “Structure Metric” for each treatment condition (equation Fig. 4B). An increased Structure Metric reflects more frequent or severe structural defects within the microtubule lattice. We found that increasing concentrations of H₂O₂ led to a coordinate increase in the Structure Metric values, indicating that, consistent with our qualitative observations, oxidative stress causes damage along the microtubule lattice (Fig. 4B, right; p=0.00145, ANOVA). We were curious as to whether longer incubation times and higher H₂O₂ concentrations would lead to larger defects. Indeed, examination of Taxol-stabilized microtubules incubated overnight in 0 – 2 mM H₂O₂ had an increasing fraction of very long (>>100 nm) open microtubule sections (Fig. S2A,B). Thus, oxidative stress acts to disrupt GDP-tubulin stability within the microtubule lattice, leading to nanoscale structural defects within the microtubule itself.

Figure 4: Oxidative stress leads to structural damage in the GDP-tubulin microtubule lattice.

(A) Representative electron microscopy images of GDP-tubulin microtubules treated with 0 mM H₂O₂ (top) and 0.5 mM H₂O₂ (bottom) for 1h. Red arrows, defects and openings in the microtubule lattice. (B) Left: Method of microtubule structural alteration quantification, adapted from Reid et al. 2017. Width deviation (top) and curvature (middle) were calculated for microtubules, and a Structure Metric (S) was quantified (bottom equation). Right: Structure Metric quantified for 0, 0.25, and 0.5mM H₂O₂, all values normalized to the control grand mean. p=0.00145, calculated via single-factor ANOVA. Crosses: Mean; Bar: Median; Box: first-third quartiles; sample sizes represent frames analyzed, many microtubules per frame. Total length of analyzed microtubules from left to right on graph: 75 μm, 44 μm, and 52 μm. Individual comparisons calculated via t-test: 0 and 0.25mM H₂O₂ (p<0.0001), 0 and 0.5mM H₂O₂ (p<0.0001), 0.25 and 0.5mM H₂O₂ (p<0.0001).

Oxidative stress-induced structural defects can be repaired via the incorporation of GTP-tubulin subunits

Paradoxically, we found that while oxidative stress disrupts GDP-tubulin interactions within the microtubule lattice, leading to defects within the microtubule itself, oxidative stress also increases microtubule rescue frequency, leading to a suppression of microtubule shortening. How could openings and holes within the microtubule lead to a suppression of dynamic microtubule shortening? Interestingly, recent publications have shown that, in the presence of free tubulin, open defects within a microtubule can be “repaired” via the incorporation of GTP-tubulin subunits into the microtubule itself, both in cells and in purified systems (Aher et al., 2020; Aumeier et al., 2016; Dimitrov et al., 2008; de Forges et al., 2016; Schaedel et al., 2015, 2019; Tropini et al., 2012; Vemu et al., 2018). Further, it has been shown that the incorporation of GTP-tubulin at damaged areas within the GDP-tubulin lattice can stabilize the microtubule in these locations, leading to an increased frequency of rescue events at the “repaired” sites (Aher et al., 2020; Aumeier et al., 2016; Dimitrov et al., 2008; de Forges et al., 2016; Rai et al., 2019; Tropini et al., 2012; Vemu et al., 2018).

Thus, we tested whether oxidative stress could lead to the incorporation of GTP-tubulin into the GDP-tubulin microtubule lattice, due to openings within the microtubule itself. We used our previously published “repair assay” to visualize the incorporation of GTP-tubulin into the microtubule lattice (Reid et al., 2017). Specifically, rhodamine-labelled, Taxol-stabilized microtubules were subjected to increasing concentrations of H₂O₂, and then the microtubules were washed, treated with green-labelled GTP-tubulin, and imaged using TIRF microscopy.

We observed the growth of green extensions from the microtubule ends, as would be expected (Fig. 5B, left, white arrows). However, green puncta were also observed along the length of the red microtubule lattice, especially in microtubules subjected to H₂O₂ treatment (Fig. 5B, left, yellow arrows). To quantify this observation, we measured the fractional coverage area of green tubulin along the red microtubule lattice, excluding green extensions at the microtubule ends (Fig. 5B, right (see methods)) (Reid et al., 2017). We found that higher H₂O₂ concentrations led an increase in incorporation of green GTP-tubulin along the length of the microtubule (Fig. 5B, right, p<<0.0001, ANOVA), suggesting that GTP-tubulin had incorporated into damaged areas along the length of the microtubule lattice.

Figure 5: Oxidative-stress induced damage to the microtubule lattice is repaired via the incorporation of GTP-tubulin.

(A) Schematic of microtubule repair assay to measure the incorporation of GTP-tubulin (green) into damaged areas on Taxol-stabilized GDP-tubulin microtubules (red). (B) Left: TIRF microscopy images of GFP-Tubulin (green) and Taxol-stabilized microtubules (red). White arrows: microtubule growth via addition of GTP-tubulin to microtubule ends (excluded from analysis). Yellow arrows: incorporation of GTP-tubulin (green) into damaged areas on the GDP-tubulin lattice (red). Right: Quantification of the coverage area of GTP-tubulin (green) lattice incorporation divided by the total GDP-tubulin microtubule lattice area (red) (p<<0.0001, calculated from single-factor ANOVA). Green tubulin growth from microtubule ends is excluded from the analysis. Individual comparisons calculated via t-test: 0 and 0.25mM H₂O₂ (p<0.005), 0 and 0.5mM H₂O₂ (p<0.0001), 0.25 and 0.5mM H₂O₂ (p<0.0001). Crosses: Mean; Bar: Median; Box: first-third quartiles; sample sizes represent number of images, many microtubules per image. (C) Representative electron microscopy images of Taxol-stabilized microtubules treated with 0 mM H₂O₂ (top) and 0.5 mM H₂O₂ (bottom) followed by repair via GTP-tubulin. Structure Metric quantification shown in Fig. S2C. (D) Left: TIRF microscopy image of 0.5 mM H₂O₂-treated microtubule (blue), GTP-tubulin (green), and DCP-Rho1 (red). White arrows: incorporation of GTP tubulin (green) into microtubule lattice co-localizes with DCP-Rho1 binding (red) within the microtubule. Right: Cross-correlation function of GTP-tubulin repair incorporation and DCP-Rho1 (black), and cross-correlation function of DCP-Rho1 and the microtubule signal (grey). Cross-correlation function plotted as function of absolute value of lags, dashed lines represent 95% confidence intervals, n=83 microtubules. (E) Left: TIRF microscopy image of 0.5 mM H₂O₂-treated microtubule (blue), GTP-tubulin (green), and Mal3-mCherry (red). Yellow arrows: incorporation of GTP tubulin (green) into microtubule lattice co-localizes with Mal3-mCherry binding (red) within the microtubule. Right: Cross-correlation function of GTP-tubulin repair incorporation and Mal3-mCherry (black), and cross-correlation function of Mal3-mCherry and the microtubule signal (grey). Cross-correlation function plotted as function of absolute value of lags, dashed lines represent 95% confidence intervals, n=100 microtubules.

