The extracellular loop of pendrin and prestin modulates their voltage-sensing property

Pendrin and prestin belong to the solute carrier 26 (SLC26) family of anion transporters. Prestin is unique among the SLC26 family members in that it displays voltage-driven motor activity (electromotility) and concurrent gating currents that manifest as nonlinear cell membrane electrical capacitance (nonlinear capacitance (NLC)). Although the anion transport mechanism of the SLC26 proteins has begun to be elucidated, the molecular mechanism of electromotility, which is thought to have evolved from an ancestral ion transport mechanism, still remains largely elusive. Here, we demonstrate that pendrin also exhibits large NLC and that charged residues present in one of the extracellular loops of pendrin and prestin play significant roles in setting the voltage-operating points of NLC. Our results suggest that the molecular mechanism responsible for sensing voltage is not unique to prestin among the members of the SLC26 family and that this voltage-sensing mechanism works independently of the anion transport mechanism.

The solute carrier 26 (SLC26) 4 proteins are ubiquitously expressed in a wide variety of species from bacteria to humans and constitute the SLC26/SulP family of ion transporters and channels (1). Although most of the members of this large gene family function as anion exchangers or channels, only prestin (SLC26A5) exhibits voltage-dependent motor activity referred to as electromotility. Electromotility of prestin and the anion transport mediated by pendrin (SLC26A4) are essential for normal development and operation of mammalian inner ears (2)(3)(4). Prestin-mediated electromotility confers a rapid voltageinduced force-generating cell length change on cochlear outer hair cells (5), which is indispensable for the frequency selectivity and sensitivity of mammalian hearing (4). Recent structural and biochemical studies (6,7) have provided significant insights into the ion transport mechanism of the SLC26 proteins. However, the molecular mechanism of the voltage-driven motor activity, which is thought to be unique to prestin among the SLC26 family members, still remains largely unknown. Electromotility accompanies a movement of prestin-associated charge. Because this charge movement is rapid, it manifests as nonlinear electrical capacitance of the cell membrane (nonlinear capacitance (NLC)) (8,9). Thus, NLC is regarded as the electric signature of electromotility, and its measurement often substitutes for direct electromotility measurement.
Because of the high degree of similarity in amino acid sequences among the SLC26 family members, it is conceivable that the tertiary structures of SLC26 proteins are quite similar among the family members and that a common molecular mechanism underlies their diverse physiological functions. Both pendrin and prestin are expressed in the mammalian inner ear, and they share 36% homology at the amino acid level. Using SLC26 models that were generated based on the solved structure of a bacterial SLC26 protein (see Fig. 1) (7), we carried out a comprehensive mutational analysis combined with whole-cell patch-clamp recordings and anion transport assays to understand the molecular basis for the functional difference between pendrin, a well-characterized anion transporter, and prestin, a voltage-driven motor. We found that pendrin also retains voltage-sensing ability, which becomes evident with addition of cyclodextrin to depolarize the peak voltage-operating point. We also found that the electrostatic property of the characteristic extracellular loop of pendrin and prestin (termed "L 2 " in this study; see Fig. 1) influences the voltage-operating points of these proteins. Furthermore, the anion transport activity of pendrin was not affected by the alterations in operating voltage range, suggesting that the voltage-sensing and anion transport mechanisms are independent of each other. Our results suggest that the extracellular loop of prestin has adopted the optimal electrostatic property, which affects the membrane surface electric potential, so that the voltage-operating point resides within the physiologically relevant voltage range, and that the electromotility may have evolved independently of its anion transport mechanism.

The extracellular loops of pendrin and prestin
Recent structural and biochemical studies have provided significant insights into the ion transport mechanism of the SLC26 proteins (6,7,10). Because the basic architecture of the transmembrane (TM) region that accommodates the putative ion translocation pathway is likely quite similar among the SLC26 family members, it is possible that the diverse physiological functions of the SLC26 family of proteins have arisen from modifications of a common ancestral ion transport mechanism. Structural information often facilitates efforts to identify regions in proteins responsible for the expressions of their diverse physiological functions. Because the structure of the TM region has only been solved for a bacterial SLC26 protein, SLC26Dg (Fig. 1A) (7), we modeled the TM structures of pendrin and prestin based on the solved structure of SLC26Dg (Fig.  1, B and C). Two prominent extracellular loops, which are referred to as loop-1 (L 1 ) and loop-2 (L 2 ) in this study, are present in pendrin and prestin but not in the bacterial SLC26 protein (Fig. 1, B and C, magenta). Because many disease-associated pendrin mutations have been identified in the L 1 (F161I, V163I, V163L, S166N, and T178P) and L 2 (V250A, S252P, I253V, L257P, and D266N) regions (11)(12)(13)(14)(15)(16)(17)(18)(19), these extracellular loops may play significant roles in defining the physiological functions of pendrin (HCO 3 Ϫ /I Ϫ /Cl Ϫ exchanger) and prestin (voltage-operated motor).

