(Received for publication, November 7, 1996, and in revised form, December 15, 1996)
From the Department of Physiology, McGill University, 3655 Drummond Street, Montréal, Québec H3G 1Y6, Canada
Protein kinase A (PKA) stimulates Cl secretion by
activating the cystic fibrosis transmembrane conductance regulator
(CFTR), a tightly regulated Cl channel in the apical
membrane of many secretory epithelia. The CFTR channel is also
modulated by protein kinase C (PKC), but the regulatory mechanisms are
poorly understood. Here we present evidence that PKA-mediated
phosphorylation alone is not a sufficient stimulus to open the CFTR
chloride channel in the presence of MgATP; constitutive PKC
phosphorylation is essential for acute activation of CFTR by PKA. When
patches were excised from transfected Chinese hamster ovary cells, CFTR
responses to PKA became progressively smaller with time and eventually
disappeared. This decline in PKA responsiveness did not occur in the
presence of exogenous PKC and was reversed by the addition of PKC to
channels that had become refractory to PKA. PKC enhanced PKA
stimulation of open probability without increasing the number of
functional channels. Short-term pretreatment of cells with the PKC
inhibitor chelerythrine (1 µM) reduced the channel
activity that could be elicited by forskolin in cell-attached patches.
Moreover, in whole cell patches, acute stimulation of CFTR currents by
chlorophenylthio-cAMP was abolished by two chemically unrelated PKC
inhibitors, although an abrupt, partial activation was observed after a
delay of >15 min. Modulation by PKC was most pronounced when basal PKC
phosphorylation was reduced by briefly preincubating cells with
chelerythrine. Constitutive PKC phosphorylation in unstimulated cells
permits the maximum elevation of open probability by PKA to reach a
level that is ~60% of that attained during in vitro
exposure to both kinases. Differences in basal PKC activity may
contribute to the variable cAMP responsiveness of CFTR channels in
different cell types.
The regulatory domain of cystic fibrosis transmembrane conductance
regulator (CFTR)1 has nine dibasic
consensus sequences for phosphorylation by protein kinase A (PKA) and
at least seven potential sites for phosphorylation by protein kinase C
(PKC). There is direct evidence for phosphorylation of serines 660, 700, 712, 737, 753, 768, 795, and 813 by PKA (1-3) and serines 660, 686, 700, and 790 by PKC (2). In freshly excised membrane patches, CFTR
Cl channels are activated by exposure to the PKA
catalytic subunit and MgATP (4, 5). Altering the four major sites of
in vivo phosphorylation reduces forskolin-stimulated
Cl
permeability (1, 6), whereas removal of all 10 dibasic sites and the additional monobasic site Ser753 further
reduces, but does not eliminate, PKA stimulation of CFTR in both intact
cells and excised patches (6, 7). There is little doubt that
PKA-mediated phosphorylation is the primary mechanism controlling CFTR
Cl
channel activity, consistent with early evidence
that cAMP stimulates transepithelial secretion, in part by increasing
apical membrane Cl
conductance (8-12).
The role of PKC in regulating the CFTR channel is less certain.
Exogenous PKC causes a small stimulation of channel activity when added
during the first few minutes after patches are excised (4, 13). PKC
enhances both the rate and magnitude of subsequent PKA stimulation of
open probability (po) by approximately 2-fold and 50%, respectively (4). Potentiation of cAMP responses has also
been observed in phorbol-stimulated Xenopus oocytes
expressing CFTR (14), HT-29 cells (15), and pancreatic duct cells, in which the PKC agonist -phorbol 12,13-dibutyrate increased
cAMP-induced whole cell CFTR Cl
current by 31% although
having no effect on basal currents (16). Potentiation has received
several different interpretations. The effects of PKC on excised
patches (4) and phorbols on T84 and C127 were suggested to involve PKC
regulation of the CFTR channel itself (17, 18). However, to explain the
apparent increase in Cl
current by PKC in pancreatic duct
cells (expressed as pA/pF), it was proposed that PKC reduces cell
membrane area rather than increasing channel activity (16). By
contrast, exocytotic addition of new channels to the plasma membrane
was suggested to explain the potentiation in HT-29 cells (15).