Finally, we collected TEM images of the repaired microtubules to ensure that the GTP-tubulin puncta did not represent non-specific aggregates along the microtubule lattice, rather than repair. We found that, after repair, TEM images of control and H₂O₂ treated microtubules appeared similar (Fig. 5C), with no significant increase in the Structure Metric for increasing H₂O₂ concentrations (p=0.969, ANOVA, Fig. S2C).

GTP-tubulin repair is co-localized with oxidized tubulin in the microtubule lattice

We found that H₂O₂ treatment led to tubulin cysteine oxidation (Fig. 3), and to damage along the microtubule lattice (Fig. 4). These damaged areas were likely repaired by GTP-tubulin in our repair assay (Fig. 5B). Thus, we asked whether tubulin cysteine oxidation was directly connected to the microtubule lattice damage, and thus to the GTP-tubulin repair signal. To test this idea, we repeated the repair experiment described above (Fig. 5A,B), but now included DCP-Rho1 in the assay (Fig. 5D, left). Line scans of the green GTP-tubulin signal and red DCP-Rho1 signals were generated for each microtubule, and the cross-correlation function was calculated using two normalized series (see methods and (Jaqaman et al., 2010a)). We found that there was a strong, statistically significant cross correlation between the green GTP-tubulin repair signal and the red DCP-Rho1 signal at 0 μm lags (Fig. 5E, right-black, 95% confidence intervals in dotted lines, n=83 microtubules). The DCP-Rho1 signal was not correlated to the microtubule-only signal, confirming that the observed cross-correlation is not due to non-specific shifts in the TIRF field intensity (Fig. 5E, right-grey). Thus, tubulin cysteine oxidation is strongly and specifically co-localized with areas of GTP-tubulin repair, suggesting that H₂O₂ treatment leads to tubulin cysteine oxidation in microtubules, which causes structural damage, allowing for GTP-tubulin repair.

GTP-tubulin repair is co-localized with the GTP-tubulin recognizing protein Mal3

The microtubule binding protein Mal3 and its human homologue EB1 binds with high affinity to GTP-tubulin, and with lower affinity to GDP-tubulin (Maurer et al., 2011; Roostalu et al., 2020; Zanic et al., 2009). In addition, EB1 directly targets microtubule structural defects (Reid et al., 2019), thus increasing the efficiency at which EB1 (or Mal3) may detect and stably bind to GTP-tubulin that is associated with openings and holes within the microtubule itself (Aher et al., 2020; Aumeier et al., 2016; Dimitrov et al., 2008; de Forges et al., 2016; Rai et al., 2019; Schaedel et al., 2015, 2019; Tropini et al., 2012; Vemu et al., 2018). Thus, we asked whether Mal3 was targeted to oxidative stress-mediated damage and GTP-tubulin repair within the microtubule lattice. To do this, we repeated the repair experiment described above (Fig. 5A,B) but included Mal3-mCherry in the assay (Fig. 5E, left). Line scans of the green GTP-tubulin signal and red Mal3-mCherry signals were generated for each microtubule, and a cross-correlation function was calculated using two normalized series (see methods and (Jaqaman et al., 2010a). We found that there was a strong cross correlation between the green GTP-tubulin repair signal and the red Mal3-mCherry signal (Fig. 5E, right-black, 95% confidence intervals in dashed lines, n=100 microtubules). Further, the Mal3-mCherry signal was not correlated to the microtubule-only signal, confirming that the observed cross-correlation was not due to non-specific shifts in the TIRF field intensity (Fig. 5D, right-grey). Thus, Mal3 localization provides a readout for GTP-tubulin repair along the length of the microtubule lattice.

Oxidative stress-induced structural defects are repaired via the incorporation of GTP-tubulin subunits in dynamic microtubules

To allow for high-resolution microscopy results, both the electron microscopy experiments (Fig. 4) and the GTP-tubulin repair experiments (Fig 5) were performed by treating Taxol-stabilized microtubules with high concentrations of H₂O₂, leading to a substantial increase in ROS relative to the dynamic microtubule assays. However, to verify that oxidative stress-mediated damage, and GTP-tubulin repair, occurs in short-lived dynamic microtubules, and in the presence of oxygen scavenging enzymes that limit the residual ROS concentration to a 0–50% increase over no treatment conditions (< 30 μM ROS; Fig. S1B), we repeated our cell-free dynamic microtubule assay with oxygen scavenging buffers (Fig. 2), but in the presence of Mal3. As demonstrated above (Fig. 5D), Mal3 localization provides a readout for GTP-tubulin repair along the length of the microtubule lattice (Fig. 6A).

As expected, Mal3-GFP was observed to track the tips of growing microtubules, where GTP-tubulin is concentrated (Fig. 6B, left, cyan arrows; Fig. S4). However, less frequently, we also observed Mal3-GFP bound along the length of the dynamic microtubules, especially in the presence of H₂O₂ (Fig. 6B, left, white arrow; Fig. S4). To quantify the relative amounts of Mal3-GFP binding along the lattice on dynamic microtubules, we measured the area of green Mal3-GFP occupancy along the length of each microtubule and divided this value by the total area of the red microtubules, for each image (coverage fraction) (Reid et al., 2017, 2019). We found that the mean coverage fraction of Mal3-GFP along the length of dynamic microtubules was significantly and substantially increased with increasing H₂O₂ concentration (~87% increase at 1 mM H₂O₂ and oxygen scavengers; Fig.6B, right, p<<0.0001, ANOVA). These results suggest that oxidative stress acts to damage the dynamic microtubule lattice on a rapid time scale (< 5 minutes), and at an increase in ROS concentration that is similar to our cellular experiments (0–50% ROS increase), thus allowing for swift GTP-tubulin repair at these sites.

Oxidative stress leads to rescue “hot spots” along the microtubule lattice

Lattice targeting of Mal3-GFP suggests that damaged areas along the dynamic microtubule length are repaired with GTP-tubulin (Fig. 6C, left). Because GTP-tubulin is more stable in the lattice than hydrolyzed GDP-tubulin, these repaired areas may become “hot spots” for rescue events along the microtubule lattice (Fig. 6C, right), as has been previously reported for Taxol incorporation into damaged microtubule lattice locations (Rai et al., 2019). Thus, we asked whether consecutive rescue events on individual microtubules occurred in similar locations, suggesting that repaired “hot spots” had developed along the length of the microtubule. Here, we measured the distance between two consecutive rescue events along the length of individual dynamic microtubules (Fig. 6D, left, Δr=distance between consecutive rescue events; yellow arrows=rescue event), and compared these distances to those in which we selected a random second rescue position. We found that the distance between two sequential rescue events was 89% smaller than if the second rescue occurred in a random location (Fig. 6D, right, p<0.0001, t-test; see methods).

The stabilization of dynamic microtubule sections beyond rescue “hot spots” was also apparent in examining the photobleaching pattern along the microtubule lengths: photobleaching was readily apparent in stabilized sections between the rescue “hot spots” and the red microtubule seed (Fig. 6D, left, magenta double arrow). Consistent with this observation, multiple rescue events frequently occurred at a given “hot spot” within the microtubule lattice (Fig. 6D, left and Figure 2C, bottom right), suggesting that GTP-tubulin at sites of repair could persist for several minutes prior to hydrolysis to GDP-tubulin. It may be that, because GTP-tubulin hydrolysis is likely catalyzed by the addition of longitudinal subunits, the GTP-tubulin “hot spots” persist until repair of a damaged lattice area is complete. Thus, we conclude that H₂O₂ treatment leads to rescue “hot spots” along the length of the microtubule lattice, which we attribute to damage and GTP-tubulin repair within the dynamic microtubule lattice.