The extracellular loops modulate the voltage-operating point of prestin
To examine the contributions of L 1 and L 2 to the function of prestin, we generated several prestin-based constructs whose L 1 and L 2 are replaced with those of pendrin individually (A5-L 1 and A5-L 2 ) or in combination (A5-L 1 /L 2 ) (Figs. 1 and 2) and measured NLC in HEK293T cell lines expressing these constructs (Fig. 2). Replacement of L 1 did not abrogate NLC but resulted in significant hyperpolarization of the voltage-operating points (V pk ) (Fig. 2, B and E). Likewise, replacement of L 2 resulted in significant V pk hyperpolarization but to a much greater degree compared with L 1 replacement (Fig. 2, C and E). No further hyperpolarization was observed upon double (L 1 / L 2 ) replacement (Fig. 2, D and E). The voltage sensitivity (␣) was not significantly affected by any of these loop replacements (one-way analysis of variance (ANOVA), p ϭ 0.14) (Fig. 2F). These results suggest that the extracellular loops of prestin modulate its operating voltage range without affecting its voltage sensitivity.

Pendrin possesses voltage-sensing ability
The voltage-driven motor function of prestin is thought to be unique among the members of the SLC26 family; however, the veracity of this presumption has not been tested rigorously. Given the high degree of similarity in amino acid sequences among the SLC26 family members, it is possible that many, if not most, other SLC26 proteins also retain a voltage-sensing mechanism, but it is not readily detectable within the experimentally measurable voltage range (typically Ϯ150 mV). We noticed that some HEK293T cells expressing the WT pendrin construct tend to exhibit a gradual increase in cell membrane capacitance below ϳ0 mV, hinting at the presence of large NLCs (Fig. 3A). We speculated that pendrin also retains voltage-sensing ability, but its V pk is extremely hyperpolarized and thus not easily detected. To test this possibility, we repeated capacitance measurement for the WT pendrin construct in the presence of methyl-␤-cyclodextrin (M␤CD), which is known to depolarize V pk of prestin by removing cholesterol from the cell membrane (20). As shown in Fig. 3B, NLCs of WT pendrin became quite evident upon application of M␤CD due to its anticipated depolarizing effect, which brought the otherwise barely detectable NLC of pendrin into the experimentally measurable voltage range. We also performed NLC measurement in Sf9 cells (Fig. S1). The low membrane cholesterol level in this cell line (21) allowed detection of large NLC of pendrin without the use of M␤CD.

The importance of the extracellular loops of pendrin for membrane targeting and modulation of the voltageoperating point
We examined the contributions of L 1 and L 2 to the voltagesensing function of pendrin as we did for prestin. To this end, we generated several pendrin-based constructs whose L 1 and L 2 are replaced with those of prestin individually (A4-L 1 and A4-L 2 ) or in combination (A4-L 1 /L 2 ) ( Figs. 1 and 3). We anticipated finding large NLCs for A4-L 1 because the L 1 region contains the "14-aa" segment that was reported to confer large NLC and electromotility on pendrin (22). However, the effect of L 1 , which are not present in the bacterial SLC26 structure, are highlighted in magenta. D and E, partial amino acid sequences of SLC26A4 (pendrin) and SLC26A5 (prestin) that contain the L 1 (D) and L 2 (E) regions. Characters shown in cyan and magenta indicate acidic and basic residues, respectively. L 1 , L 2 , 14aa, and sL 2 indicate the segments that were swapped between pendrin and prestin in this study.

