Inflammation alters regional mitochondrial Ca²⁺ in human airway smooth muscle cells

Abstract

Regulation of cytosolic Ca²⁺ concentration ([Ca²⁺]_cyt) in airway smooth muscle (ASM) is a key aspect of airway contractility and can be modulated by inflammation. Mitochondria have tremendous potential for buffering [Ca²⁺]_cyt, helping prevent Ca²⁺ overload, and modulating other intracellular events. Here, compartmentalization of mitochondria to different cellular regions may subserve different roles. In the present study, we examined the role of Ca²⁺ buffering by mitochondria and mitochondrial Ca²⁺ transport mechanisms in the regulation of [Ca²⁺]_cyt in enzymatically dissociated human ASM cells upon exposure to the proinflammatory cytokines TNF-α and IL-13. Cells were loaded simultaneously with fluo-3 AM and rhod-2 AM, and [Ca²⁺]_cyt and mitochondrial Ca²⁺ concentration ([Ca²⁺]_mito) were measured, respectively, using real-time two-color fluorescence microscopy in both the perinuclear and distal, perimembranous regions of cells. Histamine induced a rapid increase in both [Ca²⁺]_cyt and [Ca²⁺]_mito, with a significant delay in the mitochondrial response. Inhibition of the mitochondrial Na⁺/Ca²⁺ exchanger (1 μM CGP-37157) increased [Ca²⁺]_mito responses in perinuclear mitochondria but not distal mitochondria. Inhibition of the mitochondrial uniporter (1 μM Ru360) decreased [Ca²⁺]_mito responses in perinuclear and distal mitochondria. CGP-37157 and Ru360 significantly enhanced histamine-induced [Ca²⁺]_cyt. TNF-α and IL-13 both increased [Ca²⁺]_cyt, which was associated with decreased [Ca²⁺]_mito in the case of TNF-α but not IL-13. The effects of TNF-α on both [Ca²⁺]_cyt and [Ca²⁺]_mito were affected by CGP-37157 but not by Ru360. Overall, these data demonstrate that in human ASM cells, mitochondria buffer [Ca²⁺]_cyt after agonist stimulation and its enhancement by inflammation. The differential regulation of [Ca²⁺]_mito in different parts of ASM cells may serve to locally regulate Ca²⁺ fluxes from intracellular sources versus the plasma membrane as well as respond to differential energy demands at these sites. We propose that such differential mitochondrial regulation, and its disruption, may play a role in airway hyperreactivity in diseases such as asthma, where [Ca²⁺]_cyt is increased.

mitochondria have been known for several decades to have a role in Ca²⁺ sequestration (76, 82). Metabolism or ATP production could be enhanced by mitochondrial Ca²⁺ uptake, and perinuclear mitochondria might be involved in shaping nuclear Ca²⁺ signals (9, 10, 23, 64). Whether mitochondria contribute to regulate cytosolic Ca²⁺ concentration ([Ca²⁺]_cyt) is subject to debate (10, 48, 64).

The importance of defining the role of mitochondrial Ca²⁺ buffering in human airway smooth muscle (ASM) lies in understanding the mechanisms by which [Ca²⁺]_cyt is enhanced in airway diseases such as asthma, which contribute to airway hyperresponsiveness. Mitochondrial dysfunction and some ultrastructural changes in mitochondria (swelling and loss of cristae) have been observed with airway inflammation (1, 39). In human ASM, mitochondria are present in large numbers (79), and, interestingly, mitochondrial biogenesis is enhanced in asthma (28). Whether such changes can contribute to altered Ca²⁺ homeostasis or other features of asthma, such as airway remodeling, is not known.

Mitochondrial Ca²⁺ uptake during agonist stimulation has been observed in different cell types (13, 25, 33, 38, 48, 55, 56, 65) and is thought to decrease local Ca²⁺ gradients for sarcoplasmic reticulum (SR)/endoplasmic reticulum Ca²⁺ refilling while maintaining or enhancing store-operated Ca²⁺ entry (SOCE) (24, 37, 41, 45, 50, 81). Some theoretical models of Ca²⁺ signaling in smooth muscle that include mitochondria also suggest a role for modulating [Ca²⁺]_cyt regulatory mechanisms (66, 67). In this regard, mitochondria are often located near the SR and plasma membrane (PM), and this physical proximity may be important (25, 33, 63, 69), especially to buffer locally high levels of Ca²⁺ during stimulation (27, 53, 56, 57). These studies raise the question of the relation between the spatial localization of mitochondria and their ability to buffer or release Ca²⁺. Recent work has established that mitochondria contain sophisticated Ca²⁺ uptake and release mechanisms that may alter [Ca²⁺]_cyt either globally or in the vicinity of the SR and PM (10, 13, 16, 22, 25, 33, 38, 43, 48, 52, 60, 62, 63, 65, 70). There is currently little information on these roles for mitochondria in ASM cells.

Several studies have suggested that airway inflammation increases airway contractility by modulating Ca²⁺ regulatory mechanisms (2–4, 61, 71, 72, 74, 77, 84). Proinflammatory mediators, including cytokines such as TNF-α and IL-13, increase agonist-induced [Ca²⁺]_cyt responses in ASM cells (74, 75, 84). The role of mitochondria in shaping these signals has not been examined. Furthermore, little is known about the effects of inflammation on mitochondrial buffering of [Ca²⁺]_cyt in human ASM. In other smooth muscle types, inflammatory cytokines disrupt mitochondria-SR interactions and Ca²⁺ buffering (64, 65). Studies have also shown that short-term exposure to TNF-α temporary increases basal mitochondrial Ca²⁺ in the lung endothelium, leading to an increase of mitochondrial ROS generation (19, 51, 68). A recent study (13) has suggested that H₂O₂ may modulate both [Ca²⁺]_mito and [Ca²⁺]_cyt in rat ASM cells. Accordingly, we hypothesized that mitochondrial Ca²⁺ transport modulates the [Ca²⁺]_cyt increase induced by agonist stimulation in human ASM cells and that inflammation affects mitochondrial Ca²⁺ transport. As compartmentalization may play an important role in the Ca²⁺ buffering capacity of mitochondria, mitochondrial Ca²⁺ responses were investigated in perinuclear area and distal (perimembranous) regions of cells. The role of mitochondria transport was examined using inhibitors of the mitochondrial uniporter (MCU) (7, 21, 30–32) and mitochondrial Na⁺/Ca²⁺ exchanger (NCX) (22, 47, 56, 83) and by simultaneously monitoring [Ca²⁺]_cyt and mitochondrial Ca²⁺ concentration ([Ca²⁺]_mito) in the presence of TNF-α or IL-13.

