The mechanism of osmotically induced sealing of cardiac t tubules

Abstract

Cardiac t tubules undergo significant remodeling in various pathological and experimental conditions, which can be associated with mechanical or osmotic stress. In particular, it has been shown that removal of hyposmotic stress can lead to sealing of t tubules. However, the mechanisms underlying the sealing process remain essentially unknown. In this study we used dextran trapping assay to demonstrate that in adult mouse cardiomyocytes, t-tubular sealing can also be induced by hyperosmotic challenge and that both hypo- and hyperosmotic sealing display a clear threshold behavior requiring ≈100 mosmol/L minimal stress. Importantly, during both hypo- and hyperosmotic challenges, the sealing of t tubules occurs only during the shrinking phase. Analysis of the time course of t-tubular remodeling following removal of hyposmotic stress shows that t tubules become sealed essentially instantly, well before any significant reduction in cell size can be observed. Overall, the data support the hypothesis that the critical event in the process of t-tubular sealing during osmotic challenges is detachment (peeling) of the membrane from the underlying cytoskeleton due to suprathreshold stress.

NEW & NOTEWORTHY This study provides new insights into how t-tubular membranes respond to osmotic forces. In particular, the data show that osmotically induced sealing of cardiac t tubules is a threshold phenomenon initiated by detachment of t-tubular membrane from the underlying cytoskeleton. The findings are consistent with the hypothesis that final sealing of t tubules is driven by negative hydrostatic intracellular pressure coincident with cell shrinking.

INTRODUCTION

Cardiac t-tubular remodeling, characterized by loss or dilation of t tubules resulting in defects in excitation-contraction coupling, is a prevalent observation in many pathophysiological conditions [for review (12, 18)]. It strongly correlates with abnormal mechanical stress of the myocardium (10, 16, 17, 29), although it remains unclear how mechanical forces affect the t-tubular structure.

In vitro models of t-tubular remodeling may provide insight into how t-tubular membranes respond to mechanical forces. In this regard, the phenomenon of sealing of t tubules in both skeletal and cardiac muscles in response to osmotic challenges has been known for a long time (7, 9, 14, 15, 20). Yet, there is still no clear mechanistic understanding of the exact nature of the underlying membrane transformations. Early studies suggested that sealing of t tubules may be due to stretch of t-tubular membrane during rapid expansion of cells in response to osmotic swelling (11, 20). However, we recently showed that t tubules, in fact, seal upon shrinking of the cells, albeit after swelling, in hyposmotic solutions (24).

In this study, we provide further evidence that osmotically induced cell shrinking is indeed the key step in detubulation process and that shrinking per se is sufficient to seal cardiac t tubules even in the absence of prior cell swelling. A novel principal finding is that t-tubular sealing demonstrates clear threshold phenomenon with regard to the magnitude of applied osmotic stress. The sealing process following priming of t tubules by hyposmotic swelling is essentially instant and nearly finished before any significant shrinking is practically observed.

Overall, the data strongly support a hypothesis that t-tubular sealing during osmotic challenges proceeds through detachment (peeling) of the membrane from the underlying cytoskeleton.

MATERIALS AND METHODS

Animals

All experiments involving mice were carried out in accordance with the Guide for the Care and Use of Laboratory Animals, and protocols were approved by the veterinary staff of the University Committee on Use and Care of Animals at the University of Michigan. Both male and female C57BL/6 mice, 3–6 mo in age, were included in this study.

Solutions

All solutions were filtered using a 0.22-µm filter and pH adjusted with NaOH. Except for the data in Fig. 3E and Fig. 4C, the strength of hypo- and hyperosmotic NaCl solutions is indicated as the [NaCl] relative to that in modified Tyrode solution (below). For example, a hyperosmotic solution containing twice the [NaCl] of Tyrode solution is denoted 2.0 Na.

Fig. 3.

Hyperosmotic stress is sufficient to seal t tubules. A: representative confocal images of cardiomyocytes detubulated in the presence of extracellular fluorescent dextran using the indicated osmotic stresses. Cell edges are outlined in white. Cardiomyocytes detubulated with the standard hyposmotic 0.6 Na solution (dotted box) and those exposed to no osmotic changes during solution transitions (1.0 Na) served as controls. See materials and methods for detailed protocol for control cells. B: representative confocal images of cardiomyocyes detubulated using hyperosmotic solutions with increasing sucrose concentrations. A and B: scale bar is 20 µm. C, top, inset: protocol used to detubulate cardiomyocytes with hyperosmotic NaCl solutions for cardiomyocytes in A. Quantification of hyperosmotic dextran trapping relative to that in cells following hyposmotic detubulation with 0.6 Na solution (n = 20–30 cells each condition). Tyr, Tyrode solution. D: quantification of dextran trapping relative to that in cells following detubulation with hyposmotic 0.6 Na (n = 30 cells each condition). D, top, inset: protocol used to detubulate cardiomyocytes with hyperosmotic sucrose containing solutions depicted in B. E: data in C and D plotted against the measured increase in solution osmolarity. Note that both NaCl (●) and sucrose (○) osmolarity dependence relationships display clear thresholds for dextran trapping (▲ and Δ). In C and D, statistics comparing individual “hyperosmotic data” relative (Rel) to “0.6 standard” is performed using t-test. *P < 0.05; ***P < 0.001.

Fig. 4.