Discussion

In this work, we found that reactive oxygen species, which are generated under conditions of oxidative stress, oxidize microtubules (Fig. 3). To our knowledge, this is the first demonstration of direct oxidation of polymerized microtubules, in the absence of glycolytic enzymes or other microtubule-associated proteins, via oxidative stress. Our results indicate that oxidative stress leads to the irreversible cysteine oxidation of tubulin subunits within a microtubule, both in cells and in cell-free reconstitution experiments. We found that four cysteine residues were oxidized in the H₂O₂ treated microtubules, and that cysteine oxidation was the only substantial post-translational tubulin modification.

Cysteine oxidation has been implicated in weakened tubulin-tubulin interactions, and thus likely increases the susceptibility of microtubules to damage (Clark et al., 2014; Landino et al., 2002, 2011; Luduena and Roach, 1991; Mellon and Rebhun, 1976). Consistent with this implication, we found that oxidized microtubules were characterized by regions of damage along the length of the microtubule, specifically, by holes and defects within the GDP-tubulin microtubule lattice (Fig 3). This oxidation and damage occurs on a rapid time scale (< 5 minutes) (Fig. 4), thus leading to significant alterations in the length regulation of dynamic microtubules (Fig. 2).

Previous studies using alternative oxidizing reagents have found that tubulin oxidation decreases bulk microtubule polymerization (Clark et al., 2014; Landino et al., 2002, 2011), or produces divergent results depending on the oxidizing agent (Livanos et al., 2014). We note that, in order to match cellular effects, our studies used low concentrations of ROS. Thus, the fraction of soluble and polymerized tubulin with oxidized Cysteine residues in our in vitro assays is likely low. Interesting future studies using recombinant tubulin with a higher concentration of oxidized Cysteine residues could shed light on whether the weakened affinity of oxidized tubulin subunits for the microtubule lattice could suppress microtubule growth and stability, in addition to causing damage along the microtubule length.

Paradoxically, we found that oxidative stress leads to the suppression of microtubule shortening, via an increase in the frequency of rescue events (Fig. 2). The suppression of microtubule shortening is consistent with our result that oxidative stress produces increased microtubule density in primary cardiac myocytes, as well as cardiac myocyte stem cells (Fig. 1). It has been previously demonstrated that rescue events can occur as a result of GTP-tubulin incorporation at damaged microtubule lattice sites (Aher et al., 2020; Aumeier et al., 2016; Dimitrov et al., 2008; de Forges et al., 2016; Rai et al., 2019; Schaedel et al., 2015, 2019; Tropini et al., 2012; Vemu et al., 2018). Here, sites of GTP-tubulin incorporation along the length of the microtubule may slow down and ultimately stop microtubule depolymerization, allowing the microtubule to begin growing again. Our results are consistent with the recent finding that low levels of microtubule severing enzymes are able to increase microtubule length and amplify the cytoskeleton by causing nanoscale damage along the microtubule length, leading to GTP-tubulin islands that stabilize the microtubules against depolymerization (Vemu et al., 2018).

Importantly, in cardiac myocytes, the lengthening of dynamic microtubules increases the likelihood of their direct interaction and binding to regularly spaced Z-disks (Fig. 6E: red: dynamic microtubules, blue: Z-disks) (Robison et al. 2016; Robison and Prosser 2017). It has been previously shown that the binding of microtubules to Z-disks inhibits microtubule depolymerization (Drum et al. 2016). Thus, the structure of the cardiac myocyte itself may exacerbate the effects of oxidative stress: lengthened dynamic microtubules are likely to be stabilized against depolymerization when they grow long enough to interact with the regularly spaced Z-disks (Fig. 6E). This model of dynamic microtubule lengthening, combined with the association of lengthened microtubules with the Z-disk, predicts that oxidative stress could generate a dense, static microtubule network in cardiac myocytes, likely increasing cellular stiffness and disrupting contractility (Fig. 6E) (Caporizzo et al., 2018; Koide et al., 2000; Robison and Prosser, 2017; Swiatlowska et al., 2020; Tsutsui et al., 1993). Consistent with this model, previous work by Drum et al. observed a reduction in dynamic microtubule tips, as evidenced by the absence of EB3-GFP microtubule growth and shortening events, after 30 min of H₂O₂ exposure (Drum et al., 2016). This loss of dynamic microtubule tips under conditions of oxidative stress is suggestive of a shift in the balance between static and dynamic microtubules, as the lengthened dynamic microtubules are stabilized over time via interactions with the Z-disk.

Microtubule dynamics have previously been measured in ventricular myocytes, using Eb3-GFP as a marker of growing microtubule plus-ends (Drum et al., 2016). Similar to our in vitro result, Drum and colleagues found that growth and shortening rates were relatively unaffected under conditions of oxidative stress. However, large increases in both catastrophe and rescue rates were observed. In our cell-free reconsitution studies, while we also observed a substantial increase in rescue frequency (Fig. 1G), we did not observe an increase in catastrophe rate (Fig. 1F). This difference suggests that cellular factors could perhaps sensitize growing microtuble plus-ends to catastrophe in the presence of oxidative stress. Specifically, while our single microtubule studies in cell-free experiments permit high resolution measurements of the direct effect of H₂O₂ on dynamic microtubules, we cannot exclude the possibility that microtubule-associated proteins and other factors inside of cells could have a confounding effect on microtubule dynamics, over and above the direct effect of H₂O₂ as was assessed in our cell-free assay. In our cellular studies of H9c2 cells, we observed long “pause” periods with very rapid switching between short growth and shortening events (Fig. 1D). Thus, we note that increases in the rate of catastrophe could also potentially be compensated by a coordinate increase in the probability of rescue after a catastrophe event, thus leading to a net lengthening and densification of microtubules in cells.

We report a substantial increase in cellular microtubule density under conditions of oxidative stress in both H9c2 cells, as well as in primary cardiac myocytes. We did not find a significant effect of oxidative stress on the rate of microtubule nucleation, and so we posit that the most important effect of oxidative stress in cardiac myocytes is on the lengthening of microtubules, rather than on a direct increase in microtubule number (Fig. 6). By increasing microtubule length and, therefore, increasing the density of the microtubule network, oxidative stress could act to increase cardiac myocyte stiffness and impair contractility. Importantly, elevated microtubule density has been previously shown to contribute to increased cellular stiffness and contractile dysfunction in the failing heart, while microtubule depolymerization restores contractility (Caporizzo et al., 2018; Koide et al., 2000; Swiatlowska et al., 2020; Tagawa et al., 1998; Tsutsui et al., 1994; Zile et al., 2001).