Studying CFTR function after phorbol treatment is problematic when both its expression (19) and degradation (20) are strongly influenced by PKC activity. Another problem has been the difficulty of distinguishing acute PKC-induced changes in po from changes in the number of functional channels (N). Finally, the responsiveness of CFTR to PKA stimulation varies widely among different cell types even when channels are studied in excised patches. The reasons for this variation are not known.
Here we present evidence that PKC-mediated phosphorylation not only potentiates CFTR responses but is, in fact, essential for acute PKA activation of CFTR channels. This requirement for PKC only became apparent when the PKA response was studied after prolonged rundown in excised patches or after short-term pretreatment with PKC inhibitors, presumably because constitutive PKC phosphorylation was slowly removed by phosphatase activity under these conditions (4, 21, 22).
Chinese hamster ovary (CHO) cells and baby hamster kidney (BHK) cells that had been stably transfected with plasmids directing the expression of wild-type CFTR were provided by Drs. X.-B. Chang and J. R. Riordan (Mayo Clinic, Scottsdale, AZ; Ref. 6) and F. Siebert (University of Toronto, Toronto, Ontario, Canada; Ref. 7). CHO and BHK cells were cultured as described previously (4, 42), subcultured on glass coverslips in 35-mm Petri dishes, and used in patch clamp experiments 2-5 days after plating.
Single Channel StudiesSingle channels were recorded using
the cell-attached and inside-out patch clamp configurations (23; also
see Ref. 4). The bath and pipette solutions contained 150 mM NaCl, 10 mM TES, and 2 mM
MgCl2 (pH 7.2). Fresh MgATP (1 mM final
concentration) and forskolin (5 µM) were added to the
bath before each experiment. Experiments were carried out at room
temperature (~22 °C). Single channel currents were amplified
(Axopatch 200; Axon Instruments Inc., Foster City, CA), recorded on
video tape (PCM 418; Medical Systems Corp, Greenvale, NY), and analyzed
as described previously (4). The pipette voltage was held at 30 mV
during cell-attached recordings and then switched to +30 mV after
excision so that the potential across the membrane patch would be
similar in both configurations (approximately
30 mV). Channel
activity was allowed to run down for 10 min after excision to allow
dephosphorylation by membrane-associated phosphatase activity prior to
the addition of protein kinases. To estimate the N in the
patch, AMP-PNP (1 mM) was added to the bath at the end of
each experiment to lock CFTR channels in the open state so they could
be counted (24). po was calculated as the mean
number of channels open during segments of the record
(Npo) divided by the value of N
determined using AMP-PNP. The catalytic subunit of PKA was obtained
from Promega (Madison, WI). Rat brain PKC II, predominantly
Ca2+- and phospholipid-dependent PKC isoforms,
was prepared in the laboratory of M. P. Walsh (University of Calgary,
Alberta, Canada; see Ref. 4). MgATP and the lipid activator
1,2-dioctanoylglycerol (8:0) (DiC8) were purchased from
Sigma. ATP was added from a 100 mM stock
solution in buffer. DiC8 was dissolved in chloroform at 5 mM and stored under nitrogen at
20 °C before use.
Chelerythrine chloride (Biomol, Plymouth Meeting, PA) was prepared as a
10 mM stock in ME2SO, stored at
20 °C, and
diluted in bath solution immediately before use (1 µM
final concentration).
In whole cell experiments, the bath
(extracellular) solution contained 150 mM NaCl, 10 mM TES, and 2 mM MgCl2 (pH 7.40)
and was supplemented with 50 mM sucrose to minimize
swelling-activated Cl currents. The pipette (intracellular) solution
contained 110 mM
N-methyl-D-glucamine-aspartate, 30 mM N-methyl-D-glucamine-Cl, 1 mM MgCl2, 10 mM TES, and 0.1 mM
1,2-bis-(o-aminophenoxy)ethane-N,N,N,N
-tetraacetic acid tetrasodium salt; (Biomol) (pH 7.20). MgATP (1 mM) was
added to the pipette solution from a 100 mM stock in TES
buffer. Stock cpt-cAMP and 8-bromo-cAMP solutions (50 mM)
were prepared in distilled water using their sodium salts and used at a
final concentration of 100 µM. The PKC inhibitor
Gö6976 (Calbiochem) was prepared in ME2SO and diluted
in bath solution immediately before use (final concentration, 500 nM). It was also included in the pipette solution, and
experiments were begun 4 min after obtaining the whole cell configuration to allow its diffusion into the cell (25). When chelerythrine was used, it was prepared as described above and added at
a final concentration of 1 µM for 30 min prior to
recording whole cell currents.