Hyposmotic detubulation displays threshold phenomenon. A: cardiomyocytes were exposed to 0.6 Na hyposmotic solution for 1, 2, 4, or 7 min (n = 20–30 cells each) using a modified 0.6 Na detubulation protocol (inset). Fluorescent dextran is present throughout the protocol until after the removal of hyposmotic stress. Single exponential fit to the data (gray line) reveals the minimum time (threshold; ·) necessary for the t-tubular sealing to occur. Tyr, Tyrode solution. B: averaged relative (Rel) change in the cell width over time during perfusion with hyposmotic 0.6 Na solution (n = 7 cells). The time axis is aligned with that in A above. The threshold time in A is presented as a vertical dotted line in B. The cell width at the threshold time corresponds to ≈6.6% change in cell width. C: cardiomyocytes were exposed to hyposmotic solutions of varying strengths for 7 min (n = 20–30 cells each) using standard detubulation protocol (inset) and the amount of trapped dextran quantified relative to that obtained using 0.6 Na solution. Linear fit to 0.8 Na, 0.7 Na, and 0.6 Na data reveals a clear threshold for dextran trapping at ≈46 mosmol/L.

Modified Tyrode (Tyr) consisted of (in mM) 137 NaCl, 5.4 KCl, 0.5 MgCl₂, 0.3 CaCl₂, 0.16 NaH₂PO₄, 3 NaHCO₃, 5 HEPES, and 10 glucose (pH = 7.35).

Myocyte storage solution (C solution) consisted of (in mM) 122 NaCl, 5.4 KCl, 4 MgCl₂, 0.16 NaH₂PO₄, 3 NaHCO₃, 15 HEPES, and 10 glucose and (in mg/mL) 5 bovine serum albumin and 1.38 taurine (pH = 7.35).

Hyposmotic Tyrode (0.6 Na) was prepared as Tyr but with 60% of NaCl (82.2 mM). Milder hyposmotic solutions, 0.9 Na, 0.8 Na, and 0.7 Na, were prepared by mixing 0.6 Na and Tyr solution in 1:3, 1:1, and 3:1 ratios, respectively; pH = 7.35.

Chemicals

Chemicals were obtained from the following sources: HEPES (Calbiochem); KCl, NaHCO₃, and NaH₂PO₄ (Mallinckrodt Chemicals), sucrose (Acros Organics), 3 kDa tetramethylrhodamine-labeled dextran (anionic, lysine fixable; Thermo Fisher Scientific, Inc., Waltham, MA), and collagenase (type 2; Worthington Biochemical Corp., Lakewood, NJ. All other chemicals and reagents were from Sigma or Sigma-Aldrich (St. Louis, MO).

Myocyte Isolation

Left ventricular myocytes were isolated essentially as described in Moench et al. (24).

Dextran Trapping Assay

Detailed methodology has been previously published (24). Briefly, cardiomyocytes were centrifuged and pelleted in 1.5-mL microcentrifuge tubes to perform the solution exchanges. Cells were swollen in hyposmotic Tyr solution containing 60% NaCl (0.6 Na solution) for 7 min, and 1 mg/mL (final concentration) 3 kDa tetramethylrhodamine-labeled dextran was added ≈2 min before washout with Tyr, also containing dextran that was removed 5 min later. Correction for background fluorescence was performed using cells that were treated identically but in constant presence of normal Tyr solution. Specific protocols are indicated in corresponding figures. In experiments using hyperosmotic detubulation (Fig. 3), 3 kDa tetramethylrhodamine dextran is added to cells ≈2 min before applying dextran containing hyperosmotic solutions for 5 min. The data are quantified as dextran fluorescence (background corrected) relative to that in cells detubulated using the 0.6 Na solution.

Confocal Imaging

Cardiomyocytes with trapped dextran were imaged at the Microscopy and Image Analysis Laboratory (Univ. of Michigan, Ann Arbor, MI) on a Nikon A-1 confocal microscope using 60× 1.4 numerical aperture (NA) oil objective and lateral pixel size of ≈140 nm. Images of myocytes were manually outlined, and mean intracellular fluorescence of trapped dextran per unit area was calculated using Fiji (https://imagej.net/Fiji). Additional data analysis, e.g., correction for background fluorescence, was performed in Microsoft Excel. Since cardiomyocytes recover their size/volume after osmotic challenges to within few percentage of their original state (5, 24), no further adjustments for cell area are necessary.

Three-dimensional (3D) confocal images of cardiomyocytes with trapped dextran were obtained using a z-step size of 300 nm and lateral pixel size of 68 nm. Z stacks were deconvolved in Fiji using Richardson-Lucy algorithm (DeconvolutionLab2 plugin (30) and an artificially generated PSF (PSF Generator plugin). Local autothresholding using Phansalkar algorithm was performed on the deconvolved image, the cell border was manually outlined, and all extracellular particles were removed. Two-dimensional (2D) particle analysis was performed in Fiji with circularity threshold set to 0.7 to exclude clusters of t-tubular dilations from the analysis. A stack of masks of the 2D particles was then used to perform the 3D objects analysis.

In all experiments, cells were imaged within ~15 min–2 h after detubulation.

Time-lapse imaging of t-tubular vacuolation (Fig. 1) was performed using a RC-20 small-volume perfusion chamber (Warner Instruments, Hamden, CT) and 60× 1.4 NA oil objective at a frame rate of 10 s.

Fig. 1.

Formation of t-tubular dilations following hyposmotic swelling. A: selected confocal images (uniformly contrast adjusted for clarity) from a representative experiment (Supplemental Material, time series 1) of cardiomyocyte bathed in extracellular solution containing fluorescent dextran at specific times (frames 1, 7, 8, and 32 in time-lapse confocal microscopy series) during transition from a swollen state in 0.6 Na solution to a normal state in control Tyrode (Tyr) solution. Images were filtered using Gaussian Blur (ImageJ) and contrast adjusted to highlight the features of t-tubular system. V₁ and V₂, 2 selected locations indicated in frames 8 and 37, were used for quantification of the time course of vacuolation of t tubules. Changes in the relative width of the cardiomyocyte (w) were quantified by measuring the distance (d) between 2 reference locations p₁ and p₂ and correcting for small deviation of the direction of the line d relative to cross-striation direction of the cell as indicated in frame 7 image in A. B: timing of solution exchange. C: time course (in frame units; 10 s/frame, and real time) of average fluorescence (F) from 2 small areas containing t-tubular vacuoles (V₁ and V₂ in A; circles are not drawn to scale). D: time course of relative cell width (w). Time constant (τ) was estimated by fitting the data starting from frame 8 using the following formula: w = A·exp(−t/τ) + C. Single exponential fit to the shrinking phase is highlighted by gray thick line.