MATERIALS AND METHODS

Isolation of human ASM cells.

These techniques have been previously described (58, 78). Briefly, ASM cells were enzymatically dissociated from third- to sixth-generation bronchi of lung samples incidental to patient surgery at Mayo Clinic (approved and considered exempt by the Institutional Review Board of Mayo Clinic). Normal lung areas were identified, and dissected tissues were placed in ice-cold HBSS (Invitrogen, Carlsbad, CA) supplemented with 10 mM HEPES and 2 mM Ca²⁺. Under light microscopy, bronchioles were freed of cartilage, the epithelium, and surrounding tissues, and ASM tissue was dissected out. Samples were minced, and cells were dissociated using papain and collagenase with ovomucoid/albumin separation as per the manufacturer's instructions (Worthington Biochemical, Lakewood, NJ). Cell pellets were resuspended and plated in sterile culture flasks or eight-well glass-bottomed Lab-Tek chambers (Nalge Nunc, Rochester, NY). Cells were maintained at 37°C (5% CO₂-95% air) using phenol red-free DMEM-F-12 medium (Invitrogen) supplemented with 10% FBS until ∼80% confluent. Before experiments, cells were washed in PBC (Invitrogen), and medium was changed to DMEM-F-12 lacking serum for 48 h. All experiments were performed in cells from passages 1–3 of subculture. Periodic assessment of the ASM phenotype (smooth muscle actin and myosin and agonist receptors as well as lack of fibroblast markers) was performed. Cell viability was tested by the exclusion of trypan blue.

Real-time fluorescence imaging.

ASM cells plated on eight-well Lab-Tek chambers were incubated in 2.5 μM fluo-3 AM (Invitrogen, [Ca²⁺]_cyt indicator, excitation: 488 nm and emission: 510 nm) with 2.5 μM rhod-2 AM (Invitrogen, [Ca²⁺]_mito indicator, excitation: 568 nm and emission: 590 nm) for 40 min at room temperature followed by an extensive wash in dye-free solution. In control experiments, ASM cells were loaded with either fluo-3 AM or rhod-2 AM. In other experiments, cells were loaded with 2.5 μM rhod-2 AM for 40 min and then with 500 nM MitoTracker green (Invitrogen) for 5 min at room temperature to verify that rhod-2 was present specifically in mitochondria. To simultaneously visualize the SR and mitochondria, ASM cells were loaded with 1 μM BODIPY-FL-thapsigargin (endoplasmic reticulum Ca²⁺ pumps, Invitrogen) for 30 min and then with 500 nM MitoTracker red CMXRos (Invitrogen) for 5 min at room temperature.

The techniques for [Ca²⁺]_cyt imaging of human ASM cells have been previously described (58, 73, 78). Cells were visualized using an epifluorescence imaging system (MetaFluor, Universal Imaging, Downingtown, PA) on a Nikon Diaphot inverted microscope (Fryer Instruments, Edina, MN) with a ×40/1.3 numerical aperture oil-immersion lens (or a ×60/1.3 numerical aperture oil-immersion lens) and a 12-bit Photometric Cascade digital camera system (Roper Scientific, Tucson, AZ). Images were collected at 1 Hz. Cells were initially perfused with 2 mM Ca²⁺ HBSS, and baseline fluorescence was established. Changes in fluorescence were analyzed by selecting individual software-defined regions of interest (ROIs; 3 × 3 pixels) delineating mitochondrial versus cytoplasmic areas.

Calibrations.

We performed empirical calibrations of fluorescence values for fluo-3 AM and rhod-2 AM to be able to compare [Ca²⁺]_cyt and [Ca²⁺]_mito. Although these are not ratiometric dyes, Ca²⁺ can be estimated by essentially mapping fluorescence intensity to known Ca²⁺ levels. Calibrations were performed with ASM cells loaded with either fluo-3 AM or rhod-2 AM (2.5 μM) perfused successively with 11 K₂EGTA/CaEGTA buffers (free Ca²⁺ ranging from 0 to 39 μM, Ca²⁺ calibration buffer kit no. 1, Invitrogen). Gray-level values of fluorescence intensities at various settings used with the epifluorescence imaging system and for each Ca²⁺ concentration buffer were converted to a family of calibration curves as previously described (54, 59, 80).

Statistical analysis.

Four bronchial samples were used to obtain ASM cells. [Ca²⁺]_cyt responses of ∼6–8 cells/visual field and [Ca²⁺]_mito responses of at least 40 mitochondria/cell were measured, and all experiments were repeated at least 3 times for each sample obtained; n denotes the number of cells that were analyzed. In experiments where Ca²⁺ responses were compared in the presence or absence of specific drugs, a paired t-test was used, whereas population studies were compared using an unpaired t-test or one-way ANOVA with repeated measures as appropriate. A Bonferroni correction was applied for multiple comparisons. Statistical significance was established at P < 0.05. All values are expressed as means ± SE.

RESULTS

Dye loading of ASM cells.

To simultaneously measure [Ca²⁺]_cyt and [Ca²⁺]_mito, human ASM cells were loaded with fluo-3 AM and rhod-2 AM (Fig. 1). The distribution of fluo-3 AM was diffuse, whereas rhod-2 AM was localized to punctate areas typical of mitochondria. Ca²⁺ responses were measured in cells that exhibited a fusiform shape with a centrally located nucleus typical of healthy human ASM cells and showing homogenous loading with fluo-3 AM and punctate loading of rhod-2 AM (Fig. 1, A–C). In some experiments, cells were loaded with rhod-2 and MitoTracker green to confirm that the pattern of rhod-2 staining was restricted to mitochondria (Fig. 1, D–F). In a second set of experiments, ASM cells were loaded with BODIPY-FL-thapsigargin and then MitoTracker red CMXRos (Fig. 1, G–I). The pattern of MitoTracker red CMXRos staining was similar to that of rhod-2 and MitoTracker green (not shown). We observed that mitochondria are present in large numbers within human ASM cells, primarily around the nucleus, where they form a dense interconnected network. The SR Ca²⁺ pumps visualized with BODIPY-FL-thapsigargin appeared to be highly localized around the nucleus (Fig. 1G), where a dense network of mitochondria was also observed (Fig. 1, H and I). BODIPY-FL-thapsigargin fluorescence was far less intense elsewhere within the cells. Image analysis of 10 ASM cells revealed that the network of mitochondria juxtaposed with the SR Ca²⁺ pumps was within 40 μm of the centroid of the nucleus. These mitochondria were defined as perinuclear for subsequent analyses. Mitochondria located >50 μm from the centroid of the nucleus and within 10 μm of the cell membrane were defined as distal for subsequent analyses.

