Identification of Channels Promoting Calcium Spikes and Waves in HT1080 Tumor Cells

Cell Preparation.

HT1080 cells were incubated in DMEM supplied with 10% FCS and 1% antibiotics. Cells were passaged by detachment from culture flasks with a trypsin/EDTA solution (Sigma Chemical Co., St. Louis, MO) in PBS for 5 min at 37°C. For high-speed imaging and emission microfluorometry, cells were prelabeled with indo-1-AM at 5 μg/ml for 20 min at 37°C (29) .

Materials.

Tetramethylrhodamine isothiocyanate, FITC, BAPTA-AM, and indo-1-AM were obtained from Molecular Probes, Inc. (Eugene, OR). Fibronectin, verapamil, mibefradil, GdCl₃, and nicardipine were purchased from Sigma. EDTA was obtained from Fischer Scientific (Pittsburgh, PA). CAI was the generous gift of Dr. R. Schultz, Drug Synthesis & Chemistry Branch, Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis, National Cancer Institute (Bethesda, MD). A polyclonal rabbit anti-TRP-1 antibody was purchased from Chemicon International, Inc. (Temecula, CA).

Anti-T-type Ca²⁺ Channel Antibody.

Several previous workers have prepared polyclonal antibodies directed against the T-type Ca²⁺ channel (30 , 31) . The antibody was prepared as described by using a peptide corresponding to amino acids 1–22 of the NH₂-terminal region of human α1G subunit of the low-voltage T-type Ca²⁺ channel with the addition of a COOH-terminal cysteine: NH₂-MDEEEDGAGAEESGQPRSFMRL(C)-COOH (30) . The purified peptide was conjugated to keyhole limpet hemocyanin and then used to immunize rabbits (Bio-Synthesis, Inc., Lewisville, TX). The labeled IgG fraction was found to specifically label cells.

Cell Motility and Phagokinetic Tracks.

Cell motility was assessed by the Albrecht-Buehler method (32) as modified by Scott et al. (33) . Briefly, coverslips were dipped into a 0.17% gelatin solution and then coated with a colloidal gold layer. A suspension of 5000 cells/ml was treated with or without Ca²⁺ channel blockers and then seeded onto the gold-coated coverslips. After 22 h of incubation, the samples were fixed with 3.7% glutaraldehyde. The tracks produced by the cells were viewed by brightfield microscopy with a ×10 objective. Thirty randomly selected tracks were measured for each slide.

Cell Invasion.

In vitro tumor cell invasion was assessed using a Biocoat Matrigel invasion chamber (BD Biosciences, Bedford, MA) with cell culture inserts containing an 8-μm pore size membrane with a thin Matrigel basement membrane matrix as described previously (34) . One-half milliliter of cells (5 × 10⁴ cells/ml) in serum-free DMEM with or without Ca²⁺ channel blockers was added to the cell culture insert of a Biocoat Matrigel invasion chamber. Fibronectin (25 μg/ml) was added in the outer chamber as a chemoattractant. The cells were then incubated at 37°C in humidified 5% CO₂ conditions for 22 h. To quantitate tumor cell invasion, noninvading cells were removed from the top surface of the membrane by scrubbing gently with a cotton-tipped swab. The cells on the bottom surface of the membrane were fixed with Diff-Quik stain set (Dade Behring, Newark, DE) and counted to determine the number of cells that passed through the Matrigel and membrane layers.

Microfluorometry.

Quantitative microfluorometry was performed on single cells using a Peltier-cooled photomultiplier tube D104 system (Photon Technology, Lawrenceville, NJ) attached to a Zeiss Axiovert 35 (Carl Zeiss, New York, NY) fluorescence microscope (29) . A monochromator and fiber-optically coupled xenon lamp were controlled by FeliX software (Photon Technology). For indo-1-labeling experiments, the excitation was set to 350 nm (10-nm bandpass). Emission was detected using a 390LP dichroic reflector and 405DF43 emission filter. The photomultiplier tube output was plotted as a function of time.

Fluorescence Microscopy.