Whole cell currents were recorded using pipettes pulled from
thick-walled borosilicate glass (1.5-3 M). After obtaining seals, pipette capacitance was canceled using the internal circuitry of the
patch clamp (Axopatch 1C; Axon). The whole cell recording configuration
was obtained by applying excess suction, and cell capacitance was
measured and then canceled. In most experiments, Vm
was held at 0 mV for 1 s between alternating pulses to ± 60 mV (1-s duration). Current-voltage relationships were generated by
stepping from
100 to +100 mV in 10-mV increments, returning to the
holding potential of 0 mV between pulses. Currents are expressed per
unit cell capacitance (pA/pF) to allow comparison of data from
different cells. Outward (positive) current corresponds to
Cl
ions entering the cell. The Nernst prediction for a
perfectly Cl
-selective membrane was
39 mV under the
conditions used. The leak current at +60 mV (+63.4 ± 14.7 pA/pF)
was small compared with the mean cAMP-stimulated CFTR current (8.6 ± 1.4%), and therefore was not subtracted. Mean pipette capacitance
was 7.87 ± 0.08 pF (n = 46), mean whole cell
capacitance was 13.1 ± 0.6 pF (n = 51), and the
average access resistance of the pipettes was 15.6 ± 1.0 M
(n = 47).
The number of opening transitions during each
segment of >3 min was counted and used to estimate the mean burst
(open) and interburst (
closed) durations
according to:
open = (N × po) (time)/(number of openings) and
closed = (N
N × po) (time)/(number of openings
1), where
N is the number of channels estimated by locking channels
open with AMP-PNP, and po is the single-channel open probability (31).2 Data are expressed
as means ± S.E. Paired t tests and Mann-Whitney U tests were performed using GraphPad InStat software
(Biosoft, Cambridge, United Kingdom).
Exposing transfected CHO cells to forskolin activated
CFTR Cl channels in 15 of 15 cell-attached patches (Fig.
1). The po under cell-attached
conditions was 0.32 ± 0.038 (n = 4). As in
previous studies, po declined spontaneously when
patches were excised into a bath solution containing 1 mM
MgATP. When exogenous PKA (200 units/ml) was added to the bath, the
magnitude of the CFTR channel response depended on the time that had
elapsed since excision. Robust PKA stimulation was observed ~2 min
after patches were excised, whereas PKA usually failed to activate CFTR
Cl
channels when patches were excised and continuously
exposed to 1 mM MgATP for 10 min. When patches were excised
for 10 min and allowed to become refractory to PKA, responsiveness
could be fully restored by adding PKC and the lipid cofactor
DiC8 to the cytoplasmic side in the continued presence of
PKA (Fig. 1B). This stimulation after the addition of PKC
and DiC8 was elicited by the PKA already present in the
bath rather than by PKC itself, because addition of PKC and
DiC8 after the same delay had no effect in the absence of
PKA (Fig. 1C). Moreover, CFTR responsiveness to PKA did not decline if PKC was present during the 10-min period following excision
(Fig. 1D). Taken together, these results suggest that PKA
phosphorylation alone is not a sufficient stimulus to open the CFTR
channel; constitutive phosphorylation by PKC is also required.
Fig. 2 summarizes the dependence of the PKA response on
time after excision and the effect of exogenous PKC on channel
responsiveness to PKA. All channels were locked open at the end of each
experiment using AMP-PNP to determine N. This number was
then used to calculate po, as described under
"Experimental Procedures." CFTR channels were consistently
reactivated if PKA was added within 2 min after excision; however, this
stimulation of po was progressively reduced with
time after excision (Fig. 2A). The mean
po values during the PKA responses are shown in
Fig. 2B. The interval between excision and PKA addition was
inversely related to the po during PKA
stimulation, which eventually declined to 0. By contrast,
responsiveness to PKA did not decline when patches were exposed to PKC,
DiC8, and MgATP throughout the 10-min period after
excision. Single-channel po in the presence of
PKA and PKC was 0.46 ± 0.07 (n = 3), similar to
that observed for the same patches during forskolin stimulation prior
to excision (po = 0.42 ± 0.02;
p > 0.05; Fig. 2C). This po is higher than that calculated for patches
excised from unstimulated cells into the bath solution containing PKA
and ATP (0.22 ± 0.04; Ref. 24), when the number of channels in
each patch was corrected using AMP-PNP. Thus, in addition to having a
permissive effect on po (Fig. 1B),
PKC also potentiated the PKA response under these conditions by
increasing the maximum po elicited by PKA
approximately 2-fold, as reported previously (4).