Fig. 1.

Dye loading of human airway smooth muscle (ASM) cells. A and B: fluorescence images of human ASM cells loaded with 2.5 μM fluo-3 AM (A) and 2.5 μM rhod-2 AM (B) to estimate cytosolic ([Ca²⁺]_cyt) and mitochondrial Ca²⁺ concentration ([Ca²⁺]_mito), respectively. C: relative distribution of fluo-3 AM and rhod-2 AM fluorescence with the merged image of the two fluorescence images. D and E: in a separate set of experiments, human ASM cells were loaded simultaneously with rhod-2 AM (D) and then with 500 nM MitoTracker green (E) for 5 min to visualize mitochondria and showed consistent and matching patterns of mitochondria in the perinuclear areas and more distal (potentially perimembranous) areas of cells. F: an overlay of the two fluorescence images confirmed that the pattern of rhod-2 staining was mainly within mitochondria. G and H: in additional experiments, ASM cells were loaded with 1 μM BODIPY-FL-thapsigargin (G; endoplasmic reticulum Ca²⁺ pumps) and then with 500 nM MitoTracker red CMXRos (H). I: overlays of the two fluorescence images along with image analysis of the mitochondrial network were used to define the perinuclear regions. N, nucleus. Bars = 10 μm.

As shown in Fig. 1, B and E, some weaker rhod-2 AM fluorescence could be observed in the cytosol. Therefore, it was important to verify that the rhod-2 readings within the ROIs did not reflect cytosolic Ca²⁺. ASM cells were loaded with MitoTracker green and rhod-2 AM (Fig. 2, A–C), and ROIs (3 × 3 pixels) used to measure change in rhod-2 AM fluorescence were placed within the mitochondria (identified with MitoTracker green), in the cytosol or near mitochondria where some weaker rhod-2 AM background fluorescence was observed (Fig. 2C). Transient increases in rhod-2 fluorescence induced by 10 μM histamine completely overlapped with regions of MitoTracker green fluorescence (region 1 in Fig. 2C). In ROIs outside of the mitochondria (Mitotracker green fluorescence), histamine did not induce a rhod-2 fluorescence response (e.g., regions 2 and 3 in Fig. 2C). These data confirmed that the responses obtained with rhod-2 AM were specific to change in [Ca²⁺]_mito and not [Ca²⁺]_cyt (Fig. 2D).

Fig. 2.

Transient increases in rhod-2 fluorescence are localized within mitochondria. A and B: fluorescence images of human ASM cells loaded with MitoTracker green (A) and rhod-2 AM (B). C: overlay of the two fluorescence images with examples of regions of interest (ROIs) chosen within the mitochondria (region 1), in the cytosol (region 2), and near mitochondria (region 3), where some background fluorescence could be detected. Bar = 10 μm. D: measurement of rhod-2 fluorescence expressed as [Ca²⁺] (in nM) in the ROIs shown in C. A transient increase in [Ca²⁺] in response to histamine (His; 10 μM) was observed for the ROI chosen within the mitochondria (region 1, solid line) and not for the ROIs placed in either the cytosol (region 2, dotted line) or near the mitochondria (region 3, dashed line). E and F: MitoTracker green-loaded cells excited at 488 nm showed significant specific fluorescence at 501 nm, corresponding to MitoTracker green (E), whereas excitation of the cells at 568 nm (rhod-2 AM excitation wavelength) did not show significant (above background) fluorescence at 590 nm (F). G and H: similarly, excitation of ASM cells loaded with 2.5 μM rhod-2 AM at 488 nm (G; Mitotracker green excitation wavelength) did not result in any nonspecific fluorescence at 501 nm (above background), whereas excitation at 568 nm resulted in specific fluorescence at 590 nm, corresponding to rhod-2 AM (H). Ex, excitation; Em, emission.

To confirm the specificity of our fluorescent imaging, ASM cells were loaded with only MitoTracker green or rhod-2 AM. MitoTracker green-loaded cells excited at 488 nm showed specific fluorescence at 501 nm, corresponding to MitoTracker green, whereas excitation at 568 nm (the rhod-2 AM excitation wavelength) did not show significant (above background) specific fluorescence at 590 nm (Fig. 2, E and F). Similarly, the excitation of rhod-2 AM-loaded cells with 488 nm (the Mitotracker green excitation wavelength) did not result in any nonspecific fluorescence at 501 nm (above background; Fig. 2G), whereas excitation at 568 nm resulted in specific fluorescence at 590 nm, corresponding to rhod-2 AM.

In selected experiments, ASM cells were loaded with either fluo-3 AM or rhod-2 AM, and the responses to an agonist obtained for each Ca²⁺ indicator were compared with those obtained with simultaneous loading of the cells. No differences were found in terms of delays, amplitudes, etc., indicating a lack of interference between the two dyes. The ROIs (3 × 3 pixels) were placed into the cytosol to measure changes in [Ca²⁺]_cyt (Fig. 3, A and B) or were placed at the center of well-delineated mitochondrial groups to measure changes in [Ca²⁺]_mito (Fig. 3, C and D).

Fig. 3.

Measurement of [Ca²⁺]_cyt and [Ca²⁺]_mito in human ASM cells. A: fluorescence image of human ASM cells loaded with fluo-3 AM with an example of the ROI used to measure [Ca²⁺]_cyt. B: 10 μM His first induced a rise in [Ca²⁺]_cyt. C: the same ROI was used to measure the change in [Ca²⁺]_mito with the same cells simultaneously loaded with rhod-2 AM. Examples of the ROIs used to measure [Ca²⁺]_mito in the perinuclear and distal regions are also shown. D: traces obtained from the ROIs in C showed that the changes in fluorescence only occur when the ROIs are placed within the mitochondria in either the distal (dotted line) or perinuclear (dashed line) region of the cells. E and F: an overlay of the two fluorescence images (E) along with a time-extended view of [Ca²⁺] responses (F) from the graphs in B and D showed a significant delay between [Ca²⁺]_cyt and [Ca²⁺]_mito responses. Histamine (His) was added at 30 s.