Dextran Diffusion Assay

A detailed description of this assay has been recently published (32). Briefly, individual cardiomyocytes are placed in a rapid perfusion chamber, and the t-tubular network is locally filled with fluorescent dextran. Extracellular dextran is then rapidly removed, and the t-tubular fluorescence is monitored from a small intracellular spot using a wide-field fluorescence microscope as the t-tubular luminal dextran gradually diffuses out.

Measurement of Cardiomyocyte Width

Cardiomyocytes were plated onto a RC-20 small-volume perfusion chamber (Warner Instruments, Hamden, CT), and a 20× objective was used for imaging the cell. Time-lapse imaging was performed using a MD500 microscope eyepiece camera and AmScope 3.7 software (AmScope, Irvine, CA). Transmitted light images with pixel sizes ≈100 nm × 100 nm were collected at a sampling interval of 10 s. Image series were analyzed using custom-made edge-detection scripts in Fiji and MatLab (MathWorks, MA) to obtain cell width over time.

Electrophysiology

Electrophysiological experiments and corresponding data analysis were performed essentially as previously described (33). In particular, I_K1 tail current (I_K1,tail) was measured in the following way. Membrane potential was stepped to +50 mV for 400 ms and then stepped back to holding potential of −75 mV. Potassium current induced by repolarization, called I_K1,tail (Fig. 6C, inset), was fitted using a single exponential function A·exp(−t/τ) + C. Approximately equal to 15 ms of the current was excluded from the fit to minimize a contribution from capacitative (C) currents and the amplitude of exponential component (A) was then recalculated to zero time using measured time constant (τ). Normalized I_K1,tail ($I_{\text{K1,tail}}^{\text{N}}$) is calculated as the ratio of the magnitude of the I_K1,tail to the magnitude of the outward current at the end of the depolarizing voltage step, I_K,end.

Fig. 6.

Effects of swelling during whole cell patch-clamp recordings. Cardiomyocytes were voltage-clamped using whole cell configuration of patch-clamp and various parameters were measured during perfusion with Tyrode solution (Tyr; ●) and 0.6 Na hypoosmotic solution (○; gray area). Cells are perfused with Tyr before applying 0.6 Na solution. The average data from 4 cells for each condition are presented relative (Rel) to that measured just before the application of 0.6 Na hyposmotic solution. A: changes in the cell width. The time course of changes in the cell width in 0.6 Na solutions was fit using single exponential function (gray line). B: changes in the membrane capacitance (C_m). C: changes in the I_K1 tail current (I_K1,tail) time constant. C, inset: measured parameters (see materials and methods). τ₁ and τ₂ correspond to the initial and final time constants, respectively. D: changes in normalized I_K1 tail current ($I_{\text{K1,tail}}^{\text{N}}$) amplitude. a₁ and a₂ correspond to the initial and final amplitudes (C, inset), respectively. E: changes in the I_K1 amplitude. F: changes in the magnitude of the outward current at the end of depolarizing voltage step (I_K,end) (C, inset). NS, not significant. Statistics by two-way ANOVA.

Statistics

Statistical significance was determined using a two-sample t-test (two tail), assuming equal variances or ANOVA, and differences were considered significant if P < 0.05. Data are presented as means ± SE.

RESULTS

Dynamics of T-Tubular Sealing Following Hyposmotic Swelling

In this study, the structural remodeling of cardiac t tubules was first characterized by time-lapse confocal microscopy of cardiomyocytes during hyposmotic detubulation using 0.6 Na solution containing fluorescent dextran. Figure 1A shows selected images (from a representative experiment; Supplemental Material, time series 1) at four specific time points highlighting the major events occurring during washout of 0.6 Na hyposmotic solution (Fig. 1B). Dim striations are observable throughout the swollen cardiomyocyte before removal of hyposmotic solution (frames 1 and 7), indicative of the accessibility of the t-tubular lumens from the extracellular space. Following washout with normal Tyr solution, formation of large vacuole-like dilations of t tubules throughout the cell becomes evident (frames 8 and 37). Strikingly, t-tubular dilations are formed nearly instantaneously and reach their nearly stable size within seconds (compare frames 7 and 8) after removal of hyposmotic stress. Figure 1C shows that most of the changes occur within one to two frames (10–20 s), which is then followed by further gradual increase in fluorescence. Importantly, comparing the data in Fig. 1C with that in Fig. 1D shows that significant vacuolation of t tubules occurs when the shrinking of the cell is barely detectable (≤2%), and t-tubular dilations remain essentially unchanged after that as the cell continues to shrink, reaching the normal size within several minutes. As cells shrink, many individual vacuoles move in and out of focus (see Supplemental Material, time series; https://doi.org/10.6084/m9.figshare.12580727), so positions of only few vacuoles that remain in focus for a reasonable time were amendable for analysis. Since the change in the fluorescence is fast relative to the frame rate (10 s), the time course cannot be reliably fit with an exponential function. Accordingly, to further quantify this important phenomenon, we simply calculated the half-time (t_1/2) of the rise of luminal dextran fluorescence, which represents t-tubular luminal volume, since fluorescence is proportional to the volume, from its minimum to the maximum observed within 0.5–1.5 min. The analysis of the data from eight successful experiments shows that t_1/2 = 10 ± 2 s (and likely even less), which is approximately equal to fourfold smaller than t_1/2 ≈ 40 s of shrinking (swelling) of cardiomyocytes in 0.6 Na solution calculated as −τ ·ln(1/2), where the τ ≈ 60 s [e.g., Drewnowska and Baumgarten (5), Fig. 1], confirming that these t-tubular dilations form much faster than the cardiomyocyte shrinks.