Mechanism of PKC Dependence
To further investigate whether
PKC regulates the number of channels opened by PKA or the
po of individual channels, patches containing a
small number of CFTR channels were excised from CHO cells and exposed
to PKA after a delay of 5 min. Adding PKC and DiC8 to the
same patches 5 min after PKA addition did not increase the maximum
number of channels open simultaneously, which was identical to the
value of N determined using AMP-PNP at the end of the
experiment (Fig. 3A). By contrast,
po of single channels calculated from the mean
number of channels open and N was approximately doubled
(p < 0.005; Fig. 3B). When BHK cells, which
express more than a 10-fold higher density of CFTR channels, were
pretreated with the PKC inhibitor chelerythrine (1 µM)
for 5 min, adding PKA to the excised patches caused only a small
increase in channel activity, although the number of channels locked
open by AMP-PNP was high, indicating the channels were active when
exposed to PKA alone but simply had low po. For
example, the patch shown in Fig. 3D contained at least 40 active channels, each with po < 0.05. Addition
of PKC did not increase the number of channels locked open by AMP-PNP
(Fig. 3D), further evidence that channels were already
functional after PKA alone. Thus, PKC acts by enhancing PKA stimulation
of po rather than by increasing N.
PKC shortened the mean closed time (p < 0.01) but had
no effect on mean burst duration (open time; p > 0.05)
during PKA stimulation (Fig. 3C). The lower
po in Fig. 3D (<0.05) compared with
Fig. 3B (~0.2) is explained by pretreatment of the former
with chelerythrine to reduce constitutive PKC phosphorylation. These
results suggest there is some constitutive PKC activity in unstimulated
cells, which permits submaximal responses to PKA stimulation. This
possibility is examined below using cell-attached and whole cell
patches.
Effect of Inhibiting Constitutive PKC Activity in Intact Cells and Whole Cell Patches
To investigate whether endogenous PKC activity
is required for cAMP activation of CFTR in intact cells, we tested the
effect of preincubating cells with chelerythrine, a specific PKC
inhibitor (26, 27). Pretreatment with chelerythrine abolished forskolin activation of CFTR channels in 13 of 16 cell-attached patches (Fig.
4). In parallel control experiments forskolin activated CFTR Cl channels in 15 of 15 cell-attached patches.
Similarly, under control conditions, cpt-cAMP (100 µM)
activated whole cell CFTR chloride currents in 6 of 6 cells with a time
to half-maximal activation of 2.1 ± 0.6 min under control
conditions but in other experiments had no effect within 12 min when
cells were pretreated with chelerythrine (Fig. 5).
Moreover, in further studies, a selective inhibitor of the
Ca2+-dependent
and
1
isoforms of PKC (Gö6976; Refs. 28-30) also prevented acute
activation of CFTR whole cell Cl
currents by cpt-cAMP
(Fig. 5), although partial activation of CFTR was observed after
prolonged treatment with cpt-cAMP in the presence of either
chelerythrine or Gö6976. The time to half-maximal activation of
this delayed response was 23.9 ± 6.2 min in chelerythrine-treated cells and 21.5 ± 5.1 min in Gö6976-treated cells, a lag
that was approximately 10-fold longer than for control cells, which had
been treated with cpt-cAMP in the absence of PKC inhibitors (p = 0.0022 and 0.0022, respectively; Mann-Whitney
U test; Fig. 5, D and E). The delayed
response was abrupt, suggesting it may be caused by sudden activation
of an alternative pathway that circumvents inhibition of PKC activity.
Similar results were obtained when chelerythrine-treated cells were
exposed to 8-bromo-cAMP rather than cpt-cAMP (data not shown). By
contrast, a delayed response did not develop during very long
recordings in the absence of cAMP analogs (range, 31-50 min;
n = 3). These results suggest that the inhibition of
PKC can eventually be overcome by activation of an alternative pathway,
which restores the responsiveness of CFTR channels to PKA.