[Ca²⁺]_cyt and [Ca²⁺]_mito responses to histamine.

In unstimulated cells, basal [Ca²⁺]_cyt (181 ± 39 nM) and [Ca²⁺]_mito measured either from mitochondria localized in the distal or perinuclear regions of the cells (174 ± 36 and 166 ± 30 nM, respectively, n = 31) were not significantly different (P > 0.05; Figs. 3, B and D, and 4A). In the presence of 2 mM extracellular Ca²⁺, exposure to 10 μM histamine induced a characteristic large increase in [Ca²⁺]_cyt, which was followed by a rapid increase in [Ca²⁺]_mito (Fig. 3, B and D). When the ROIs to measure changes in rhod-2 fluorescence were placed within the cytosol, we did not observe any response to histamine stimulation (Fig. 3D).

Fig. 4.

Comparison of [Ca²⁺]_cyt and [Ca²⁺]_mito in human ASM. A: basal [Ca²⁺]_cyt and [Ca²⁺]_mito from mitochondria in the distal versus perinuclear regions of cells and the His effect on the amplitude of [Ca²⁺] responses (peak amplitude, n = 31). Values are means ± SE. B: effect of inhibited mitochondrial transport on the delay (in s) between [Ca²⁺]_cyt and [Ca²⁺]_mito. When measured at the initiation of either response, the delay between [Ca²⁺]_cyt and [Ca²⁺]_mito was estimated at 11 ± 3 and 9 ± 3 s for the mitochondria localized in the perinuclear or distal regions of the cells, respectively. Inhibition of the mitochondrial uniporter (MCU) with Ru360 increased this delay for both mitochondria in the perinuclear regions (15 ± 2 s) and distal regions of the cells (16 ± 3 s). CGP-37157 (CGP) had no effect on the delays between [Ca²⁺]_cyt and [Ca²⁺]_mito. *Significant effect (P < 0.05).

The [Ca²⁺]_mito responses were delayed by 10 ± 3 s from [Ca²⁺]_cyt responses when measured at their initiations (P < 0.05, n = 31; Figs. 3, E and F, and 4B). This observed delay was another strong indication that we measured [Ca²⁺]_mito and not [Ca²⁺]_cyt. After a peak, [Ca²⁺]_cyt rapidly decreased toward baseline, whereas the decrease in [Ca²⁺]_mito was much slower (Fig. 3, B and D). A significant delay between [Ca²⁺]_cyt and [Ca²⁺]_mito peaks was estimated at 12 ± 5 s (P < 0.05, n = 31; Fig. 3F). No significant delay was found between the peaks of [Ca²⁺]_mito responses from distal or perinuclear regions (Fig. 3D). The amplitude of [Ca²⁺]_cyt and [Ca²⁺]_mito responses induced by histamine were not significantly different (P > 0.05; Fig. 4A). The amplitude of [Ca²⁺]_mito in the distal region was slightly lower than for the perinuclear region; however, there was substantially greater variability in the [Ca²⁺]_mito responses in the distal regions (Fig. 4A). The fall of both [Ca²⁺] responses (starting at the peak until the return to the basal Ca²⁺ level) was measured by calculating the areas under the curves using SigmaPlot (Systat Software, Chicago, IL) and were significantly higher for [Ca²⁺]_mito responses than for [Ca²⁺]_cyt responses (69,041 ± 1,025 and 51,076 ± 1,236 nM³, respectively, P < 0.05, n = 31).

Effect of inhibited mitochondrial Ca²⁺ transport on [Ca²⁺]_cyt.

To test the hypothesis that mitochondrial Ca²⁺ transport regulates [Ca²⁺]_cyt, ASM cells loaded with fluo-3 AM and rhod-2 AM were perfused with 1 μM of the mitochondrial NCX inhibitor 7-chloro-5-(2-chlorophenyl)-1,5-dihydro-4,1-benzothiazepine-2(3H)-one (CGP-37157; Calbiochem, San Diego, CA) (17, 18). CGP-37157 did not significantly influence baseline [Ca²⁺]_cyt or [Ca²⁺]_mito (Fig. 5). However, compared with control cells, the amplitude of the [Ca²⁺]_cyt response to 10 μM histamine was almost doubled in ASM cells that were exposed to CGP-37157 (∼195% of control, P < 0.05, n = 48; Figs. 5A and 6A), and [Ca²⁺]_mito from perinuclear regions was also increased (∼152% of control, P < 0.05, n = 48; Figs. 5B and 6B). CGP-37157 did not increase the delay between [Ca²⁺]_cyt and [Ca²⁺]_mito responses (Fig. 4B). In contrast, CGP-37157 did not significantly influence the amplitude of [Ca²⁺]_mito responses to histamine in distal regions of ASM cells (Figs. 5C and 6B). In addition, CGP-37157 significantly increased the rate of fall of the cytosolic Ca²⁺ response to histamine (∼60% of control), whereas the rate of fall of [Ca²⁺]_mito responses from perinuclear or distal regions was decreased (∼20%, P < 0.05, n = 48; Figs. 5 and 6C).

Fig. 5.

Effect of inhibited mitochondrial transport on [Ca²⁺]_cyt and [Ca²⁺]_mito. ASM cells loaded with fluo-3 AM and rhod-2 AM were exposed to 10 μM His. A: compared with control cells (solid line), inhibition of MCU with Ru360 (dotted line) increased [Ca²⁺]_cyt induced by His (10 μM). A similar result was obtained when the mitochondrial Na⁺/Ca²⁺ exchanger (NCX) was inhibited with CGP (dashed line). B and C: compared with control cells (solid line), Ru360 (dotted line) decreased [Ca²⁺]_mito induced by His (10 μM) in the perinuclear regions (B) or distal regions (C) of the cells. CGP (dashed line) increased [Ca²⁺]_mito in the perinuclear regions (B) but not in the distal regions (C) of the cells.

Fig. 6.

A–C: summary of the effects of inhibited mitochondrial transport on the amplitude of the [Ca²⁺]_cyt response (expressed as a percentage of control; A), the amplitude of the [Ca²⁺]_mito amplitude (B), and the rate of fall (C; expressed as a percentage of control and corresponding to the areas under the curves, starting at the peak until the return to the basal Ca²⁺ level). Ru360 did not affect the decay of the [Ca²⁺]_cyt responses but decreased the associated rate of fall of [Ca²⁺]_mito for both the perinuclear and distal regions of the cells. CGP increased the rate of fall of [Ca²⁺]_cyt responses, which was accompanied by a decrease in the decay of [Ca²⁺]_mito responses. Values are means ± SE. *Significant effect (P < 0.05).