Fluorescence microscopy was performed with an Axiovert 135 fluorescence microscope with a high numerical aperture condenser, quartz objective, and an AttoArc mercury lamp (Zeiss). For immunfluorescence microscopy, cells were fixed with 3.7% glutaraldehyde for 20 min at room temperature. Samples were washed and then labeled with the first and second step antibodies. To minimize nonspecific binding, 1% BSA was included in these solutions. Images from multiple focal planes were collected using a z-scan apparatus (Veytek, Inc., Fairfield, IA) and cooled Retiga 1300 camera (Q-Imaging, British Columbia, Canada) or an intensified charge-coupled device camera (Hamamatsu, Hamamatsu City, Japan). Images were managed, processed, and deconvoluted using Supermicro Ultra 320 workstation with the software packages: Image-Pro, Microtome, Image-Scan, and Vox-Blast (Veytek). A narrow bandpass filter set with excitation at 485DF20 and emission at 530DF30 nm and a long-pass dichroic mirror at 510 nm were used. Representative sections of immunostained fluorescent samples were collected.

High-Speed Fluorescence Microscopy.

High-speed microscopy was performed using an Axiovert 135 fluorescence microscope with a quartz condenser, quartz objective, and an AttoArc mercury lamp (Zeiss). To detect fluorescence changes in the short wavelength emission region of indo-1, a 355HT15 exciter, 390LP dichroic reflector, and 405DF43 emission filter were used. To increase light collection efficiency, the microscope’s bottom port was used. This port was fiberoptically coupled to the input of an Acton-150 (Acton Instruments, Acton, MA) imaging spectrophotometer. The exit side was connected to a liquid N₂-cooled intensifier attached to a Peltier-cooled I-MAX-512 camera (∼−20°C; Princeton Instruments, Trenton, NJ; Refs. 34, 35, 36 ). A Gen-II tube was used to provide maximal efficiency in the violet-blue region of the spectrum (29) , where indo-1 emits. The camera was controlled by a high-speed Princeton ST-133 interface and Stanford Research Systems (Sunnyvale, CA) DG-535 delay gate generator (37) . A custom-built computer with dual 3.06 GHz Xenon processors with 1 MB onboard cache each, 3-Gb RAM, 3.2 Tb of hard drive space with 64 MB cache, and a RAID-5 hard drive management system was used. For experiments, 2 Gb of RAM was allocated as a RAM disk. Winspec-32 (version 2.5.14.1; Princeton Instruments) software was used with a PCI communication accelerator. To improve computer acquisition times, the size of the pixel detection array of the charge-coupled device (CCD) chip was adjusted using a Virtual-Chip plug-in. Winspec CPU calls were given system priority to enhance the instrument’s duty cycle. Data were acquired without reporting to the monitor to further improve system speed. Data capture used a software-allocated RAM disk.

Immunofluorescent Staining of T-Type Ca²⁺ Channels and TRP1 Cation Channels on HT1080 Cells.

In the present study, we test the hypothesis that LVA (T-type) and NVG channels participate in the signaling mechanism(s) necessary for HT1080 cell migration. Therefore, we sought to confirm that structural elements of these channels are expressed by HT1080 cells using immunofluorescence microscopy. HT1080 cells were stained with a rabbit anti-T-type Ca²⁺ channel antibody followed by a FITC-conjugated antirabbit IgG antibody. Stacks of fluorescence micrographs were obtained from individual cells followed by deconvolution analyses of the z-scans. Fig. 1, A and B ⇓ show differential interference contrast microscopy and fluorescence micrographs of an HT1080 cell. The T-type channel is located primarily about the cell’s circumference. Fig. 1, C and D ⇓ show a similar experiment using a rabbit anti-TRP1 antibody/FITC-conjugated antirabbit IgG staining. Although TRP1 staining was largely found at the cell periphery, the labeling was punctate in appearance. Labeling was often absent from pseudopods and may include a granule fraction near the plasma membrane. In addition, the z-scan/deconvolution image of Fig. 1D ⇓ suggests that minor labeling of the endoplasmic reticulum is also observed with the anti-TRP1 reagent. This is consistent with the fact that the fixation protocol results in permeabilization of the cell. Although previous studies have shown that tumor cells express T-type channel message, these studies now identify the structural expression of TRP1 and T-type channels by HT1080 cells.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Immunofluorescence staining of T-type Ca²⁺ channels and transient receptor potential 1 cation channels of HT1080 cells. Adherent HT1080 cells were fixed with glutaraldehyde and then labeled with rabbit anti-T-type Ca²⁺ channel polyclonal antibody (2 μg/ml; A and B) or rabbit antitransient receptor potential 1 cation channel polyclonal antibody (2 μg/ml; C and D), respectively, followed by labeling with a tetramethylrhodamine isothiocyanate-conjugated goat antirabbit antibody (1 μg/ml). Representative differential interference contrast microscopy (A and C) and deconvoluted z-scan fluorescence (B and D) images are shown. Although the T-type channels are largely distributed randomly at the cell periphery, the transient receptor potential 1 channels are found to have a more punctate distribution about the periphery (n = 3; ×980).