Pendrin retains voltage-sensing ability
replacement was ambiguous (Fig. 3C) as NLCs of A4-L 1 (and WT pendrin (WT-A4)) were only partially measurable to confidently estimate the NLC parameters. Qualitatively, the L 1 replacement did not seem to significantly affect V pk of pendrin (Fig. 3, A and C). We also generated a construct in which only the 14-aa segment used in the previous study (22) was replaced (termed "A4-L 1 (14aa)" in this study; Fig. 1D). Unexpectedly, however, we could not detect NLC for this construct (Fig. 3D). The reason for the discrepancy is unclear. At the least, the absence of NLC in A4-L 1 (14aa) should not be solely ascribed to impaired membrane targeting of the construct because the HCO 3 Ϫ /Cl Ϫ antiport activity was detected for A4-L 1 (14aa) (Fig.  4A). Although the transport activity of A4-L 1 (14aa) was smaller than that of WT-A4 (p ϭ 0.0016) (Fig. 4A), it was comparable with that of A4-sL 2 (p Ͼ 0.999; Fig. 4A), which exhibited a large NLC ( Fig. 3F; see below).

The role of the charged residues in loop-2 in establishing V pk
The L 2 swapping between prestin and pendrin resulted in hyperpolarization for prestin and depolarization for pendrin (Figs. 2 and 3). This inverse effect of the L 2 swapping can be explained by the electrostatic property of the L 2 region that is presumably expected to be exposed to the extracellular surface of the membrane. For example, the presence of acidic residues would decrease the surface charge density ( o ), which should result in concomitant drops (hyperpolarization) of the membrane surface potential (⌽ o ) (Fig. 5A). The opposite (depolarization of ⌽ o ) would be expected if basic residues were present in L 2 (Fig. 5B). If the inner surface potential (⌽ i ) remained unchanged, hyperpolarization of ⌽ o should result in hyperpolarization of V pk and vice versa. This interpretation is compatible with the fact that the net calculated charges of L 2 are Ϫ1 and ϩ4 for pendrin and prestin, respectively (Fig. 1E).
Glu 259 is the sole charged residue present in sL 2 of pendrin, whereas none is present in that of prestin (Fig. 1E). To test whether the significant V pk depolarization found in A4-sL 2 ( Fig.  3F) can be attributed to neutralization of the negative charge pertaining to Glu 259 , additional pendrin constructs harboring E259Q, E259R, or E259K mutation were generated, and their NLCs were measured. NLCs became clearly measurable upon neutralization of Glu 259 (E259Q) (Fig. 5, C and J) and resemble

Pendrin retains voltage-sensing ability
those found in A4-sL 2 (Fig. 3F). Polarity reversal of Glu 259 (E259R and E259K) also made NLC of pendrin clearly detectable within the measurable voltage range due to V pk depolarization (Fig. 5, D, E, and J). We also examined the effect of the negative charge pertaining to Asp 266 , which is not located within the sL 2 region but located in L 2 (Fig. 1E). We found that neutralization of this negative charge (D266N) also depolarizes V pk and makes large NLC of pendrin readily detectable (Fig. 5, F and J). Simultaneous introduction of charge-affecting mutations to Glu 259 and Asp 266 (A4-E259Q/D266N, A4-E259R/ D266N, and A4-E259K/D266N) also depolarized V pk but to greater extents (Fig. 5, G-J). These results suggest that the effects of these charge manipulations are additive. In fact, we found a statistically significant linear correlation (r ϭ 0.85, p ϭ 0.008) between the calculated net charge in L 2 versus V pk (Fig.   5K). The ␣ value and its variation were similar among the pendrin constructs (one-way ANOVA, p ϭ 0.39) (Fig. 5L), which argues against the possibility that the V pk values were significantly over-or underestimated. These observations are uniformly explained by depolarization of ⌽ o as a result of reduction of the negative surface charge present in sL 2 or L 2 (Fig. 5,  A and B).
Charge manipulations of Asp 271 (D271N, D271R, and D271K), which is located in the transmembrane helix that immediately follows L 2 (Fig. 1), significantly impaired the targeting of pendrin to the cell membrane (Fig. S2A). Consistently, these mutants showed little or no NLC and transport activity (Fig. S2, B and E), implying the importance of Asp 271 for maintaining the structural integrity of the transmembrane region.