In future work, it will be interesting to study the downstream effects of oxidative stress on hypertrophy, sarcomere organization, and viability. Oxidative stress activates signaling pathways associated with a range of cardiopathologies, including Hypertrophic Cardiomyopathy (Aikawa et al., 2001; Sabri et al., 2003; Tu et al., 2002), impairs excitation-contractile coupling via oxidation of the contractile apparatus (Fearon et al., 1999; Kawakami and Okabe, 1998; Xu et al., 1997), and promotes cell death (Xie et al., 2014; Zhang et al., 2018; Zhao et al., 2019). Additionally, heart failure accompanies a decreased density of sarcomeric proteins in cells (Bollen et al., 2017; Chen et al., 2018; Witjas-Paalberends et al., 2014). Thus, important future work will examine the effects of H₂O₂-dependent microtubule remodeling on cardiomyocyte contractility, and on the trafficking of key components of the contractile apparatus to the sarcomere, where they are required for cellular contractility.

In summary, we find that oxidative stress contributes to a dramatic, pathogenic shift from a dynamic, versatile microtubule network into a dense and static microtubule network inside of cardiac myocytes (Fig. 6D), which likely contributes to increased cellular stiffness and contractile dysfunction in failing heart cells. This deleterious effect of oxidative stress in contractile cells may have important broader health impacts, such as in muscle dystrophies, ischemic heart disease, systolic and diastolic heart failure, and a whole host of cardiomyopathies.

Limitations of the Study

The goal of this work was to study mechanisms through which oxidative stress influences microtubule length dynamics, both in cells and in reconstituted microtubules. Ideally, the ROS concentrations in the cell-free reconstitution assays would be identical to those in the intracellular environment. Thus, the baseline absolute intracellular ROS concentration would be measured, as well as the intracellular ROS concentration with H₂O₂ treatment(s). Then, these concentrations would be exactly matched by titrating H₂O₂ into a microtubule dynamics reconstitution assay that utilized a non-fluorescent imaging technique such as Differential Interference Contrast (DIC) microscopy, which would therefore not require oxygen scavenging enzymes. However, there is currently no reliable means to measure the absolute ROS concentration in cells. Thus, in order to compare our cellular experiments to in vitro assays, we quantified the relative increase in intracellular ROS content after H₂O₂ treatment, and then matched this relative increase in our microtubule dynamics reconstitution experiments. We note that because the reconstituted microtubule dynamics experiments were performed using fluorescence microscopy in the presence of oxygen scavenging enzymes, the concentration of H₂O₂ that was added to the in vitro assays was significantly higher than the concentration added to the media in the cellular assays. However, the oxygen scavenging buffers that were included in the reconstitution assay reduced the absolute amount of residual ROS present during the experiment to ~0–30 μM – similar to or less than the 100 μM H₂O₂ that was added to the media in the cell experiments (Fig. S1B).

STAR Methods

RESOURCE AVAILABILITY

Lead Contact:

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Melissa Gardner (klei0091@umn.edu).

Materials Availability:

This study did not generate new unique reagents.

Data and Code Availability:

All data reported in this paper will be shared by the lead contact upon request. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

No new cell lines were generated for these experiments. Primary cardiomyocytes and H9c2 cell lines were incubated in media described under each experiment within Method Details.

METHOD DETAILS

Ventricular Cardiac myocyte Isolation and Primary Culture

Adult rat ventricular myocytes were isolated as previously described (Thompson et al., 2016). Briefly, adult female rats were anaesthetized (isoflurane) and injected with heparin (15,000 U/kg) and Fatal Plus (150 mg/kg). Enzymatic digestion of hearts was achieved via the Langendorff procedure. Ventricles were triturated and cells were plated on laminin-coated glass coverslips and cultured overnight at 37°C in M199 media (Sigma) supplemented with 25mM HEPES, 26.2mM sodium bicarbonate, 0.02% BSA, 50U/mL penicillin-streptomycin, 5ug/mL insulin, 5ug/mL transferrin, and 5ng/mL selenite and pH-adjusted to 7.4.

H₂O₂ exposure, immunostaining, and imaging of cardiac myocytes

Coverslip-adherent cardiac myocytes were treated with pre-warmed H₂O₂ (0–1mM) in M199 media for 1 h at 37°C, washed in pre-warmed PBS, and fixed by incubating in 1% paraformaldehyde in pre-warmed PBS for 1h. Coverslips were blocked in blocking buffer containing 3% BSA and 0.1% Triton X-100 in PBS for 30min at room temperature. For immunostaining of α-tubulin, coverslips were probed with mouse anti-α-tubulin antibody (DM1A, VWR; #PI62204; 1:1000 in blocking buffer) for 1h at room temperature, washed twice in blocking buffer for 10min each, and treated with FITC anti-mouse IgG antibody (F0257, Sigma; 1:1000 in blocking buffer) overnight at 4°C. Coverslips were again washed twice in blocking buffer for 10 min and mounted on slides with ProLong Diamond Antifade (Thermo Fisher #P36965) for imaging. Immunostained cardiac myocytes were imaged with a laser scanning confocal microscopy (Nikon Ti2, 488nm laser line) fitted with a 100x oil objective (Nikon N2 Apochromat TIRF 100X Oil, 1.49 NA), which allowed for a 0.2 μm pixel size. The microtubule density was estimated using a custom Matlab code to first detect microtubule area, and then detect the entire cell area. The microtubule area was divided by the total cell area to compute a microtubule density. For each cell, a single slice was selected for analysis just below the nucleus.

Detection of ROS in Live H9c2 Cells

Live H9c2 cells were loaded with Dihydroethidium (DHE) as previously described (Carter et al., 1994). H9c2 cells were passaged and grown in DMEM media with 10% fetal bovine serum for 3d (37°C, 5% CO₂). Cells were incubated in 10uM DHE in DMEM media with 10% fetal bovine serum for 1h (37°C, 5% CO₂) followed by two washes in warm PBS. Cells were then incubated in 100 μM H₂O₂ or media alone for 1h, then imaged with laser scanning confocal microscopy. Imaging was performed with a 0.2μm pixel size and images were acquired with 0.2μm along the z-axis. Maximum projection images were created for each z-stack using ImageJ software. Images were analyzed using automated object detection code in Matlab, in which filtering parameters were adjusted to first detect the nuclei, after which the median nuclear DHE fluorescence was recorded. Then, the entire cell except the nucleus was detected, and the median cytoplasmic DHE fluorescence was recorded. Finally the median DHE fluorescence in the nucleus was normalized to the median cytoplasmic DHE fluorescence for each cell.

Detection of Dynamic Microtubules in Live H9c2 Cells

Cells were passaged, counted, and diluted in DMEM media with 10% fetal bovine serum/pen/strep such that 1000 cells in 200uL suspension were adhered to glass-bottom dishes. 1 μL of CellLight Tubulin-GFP (C10613, Invitrogen) was added to each dish, and cells were transfected overnight (37°C, 5% CO₂) before 500uL of DMEM media with 10% fetal bovine serum/pen/strep was added to each dish. After 5d of treatment in CellLight Tubulin-GFP, media was replaced with CO₂-independent media (18045088, ThermoFisher), with or without H₂O₂, and incubated for 1h (37°C, 5% CO₂) before confocal imaging. H9c2 Cells were imaged with a laser scanning confocal microscopy (Nikon Ti2, 488nm laser line) fitted with a 100x oil objective (Nikon N2 Apochromat TIRF 100X Oil, 1.49 NA), which allowed for a 0.2μm pixel size, and acquisition rate of 0.1fps.