HT1080 Cells Exhibit Ca²⁺ Oscillations that Are Blocked by Mibefradil and Gd³⁺.

We next evaluated Ca²⁺ signal oscillations in HT1080 cells. Cells were labeled with indo-1-AM as described in “Materials and Methods.” HT1080 cells were suspended in HBSS and then incubated at 37°C for 20 min to allow the labeled cells to adhere to glass coverslips. Microfluorometry of untreated polarized cells showed a sustained series of Ca²⁺ spikes at 20-s intervals (Fig. 2 ⇓ , trace 1). To observe the longitudinal properties of these Ca²⁺ spikes, a long-time base was used in Fig. 2 ⇓ . However, to observe the temporal details of each Ca²⁺ spike, data were also acquired using a short-time base. Fig. 3 ⇓ , trace 1 shows a Ca²⁺ spike of an untreated HT1080 cell. This shows a typical Ca²⁺ spike profile consisting of a low-intensity shoulder on the leading edge, brief peak, and decay in indo-1 intensity. To confirm that Ca²⁺ entry participates in Ca²⁺ spikes, cells were treated with BAPTA-AM or EDTA, which are intracellular and extracellular Ca²⁺ chelators, respectively. Fig. 2 ⇓ , trace 2 shows an experiment identical to that of trace 1, except that BAPTA-AM was injected into the sample holder to give a final concentration of 50 μm. Within 3 min, the Ca²⁺ spike amplitude began to decrease, similar to that we have described previously in leukocytes (29) . Similarly, rapid removal of extracellular Ca²⁺ by the addition of 5 mm EDTA in Ca²⁺-free HBSS, to remove Ca²⁺ in the buffer, also blocked Ca²⁺ spikes (trace 3). Hence, extracellular Ca²⁺ participates in the formation of intracellular Ca²⁺ spikes. No changes in Ca²⁺ spike frequency or amplitude were observed when cells were treated with 15 μm SKF96365, 1 μm diltiazem, 4 μm verapamil, and 1 μm nicardipine (Fig. 2 ⇓ , traces 4–7, respectively). These concentrations were chosen as they correspond to the levels associated with specific Ca²⁺ channel blockage; e.g., nicardipine is a HVA (L-type) Ca²⁺ channel blocker effective in vitro at this concentration (38) . The NVG channel blocker and antimetastatic agent CAI, used at 10 μm for these in vitro studies (39) , reduced Ca²⁺ spikes to Ca²⁺ “bumps,” which was then followed by elimination of the Ca²⁺ signal (Fig. 2 ⇓ , trace 8). In the experiment of Fig. 2 ⇓ , trace 9, 1.5 μm mibefradil, a T-type channel blocker, was added at the time indicated by the arrow. Within ∼3–4 min, the Ca²⁺ spikes were replaced by very low-intensity bumps, which then disappeared. Similar results were obtained using Gd³⁺, a specific TRP (40) , and stretch-activated cation channel blocker (41) . Again, Ca²⁺ spikes diminished and then disappeared (Fig. 2 ⇓ , trace 10). The Ca²⁺ bumps were then evaluated using the short-time base settings (Fig. 3 ⇓ , trace 2). These experiments revealed that Ca²⁺ bumps closely resemble the intensity shoulder noted previously for untreated HT1080 cells (Fig. 3 ⇓ , trace 1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Effect of SKF96365, diltiazem, verapamil, nicardipine, carboxyamido-triazole (CAI), mibefradil, and Gd³⁺ on Ca²⁺ spikes in HT1080 cells. Indo-1-labeled HT1080 cells were allowed to polarize on coverslips. Ca²⁺ spikes were followed over time using a microfluorometry apparatus. Experiments were performed using long-time base to reveal changes in Ca²⁺ spikes over a period of minutes after the addition of reagents (arrows). The addition of PBS (trace 1), BAPTA (50 μm; trace 2), EDTA (5 mm; trace 3), SKF96365 (15 μm; trace 4), diltiazem (1 μm; trace 5), verapamil (4 μm; trace 6), and nicardipine (1 μm; trace 7) had no effect on Ca²⁺ spikes of HT1080 cells. However, when CAI (10 μm; trace 8), mibefradil (1.5 μm; trace 9), or Gd³⁺ (1.5 μm; trace 10) were added to cells, the fluorescence intensity of spikes diminished greatly, thus forming small Ca²⁺ “bumps.” The Ca²⁺ signals were eliminated within minutes. Bar = 2000 counts (n = 3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Evaluation of Ca²⁺ spikes using a short-time base for data acquisition. Experiments were performed as described in the previous figure, except that a short-time base was used to provide the fine structure of the Ca²⁺ spikes. A typical Ca²⁺ spike profile of an untreated HT1080 cell consists of a low-intensity shoulder on the leading edge and brief peak, followed by a decay in indo-1 intensity (trace 1). However, when cells were treated with Gd³⁺ (1.5 μm), the Ca²⁺ spike was replaced within minutes by a very low-intensity bump, which resembles the shoulder of a typical spike. Parallel results were obtained with the various inhibitors described in Fig. 2 ⇓ (data not shown; n = 3).