In a separate set of experiments (see materials and methods), a batch of cardiomyocytes was detubulated in dextran containing solution and the size of t-tubular dilations was quantified using particle analysis performed on confocal images (see Supplemental Material for a representative image) of cardiomyocytes with trapped fluorescent dextran taken at significantly later times. The average (apparent) diameter of dilations, assuming that they can be represented as circular structures, was 552 ± 52 nm (Fig. 2), which is significantly larger than the reported diameter of intact mouse t tubules of ≈170 nm (21). It should be noted that although the above average diameter is well above the resolution limit of confocal microscopy, quantification of objects smaller than ≈250 nm becomes less quantitative. The increase in the size of t tubules at certain locations (vacuoles) is also visually confirmed by significant increase in dextran fluorescence.

Fig. 2.

Histogram of apparent diameters of t-tubular dilations obtained from 2-dimensional confocal images. N = total 18,301 particles from 3 cardiomyocytes. The data (means ± SE) are averages of 3 individual histograms. Representative image of cardiomyocyte can be found in Supplemental Material.

Overall, the results demonstrate that removal of hyposmotic stress leads to rapid formation of large dilations of t tubules well before any significant shrinking of the cardiomyocytes.

Hyperosmotic Stress also Leads to Sealing of T Tubules

We have previously shown that sealing of t tubules exclusively occurs during shrinking of the cardiomyocytes from the swollen state induced by prior exposure to hyposmotic solutions (24). To determine whether cell shrinking from a normal state is sufficient to seal t tubules, hyperosmotic solutions of various strengths were applied to normal cardiomyocytes in the presence of fluorescent dextran, which was removed before the removal of hyperosmotic stress. Cardiomyocytes detubulated using the previously described hyposmotic stress protocol with 0.6 Na solution (Fig. 1B), which served as control. Figure 3, A and C, shows that application of NaCl containing hyperosmotic solutions leads to dose-dependent sealing of t tubules, with 2.0 Na solution inducing in a nearly uniform trapping of dextran throughout the t-tubular network of the cardiomyocyte, similar in magnitude to that observed in hyposmotic 0.6 Na detubulation. Weaker hyperosmotic stresses, e.g., using 1.5 Na solution, resulted in “patchy” patterns of dextran trapping, suggesting that some t-tubular segments within the network fully seal while others remain open, and essentially no trapping was observed with hyperosmotic 1.25 Na solution.

Since detubulation protocols involve manipulations of extracellular Na⁺, it can be argued that the changes in its concentration may lead to changes in intracellular Ca²⁺ through Na⁺/Ca²⁺ exchanger (NCX) or other Ca²⁺ handling pathways, which in turn can be considered as a major reason of t-tubular sealing. However, we previously reported that neither blockade of Ca²⁺ channels by nicardipine nor inclusion of intracellular EGTA in patch-clamp experiments, as well as other relevant manipulations, significantly affected t-tubular sealing (23, 24). In addition, t-tubular sealing is not affected by removing extracellular Ca²⁺ (unpublished data), thus essentially eliminating NCX and intracellular Ca²⁺ as major players in osmotically induced detubulation.

To exclude the possibility of ionic effects being a primary reason of membrane remodeling, a non-ionic osmolyte, sucrose, was tested. Principally similar results were obtained with solutions made hyperosmotic by the addition of sucrose (Fig. 3, B and D). Consistent with our previous findings (24), application of hyperosmotic solution containing 100 mM sucrose led to little dextran trapping. When greater concentrations of sucrose are used, hyperosmotic solutions detubulate cardiomyocytes in a dose-dependent manner, albeit with less efficiency compared with that when using hyperosmotic NaCl solutions. In particular, the maximal amount of trapped dextran at ≈600 mosmol/L of extra osmolarity in sucrose containing solution reaches only ≈65% of that observed in NaCl containing solutions having only ≈230 mosmol/L extra osmolarity (Fig. 3E). It was also noted that the pattern of dextran trapping in sucrose containing solutions appeared to show less dextran fluorescence in the center of the cells (Fig. 3B).

It is hard to predict the magnitude of the difference between the efficiencies of detubulation by NaCl and sucrose solutions of the same osmolarity, although this difference is surely highly expected. In particular, ionic strength, which likely has significant effect on the stability of the lipid bilayer, is dramatically different in those solutions. In this regard, it is known that modification of the lipid bilayer may significantly affect detubulation (25, 33). Viscosity of sucrose solutions is also much higher than that of NaCl-based solutions (>70% higher for 500 mM sucrose). Since significant water movements are expected, the latter in part may explain the specific pattern of dextran trapping in sucrose containing solutions. In all cardiomyocytes detubulated with hyperosmotic solutions-sealed t tubules appear swollen similar to that observed in cells detubulated using hyposmotic stress (data not quantified).

The most important finding is that hyperosmotic detubulation by both types of solution displays a clear threshold phenomenon. Specifically, for both NaCl and sucrose containing solutions, the dose-response relationships do not originate at zero gradient of osmolarity but rather at ≈65 and ≈110 mosmol/L for NaCl and sucrose containing solutions, respectively (Fig. 3E).

Overall the results demonstrate that cell shrinking in hyperosmotic solutions is sufficient to induce t-tubular sealing, albeit at increased osmotic pressures relative to that employed in detubulation using hyposmotic solutions.

Hyposmotic Detubulation also Displays Threshold Phenomenon

To test whether threshold behavior of hyperosmotic detubulation described above is unique for this type of stress, we performed a more detailed analysis of hyposmotic detubulation.