A custom Matlab script was used to manually track the position of individual microtubule tips in each frame by clicking on individual microtubule tips. The total tip displacement from the first frame was recorded over time for each microtubule. For each microtubule, periods of elongation, shortening, and pausing were identified by examination of the traces: a shortening event required three consecutive points in which the microtubule length was decreased, and, similarly, a growing event required three consecutive points in which the microtubule length was increasing. Length changes of a duration shorter than three consecutive time points were considered to be pauses, rather than lengthening or shortening events.

Tubulin Purification and Labeling

Tubulin was purified from pig brain extract through repeated cycles of polymerization-depolymerization and labeled with rhodamine, Alexa-488, or Alexa-647 as previously described (Castoldi and Popov, 2003; Gell et al., 2010).

Construction and preparation of flow chambers for TIRF microscopy imaging

Flow chambers were assembled for TIRF microscopy as described (Gell et al., 2010) with the following adaptation: To create a ‘lane’ for the unidirectional flow of samples, two narrow strips of Parafilm were arranged parallel to each other in between two hydrophobic silanized coverslips. Chambers were subjected to heat in order to melt the Parafilm strips and create a seal between the coverslips. Before use, the chamber was treated with rabbit anti-rhodamine antibody diluted 1:50 (A6397, Thermo Fisher), or a mixture of streptavidin (20μg/mL) and neutravidin (25μg/mL), in Brb80 for 30 min. followed by blocking with pluronic F127 for at least 20 min.

Construction of stabilized microtubule seeds

Stable GMPCPP microtubule seeds were prepared from a mixture of 3.9 μM tubulin (25% rhodamine-labeled, 75% unlabeled), 1mM GMPCPP, and 1.2 mM MgCl₂ in Brb80 and incubated first on ice for 5 min, followed by 2 h at 37°C (Reid et al., 2017). Following incubation, GMPCPP microtubules were diluted in 400 μL warm Brb80, spun via air-driven centrifuge (Airfuge, Beckman Coulter, 20psi, 5min), and re-suspended in 400 μL of warm 10 μM Taxol in Brb80. The GMPCPP microtubules were stored at 37°C and used for microtubule dynamics experiments up to 5 days after preparation.

Detection of Reactive Oxygen Species and TIRF Microscopy

Rhodamine-labeled GMPCPP microtubule ‘seeds’ were adhered to an anti-rhodamine antibody-coated flow chamber (Gell et al., 2010). For experiments using ROS scavengers, chambers were washed with 80 μL pre-warmed Imaging Buffer (20μg/mL glucose oxidase, 10μg/mL catalase, 20mM D-Glucose, 10mM DTT, 80μg/mL casein, and 1% tween-20). A reaction mixture containing 0.15% methyl cellulose, 50μM Amplex Red, 1U/mL Horseradish Peroxidase, H₂O₂ (0–1mM), and Imaging Buffer was prepared. For experiments performed in the absence of oxygen scavenging enzymes, glucose oxidase and catalase was omitted from the Imaging buffer. The reaction mixture was then introduced into the imaging chambers and movies were acquired at 28 °C for 1h at 0.2fps using a TIRF microscope (Nikon Eclipse Ti TIRF) fitted with a 100x oil objective (Nikon CFI Apochromat TIRF 100XC Oil, 1.49 NA) and CCD camera (Andor, iXon3), allowing for a 160nm pixel size. Images acquired using the 561nm laser line were analyzed using a custom MATLAB script, and the median Amplex Red intensity across each frame was quantified.

Dynamic Microtubule Assay and TIRF Microscopy

Rhodamine-labeled GMPCPP microtubule ‘seeds’ were adhered to an anti-rhodamine antibody-coated flow chamber (Gell et al., 2010) and washed with 80 μL pre-warmed Imaging Buffer (20μg/mL glucose oxidase, 10μg/mL catalase, 20mM D-Glucose, 10mM DTT, 80μg/mL casein, and 1% tween-20) to minimize photobleaching. A reaction mixture containing 10.3μM tubulin (15.8% Alexa488-labeled, 84.2% unlabeled), 55mM KCl, 1 mM GTP, 0.15% methyl cellulose and Imaging Buffer was prepared. For oxidative stress experiments, H₂O₂ (0–1000 μM) was included in the reaction mixture. The reaction mixture was centrifuged for 5 min at 4°C to remove protein aggregates and the supernatant was introduced into the imaging chamber.

Movies of dynamic microtubules were acquired at 28°C for 1h at 0.2fps using 488nm and 561nm laser lines with a TIRF microscope (Nikon Eclipse Ti TIRF) fitted with an 100x oil objective (Nikon CFI Apochromat TIRF 100XC Oil, 1.49 NA) and CCD camera (Andor, iXon3). This TIRF microscopy imaging system allowed for a 160nm pixel size. With this experimental protocol, multiple parameters of length dynamics (elongation rate, shortening rate, rescue frequency, and catastrophe frequency) were quantified using ImageJ (Coombes et al., 2013; Gardner et al., 2011b, 2011a; Reid et al., 2016).

Dynamic Microtubule Assay Image Analysis

Kymographs for each dynamic microtubule extension were generated from movies using ImageJ. For each microtubule, the growth rates, shortening rates, catastrophe frequency, and rescue ratio was quantified. For the growth rate analysis, only microtubules in which the entire lifetime of microtubule elongation was observed were analyzed. Growth rate was calculated as the microtubule length at catastrophe divided by the duration of microtubule growth, from growth initiation from the seed until catastrophe. For the catastrophe frequency analysis, the catastrophe frequency was calculated as the inverse of the duration of microtubule growth. For the shortening rate analysis, only complete shortening events, defined as the period following catastrophe until either complete depolymerization or rescue, were analyzed. Shortening rate was calculated as the length a microtubule shortened after catastrophe divided by the shortening time. For the rescue frequency analysis, because microtubule rescue only occurs after a catastrophe event, the probability of rescue is first dependent on a catastrophe event. Therefore, the rescue frequency was calculated as number of rescue events in a movie, divided by the number of catastrophe events, and then normalized to the mean microtubule length at catastrophe, to account for the number of potential rescue locations per unit length.

Mass Spectrometry for the Detection of Post-Translational Modifications in Reconstituted Microtubules

Sample Preparation:

Samples were reconstituted in 50 μl denaturing buffer (7 M urea, 2 M thiourea, 0.5 triethlyammonium bicarbonate pH8.5, 20% acetonitrile). A Bradford assay was performed to determine protein concentration. Two 3.5 μg aliquots for each sample were made for proteolytic digestion. All the samples were brought to the same volume with denaturing buffer and incubated at room temperature for 15 min. The samples were then diluted 4x with water. Trypsin and chymotrypsin were reconstituted with 10mM calcium chloride in water. Trypsin was added to the first aliquot and chymotrypsin was added to the second aliquot. Both proteolytic enzymes were added in a 1:40 enzyme to total protein ratio. Samples were incubated 16 h at 37°C. After incubation, samples were frozen at −80°C and dried in a Speedvac. Once dried, samples were reconstituted in 98/2 water/acetonitrile with 0.1% formic acid. The chymotrypsin digested aliquots were cleaned up with a C18 STAGE tips (Rappsilber et al., 2003), and the trypsin aliquots were cleaned up with a MCX-like STAGE tips following the same protocol but using Empore SDB-RPS extraction disks material from 3M. The eluted samples were Speedvaced to dryness.