Pendrin retains voltage-sensing ability The relationship between the voltage-sensing ability of pendrin and its anion transport function
The voltage-driven motor function of prestin has been presumed to stem from a common anion transport mechanism that is likely shared among the members of the SLC26 family (23,24). To define the molecular mechanism underlying electromotility, it is important to examine whether the anion transport activity of prestin is affected by voltage. Unfortunately, prestin is not suitable for pursuing this possibility because we could not detect HCO 3 Ϫ /Cl Ϫ antiport activity for prestin (WT-A5) or any of our prestin-based constructs. We therefore examined the NLC-anion transport relationship in pendrin because we identified large NLC in pendrin (Fig. 3). The resting mem-brane potentials of HEK293T cells expressing WT pendrin (mean Ϯ S.D.) were Ϫ15.0 Ϯ 5.8 (n ϭ 12) and Ϫ12.3 Ϯ 4.4 mV (n ϭ 16) in the presence and absence of M␤CD, respectively, with no statistically significant difference among the groups (p ϭ 0.17). If the voltage-sensing mechanism were intimately related to the anion transport mechanism, augmentation of the anion transport activity would be expected upon application of M␤CD due to a significant shift of V pk toward the resting membrane potential of the pendrin-expressing cells (Fig. 3, A and B). However, we did not find a statistically significant change in the anion transport activity upon application of M␤CD (p Ͼ 0.999; Fig. 4A), suggesting independence of the transport and voltagesensing mechanisms.

NLC1 and NLC2 segments of pendrin and prestin are not mutually interchangeable
Previous efforts to identify a segment(s) in the prestin molecule that is essential for the expression of electromotility have been inconsistent. One study (25) identified such segments, termed "NLC1" and "NLC2," in the region forming the presumed anion translocation pathway (6), whereas other studies identified a short segment (termed "14aa" in the present study) located in L 1 (22,26). These results are apparently irreconcilable because the short 14-aa segment is not a part of the large NLC1/NLC2 segments. Because we could not confirm the importance of the 14-aa short segment for the expression of NLC (Fig. 3D), we also generated eight chimeric constructs in which the NLC1 and NLC2 segments were swapped between pendrin and prestin (A4A5Ch1-A4A5Ch8) (Fig. 6A), which are similar to those generated by Schaechinger et al. (25) using electromotile and nonelectromotile prestin orthologues. Unfortunately, none of our pendrin/prestin chimeras showed NLC or HCO 3 Ϫ /Cl Ϫ antiport activity (Fig. S2, C-E), probably due to the lack of membrane targeting (Fig. 6B), suggesting a difference in the intramolecular interactions within the TM domain between pendrin and prestin.

Discussion
The electromotile function of prestin (SLC26A5) is thought to be unique among the members of the SLC26 family.
Although not all prestin orthologues exhibit motor activity, they all show NLC (27)(28)(29)(30). NLC indicates the presence of a rapid voltage-sensing mechanism, and it is directly coupled to concurrent motor activity in electromotile prestin orthologues (31)(32)(33)(34). In the present study, we show that pendrin (SLC26A4) also exhibits NLC but with extremely hyperpolarized V pk . Given the high degree of similarity in amino acid sequences among the SLC26 proteins, the presence of NLC in other members in the family would not be surprising, but this possibility has not been thoroughly examined heretofore. A previous study also reported NLC parameters for pendrin-expressing cells (22). However, the magnitude of the NLCs were much smaller than those observed in the present study despite their apparent V pk (ϳϪ40 mV) located within an easily measurable voltage range (22). In addition, the small NLCs found in the previous study were qualitatively very different from those reported in this study as the V pk values were depolarized, but not hyperpolarized, with respect to V pk of prestin (22). We claim that the

Pendrin retains voltage-sensing ability
NLCs detected in the present study for the pendrin-expressing cells are indeed mediated by pendrin because we found that the V pk of the NLCs responds to cholesterol depletion by M␤CD (Fig. 3) in the same way as previously demonstrated for prestin (20) and that swapping of L 2 that contains opposite electrostatic property between pendrin (net negative) and prestin (net positive) results in V pk shifts toward mutually opposing directions for pendrin (depolarization; Fig. 3) and prestin (hyperpolarization; Fig. 2). Successful identification of large NLCs in pendrinexpressing Sf9 cells without the use of any genetic or chemical manipulations further authenticates our claim (Fig. S1).
There is controversy over the identity of the voltage-sensing charge (6, 35-38). An early study proposed an incomplete transport model to explain the voltage-sensing mechanism of prestin where intracellular Cl Ϫ or HCO 3 Ϫ is used as an extrinsic voltage sensor (35). The basic notion of this model does not disagree with the fact that the HCO 3 Ϫ /Cl Ϫ antiport activity of prestin is very low (39) or virtually undetectable. However, later studies found that electromotility can be conferred on nonelectromotile SLC26 proteins without diminishing or abrogating the transport activities (25,26). Our finding that pendrin exhibits both significant HCO 3 Ϫ /Cl Ϫ antiport activity and NLC (Figs. 3 and 4) underscores that the anion transport function and voltage-sensing ability are not mutually exclusive. The identification of large NLC in pendrin provided a fortuitous opportunity to examine the functional relationship between voltage-