In a second set of experiments, MCU was inhibited with 1 μM Ru360 (Sigma-Aldrich, St. Louis, MO) (18, 43), which also did not significantly influence baseline [Ca²⁺]_cyt or [Ca²⁺]_mito (Fig. 5). Compared with control ASM cells, Ru360 significantly increased the amplitude of [Ca²⁺]_cyt responses induced by histamine (∼169% of control; Figs. 5A and 6A). The corresponding amplitude of [Ca²⁺]_mito responses from distal, and to a lesser extent perinuclear, regions were decreased (∼40–20% of control, n = 48, P < 0.05; Figs. 5, B and C, and 6B). We also observed that Ru360 increased the delay between [Ca²⁺]_cyt and [Ca²⁺]_mito responses (measured at the initiation of the responses) by ∼5 ± 2 s, for both the perinuclear regions and distal regions (P < 0.05; Fig. 4B). Ru360 did not change the fall of the cytosolic Ca²⁺ response to histamine, but the fall of [Ca²⁺]_mito response from perinuclear or distal regions was decreased (∼20%, P < 0.05, n = 48; Fig. 6C).

Effect of TNF-α and inhibited mitochondrial Ca²⁺ transport on [Ca²⁺]_cyt.

ASM cells were incubated for 48 h with TNF-α (20 ng/ml) (72) and then loaded with fluo-3 AM and rhod-2 AM. Inhibitors such as CGP-37157 (mitochondrial NCX inhibitor) or Ru360 (MCU inhibitor) were subsequently added. Basal [Ca²⁺]_cyt was moderately increased after the exposure to TNF-α (∼50 nM, P < 0.05, n = 40; Fig. 7A). In contrast, basal [Ca²⁺]_mito was unaffected by TNF-α (Fig. 7, B and C). Consistent with previous observations (74, 75, 84), TNF-α significantly increased the amplitude of the [Ca²⁺]_cyt response to histamine (10 μM) compared with the vehicle control (∼330% of control, P < 0.05, n = 40; Figs. 7A and 8A). In contrast, the amplitude of the [Ca²⁺]_mito responses was substantially smaller in TNF-α-exposed cells. However, the effect of TNF-α was only observed in mitochondria localized to the perinuclear region of the cells (∼60% of control, P < 0.05, n = 32; Figs. 7, B and C, and 8B). CGP-37157 slightly reduced the effect of TNF-α on the [Ca²⁺]_cyt response to histamine (∼277% of the amplitude of control, P < 0.05, n = 35; Figs. 7A and 8A), and this was correlated with a corresponding increase in [Ca²⁺]_mito within the perinuclear region (∼155% of the amplitude of control, P < 0.05, n = 37; Figs. 7, B and C, and 8B). As in control cells, CGP-37157 did not influence the amplitude of [Ca²⁺]_mito from the distal regions of ASM cells treated with TNF-α. In contrast to CGP-37157, Ru360 had no effect on the increase of [Ca²⁺]_cyt (Figs. 7A and 8A) or the decrease of the [Ca²⁺]_mito response to histamine (10 μM) induced by TNF-α (Figs. 7, B and C, and 8B). The delay between [Ca²⁺]_cyt and [Ca²⁺]_mito responses was not significantly changed by TNF-α and the subsequent inhibition of MCU or NCX (P > 0.05, n = 37). The fall of [Ca²⁺]_cyt was increased by TNF-α (∼40–50% of control, P < 0.05, n = 37; Fig. 8C). Ru360 or CGP-37157 did not significantly further increase the fall of [Ca²⁺]_cyt, which was found to be more variable than for control cells. TNF-α decreased the fall of [Ca²⁺]_mito from perinuclear regions (∼20% of control, P < 0.05, n = 37; Fig. 8C), which was, in contrast, significantly increased by CGP-37157 (∼40% of control, P < 0.05, n = 37; Fig. 8C). TNF-α with or without CGP-37157 did not change the fall of [Ca²⁺]_mito from distal regions. Ru360 did not affect the fall of [Ca²⁺]_mito (P > 0.05; Fig. 8C).

Fig. 7.

Effect of the proinflammatory cytokine TNF-α and inhibited mitochondrial transport on [Ca²⁺]_cyt and [Ca²⁺]_mito. ASM cells incubated for 48 h with TNF-α and then loaded with fluo-3 AM and rhod-2 AM were exposed to 10 μM His. A: compared with the vehicle control, TNF-α increased [Ca²⁺]_cyt responses induced by His (10 μM). A similar result was obtained when MCU was inhibited with Ru360 (dotted line). Inhibition of mitochondrial NCX with CGP (dashed line) decreased [Ca²⁺]_cyt in TNF-α-exposed cells. B and C: TNF-α decreased [Ca²⁺]_mito responses to His measured from mitochondria in the perinuclear regions (B) but not from mitochondria in the distal regions of cells (C). Ru360 did not further decrease [Ca²⁺]_mito in TNF-α-exposed cells. In contrast, CGP significantly increased [Ca²⁺]_mito measured from mitochondria in the perinuclear regions of cells but not from mitochondria in the distal regions.

Fig. 8.

A and B: summary of the effect of TNF-α and inhibited mitochondrial transport on the amplitude of [Ca²⁺]_cyt (A) and [Ca²⁺]_mito (B) responses. C: the rate of fall expressed as a percentage of the vehicle control and corresponding to the areas under the curves (starting at the peak until the return to the basal Ca²⁺ level). TNF-α increased the decay of [Ca²⁺]_cyt and decreased the decay of [Ca²⁺]_mito responses from the distal regions of the cells. Ru360 increased the decay of [Ca²⁺]_cyt responses but not [Ca²⁺]_mito responses. Compared with the vehicle control, CGP increased the decay of [Ca²⁺]_cyt and [Ca²⁺]_mito responses from the perinuclear regions of the cells. Values are means ± SE. *Significant effect (P < 0.05).

Effect of IL-13 and inhibited mitochondrial Ca²⁺ transport on [Ca²⁺]_cyt.