Orbitrap Fusion Liquid Chromatography-Mass Spectometry:

We reconstituted the dried peptide mixtures in 97.9:2:0.1, H2O:acetonitrile (ACN):formic acid (FA) and analyzed ~0.125 microgram of each sample by capillary LC-MS. We used a Thermo Fisher Scientific (Waltham, MA) Easy NanoLC 1100 system with direct column load in 97.9:2:0.1, H2O:ACN:FA onto a 100 um inner diameter 55 cm self-packed LC column with 1.9 um 120 Å C18-aq Dr. Maisch GmbH ReproSil-PUR resin. We performed gradient elution of peptides at 325 nl/min using a column heater (Model PRSO-V2-ES72, Sonation, Biberach an der Riss, Germany) set to 55 °C with 5 – 22% solvent B (0.1% FA in ACN) in 45 minutes, 22 – 35% B from 45 – 70 min, and 35 – 90% B from 70 – 80 min, where solvent A was 97.9:2:0.1, H2O:ACN:FA. We acquired data on an Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific, Inc., Waltham, MA) in data dependent mode with the following parameters: ESI voltage 2.1kV, ion transfer tube 275 °C; Orbitrap MS1 scan 120k resolution in profile from 360 – 1580 m/z with 50 msec injection time, 125% (5 × 10⁵) normalized automatic gain control (AGC); MS2 triggered on the top 15 most abundant ions above 2 × 10⁴ counts with ion trap collision induced dissociation (CID) activation with 35% collision energy, 10 msec activation time and 0.25 activation Q, 1.6 Da quadrupole isolation window, ion trap detection with 35 msec injection time, 1 × 10⁴ AGC, dynamic exclusion duration 15 sec with +/− 10 ppm mass tolerance.

Mass Spectral Database Search:

We used Peaks^® Studio v10 (Ma et al., 2003) (Bioinformatics Solutions, Inc, Waterloo, ON CA) for interpretation of tandem MS (mass spectra) and protein inference. Search parameters were: porcine (taxon ID 9823) universal proteome sequence database (UP000314985) from UniProt.org (Sept 18, 2019) merged with common lab contaminant proteins from http://www.thegpm.org/crap/; parent mass error tolerance 20.0 ppm; fragment mass error tolerance 0.6 Da; precursor mass search type monoisotopic; enzyme trypsin or chymotrypsin (sample dependent) with maximum 2 missed cleavage sites and semi-specific digest mode; variable modifications: methionine oxidation, pyroglutamic acid, deamidation of asparagine and glutamine, protein N-terminal acetylation, carbamidomethyl (CAM) cysteine, dihydroxylation of YWFRKPC, cysteine oxidation to cysteic acid, oxidation of CKNPFYRHWGEILQSTV residues; maximum variable modifications per peptide 3; false discovery rate calculation On; spectra merge option OFF, charge correction on for charge states 2 – 8; spectral filter quality >0.65; the de novo sequencing parameters were: parent mass tolerance 20 ppm, fragment mass tolerance 0.6 Da, variable modification oxidized methionine. The peptide and protein lists were filtered at 1% false discovery rate.

Identification of Post-Translationally Modified Residues from Mass Spectrometry Data:

To find the tubulin residues with post-translational modifications (PTM’s), a custom-built python platform was used, which analyzed each identified peptide and queried the protein list to associate each peptide with the corresponding tubulin protein. If a peptide was found to belong to multiple proteins, the peptide was classified as being found in all proteins. Next, the mass shift in the peptide was used to identify a PTM that occurred and the specific residue location for the corresponding tubulin protein was noted. This process was repeated for all peptides and all PTMs separately for both the trypsin and chymotrypsin-digested samples. Next, a custom-built R script was used to determine each residue in which PTMs were found for each tubulin protein, for both the control and H₂O₂-treated microtubule samples. A PTM was determined to be unique to the control microtubule sample if at least one of the peptides from the control sample contained the PTM, and no peptides from the H₂O₂-treated microtubule sample contained the PTM. The opposite was true for the H₂O₂-treated microtubule samples. Finally, we classified a PTM as shared if there was at least one peptide with the PTM found in one of the protein digests (trypsin or chymotrypsin) for both the control and H₂O₂ treated samples.

Western Blotting to Detect Tubulin Cysteine Oxidation in Reconstituted Microtubules

A mixture of 36.5 μM unlabeled tubulin, 1mM GTP, and 50μM DCP-Rho1 in Brb80 was prepared on ice and incubated for 2.5h at 37°C in the presence or absence of H₂O₂ (0–1.5mM). The mixture was diluted 30-fold with warm 10μM Taxol in Brb80. 240μL of the microtubule sample were spun via air-driven centrifuge (20psi, 5min), and the pellet was re-suspended in 100μL of warm 10μM Taxol in Brb80 and stored at −80°C. Reducing electrophoresis buffer was added and heated to denature proteins, and immunoblotting was performed. Samples were resolved in 10% SDS-PAGE and transferred to PVDF membranes. Membranes were blocked in 1% BSA in Tris-buffered saline with 0.15% Tween. For the detection of DCP-Rho1, lysates were probed with primary antibody against rhodamine (A6397, Fisher Scientific) followed by HRP-conjugated anti-rabbit secondary antibody (sc-2004, Santa Cruz Biotechnology). For detection of tubulin, membranes were probed with primary antibody against alpha-tubulin (clone DM1A, #PI34095, VWR) followed by HRP-conjugated anti-mouse secondary antibody (SC-205, Santa Cruz BioTechnology). Western blots were developed using chemiluminescence (PI34095, Fisher) and densitometry analysis was performed using ImageJ software.

Western Blotting to Detect Tubulin Cysteine Oxidation in H9c2 Cells

H9c2 cells were grown in DMEM+ media for 3d, then passaged into 6-well dishes. Cells were incubated in 0–1mM H₂O₂ for 1h. During the final 10 min of the incubation in H₂O₂, 10 μM DCP-Rho1 in DMSO was added. Cells were released and lysed in hot, reducing electrophoresis buffer for 6min. Samples were resolved in 10% SDS-PAGE and transferred to PVDF membranes. Membranes were then blocked in 1% BSA in Tris-buffered saline with 0.15% Tween. For detection of DCP-Rho1, lysates were probed with primary antibody against rhodamine (A6397, Fisher Scientific) followed by HRP-conjugated anti-rabbit secondary antibody (sc-2004, Santa Cruz Biotechnology). For detection of tubulin, membranes were probed with primary antibody against alpha-tubulin (clone DM1A, #PI34095, VWR) followed by HRP-conjugated anti-mouse secondary antibody (SC-205, Santa Cruz BioTechnology). Western blots were developed using chemiluminescence (PI34095, Fisher) and densitometry analysis was performed using ImageJ software.

Microtubule Turbidity Nucleation Assay

Bulk tubulin polymerization was assessed by measuring sample turbidity, as the absorbance at 350 nm (Nanophotometer NP80, Implen), as previously described (Portran et al., 2017). Briefly, the spectrophotometer was pre-warmed in a 30°C room for 2h before use, and blanked on dH₂O. A mixture of 42μM tubulin, 1mM GTP, 5mM DTT, and 5% glycerol was prepared in Brb80 on ice. For H₂O₂ experiments, this mixture additionally contained 0.5mM H₂O₂. The tubulin mixture was added to a pre-warmed cuvette and absorbance measurements were acquired every 30sec for 30min.