The CFTR chloride channel is regulated by PKA- and PKC-mediated phosphorylation; however, the relationship between these kinases has not been studied in detail. Previous work suggested that PKC increases the stimulation of po by PKA approximately 2-fold (4); however, no precautions were taken in those early experiments to minimize PKC phosphorylation prior to testing kinase effects. A major finding of the present study is that PKC regulation of CFTR is mainly constitutive, at least in the two mammalian expression systems studied. Using somewhat different protocols we found that PKC not only potentiates responses but is required for acute stimulation of CFTR by PKA. This dependence on PKC was observed in excised and whole cell patches and in intact cells, but only when protocols were designed to allow dephosphorylation of PKC sites, which may explain why a strict dependence on PKC has not been observed previously. Although PKC is permissive and modulates the PKA responses over a wide range in excised patches (po = 0-0.7), basal PKC activity in resting CHO cells appears sufficient to maintain PKA responsiveness near its maximal level.
The present results confirm that forskolin stimulates CFTR channels in
CHO and BHK cells, and that CFTR channel activity declines rapidly when
patches are excised in the absence of PKA, consistent with
dephosphorylation of PKA sites by membrane-associated phosphatases (4).
Moreover, PKA reactivates CFTR chloride channels when added soon after
patches are excised but not when more than 10 min have elapsed. We
attribute this decline in responsiveness to PKA to slow
dephosphorylation of PKC sites, because it did not occur in the
presence of PKC and was reversed by the addition of PKC to the bath.
Although PKC alone caused an apparent, partial (10%) activation in
this and previous studies, the small PKC response may be a
manifestation of the permissive effect of PKC. The early experiments
suggesting activation by PKC alone were not preceded by a lengthy
rundown period such as the one shown in Fig. 1C. Thus the
stimulation by PKC may have been due to the reactivation of channels
that had become dephosphorylated at a permissive PKC site but retained
some residual PKA phosphorylation. Consistent with this explanation,
some CFTR activity was observed in the present work when patches were
excised directly into PKC-containing buffer lacking PKA, comparable
with that reported previously when PKC was added immediately after
channel rundown (po = 0.04, (4) and 0.034 ± 0.0125, (31)). Regulation of Cl secretion in native
epithelia is complicated because PKC has multiple sites of action,
including inhibition of basolateral potassium conductance.
Nevertheless, there are several reports of phorbol ester stimulation of
Cl
secretion in airway and other epithelia (32-36) that
might be explained by basal PKA activity combined with the permissive
effect of PKC described here.
PKA reactivated CFTR channels when added 2 min after excision but not when added 10 min after excision; therefore, the permissive PKC site(s) may be dephosphorylated more slowly than the PKA sites mediating initial rundown of channel activity. The rundown caused by dephosphorylation of PKA sites occurs within 100 s when patches are excised from these cells at room temperature (22). Biochemical studies will be needed to assess whether constitutive PKC phosphorylation is required for phosphorylation of CFTR by PKA as in hierarchical phosphorylation (37) or whether PKA phosphorylation of CFTR is normal in the absence of PKC and only the functional response of the channel to PKA phosphorylation is attenuated. Prestimulating CHO cells with phorbol ester did not noticeably increase phosphorylation of CFTR protein during forskolin stimulation (6); however, the functional responses may depend on trace phosphorylation at cryptic sites, which could escape detection. Interdependent PKC and PKA regulation may allow more precise control of CFTR activity, as would regulation by multiple phosphatases acting at PKA or PKC sites.
Permissive control of cAMP stimulation by PKC was also observed in intact cells; PKC inhibitors greatly reduced forskolin-stimulated CFTR activity in cell-attached patches and abolished acute stimulation of whole cell CFTR currents by cpt-cAMP. Thus constitutive, endogenous PKC activity is sufficient for activation of CFTR by forskolin or cAMP. The fact that exogenous PKC needs to be added to excised patches implies that the endogenous PKC activity responsible for priming CFTR is not membrane-associated, although it remains a formal possibility that PKC activity in the patches is membrane-associated but requires an additional cytoplasmic factor for activity. A membrane-associated PKC isoform has been shown previously to stimulate CFTR channels when added to patches excised from NIH-3T3 cells (13).