ASM cells were incubated for 48 h with IL-13 (20 ng/ml) (72), and experimental protocols similar to those with TNF-α were conducted. IL-13 also induced an increase of the amplitude of the [Ca²⁺]_cyt response to histamine (10 μM) compared with the vehicle control, albeit not as substantial as with TNF-α (∼220% of control, P < 0.05, n = 20; Figs. 9A and 10A). Basal [Ca²⁺]_cyt and [Ca²⁺]_mito were not affected by IL-13 (Fig. 9, A–C). No significant change of the amplitude of the [Ca²⁺]_mito response to histamine was observed in the distal or perinuclear regions of the cells exposed to IL-13 (P > 0.05, n = 20; Fig. 9, B and C, and 10B). The delay between [Ca²⁺]_cyt and [Ca²⁺]_mito responses was also not significantly changed by IL-13 (P > 0.05, n = 20). However, the delays observed after the exposure to IL-13 or TNF-α were much variable than in control cells. Inhibition of MCU with Ru360 increased the [Ca²⁺]_cyt response to histamine, whereas inhibition of mitochondrial NCX with CGP-37157 reduced the effect of IL-13 (∼272% and ∼177% of control, respectively, P < 0.05, n = 20; Figs. 9A and 10A). In ASM cells treated with IL-13, Ru360 decrease the amplitude of the [Ca²⁺]_mito response to histamine in the distal regions (∼46% of control, P < 0.05, n = 23) and perinuclear regions of the cells (∼38% of control, P < 0.05, n = 23; Figs. 9, B and C, and 10B). CGP-37157 increased the amplitude of the [Ca²⁺]_mito response to histamine in the perinuclear regions of the cells treated with IL-13 (∼165% of control, P < 0.05, n = 19; Figs. 9B and 10B). In the distal regions, the effect of CGP-37157 was found to be more variable and overall not significant (Fig. 10B). The fall of [Ca²⁺]_cyt was increased by IL-13 (∼50% of control, P < 0.05, n = 23), whereas the fall of [Ca²⁺]_mito was decreased (∼40% of control, P < 0.05, n = 23; Fig. 10C). Ru360 did not further increase the fall of [Ca²⁺]_cyt or [Ca²⁺]_mito compared with IL-13-treated cells. In contrast, CGP-37157 significantly increased the fall of [Ca²⁺]_mito (∼35% of control, P < 0.05, n = 23) but not the fall of [Ca²⁺]_cyt compared with IL-13 alone (Fig. 10C).

Fig. 9.

Effect of IL-13 and inhibited mitochondrial transport on [Ca²⁺]_cyt and [Ca²⁺]_mito. A: compared with the vehicle control, IL-13 increased [Ca²⁺]_cyt responses to His. Inhibition of MCU with Ru360 (dotted line) further increased [Ca²⁺]_cyt responses in IL-13-exposed cells, whereas inhibition of mitochondrial NCX with CGP (dashed line) decreased [Ca²⁺]_cyt responses. B and C: effects of IL-13 and inhibited mitochondrial transport on [Ca²⁺]_mito responses. Compared with the vehicle control, IL-13 had no effect on [Ca²⁺]_mito responses to His measured from both mitochondria in the perinuclear (B) and distal (C) regions of cells. Ru360 significantly decreased [Ca²⁺]_mito in IL-13-exposed cells, whereas CGP increased [Ca²⁺]_mito responses measured from mitochondria in the perinuclear regions but not from mitochondria in the distal regions of cells.

Fig. 10.

A and B: summary of the effects of IL-13 and inhibited mitochondrial transport on the amplitude of [Ca²⁺]_cyt (A) and [Ca²⁺]_mito (B) responses (expressed as a percentage of the vehicle control). Values are means ± SE. *Significant effect (P < 0.05). C: the rate of fall expressed as a percentage of the vehicle control and corresponding to the areas under the curves (starting at the peak until the return to the basal Ca²⁺ level). IL-13 increased the decay of [Ca²⁺]_cyt and [Ca²⁺]_mito responses from the distal regions of the cells, whereas it decreased the decay of [Ca²⁺]_mito responses from the perinuclear regions. Ru360 did not further increased the decay of [Ca²⁺]_cyt responses but decreased the decay of [Ca²⁺]_mito responses from both the perinuclear and distal regions of the cells. Compared with the vehicle control, CGP increased the decay of [Ca²⁺]_cyt and [Ca²⁺]_mito responses from the perinuclear regions of the cells. Values are means ± SE. *Significant effect (P < 0.05).

DISCUSSION

During agonist stimulation, mitochondrial Ca²⁺ uptake of cytosolic Ca²⁺ has been described in many cell types (13, 25, 33, 38, 48, 55, 56, 65). In this regard, spatial localization of mitochondria may play an important role. Mitochondria are often located near the SR or PM, and a common hypothesis is that mitochondria alter SR Ca²⁺ release and reuptake as well as PM Ca²⁺ influx/efflux (10, 13, 16, 25, 33, 38, 43, 48, 52, 60, 62, 63, 65, 70). Inflammation may further modulate some mitochondrial functions (19, 51, 68) and the relationship between mitochondria and the SR (64, 65). In the present study, we examined the relation between [Ca²⁺]_mito and [Ca²⁺]_cyt in human ASM cells during agonist stimulation. More importantly, we showed that inflammatory mediators, such as TNF-α, increase [Ca²⁺]_cyt by disrupting mitochondrial Ca²⁺ transport. Here, both mitochondrial NCX (mediating Ca²⁺ release in excitable cells) and MCU (mediating Ca²⁺ uptake) appear to be involved. However, an important distinction is that the contribution of these mechanisms may vary depending on the perinuclear versus more distal localization of mitochondria.

Under baseline [Ca²⁺]_cyt conditions, mitochondria have relatively “low affinity” for Ca²⁺ (32, 48, 64, 69). In accordance, we found that inhibition of mitochondrial Ca²⁺ regulatory mechanisms, such as mitochondrial NCX with CGP-37157 or MCU with Ru360, did not influence [Ca²⁺]_cyt or [Ca²⁺]_mito in unstimulated ASM cells. After agonist stimulation, rapid increases in [Ca²⁺]_cyt appeared to be buffered by mitochondria via the activation of both mitochondrial NCX and MCU. While inhibition of mitochondrial Ca²⁺ transport was expected to modulate [Ca²⁺]_mito, the changes in [Ca²⁺]_cyt support the idea that mitochondria contribute to [Ca²⁺]_cyt regulation in human ASM cells. Here, we observed a significant delay between the initiation and peak of increase in [Ca²⁺]_cyt versus [Ca²⁺]_mito. Inhibition of mitochondrial Ca²⁺ uptake with Ru360 but not mitochondrial Ca²⁺ release modulate these delays, suggesting that mitochondrial Ca²⁺ buffering is involved in shaping [Ca²⁺]_cyt responses, as predicted by some theoretical models of Ca²⁺ signaling in smooth muscle cells (66, 67).