The HT1080 Ca²⁺ Spike Is a Perimembrane Ca²⁺ Wave whose Propagation Is Affected by Mibefradil and Gd³⁺.

Recent studies from this laboratory have shown that Ca²⁺ spikes of polarized neutrophils and transfectants are associated with certain intracellular Ca²⁺ waves (29 , 34) . To visualize these waves, images must be collected using very short shutter speeds to avoid image blurring attributable to wave motion and indo-1 diffusion while the shutter is open. A high-speed imaging method was developed to detect these waves (29 , 34, 35, 36, 37) . High-speed microscopy shows that the Ca²⁺ spike (Fig. 2 ⇓ , trace 1; Fig. 3 ⇓ , trace 1) is a counterclockwise (in the microscope’s frame of reference) Ca²⁺ wave (Fig. 4A) ⇓ , e.g., in Fig. 4A ⇓ , frames 2–16, the bright spot is a region of high Ca²⁺ concentration brought about by the local opening of Ca²⁺ channels. Because indo-1 has a relatively high affinity for Ca²⁺ and a finite amount of time is required for the signaling apparatus to pump down free cytosolic Ca²⁺, a low-intensity “tail” is observed next to the region of bright Ca²⁺-indo-1 fluorescence (i.e., the Ca²⁺ wave has a low-intensity tail on the side opposite to the direction of wave motion). This wave was found to travel with a velocity of ∼200 μm/s. However, after the addition of mibefradil, a brief Ca²⁺ spark was observed at the lamellipodium, although no propagating Ca²⁺ wave was found (Fig. 4B) ⇓ . Similarly, a Ca²⁺ wave was observed in cells before (Fig. 4C) ⇓ but not after (Fig. 4D) ⇓ the addition of Gd³⁺. Thus, the regular Ca²⁺ spikes observed in migrating tumor cells are counterclockwise waves, when viewed from the bottom of the cell on a coverslip. Furthermore, the shoulder of the temporal Ca²⁺ spike likely corresponds to the temporal bump and spatiotemporal spark noted above. These waves are blocked by mibefradil and Gd³⁺, which suggests that their binding sites, including LVG and NVG channels, participate in wave propagation around the cell or ignition of the wave at the level of the plasma membrane.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