In the first approach, we analyzed the kinetics of dextran trapping using a fixed hyposmotic stress applied for varying durations (Fig. 4A). Since dextran is present in solution during both swelling and shrinking phases, the data are not affected by dextran diffusion rates in t tubules. When the data were fit using a single exponential function (Fig. 4A; gray highlight), the fitted curve intercepted the time axis at ≈0.7 min, indicating that a certain threshold value in cell swelling must be reached for the later dextran trapping to occur. When translated to the cell size, the data show that detubulation threshold corresponds to ≈6% change in the cell width (Fig. 4B).

In the second approach, the strength of hyposmotic solution was varied while keeping the exposure time fixed and long enough for the cell swelling to reach steady-state values (Fig. 4C, inset). The data show that dextran trapping determined this way also shows a clear threshold behavior requiring ≈46 mosmol/L for the t-tubular sealing to occur.

Hyposmotic Stress Primes T Tubules for Sealing

The difference in the thresholds for hypo- and hyperosmotic detubulation (≈46 and 65 mosmol/L, respectively) hinted that there may be a mechanistic difference between the two processes. Accordingly, to further separate the contribution of swelling and shrinking versus direct shrinking to the extent of detubulation, cardiomyocytes were initially swollen to varying magnitudes and subsequently shrunken using a constant osmotic gradient (≈90 mosmol/L; a difference between osmolarity of Tyr and 0.6 Na solutions; Fig. 5). The data show that “priming” by swelling is indeed the primary, although not exclusive, contributor to the efficiency of hyposmotic detubulation, although a significant but lesser degree of detubulation can be achieved by applying hyperosmotic solutions of similar strength directly to normal cardiomyocytes (Fig. 5; e.g., 1.0 to 1.4 Na gradient).

Fig. 5.

Prior exposure to hyposmotic stress augments dextran trapping. A: modified protocol for hyposmotic detubulation. Cardiomyocytes were treated with hyposmotic solutions of varying strength (Pre) followed by washout solutions (Post) having ≈90 mosmol/L (equivalent to ≈0.4 Na) lower osmolarity relative to that of presolutions (constant osmotic gradient). B: quantification of dextran trapping in experiments as described in A; n = 19–22 cells for each condition. Rel, relative. Statistics by one-way ANOVA.

Diffusional Aspects of Osmotic Detubulation

To get further insights into the structural rearrangements of t tubules during osmotic challenges in this study, we have employed two approaches assessing their diffusional accessibility.

Electrophysiological approach.

Diffusional properties of t tubules can be assessed by measuring so-called I_K1,tail current using whole cell patch-clamp recordings (3, 4). It has been shown that application and washout of hyposmotic 0.6 Na solution during whole cell recordings (cell membrane is not fully closed anymore) does lead to sealing of t tubules (24), although the magnitude of cell swelling (if any) was not measured at that time.

The data in Fig. 6A show that even in the normal Tyr solution, cell dialysis leads to a small gradual shrinking of the cell width by ≈6% over 10 min. In contrast, perfusion with the 0.6 Na solution leads to ≈15% increase in cell width after ≈7 min, consistent with the magnitude of hyposmotic swelling in intact cells (Fig. 5B). The rate of cell swelling is also essentially preserved, albeit being somewhat smaller (τ ≈ 2 min) when compared with that measured in intact cells (τ ≈ 1 min; Fig. 5B).

Similarly, membrane capacitance also gradually declines in Tyr solution by <10% over 10 min (Fig. 6B). However, the rate of this decline is not significantly affected by perfusion with 0.6 Na solution, consistent with previous findings showing that in intact cardiomyocytes membrane capacitance remains constant during hyposmotic stress (24).

The I_K1,tail current was next assessed (see methods for details). This current is dependent on potassium accumulation within the t-tubular lumen caused by potassium efflux through various voltage-dependent potassium channels during prolonged depolarization steps (3, 4). Upon repolarization to resting membrane potential, the accumulated potassium dissipates either by flowing back into the cell through I_K1, forming the I_K1,tail current, or by diffusing out of the t-tubular lumen.

The dissipation of luminal potassium can be tracked by measuring the amplitude and kinetics of I_K1,tail current, reflecting the volume of t tubules and the rate of K⁺ diffusion, respectively (3, 4, 25). As shown in Fig. 6, C and D, osmotic swelling in 0.6 Na solution is characterized by a greater than twofold increase in the time constant of I_K1,tail (compared with only 25% increase observed in Tyr solution) and concurrent ≈50% decrease in the amplitude of $I_{\text{K1,tail}}^{\text{N}}$.

Since both the kinetics and the amplitude of $I_{\text{K1,tail}}^{\text{N}}$ tail strongly depend on the density of I_K1, the effect of hyposmotic swelling on this current was also assessed. The data in Fig. 6E show that perfusion with 0.6 Na solution leads to ≈25% decrease in I_K1, and therefore this effect should be considered for correct quantitative interpretation of the data (see discussion). Furthermore, I_K,end, which is the primary source of potassium influx underlying I_K1,tail, is unaffected by hyposmotic swelling (Fig. 6F).

Dextran diffusion approach.

Restricted diffusion of t-tubular K⁺ (above) during hyposmotic swelling is consistent with our previous data showing restricted diffusion of 3 kDa dextran (32). Here, we further expand on this initial finding by quantifying dextran diffusion in response to varying hyposmotic stresses. Figure 7A shows that cells perfused with Tyr solution displayed a small gradual decrease in the time constant of dextran diffusion (τ_dex), consistent with our previous observation (32). Exposure to hyposmotic stress, however, caused significant osmolarity-dependent changes in τ_dex (Fig. 7A), increasing greater than twofold after 7 min exposure to 0.6 Na solution. Importantly, the increase in τ_dex is significantly greater than that expected solely due to increase in cell size (32).

Fig. 7.