CFTR channels were weakly responsive to PKA when cells were briefly
pretreated with chelerythrine (5 instead of >30 min) or when patches
were excised from cells and allowed to run down for only 5 min so that
some PKC phosphorylation would remain. To assess whether the permissive
effect of PKC involved an increase in po or the
number of functional channels under these conditions, we took the
advantage of the previous findings that the poorly hydrolyzable analog
AMP-PNP causes CFTR to become locked in an open burst state in which
po 1.0 (38, 39). Exposure to PKC increased
the po without altering the number of functional
(i.e. PKA-responsive and locked by AMP-PNP) channels in the
patches. This result was most obvious when CHO patches containing only
a few channels were used, since the probability that all the functional
channels would be open simultaneously was highest under these
conditions.
To ensure that increases in N were not missed by studying cells with low CFTR channel density, we also examined BHK cells that contained between 40 and several hundred CFTR channels per patch. When BHK cells were pretreated with chelerythrine for 5 min, subsequent exposure of excised patches to PKA caused only a small stimulation, although the number of functional channels as defined by their susceptibility to locking by AMP-PNP was high and was not increased further by PKC. If exocytotic insertion of CFTR channels is stimulated by PKC, it was not apparent in the patches we studied and did not contribute to the potentiation reported in this article. Kinetic analysis of patches from CHO cells revealed that PKC decreased the mean closed time during PKA stimulation and had no effect on the mean open time. These results are consistent with an increased rate of channel opening, such as the one that accompanies phosphorylation-induced increases in ATP affinity (24, 31). In a CFTR mutant with alterations at all 10 dibasic consensus PKA sequences (10SA; Ref. 6), the reduction in po was associated with a lower bursting rate (i.e. a longer closed time) rather than a shortening of the bursts (24). Strong potentiation by PKC was still observed in the 10SA mutant, which lacks serines at positions 660, 686, and 700; therefore, these dual PKA and PKC sites are not essential for PKC-dependent regulation of PKA responsiveness.
PKC isozymes have been classified into several distinct groups (40).
The "Ca-dependent" group (,
I,
II, and
isozymes) contains a putative
Ca2+ binding site, which is lacking in the
"Ca-independent" group (
,
,
,
, and µ isozymes).
Another "atypical" group (including the
and
isozymes) is
structurally distinct and poorly characterized. Based on its
sensitivity to Gö6976, the permissive regulation in CHO and BHK
cells appears to be mediated by endogenous Ca-dependent
or
I forms. Nevertheless, PKA responsiveness in excised
patches was restored by the addition of rat brain PKC, which consists mainly of the Ca-dependent
II isozyme, and
the role of PKC isozymes in CFTR regulation may vary according to cell
type.
Although the initial experiments in this study suggested that PKA
responsiveness is completely dependent on PKC phosphorylation, we were
surprised to find that whole Cl cell currents suddenly
became responsive to cpt-cAMP in the presence of PKC inhibitors after a
delay of more than 20 min. Since channels in excised patches did not
regain responsiveness to PKA during long exposures (>30 min) in the
absence of PKC, the surrogate pathway that overcomes PKC inhibition in
whole cell patches is probably not membrane-delimited. In whole cell
studies, similar responses were observed with two structurally
unrelated PKC inhibitors, further evidence that inhibition of acute
cAMP stimulation is a bona fide effect on PKC. Also, PKA responsiveness
did not increase gradually in whole cell patches in a manner consistent
with the loss of inhibitor potency. Rather, the delayed response
developed abruptly after 20 min, as if the dependence on PKC was
relieved by activation of an alternate signaling pathway.
In summary, these results indicate that PKA-mediated phosphorylation alone is not a sufficient stimulus to activate the CFTR chloride channel. We propose that under normal conditions, constitutive phosphorylation of CFTR by PKC has a permissive role in priming the channel for acute activation by PKA. Thus maneuvers that lower this constitutive phosphorylation reveal PKC effects that are much more profound than previously suspected. These results have implications for functional studies of CFTR phosphoforms; PKA-stimulated channel activity in membrane patches and bilayers will be dictated by residual PKC phosphorylation. Variations in constitutive phosphorylation by PKC may also help explain the variable responsiveness of CFTR channels to stimulation by PKA in different expression systems (7, 41).
We thank Jie Liao for culturing the cells.