Cytosolic Ca²⁺ is enhanced in airway diseases such as asthma, and it has been proposed that inflammation is a key factor. We and others (2–4, 61, 71, 74, 75, 77, 84) have shown that airway inflammation increases airway contractility, and this is correlated with an increase in agonist-induced [Ca²⁺]_cyt responses. Several [Ca²⁺]_cyt regulatory mechanisms have been reported to be modulated by airway inflammation, but the role of mitochondrial Ca²⁺ buffering has not been previously examined. In this study, we showed that TNF-α, which increases [Ca²⁺]_cyt, concurrently decreases [Ca²⁺]_mito by inhibiting MCU. In contrast, IL-13, while also enhancing the [Ca²⁺]_cyt response to histamine, does not alter [Ca²⁺]_mito substantially. These novel data suggest an important mechanism by which cytokines may increase [Ca²⁺]_cyt and thus enhance contractility: impaired mitochondrial Ca²⁺ buffering. There is currently very little information in this regard. A previous study (68) in the lung endothelium found that a 10-min infusion of TNF-α increases basal [Ca²⁺]_mito for up to 1 h. This was correlated with an increase of mitochondrial H₂O₂. However, the effects on [Ca²⁺]_cyt or changes in [Ca²⁺]_mito after a long exposure to TNF-α were not examined. In our study, after 48 h of exposure to TNF-α, we observed a moderate increase of basal [Ca²⁺]_cyt, but basal [Ca²⁺]_mito was unaffected. IL-13 did not change basal [Ca²⁺]_cyt or [Ca²⁺]_mito. However, the responses to an agonist were clearly affected, indicating that altered mitochondrial Ca²⁺ buffering may modulate cytokine effects in the short term (agonist responses) as well as in the long term (basal Ca²⁺ with potential downstream genomic effects). Here, a number of [Ca²⁺]_cyt regulatory mechanisms may be involved in their interactions with mitochondria. PM NCX, sarco(endo)plasmic reticulum Ca²⁺-ATPase (SERCA), and SOCE have been found to be affected by both cytokines (26, 72, 74, 84), which, in turn, may affect [Ca²⁺]_mito, but their contributions remain to be determined. Recent studies (13, 68) have indicated that inflammation could increase H₂O₂ production by mitochondria, which, in turn, could affect both [Ca²⁺]_cyt and [Ca²⁺]_mito. The role of inflammation on mitochondrial Ca²⁺ uptake and Ca²⁺ release is not clear. In our study, we found that Ru360 did not further decrease [Ca²⁺]_mito in ASM cells treated with TNF-α, suggesting that MCU was inhibited. An increase of mitochondrial Ca²⁺ uptake leads to ROS generation and ATP production, and high mitochondrial [Ca²⁺] is detrimental to mitochondrial functions. In our study, it was not clear if after 48 h of exposure to TNF-α the MCU is inhibited to prevent a high level of mitochondrial H₂O₂ or whether TNF-α is involved in regulating the energetic demand of the cells. Potential downstream genomic effects also remain to be explored, especially in ASM.

Mitochondrial Ca²⁺ transport is activated by large [Ca²⁺]_cyt changes in the micromolar range (32, 48, 64, 69). Given that global [Ca²⁺]_cyt elevations in most tissues are only in the low micromolar range, local [Ca²⁺]_cyt levels in the vicinity of the SR or PM must reach high enough levels to activate mitochondrial Ca²⁺ buffering (27, 53, 56, 57). A recent study (27) estimated that [Ca²⁺] is 5–10 times higher in these microdomains than elsewhere in the cytosol. Depending on the [Ca²⁺]_mito level, subsequent Ca²⁺ release back into the cytosol may be fast or slow (62). In non-ASM tissues, mitochondria respond to local [Ca²⁺]_cyt in the vicinity of open Ca²⁺ channels (20, 37, 62). In vascular and gastrointestinal smooth muscles, Ca²⁺ release via inositol 1,4,5-triphosphate (IP₃) receptor (IP₃R) channels (83) and ryanodine receptor (RyR) channels (83) results in large [Ca²⁺]_mito elevations, as it appears to do in ASM cells based on our results. Accordingly, the physical proximity of mitochondria to the SR and PM may be important in terms of buffering SR Ca²⁺ release or modulating influx and might explain some of the differences that we observed between [Ca²⁺]_mito from distal or perinuclear regions.

Consistent with an expected loss of mitochondrial Ca²⁺ extrusion and mitochondrial Ca²⁺ uptake (8, 10, 62), inhibition of mitochondrial NCX with CGP-37157 increased [Ca²⁺]_mito, whereas inhibition of MCU with Ru360 decreased [Ca²⁺]_mito. However, an interesting and novel observation in our study was the differential roles of mitochondrial NCX versus MCU in the perinuclear versus distal regions of ASM cells. In the distal regions, CGP-37157 had no significant effect, whereas Ru360 consistently decreased [Ca²⁺]_mito. Here, based on the typical fusiform shape of ASM cells, we propose that the distal mitochondria may be close to the PM and thus affected by and, in turn, influence Ca²⁺ influx. Accordingly, a role for the uniporter in buffering Ca²⁺ influx is likely to take precedence in these cellular areas. Mitochondrial Ca²⁺ uptake is also associated with increased ATP production. As the distal area of ASM cells is primarily associated with the contractile apparatus, the energetic demand might be higher than in the perinuclear area and favor mitochondrial Ca²⁺ uptake. In TNF-α-exposed cells, Ru360 did not further decrease [Ca²⁺]_mito, suggesting an inhibitory effect of TNF-α on MCU. These differences between the two regions remain to be further explored in the context of cytokine effects on energy availability for the contractile apparatus in the distal areas of ASM cells versus other effects, such as in the SR or nuclear Ca²⁺ reuptake in the perinuclear area.