Effects of hyposmotic stress on dextran diffusion in t tubules. A: quantification of the changes in time constant measured in single cells using the dextran diffusion assay before and during (gray background) exposure to the indicated hyposmotic solutions. Perfusion of the hyposmotic solutions began at time 0 min. The time constants are relative (Rel) to that measured at zero time. The data in 0.6 Na solution were previously published by Uchida and Lopatin (32) and included (1 outlier excluded) here with permission from Elsevier. B: time constants 7 min after perfusion with hyposmotic solutions were normalized to that measured in 0.6 Na solution. C: amplitude (Amp) of the dextran diffusion signal after 7 min perfusion with the indicated solutions was normalized (Norm) to the initial dextran diffusion signal measured in Tyrode (Tyr) solution. D: relative di-8-butyl-amino-naphthyl-ethylene-pyridinium-propyl-sulfonate (di-8-ANEPPS) fluorescence (F; ○) measured within a cardiomyocyte using the same spot illumination utilized for the dextran diffusion measurements. Control cells continuously perfused with Tyr (●) were used to normalize the data to control for photobleaching; n = 7–12 cells for each condition. Statistics by one-way ANOVA in B and C.

While τ_dex increased during hyposmotic swelling, the amplitude of dextran fluorescence, reflecting the volume of t-tubular lumen, significantly decreased (Fig. 7C), consistent with the observed changes in the amplitude of I_K1,tail (above). In particular, after 7 min swelling in 0.6 Na solution, the amplitude of dextran fluorescence decreases by ≈40%. It should be noted, however, that part of this decrease can be explained by the movement of t tubules out of the recording spot during cell swelling (see materials and methods, and Ref (32).). Accordingly, to quantify this effect, similar experiments were performed with cardiomyocytes labeled with membrane-specific dye di-8-butyl-amino-naphthyl-ethylene-pyridinium-propyl-sulfonate (di-8-ANEPPS). After correcting for the effects of photobleaching, the di-8-ANEPPS fluorescence decreased only by ≈20% during swelling in hyposmotic solution, suggesting that ≈20% of the t-tubular membranes was indeed displaced out of the recording spot (Fig. 6D). Accordingly, the data show that the true reduction in the amplitude of dextran fluorescence is ≈20%.

Altogether, the data are consistent with the likely development of significant t-tubular constrictions in response to hyposmotic stress.

DISCUSSION

Mechanistic Model of Osmotically Induced Detubulation

It was suggested in early studies in striated muscles that the rapid cell swelling, occurring upon removal of previously applied (membrane permeable) formamide, is responsible for the detachment of t tubules from the surface sarcolemma (2, 20, 27). However, we previously demonstrated that in mouse ventricular myocytes, cell swelling per se, as it occurs for example in hyposmotic solutions, does not induce detubulation (24). Instead, we showed that detubulation occurs during cell shrinking upon removal of previously applied hyposmotic solution, albeit with initial cell swelling being a key factor priming cells for ultimate detubulation (24). However, the mechanistic understanding of the processes underlying sealing of t tubules remained obscure. In this regard, we report new data that provide a closer look at the events occurring during detubulation and propose a mechanistic model consistent with the available data.

The first important observation originates from the data in Fig. 1. The results reveal an unexpected and quite counterintuitive phenomenon of nearly instantaneous vacuolation of t tubules upon removal of hyposmotic gradient. Figure 1B shows that vacuoles essentially reach their final size when little to no shrinking of the cell has occurred, as if shrinking per se has no relation to vacuolation.

These results suggest that some submicroscopic transformation of t tubules priming them for consequent sealing has already occurred during the swelling phase. We propose that this transformation is likely detachment or “peeling” of the membrane from the underlying cytoskeleton which is commonly observed at the macroscopic level as membrane blebbing in response to hyposmotic stress (see cartoon in Fig. 8 that highlights some key steps).

Fig. 8.

Mechanistic model of hyposmotically induced sealing of cardiac t tubules. A–E: sequence of events occurring in the transverse t tubule during and after application of hyposmotic solution (“priming”). Gray color represents cytoskeleton. XY plane highlights “horizontal” cross section of transverse t tubule. XZ plane highlights “longitudinal” cross section of transverse t tubule. Radial stretching of the cell is ignored in this cartoon. F: it is assumed that due to viscoelastic properties of the cytoskeleton, the positive intracellular pressure (+P) present during swelling phase is transiently reversed to a negative value (−P) before a final return to normal (zero) pressure (0 P). Norm, normal solution. See text for more detail.

Although membrane blebbing on microscopic scale requires significant time to develop, and thus is not observed in our experiments due to relatively short exposure to hyposmotic solutions (≈7 min; longer exposures do lead to observable blebbing; data not shown), multiple experimental findings strongly support the existence of submicroscopic blebbing during osmotic detubulation.

If submicroscopic peeling (blebbing) of the membrane does occur during hyposmotic swelling, it would lead to significant and characteristic changes in some measurable parameters of t tubules. First, membrane blebbing should lead to apparent slowing of molecular diffusion in t tubules due to the development of t-tubular constrictions (Fig. 8). Consistent with this, the data in Fig. 7A show that, indeed, the time constant of dextran diffusion in t tubules progressively increases during exposure to hyposmotic solutions. It can be argued though that the increase in the time constant can also be explained by 1) formation of dilations (32) or simply by 2) the increase in the size of the cell, leading to longer diffusion time. However, the data in Fig. 7C show that the luminal volume of t tubules, assessed as the magnitude of t-tubular fluorescence of dextran, decreases during cell swelling. It has also been shown that the increase in cell size contributes little (<30%) to the overall changes (>200%) in diffusion time constant (32).