GTP Hydrolysis assay

A mixture of 10 μM tubulin, 0.5mM GTP, 0.5mM DTT, and 5% glycerol in Brb80 was prepared on ice. For H₂O₂ experiments, this mixture also included 0.5 mM H₂O₂. Reactions were incubated either at 0°C or 37°C for 2.5h, after which reactions were stopped by the addition of 5% TCA on ice for 2min. Samples were centrifuged for 10min at 4°C to remove protein aggregates. 15μL of the supernatant was added to 35μL Cytophos reagent (BK054, Cytoskeleton), vortexed gently, and incubated for 10min at 22°C. To measure the amount of inorganic phosphate, the absorbance at 650nm (Nanophotometer NP80, Implen) was recorded three times per sample. A phosphate standard curve was determined using the Cytophos phosphate standard (BK054, Cytoskeleton), diluted 0–100 μM.

Transmission Electron Microscopy (TEM)

A mixture of 66μM tubulin, 1mM GTP, and 4.5mM MgCl₂ was prepared in Brb80 and placed on ice for 5min, followed by a 2h incubation at 37°C. Reconstituted microtubules were diluted 20x in warm 20μM Taxol in Brb80. Taxol-stabilized microtubules were diluted 3x in warm Taxol (10μM)/Brb80 and stored overnight at 37°C. Microtubules were then spun via air-driven centrifuge (Airfuge, Beckman Coulter, 20psi, 5min), and resuspended in 50 μL of warm 10 μM Taxol in Brb80. Microtubules were then diluted 2-fold in a warm Brb80 solution containing 10 μM Taxol and H₂O₂ (0–1mM), and incubated for 1h at 37°C. 10μL of reconstituted microtubules was placed on a pre-warmed 300-mesh carbon coated copper grid for 1min, followed by 4 drops of 1% uranyl acetate for 1min. Filter paper was then used to wick away the excess stain from the grid. The grid was left to dry for 10min and stored. Microtubule specimens were imaged using TEM (FEI Technai Spirit BioTWIN) at 6500x magnification.

To quantify structural defects within the microtubule lattice, we used our previously published Structure Metric, which is a measure of both microtubule width and curvature (Reid et al., 2017). In brief, microtubules were identified using the previously published semi-automated image analysis platform. First, for each segment within an individual microtubule, width (W) was measured and the Width Deviation was calculated as |W − W_exp| and summed across the entire microtubule (Fig. 4D, left top), where W_exp is the expected width of individual microtubules. The absolute curvature of the microtubule midline was also measured for each microtubule segment and summed for each individual microtubule (C_total). Finally, the Structure Metric (S) for each individual microtubule was calculated as the sum of the width deviation and curvature, normalized to account for the difference in scale between the width and curvature metrics, as previously described.

Mal3-GFP Purification

The pETMM11-HIS6x-Mal3-GFP plasmid with a TEV cut site after the His6x tag was a kind gift from Dr. Thomas Surrey. The plasmid was transformed into Rosetta (DE3) pLysS E. coli and grown in 800mL of LB+kan+cam at 37°C to an OD of approximately 0.4. To induce protein expression, IPTG was added to 0.2mM and the culture was mixed at 14°C for 16hr. Cells were centrifuged (30min., 4°C, 4400×g) and resuspended in 25mL lysis buffer (50mM Tris pH7.5, 200mM NaCl, 5% glycerol, 20mM imidazole, 5mM β-mercaptoethanol, 0.2% triton X-100), protease inhibitors (1mM PMSF, 10μM Pepstatin A, 10μM E-64, 0.3μM aprotinin), and DNAse I (1U/mL). The cell suspension was sonicated on ice (90% power, 50% duty, 6×1min). Cell lysates were centrifuged (1h, 4°C, 14000 ×g) and the soluble fraction was passed through 1mL of Talon Metal Affinity Resin (Clontech #635509). The resin was washed for four times with 4mL Wash Buffer (50mM Tris pH7.5, 500mM NaCl, 5% glycerol, 20mM imidazole, 5mM β-mercaptoethanol, 0.1mM PMSF, 1μM Pepstatin A, 1μM E-64, 30nM aprotinin) for 5min each. Protein was eluted from the resin by mixing with 1mL of Elution Buffer (50mM Tris pH 7.5, 200mM NaCl, 250mM imidazole, 0.1mM PMSF, 1μM Pepstatin A, 1μM E-64, 30nM aprotinin) for 15min followed by slow centrifugation through a fritted column to retrieve eluate. To cleave the HIS6x tag, 10 units of GST-tagged TEV enzyme (TurboTEV, #T0102M, Accelagen) and 14mM β-mercaptoethanol were added and the eluate was dialysed into Brb80 overnight at 4°C. To remove the TEV enzyme, the dialysate was mixed with 100ul of glutathione-sepharose (GE Healthcare #17-0756-01) for 30 min. at 4°C and spun (1min, 2000×g). The Mal3-GFP protein was quantified by band intensity on a coomassie-stained SDS PAGE protein gel.

Microtubule Repair Assay

GDP microtubules were prepared from a mixture of 36.4μM tubulin (16% Alexa647-labeled, 21% Biotinylated, 63% unlabeled), 4.5mM MgCl2, and 1mM GTP in Brb80. The mixture was incubated for 2h at 37°C, diluted 4x in warm 10μM Taxol in Brb80, and stored overnight at 37°C. 220μL of additional warm 10μM Taxol in Brb80 was added and the Taxol-stabilized microtubules were then spun via air-driven ultracentrifuge (20psi, 5min), resuspended in 50μL of warm 10μM Taxol in Brb80. To visualize the incorporation of soluble tubulin at ‘defects,’ or gaps, within the microtubule lattice, a reporter tubulin was used (Reid et al., 2017). Briefly, 15μL of the Taxol-stabilized microtubules were added to 30μL of 2x Tubulin Reporter Solution (424nM Mal3-mCherry, 1.5μM tubulin (50% Alexa488-labelled), 0.1mM GTP, 1mM MgCl₂, and 100μM taxol in Brb80), and 15μL H₂O₂ (1–2mM) in Brb80/taxol, resulting in a total H₂O₂ concentration of 0–500μM. The microtubule mixture was then incubated for 1h at 37°C. For experiments containing DCP-Rho1, Mal3-mCherry was replaced with 6μM DCP-Rho1 in the 2x Tubulin Reporter Solution, and the mixture was spun via ultracentrifugation for 5min after incubation. Further, the Taxol-stabilized microtubules were pre-treated with H₂O₂, and then spun down and resuspended in the repair mixture, so that the repair mixture was never in contact with the H₂O₂. This protocol ensured that DCP-Rho1 was detecting oxidized tubulin on the microtubules themselves, and that the repair tubulin was not oxidized. Microtubules were then processed for TEM (see above) or adhered to a streptavidin/neutravidin-treated imaging chamber and imaged via TIRF microscopy as described above.