Impaired mitochondrial Ca²⁺ buffering has been reported to enhance [Ca²⁺]_cyt release in different cell types (8, 10, 14, 62). The regulation of [Ca²⁺]_cyt by mitochondria may reflect the Ca²⁺ buffering capacity (either global or localized) of the mitochondria or/and by indirect effects of some mitochondrial function. The physical proximity of mitochondria to the SR had led several recent studies (11, 34, 36, 66) to examine possible functional interactions between the SR and mitochondria. A common hypothesis is that the mitochondria alter SR Ca²⁺ refilling by competing with SERCA pumps for the removal of cytosolic Ca²⁺ (49). In smooth muscle (as in other muscles), SERCA is the mechanism for filling the SR, with two Ca²⁺ pumped back for every ATP consumed against a gradient of ∼1 mM in the SR lumen (42). SERCA activity is directly proportional to local [Ca²⁺]_cyt. By affecting local [Ca²⁺]_cyt as well as ATP, mitochondria may affect SERCA activity. It has been proposed that mitochondria could generate microdomains of low Ca²⁺ next to SERCA pumps, and this ability was found to be dependent on mitochondrial Ca²⁺ uptake and not on mitochondrial Ca²⁺ release (29, 40). Our findings of inhibitory effects of Ru360 on [Ca²⁺]_cyt in the perinuclear areas is consistent with the above scenario. Whether SR luminal Ca²⁺ is actually affected in human ASM cells remains to be examined.

Mitochondrial Ca²⁺ fluxes can also regulate IP₃R and RyR channel activities and thus SR Ca²⁺ release (6, 11, 12, 14, 15, 34, 36). Here, the effects appear to be cell type specific, with both inhibition and promotion of SR Ca²⁺ release being reported (6, 11, 12, 15). Cheranov et al. (14) demonstrated that CGP-37157 increases the driving force for [Ca²⁺]_mito release and elevates local [Ca²⁺]_cyt near IP₃R channels or RyR channels, thereby affecting SR Ca²⁺ release. Such a scenario may explain the observed increase in [Ca²⁺]_cyt in the presence of CGP-37157, especially in the perinuclear areas, where the SR is likely to be located. Additionally, H₂O₂ production, which is dependent on mitochondrial Ca²⁺ uptake, may be induced by histamine and affect both the IP₃ signaling cascade and SOCE (13) as well as ATP production. ATP is also an important effector of IP₃Rs and is dependent on [Ca²⁺]_mito and ROS production. The potential effect of H₂O₂, notably on the fall of the [Ca²⁺]_cyt response, might explain the slow decay of [Ca²⁺]_cyt observed with CGP-37157. We cannot exclude a nonspecific effect of CGP-37157, but this is generally observed with a concentration of >10 μM (for a review, see Ref. 18), which was not used here. Smaller doses of CGP-37157 had no effect on both [Ca²⁺]_cyt or [Ca²⁺]_mito in our primary ASM cells (data not shown).

Recently, the molecular identities of MCU [MCU/mitochondrial Ca²⁺ uptake 1 (MICU1)] and NCX (also called NCLX) (7, 21, 22, 47) have been uncovered, and the development of molecular tools may help us to better understand the relation between MCU and NCX with other established Ca²⁺ regulatory mechanisms. In this regard, studies examining MICU1 and NCX have relied on the specificity of Ru360 and CGP-37157 (18). Accordingly, these pharmacological inhibitors form an appropriate first step toward examining [Ca²⁺]_mito in primary cells such as human ASM, which tend to be recalcitrant to RNA interference approaches targeting MICU1 and NCX (pilot studies).

In addition to SR Ca²⁺ release and reuptake, PM Ca²⁺ influx/efflux is a major component of [Ca²⁺]_cyt regulation in ASM. We (5, 75, 84) have previously demonstrated the importance of SOCE in ASM, triggered after SR Ca²⁺ depletion. It is important to recognize that the extent of Ca²⁺ influx via SOCE is not simply dependent on local PM Ca²⁺ gradients but is regulated by the state of SR repletion, [Ca²⁺]_cyt-dependent fast and slow inactivation, and even by other intracellular signaling cascades (e.g., diacylglycerol) (50). There is now evidence in non-ASM tissues that mitochondria can also influence SOCE (35, 44, 50). However, this effect may be indirect. By taking up Ca²⁺ released from the SR, mitochondrial Ca²⁺ buffering can enhance store depletion and hence maintain SOCE (48, 50). The available information suggests that the physical proximity of mitochondria to the PM may be important (50). A recent study (46) has shown that the inhibition of mitochondrial NCX is fundamental to SOCE regulation in endothelial cells. Whether these scenarios hold true in ASM cells has not been determined. Our finding that Ru360 reduces [Ca²⁺]_mito while elevating [Ca²⁺]_cyt is suggestive. However, a more interesting observation is that in the presence of TNF-α, Ru360 did not affect either [Ca²⁺]_cyt or [Ca²⁺]_mito. We (84) have previously shown that TNF-α increases SOCE. Therefore, we would have expected [Ca²⁺]_mito to be enhanced in the presence of TNF-α, allowing SOCE to be maintained. The reasons underlying this discrepancy are not clear but may involve cytokine-induced dysregulation of MCU itself.

In the present study, we found that the effects of TNF-α versus IL-13 on [Ca²⁺]_mito were substantially different, especially depending on the location of the mitochondria. The functional relevance of these differences remains to be determined. However, it is possible that although both cytokines, which are important in asthma, can robustly enhance [Ca²⁺]_cyt, the mechanisms by which this is achieved may differ, including mitochondrial Ca²⁺ regulation.

Overall, the present study demonstrates that mitochondria buffer [Ca²⁺]_cyt in human ASM cells and contribute to the modulation of [Ca²⁺]_cyt. Considering the importance of [Ca²⁺]_cyt regulatory mechanisms in enhancing [Ca²⁺]_cyt under conditions such as inflammation (relevant to diseases such as asthma), our study underlines the emerging importance of mitochondria, especially MCU, in inflammatory airway disease.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grants HL-74309 (to G. C. Sieck), HL-090595 (to C. M. Pabelick), HL-088029 (to Y. S. Prakash), and HL-056470 (to Y. S. Prakash and G. C. Sieck).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: P.F.D., B.Y., and M.A.T. performed experiments; P.F.D., B.Y., and M.A.T. analyzed data; P.F.D., C.M.P., Y.S.P., and G.C.S. interpreted results of experiments; P.F.D., Y.S.P., and G.C.S. prepared figures; P.F.D., C.M.P., Y.S.P., and G.C.S. drafted manuscript; P.F.D., Y.S.P., and G.C.S. edited and revised manuscript; C.M.P., Y.S.P., and G.C.S. conception and design of research; Y.S.P. and G.C.S. approved final version of manuscript.