Similarly, diffusion of t-tubular K⁺, assessed by measuring the time constant of I_K1,tail, is also significantly restricted by cell swelling (Fig. 6C). Quantitatively, however, the interpretation of the increased I_K1,tail time constant is not that straightforward since the slower dissipation of luminal K⁺ could in part be explained by reduced I_K1. In this regard, Clark et al. [(4); shown in their Table 1] estimated that the contribution of I_K1 and K⁺ diffusion in t tubules to the I_K1,tail time constant is roughly equal. Assuming that diffusion of K⁺ is unaffected and I_K1 is reduced by ≈25% (Fig. 6E), one can estimate that the time constant of I_K1,tail current should then increase only by ≈15%. This estimate is folds smaller than the measured increase in the time constant of I_K1,tail (≈70%), suggesting that K⁺ diffusion is indeed significantly restricted in t tubules during hyposmotic swelling.

A second line of support for membrane peeling hypothesis comes from the fact that the initiation of membrane detachment (blebbing) requires a certain minimum (threshold) of hydrostatic pressure (force) to overcome the adhesion of the membrane to the underlying cytoskeleton (13). In this regard, the data in Fig. 4C show that sealing of t tubules also requires a certain minimum of intracellular hydrostatic pressure as evidenced by a clear threshold dependence of dextran trapping on the strength of hyposmotic solutions. This threshold also surely depends on the original state of the lipid bilayer (cholesterol, etc.), as well as on the integrity of numerous structural proteins (e.g., BIN1, junctophilin, telethonin, etc.) involved in t-tubular organization. The proposed membrane-peeling hypothesis is generally compatible with existing theories of t-tubular remodeling in vivo. For example, in heart failure, loss of structural proteins such as junctophilin 2 is associated with loss of t-tubular integrity (34). One hypothesis may be that loss of these structural proteins may lower the threshold for t-tubular remodeling and predispose t tubules to disruption by mechanical stress.

A third line of support for membrane peeling hypothesis can be drawn from the data in Fig. 4C, showing that the amount of trapped dextran is essentially linearly dependent on the magnitude of hyposmotic stress. The latter means that stronger (and/or longer) stresses would likely lead to even larger amount of trapped dextran, and the limit is practically determined only by cells’ survival. It should be noted, however, that when applying strong hyposmotic stress, the amount of trapped dextran does not necessarily equate with the fraction (number) of sealed t tubules, which is likely quite close to maximum (i.e., whole t system is sealed and dextran cannot escape to the outside solution) when standard protocol of detubulation (7 min exposure to 0.6 Na solution) is used. Therefore, the data are more consistent with the idea that at suprathreshold hyposmotic stresses, the increasing amount of trapped dextran is associated with the increasing volume of t tubules (originally, ≈3 to 4%) as they are being transformed from highly heterogeneous in diameter cylinder-like structures to spherical vacuoles (sphere has the largest volume-to-surface ratio), initiated by formation of submicroscopic membrane blebs.

At this juncture, an alternative hypothesis would be that the elongation of radial t tubules due to swelling (increase in cell width) may lead to the creation of “bottlenecks” as a consequence of spatial heterogeneity in mechanoelastic properties of t tubules. Several findings, however, argue against the above suggestion. In particular (1), swelling of the cells is significantly faster than the development of the presumed “bottlenecks” [τ_swelling ≈ 1 min vs. change in the rate of diffusion τ > 3 min (32)], and thus the data are more consistent with the slower process of membrane detachment (growing of membrane blebs) (2). It would also be hard to explain a nearly instantaneous formation of t-tubular vacuoles upon removal of hyposmotic stress when essentially no cell shrinking is yet observed (Fig. 1) (3). Radial elongation of t tubules upon swelling would not be expected to display threshold behavior. Overall, our experimental data cannot simply be explained by macroscopic changes in cellular geometry and likely point to submicroscopic disruption in the membrane-cytoskeleton interaction similar to membrane peeling.

Do T Tubules Seal by a Different Mechanism During Hyposmotic and Hyperosmotic Shrinking?

It may seem that the mechanisms of detubulation during hyposmotic and hyperosmotic challenges are quite different. For example, in contrast to hyposmotically induced detubulation, application of equal-strength hypertonic solutions is significantly less effective (Fig. 4). Also, in the first approach, sealing of t tubules occurs after removal of hyposmotic solution, while application of hyperosmotic solutions in the other one leads to sealing of t tubules right away.

However, the difference between hyposmotic and hyperosmotic detubulation is likely not of a principal nature but rather can be explained, at least in part, by asymmetries of osmotic forces and the sarcolemmal membrane. In particular, in hyposmotic solutions, the membrane expands outward where less or little resistance to expansion occurs, likely leading to “peeling” or “blebbing” of the membrane that in turn may lead to membrane misfolding and formation of t-tubular dilations upon subsequent cell shrinking. In contrast, one may hypothesize that stronger forces may be necessary to force the membrane to “expand” inward into the crowded intracellular space upon shrinking in hyperosmotic solutions from normal state, ultimately leading to similar vacuolation and sealing of t tubules.

Therefore, it is straightforward to suggest that application of hypertonic solutions to normal cells would also lead to formation of membrane blebs, albeit inwardly directed, some of which will ultimately become sealed vacuoles. The strongest support for this view comes from the fact that hyperosmotically induced sealing of t tubules also displays a clear threshold phenomenon (Fig. 4), consistent with peeling on the membrane and formation of inward-going membrane blebs.

How Do T Tubules Seal?