Pendrin retains voltage-sensing ability
Pendrin retains voltage-sensing ability sensing and anion transport mechanisms, which could not be pursued in prestin due to its extremely small anion transport activity to be compared with NLC. We found that the HCO 3 Ϫ / Cl Ϫ antiport activity of pendrin is not affected by a significant depolarizing shift of NLC toward the resting membrane potentials of pendrin-expressing cells upon M␤CD application (Figs.  3 and 4). This observation suggests that voltage-sensing and anion transport mechanisms may not be intimately related, which is in agreement with a previous observation suggesting separation of voltage-sensing and anion transport functions for prestin (38).
It is possible that most, if not all, members of the SLC26 family retain voltage-sensing ability. It is important to note that the absence of measurable NLC does not necessarily mean inability of an SLC26 protein to sense voltage. Because the experimentally measurable range of membrane potential is limited, NLC with extremely hyper-/depolarized V pk is difficult to identify. In the present study, we demonstrate that this is, in fact, the case for pendrin (Fig. 3A). Fast voltage-sensing kinetics is another important factor for a successful NLC recording because detection of capacitive currents is the basis of NLC measurement. If voltage-sensing charge movement is very slow, it would not manifest as NLC. For example, NLC of zebrafish prestin shows significant stimulus voltage frequency dependence, and it becomes unmeasurable at high stimulus frequencies due to its relatively slow voltage-sensing kinetics (27). It is conceivable that the voltage-sensing kinetics of some SLC26 members may be too slow to exhibit its voltage-sensing ability as NLC. Collectively, absence or abrogation of NLC by genetic or pharmacological manipulation needs to be concluded carefully.
For physiological electromotile function of prestin orthologues, it is important to set their V pk within the physiological V m range. Our study demonstrates that the charged residues present in L 2 greatly contribute to set distinct V pk for prestin and pendrin (Figs. 2, 3, and 5) and that the directions of the V pk shift induced by charge-affecting mutations in L 2 can be simply explained by changes in the membrane surface charge density ( o ) that concomitantly affect membrane surface electrical potential (⌽ o ) (Fig. 5, A and B). Furthermore, all previously reported V pk shifts of prestin that were induced by charge-affecting mutations introduced to L 2 , i.e. T234K (depolarization) (40), K233Q/K235Q/R236Q (hyperpolarization) (35), S238D (hyperpolarization) (41), and K255Q (hyperpolarization) (38), can be explained by their anticipated effects on the membrane surface potential. On the contrary, L 1 does not seem to contribute to confer distinct V pk on pendrin (Fig. 3C). Although significant V pk hyperpolarization was observed in prestin upon L 1 swapping (Fig. 2B), no additive effect on V pk was found upon L 1 /L 2 double swapping (Fig. 2D). In any case, the hyperpolarizing V pk shift of prestin upon L 1 swapping (Fig. 2B) cannot be explained by the electrostatic property of L 1 because the net charge of L 1 is almost the same between prestin and pendrin (Fig. 1D). Also, hyperpolarizing V pk shifts of prestin induced by D154N and D155N mutations (35) cannot be explained by an L 1 -associated surface charge effect. These irregular and unpredictable consequences of L 1 swapping may imply a dynamic role of the L 1 region in prestin.
It is noteworthy that the voltage sensitivity of pendrin, which is reflected in its ␣ value, is 2-3 times smaller than that of prestin (Figs. 2F and 5L) and that the ␣ values of nonelectromotile prestin orthologues are also smaller than those of electromotile prestin orthologues (27,28). Because ␣ values would indicate the magnitude and extent of voltage-induced conformational change (see "Experimental procedures"), it is interesting to examine whether and how the magnitudes of ␣ correlate with the expression of electromotility. Although electrophysiological characterization has already been conducted for various prestin orthologues (27)(28)(29)(30), similar efforts have not been expanded to the other members in the SLC26 family. Thorough comparative studies across the SLC26 family would facilitate efforts in defining the molecular mechanisms underlying the uniqueness and generality of electrophysiological characteristics among the SLC26 family members.