High-speed imaging reveals Ca²⁺ waves in HT1080 cells. Micrographs of indo-1-labeled cells were acquired using a 100-ns shutter speed and 30-μs duty cycle. The Ca²⁺ spikes of Figs. 2 ⇓ and 3 ⇓ were found to be Ca²⁺ waves propagating in the perimembrane region of polarized HT1080 cells when viewed using high-speed microscopy. In A, frame 1 shows a differential interference contrast microscopy micrograph of an HT1080 cell. In A, frames 2–16 show a corresponding series of high-speed fluorescence micrographs of this cell. A Ca²⁺ wave begins at the leading edge of the cell and then propagates in a counterclockwise direction (A, frames 2–16). However, when this same cell was treated with mibefradil (1.5 μm; B), Ca²⁺ wave propagation was blocked. A brief Ca²⁺ spark was observed in mibefradil-treated cells (arrow in B, frames 4 and 5). C and D, another cell before (C) and after (D) treatment with Gd³⁺. A perimembrane Ca²⁺ wave was observed in the cell before treatment, whereas a dim Ca²⁺ spark was observed after exposure to Gd³⁺ (1.5 μm; D, arrows). Consecutive signals are shown in A and B and C and D. In B and D, the Ca²⁺ sparks soon disappeared (n = 4; ×840).

Mibefradil and Gd³⁺ Inhibit HT1080 Migration.

The central role of Ca²⁺ and Ca²⁺ oscillations in cell motility and ability of mibefradil and Gd³⁺ to influence these oscillations and waves suggest that these reagents may affect tumor cell motility. The phagokinetic track assay was used to test the ability of mibefradil and Gd³⁺ to inhibit HT1080 cell motility (32 , 33) . HT1080 cells were incubated on gold particle-coated glass coverslips for 22 h. As cells migrate on these surfaces, the gold particles are cleared from regions of the surface and frequently picked up by the cells. This makes the cells appear darker and surrounding area appear lighter when viewed by microscopy. Fig. 5 ⇓ shows representative micrographs of this assay. In the absence of added reagents, large areas of the surface are cleared by the cells, as the cells become darker (Fig. 5A) ⇓ . As shown by brightfield microscopy, we observed that the intra and extracellular Ca²⁺ chelators BAPTA (25 μm) and EDTA (5 mm) blocked cell motility. The Ca²⁺ channel blockers, 10 μm CAI (Fig. 5D) ⇓ , 1.5 μm mibefradil (Fig. 5E) ⇓ , and 1.5 μm Gd³⁺ (Fig. 5F) ⇓ , significantly decreased the ability of HT1080 cells to clear the surrounding surface. However, at concentrations of SKF96365, diltiazem, verapamil, and nicardipine expected to block the relevant Ca²⁺ channels, no inhibitory effect could be observed (Fig. 5, G–J) ⇓ . Quantitative data were also acquired by analyses of these micrographs. Because the size of HT1080 cells varies considerably, a swept area ratio, which equals to total swept area divided by cell area, was used to evaluate cell motility. These data reiterated the fact that mibefradil and Gd³⁺ significantly inhibited cell motility in a dose-dependent manner (Fig. 6) ⇓ . Six μm Gd³⁺ almost completely abolished HT1080 movement (Fig. 6B) ⇓ , and 5 μm mibefradil decreased the cell-swept area ∼90% compared with controls (Fig. 6A) ⇓ . Thus, both qualitative and quantitative data indicate that blockade of LVG and NVG channels, but not L-type Ca²⁺ channels, diminishes HT1080 cell motility.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Effect of Ca²⁺ channel blockers on HT1080 motility. A suspension of HT1080 cells with or without calcium channel blockers was seeded onto gold-coated coverslips. Brightfield microscopy was performed after incubation at 37°C. The extent of cell motility is indicated by the amount of neighboring area cleared by the cells. An untreated cell is shown in A. Control experiments using treatment with the intra and extracellular Ca²⁺ chelators BAPTA (25 μm; B) and EDTA (5 mm; C) showed an absence of cell motility. Treatment with carboxyamido-triazole (CAI; 10 μm; D), mibefradil (1.5 μm; E), or Gd³⁺ (1.5 μm; F) significantly reduced HT1080 area cleared by the cells. However, SKF96365 (15 μm; G), diltiazem (1 μm; H), verapamil (4 μm; I), and nicardipine (1 μm; J) had no effect on cell migration (n = 3; ×80).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Dose-dependent inhibition of HT1080 cell motility by mibefradil and Gd³⁺. HT1080 cells treated with various doses of mibefradil (A) or Gd³⁺ (B) were incubated on gold-coated coverslips. The area cleared by each cell was measured using Scion Image software. The swept area ratio was obtained by dividing the total swept area by the cell area. Both mibefradil and Gd³⁺ effectively inhibited HT1080 cell motility in a dose-dependent manner (n = 4; ∗, P < 0.001).