The data in Figs. 1 and 2 show that sealed t tubules appear swollen significantly beyond the average size (diameter) of original t tubules (similar effect is observed during detubulation in hyperosmotic solutions; data were not quantified). According to our “membrane peeling” hypothesis, when cells are exposed to strong hyposmotic solutions, submicroscopic membrane blebs begin to form because of increased intracellular hydrostatic pressure, but their growth in t tubules is limited because of very small t-tubular volume (Fig. 8). Simply because of geometrical reasons, even at their maximum allowed size, these blebs can only reduce the t tubules’ volume and slow the apparent diffusion in t tubules but cannot completely block (seal) t-tubular lumens. When the osmotic gradient is reversed, the available blebs are essentially instantly reversed (Fig. 1) and allowed to grow into relatively infinite, although more crowded, intracellular space. Biological membranes (lipid bilayers) are essentially nonstretchable and rupture upon less than few percent change in the area (1). Therefore, according to Laplace’s law, the largest bleb in a local region will take over its neighbors and will continue to grow to vacuole size using lipid flow from neighboring t-tubular segments (Fig. 8E). T-tubular membranes are very ruffled and may contain caveolae, and thus their submicroscopic flattening would serve as another way of providing lipid for growing vacuoles. According to the same law, the largest (yet, still small) t-tubular segments, such as dilations observed at the dyads (35), will also experience greater membrane tension and thus may also grow at the expense of the nearby narrow (constricted) t tubules, ultimately leading to their thinning to the point that diffusion of dextran is slowed enough to correspond experimentally to what is called t-tubular sealing.

The above scenario is supported by observations that constrictions in skeletal muscle t tubules is a characteristic feature of the vacuolation process, e.g., in skinned toad skeletal muscle fibers, large 500 kDa dextrans are excluded from longitudinal tubules, suggesting that there are constriction points at the junctions of transversally oriented and longitudinally oriented tubules (6). In particular, stretch of skinned fibers induced vacuolation specifically in longitudinal tubules that are flanked by these junctions, suggesting that bottlenecks within skeletal muscle t tubules provide the heterogeneity in t-tubular radii for t-tubular dilation to occur.

Some studies have also provided evidence that t tubules normally dilate at dyadic junctions. Wong et al. observed dilations of the t-tubular lumen at dyadic junctions in a serial block face-scanning EM (35). This is consistent with a previous study of freeze fracture images where distensions of t tubules appeared at transverse and axial junctions (8). The junctional SR was seen clearly at one of the junctions, suggesting that these distensions may be associated with dyads. In experiments using super-resolution 3D dSTORM imaging, dilations were observed at regions where Cav3 and RyR2 staining colocalized and the authors interpreted this colocalization to further suggest a dyadic location for the t-tubular dilations (35). It should be noted, however, that previous reports using confocal microscopy have suggested that Cav3 staining is excluded from dyads (19, 31). Thus, t tubules show regular dilations although their exact localization remains unclear. If t tubules are normally dilated at the dyads and are, as suggested above, more susceptible to vacuolation, then one would expect vacuolated t tubules to be associated with the dyad following hyperosmotic stress. In one study it has been demonstrated that cardiomyocytes fixed in hyperosmotic solutions show dilated t tubules associated with the junctional SR (26). However, the association of vacuolated t tubules with junctional SR may simply be a consequence of the expansion of distal t-tubular dilations, which brings the vacuole in close proximity to nearby dyads.

As an aside, cell types that lack t tubules (fibroblasts and neurons) also form vacuole-like dilations (VLD) of the surface membrane following removal of hyposmotic stress (22, 28). The similarity between osmotic detubulation and VLD formation (Fig. S1) suggests that both phenomena may be mechanistically related, and understanding the forces induced by osmotic volume changes may provide insights into how certain membranes aberrantly transform into vacuolar structures. The link between VLD formation and osmotic detubulation is further explored in the Supplemental Material using computational modeling.

In particular, we address one of the common views on VLD formation and detubulation that the driving force behind both phenomena is a buildup of positive pressure (either in the confined area between the cell membrane and the substrate or within narrow t tubules) due to osmotically driven water efflux. Opposite to that common view, quantitative computer analysis shows that the magnitudes of those pressures are too small (in the order of few cmH₂O) to cause the phenomena (Figs. S2–S4). Instead, the analysis points to development of strong negative intracellular pressures upon removal of hyposmotic (or application of hyperosmotic) stress, which has been assumed throughout this discussion. The latter, in turn, leads to another intriguing question as to why membrane vacuolation only occurs in t tubules if negative intracellular pressure is surely applied to all sarcolemmal membrane. In this regard, we show in Supplemental Material that the explanation can be easily found if one simply considers the difference in the topology of t-tubular and outer sarcolemma membranes (Figs. S5–S6).

Conclusions

Cardiac t tubules present a specialized domain of cellular membrane that responds in a highly specific manner to osmotic and mechanical forces. Both hyposmotic and hyperosmotic stresses lead to sealing of t tubules in a threshold-dependent manner, strongly suggesting that the peeling of the lipid bilayer from the underlying cytoskeleton is the key step underlying the threshold phenomenon. The experimental and modeling data support the hypothesis that the driving force for t-tubular sealing is negative-intracellular rather than positive-luminal hydrostatic pressure. Exclusive vacuolation of t-tubular membrane despite equal pressure experienced by outer sarcolemma is consistent with specific geometrical and topological nature of t tubules and a highly likely contribution of preexisting t-tubular dilations such as those observed at dyads.

GRANTS

This work was supported by the National Institutes of Health Grants HL-127023 (to A.N.L.) and T32-GM-008322 (to K.U.) and American Heart Association Grant 17PRE33350049 (to K.U.).

Present address of K. Uchida: Dept. of Physiology, Pennsylvania Muscle Inst., Univ. of Pennsylvania Perelman School of Medicine, Philadelphia, PA.

Present address of I. Moench: GSK, Novel Human Genetics RU, Collegeville, PA.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

K.U., A.N., and A.N.L. conceived and designed research; K.U., A.N., I.M., G.T., Y.E., and A.N.L. performed experiments; K.U., A.N., I.M., G.T., Y.E., and A.N.L. analyzed data; K.U., A.N., I.M., G.T., and A.N.L. interpreted results of experiments; K.U., A.N., and A.N.L. prepared figures; K.U. and A.N. drafted manuscript; K.U., A.N., I.M., G.T., Y.E., and A.N.L. edited and revised manuscript; K.U., A.N., I.M., G.T., Y.E., and A.N.L. approved final version of manuscript.