Pendrin retains voltage-sensing ability
A4A5Ch8). The complete DNA sequences of all these constructs, along with the PCR primers and synthesized dsDNA information, are provided in the supporting information. These constructs with C-terminally attached ECFP (for prestin-based constructs) or mTurquoise2 (for pendrin-based constructs) were cloned into a pSBtet-Pur vector (Addgene, Cambridge, MA) (43). The functions of pendrin and prestin are not abrogated by a C-terminally fused fluorescent protein (32,44). Stable cell lines that express these recombinant protein constructs were established in HEK293T cells as described previously (43). Expressions of the pendrin/prestin constructs were induced by application of doxycycline (1 g/ml) to the cell culture medium 1-3 days prior to experiments.

Electrophysiology
Whole-cell recordings were performed at room temperature using an Axopatch 200A/B amplifier (Molecular Devices, Sunnyvale, CA). Recording pipettes were pulled from borosilicate glass to achieve initial bath resistances averaging 3-4 megaohms. Whole-cell NLC recordings were performed using a 0-mV holding potential and a sinusoidal voltage stimulus (2.5-Hz, 120 -150-mV amplitude) superimposed with two higher frequency stimuli (390.6 and 781.2 Hz, 10-mV amplitude). Recording pipettes were filled with an intracellular solution containing 140 mM CsCl, 2 mM MgCl 2 , 10 mM EGTA, and 10 mM HEPES (pH 7.3 adjusted at room temperature). Cells were bathed in an extracellular solution containing 120 mM NaCl, 20 mM tetraethylammonium chloride, 2 mM CoCl 2 , 2 mM MgCl 2 , and 10 mM HEPES (pH 7.3 adjusted at room temperature). Osmolarity was adjusted to 310 mosmol liter Ϫ1 with glucose (34,46). Intracellular pressure was kept at 0 mmHg, and current data were collected by jClamp (SciSoft Co., New Haven, CT) using a fast Fourier transform-based admittance analysis to determine NLC (47). The resting cell membrane potentials were determined in Hanks' balanced salt solution (14025, Thermo Fisher Scientific, Waltham, MA) using recording pipettes filled with an intracellular solution containing 140 mM KCl, 2 mM MgCl 2 , 10 mM EGTA, and 10 mM HEPES (pH 7.3 adjusted at room temperature).

NLC data analysis
Voltage-dependent cell membrane electric capacitance data were analyzed using the following two-state Boltzmann equation, where ␣ is the slope factor of the voltage dependence of charge transfer, Q max is the maximum charge transfer, V m is the membrane potential, V pk is the voltage at which the maximum charge movement is attained, and C lin is the linear capacitance (34,46,48,49). In this two-state Boltzmann model, ␣ is defined as z app e/k B T where z app is the apparent valence of charge movement, e is electron charge, k B is the Boltzmann constant, and T is absolute temperature. The magnitude of z app reflects the valence (z) and the traveling distance (␦) along the direction of the electric field of a voltage-sensing charge.

Confocal imaging
The stable cell lines were cultured in Dulbecco's modified Eagle's medium (DMEM, high glucose, GlutaMAX, pyruvate, Thermo Fisher Scientific) with 10% FBS (French origin, Bio-West, Riverside, MO) at 37°C (5% CO 2 ) in glass-bottom dishes (D11130H, Matsunami Glass, Kishiwada, Japan). When cell confluence reached 20 -30%, expressions of the pendrin/prestin constructs were induced by 1 g/ml doxycycline hyclate (LKT Laboratories, St. Paul, MN) included in the medium for 2-3 days. Fluorescence images of the cells were acquired using an LSM710 confocal microscope (Carl Zeiss, Oberkochen, Germany) equipped with a 63ϫ Plan-Apochromat objective lens (numerical aperture, 1.40 oil) and a filter set (BP 395-440 excitation filter, FT460 dichroic mirror, and LP470 emission filter). The cells were illuminated with a 30-milliwatt 405-nm diode laser (laser power, 2%) at room temperature, and the images were taken with ZEN software (Carl Zeiss). The pinhole size and pixel dwell time were set to be 1 airy unit (490 nm) and 1.27 s, respectively. Four sequentially scanned images were averaged and then processed with a despeckle algorithm and contrast/brightness adjustment using Fiji software (50).

Statistical analyses
Statistical analyses were performed using Prism (GraphPad Software). Student's t test was used for comparisons between two groups. One-way ANOVA combined with the Tukey-Kramer test was used for multiple comparisons. p Ͻ 0.05 was considered statistically significant.