Microtubule Repair Assay Image Analysis – Coverage Fraction

To quantify incorporation of green GTP-tubulin, the fractional coverage area of green tubulin along the red microtubule was determined using a previously described custom built MATLAB analysis tool (Reid et al., 2017). First, the red microtubule channel was processed to identify GDP microtubules. The green reporter tubulin channel was then filtered to reduce high-frequency noise and smoothing was performed to correct for heterogeneity within TIRF illumination of the optical field. A green channel threshold was manually selected just above background and maintained for consistent analysis across all experiments. For each TIRF image, the fractional area of green reporter tubulin incorporation within the red microtubule was recorded as the Coverage Fraction, excluding green tubulin at the microtubule ends, which represents normal microtubule elongation.

Microtubule Repair Assay Image Analysis – Cross Correlation

Microtubule fluorescence intensity was analyzed along the length of individual microtubules as previously described (Coombes et al., 2016). Briefly, individual microtubules were cropped from TIRF microscopy images using ImageJ. Fluorescence intensity linescans were generated for each individual microtubule as previously described (Coombes et al., 2016), and plotted as a function of microtubule length. To determine whether microtubule repair co-localizes with Mal3 or DCPC-Rho1 binding, linescans generated from the two respective channels were normalized as follows (Jaqaman et al., 2010b):

$${\tilde{x}}_i(l)=\frac{x_i(l)-{\mu }_i}{{\sigma }_i}$$

Where each linescan x_i(l) has mean intensity μ_i and standard devation σ_i. The cross-correlation function for each microtubule was then generated from the normalized linescans of the two channels using the matlab built-in function, xcorr. Cross-correlation functions were averaged across all microtubules and the absolute value was reported, along with 95% confidence intervals.

Mal3 Microtubule Binding Assay in Dynamic Microtubules

The microtubule dynamics assay was completed as described above with the reaction mixture containing 10μM tubulin (12.7% Alexa647-labeled, 87.3% unlabeled), 1mM GTP, 90mM KCl, 39.4nM Mal3-GFP, imaging buffer, and H₂O₂ (0 or 0.5mM). Images of dynamic microtubules were acquired at 28°C using 488nm and 561nm lasers with a TIRF microscope (Nikon Eclipse Ti TIRF) fitted with an 100x oil objective (Nikon CFI Apochromat TIRF 100XC Oil, 1.49 NA), CCD camera (Andor, iXon3) and, 2.5x projection lens. This TIRF microscopy imaging system allowed for a 64nm pixel size.

Mal3 Microtubule Binding Assay Image Analysis

To compare relative binding of Mal3-GFP to dynamic microtubules in the presence and absence of H₂O₂, the Mal3 Coverage Fraction was calculated as the total area of green (Mal3-GFP) occupancy divided by the total area of the red microtubules for each image. A semi-automated MATLAB analysis code was used as previously described (Reid et al., 2017).

Amplex Red assay to measure residual H₂O₂ in Imaging Buffer

In order to assess the concentration of H₂O₂ in the presence of the reactive oxygen species scavengers within imaging buffer, we used the previously established Amplex Red assay (Schlieve et al., 2006). Briefly, Amplex Red, an H₂O₂-sensitive dye was used. Solutions of 50 μM Amplex Red, 1 U/mL Horseradish peroxidase, increasing dilutions of imaging buffer, and 30 μM H₂O₂ were made in a 96-well plate. Samples were incubated for 30 min in the dark. Samples were then excited at 530 nm using a fluorescence spectrophotometer (Molecular Devices, Spectramax Gemini XPS Microplate Reader) and fluorescence was measured at 590 nm emission.

To determine the absolute H₂O₂ concentration from absorbance measurements, samples were prepared with increasing imaging buffer dilutions (0.003–1x). The recovery fraction for each imaging buffer dilution was determined using the following equation:

$$\mathit{\text{Recovery}}{\mathit{\text{Fraction}}}_{\mathit{\text{sample}}}=1-\frac{RF{U}_{\mathit{\text{sample}}}}{RF{U}_{0x\mathit{\text{Imaging}}\mathit{\text{Buffer}}}}$$ (4)

To predict the remaining H₂O₂ concentration in the 0, 0.5, and 1 mM H₂O₂ microtubule dynamics experiments, the mean recovery fraction was multiplied by each starting H₂O₂ concentration.

QUANTIFICATION AND STATISTICAL ANALYSIS

Data analysis was conducted using the software, Microsoft Excel or Matlab. Student t-test was used for two-sample comparisons. For multiple comparisons, statistical significance was determined using one-way analysis of variance (ANOVA). Image analysis is described in detail in methods above. Specific statistical approaches used for each figure are indicated in the figure legends.

Supplementary Material

Supplemental Movie 1

In vitro Dynamic Microtubule Assay (0mM H₂O₂), Related toFigure 2: A dynamic microtubule in a cell-free assay, imaged via TIRF microscopy. Stabilized microtubule seeds (red) are immobilized to an imaging chamber via anti-rhodamine antibody. Free tubulin (green) was introduced into the chamber in the absence of H₂O₂. Movies acquired at 0.2 frames/sec using TIRF microscopy (0.16 μm pixel size).

Supplemental Movie 2

In vitro Dynamic Microtubule Assay (0.5mM H₂O₂), Related toFigure 2: A dynamic microtubule in a cell-free assay, imaged via TIRF microscopy. Stabilized microtubule seeds (red) are immobilized to an imaging chamber via anti-rhodamine antibody. Free tubulin (green) was introduced into the chamber in the presence of 0.5 mM H₂O₂. Movies acquired at 0.2 frames/sec using TIRF microscopy (0.16 μm pixel size).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-α-tubulin antibody (DM1A)	Thermo Fisher Scientific	Cat# 62204; RRID: AB_1965960
anti-mouse IgG antibody	Sigma-Aldrich	Cat# F0257; RRID: AB_259378
anti-rhodamine antibody	Thermo Fisher Scientific	Cat# A-6397; RRID: AB_2536196
HRP-conjugated anti-rabbit antibody	Santa Cruz Biotechnology	Cat# sc-2004; RRID: AB_631746
Bacterial and virus strains
Rosetta (DE3) pLysS E. coli	Novagen	--
Biological samples
Adult rat ventricular myocytes	Purified from rats per methods	--
Chemicals, peptides, and recombinant proteins
Dihydroethidium (DHE)	Sigma-Aldrich	Cat# D7008
Amplex Red	ThermoFisher Scientific	Cat# A12222
Experimental models: Cell lines
H9c2 cell line	ECACC	Cat# 88092904; RRID: CVCL_0286
Recombinant DNA
CellLight Tubulin-GFP, BacMam 2.0	Thermo Fisher Scientific	Cat# C10613
Plasmid: pETMM11-HIS6x-Mal3-GFP	Gift from T. Surrey	MGP131
Plasmid: HIS6x-TEVsite-MAL3-MCHERRY	Gift from T. Surrey	MGP106
Software and algorithms
MATLAB	Mathworks	RRID:SCR_001622
ImageJ - Fiji	NIH	RRID:SCR_002285

Highlights.

Oxidative stress leads to cysteine oxidation of tubulin within microtubules

Oxidized tubulin in microtubules is associated with structural damage to the lattice

Damaged lattice regions within microtubules are repaired with stabilizing GTP-tubulin

Oxidative stress may facilitate microtubule network densification in cardiomyocytes