Previous studies have shown that fibronectin mobilizes Ca²⁺ from internal and external stores to the tumor cell cytoplasm (42) . We confirmed this fact for HT1080 cells. Coverslips were coated with fibronectin as described (43) , and then indo-1-labeled HT1080 cells were allowed to attach to the surface. Ca²⁺ spikes were observed at a much higher frequency, thus indicating an enhancement of the Ca²⁺-signaling apparatus (data not shown). No significant change in cell migration was observed in the presence of 25 μg/ml fibronectin for untreated, control, or drug-treated samples (data not shown). Although fibronectin influences the Ca²⁺ signals, it does not affect the ability of channel blockers to influence cell motility.

Mibefradil and Gd³⁺ Inhibit HT1080 Cell Invasiveness.

Because cell motility is required for invasion, we tested the ability of mibefradil and Gd³⁺ to inhibit HT1080 cell invasion. Biocoat Matrigel invasion chambers were used for HT1080 cell invasion experiments. Cell suspensions were loaded to the top chamber with or without the addition of mibefradil, Gd³⁺, nicardipine, or other reagents. After a 22-h incubation period, transmigrated cells on the underside of the inserts were fixed, stained, and counted. Both mibefradil and Gd³⁺ inhibited HT1080 cell transmigration in a dose-dependent fashion (Fig. 7) ⇓ . However, nicardipine, SKF96365, verapamil, and diltiazem had no effect on cell invasiveness at pharmacologically appropriate doses (Fig. 8) ⇓ . Moreover, when the lowest doses of mibefradil and Gd³⁺ that exhibit maximal inhibition of invasiveness (5 and 6 μm, respectively; Fig. 7 ⇓ ) were used simultaneously, the drug combination further augmented inhibition (Fig. 8) ⇓ . Inhibition (95%) was seen using the combination of 5 μm mibefradil and 20 μm Gd³⁺. The complementary action of these two reagents is consistent with their ability to act on two different membrane sites contributing to the same physical wave. Similarly, CAI plus mibefradil showed a significant reduction in invasiveness in comparison with mibefradil alone. Therefore, HT1080 cell invasion was regulated by T-type and NVG Ca²⁺ channels but not L-type Ca²⁺ channels.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Dose-dependent inhibition of HT1080 invasion of Matrigel by mibefradil and Gd³⁺. HT1080 cell suspensions (10⁵/ml) were loaded onto the top Matrigel chamber with or without the addition of various concentrations of mibefradil or Gd³⁺. A final concentration of 25 μg/ml fibronectin was added to the bottom chamber medium as a chemoattractant. After a 22-h incubation, transmigrated cells on the bottom side of the inserts were fixed, stained, and counted. Both mibefradil (A) and Gd³⁺ (B) inhibited HT1080 cell invasion in a dose-dependent fashion (n = 4; ∗, P < 0.001).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 8.

The combined inhibitory effect of T-type and nonvoltage-gated channel blockers on HT1080 invasion. Matrigel chambers were used as described in Fig. 7 ⇓ . Untreated and negative control Ca²⁺ channel blockers [SKF96365 (15 μm), diltiazem (1 μm), verapamil (4 μm), and nicardipine (1 μm)] gave similar levels of cell invasiveness. Carboxyamido-triazole (CAI; 5 μm), mibefradil (5 μm), and Gd³⁺ (6 μm) showed significant inhibition of cell invasion. When 5 μm mibefradil and 6 μm Gd³⁺ were added simultaneously to cells, the inhibition was enhanced ≤95%. The combination of CAI (20 μm) and mibefradil (5 μm) also showed greater inhibition of cell invasiveness (n = 4; ∗, P < 0.001 in comparison with controls; ∗∗, P < 0.01 with respect to the comparisons indicated).