(Received for publication, January 10, 1997, and in revised form, April 3, 1997)
From the Department of Microbiology and Immunology,
National Yang-Ming University School of Medicine, Taipei 11221, Taiwan,
the
Immunology Research Center, National Yang-Ming University
School of Medicine, Taipei 11221, Taiwan, the § Department
of Pharmacology, College of Medicine, National Taiwan University,
Taipei 100, Taiwan, and the ¶ Medical Research Council
Immunochemistry Unit, Department of Biochemistry, University of Oxford,
South Parks Road, Oxford OX1 3QU, United Kingdom
The lymphotoxin- receptor (LT-
R) has been
shown to be the receptor for the membrane-bound lymphotoxin
heterotrimers LT
1/
2 and LT
2/
1. The extracellular domain of
LT-
R shows extensive similarity with members of the tumor necrosis
factor receptor family, while its cytoplasmic domain is distinct and
lacks any inherent enzymatic activity. This suggests that the
interaction of LT-
R with other molecules might be important for
signal transduction. Here we demonstrate the association of a fusion
protein, comprising glutathione S-transferase and the
cytoplasmic domain of LT-
R (GST-LT-
R(CD)), with several proteins
in the size range 29-80 kDa from HepG2 cell lysates. We present
evidence that two of these proteins are serine/threonine kinases, which
associate with amino acids 324-377 of the cytoplasmic domain of
LT-
R and phosphorylate this receptor. The characteristics of these
novel kinases indicate that they are distinct from the previously
described tumor necrosis factor receptor-associated kinases. This
suggests the presence of novel signal transduction pathway(s) for
LT-
R.
Tumor necrosis factor (TNF)1 is a
pleiotropic factor which induces an extraordinarily wide variety of
biological responses, mainly related to immune regulation and
inflammation (1, 2). Two structurally-related molecules, TNF
and
lymphotoxin (LT)
(also known as TNF
), are encoded by two genes
located just 1 kilobase apart in the class III region of the human
major histocompatibility complex (3). Both TNF
and LT
can bind to
tumor necrosis factor receptors type I and type II (TNFRI and TNFRII)
to mediate similar biological effects. Recently, a member of the TNF/LT
family known as LT
, which is encoded by a gene lying 4 kilobases
centromeric to the TNF
/LT
genes, was found to be expressed on the
surface of natural killer (NK) cells, lymphokine-activated killer (LAK) cells, and activated T and B lymphocytes (3-5). In contrast to the
homotrimeric structures of TNF
(which occurs as membrane-bound and
secreted forms) and LT
(which is secreted), LT
is a 33-kDa transmembrane protein, which associates with LT
and anchors it to
the cell surface. Stoichiometrically, membrane lymphotoxin is a
heterotrimer and is composed of either LT
1/LT
2 (major form) or
LT
2/LT
1 (minor form) (6). The LT
2/LT
1 heterotrimer has been
shown to bind to TNFRI, TNFRII, and LT-
R, while the LT
1/LT
2 heterotrimer binds only to LT-
R (also known as TNFRIII, or
TNFR-related protein (TNFRrp)) (6-8).
In contrast to TNFRI and TNFRII, the functions of LT-R are still not
well understood. In LT
knock-out mice, which lack the membrane form
of lymphotoxin, normal development of peripheral lymph nodes is ablated
(9, 10). Since this phenotype is not observed in either TNFRI or TNFRII
knock-out mice (11-13), it has been speculated that signaling through
LT-
R might be involved in the development of lymphoid organs.
Furthermore, stimulation of LT-
R by immobilized anti-LT-
R
monoclonal antibodies has been shown to trigger activation of NF-
B
in HT-29 (human adenocarcinoma) and WI38 (human lung fibroblast) cells
(14) and signaling through LT-
R in conjunction with interferon-
can induce cell death through apoptosis in some adenocarcinoma cell
lines (15). These studies provide evidence that LT-
R can transduce
signals that trigger both cell proliferation and apoptosis. The
cytoplasmic domain of LT-
R, like other members of the TNF receptor
family, does not contain consensus sequences characteristic of known
kinases or any motifs associated with signal transduction. Therefore, kinases or other proteins associated with the cytoplasmic domain of
LT-
R are likely to be involved in the signal transduction pathway.
Recently, two members of the TNF receptor-associated factor (TRAF)
family, TRAF3 and TRAF5, were found to associate with LT-
R (16, 17).
There is evidence that TRAF5 is involved in the activation of NF-
B
(17), but the role of TRAF3 in the LT-
R signaling pathway has not
been reported.
While members of the Janus kinase (JAK) family have been shown to play
a key role in signal transduction via many cytokine receptors (18-20),
the nature of the kinases involved in the signal transduction pathways
of the TNF receptor family has only recently been investigated. Several
groups have shown that the cytoplasmic domains of both TNFRI and TNFRII
can associate with serine/threonine kinases. Darnay et al.
(21, 22) have identified serine/threonine kinases that specifically
associate with fusion proteins comprising glutathione
S-transferase and the cytoplasmic domain of TNFRI (GST-p60)
and TNFRII (GST-p80). Furthermore, the TNFRI-associated kinase
(p60-TRAK) was shown to bind residues 344-397 of the TNFRI cytoplasmic
domain (death domain) and to be involved in TNF-induced signaling (23).
The TNFRI-associated kinase activity was also demonstrated in
immunoprecipitates from U937 cell lysates using an anti-TNFRI antibody
(24). Another TNFRI-associated protein, RIP, has been shown to be a
serine/threonine kinase, which contains a kinase domain in its N
terminus and a death domain in its C terminus. The
TNF-dependent recruitment of RIP to the TNFRI signal complex is mediated by a TNFRI-associated death domain-containing protein (TRADD) and has been shown to induce apoptosis and NF-B activation following receptor triggering (25). However, it is unclear
whether or not RIP can phosphorylate TNFRI. The TNFRII-associated kinase has been identified as casein kinase-1 (CK-1) (26) and CKI-7, an
inhibitor of CK-1, has been shown to inhibit the phosphorylation of
TNFRII and to potentiate TNF-induced apoptosis. This evidence suggests
that CK-1 is involved in signal transduction through TNFRII. Based on
the information above, we wished to determine whether LT-
R is
associated with kinase activity. Here we report the identification of
two serine/threonine kinases specifically associated with LT-
R and
demonstrate that these can phosphorylate LT-
R, but not TNFRI. These
novel kinases are likely to be important in LT-
R signal transduction
pathway(s).
Mammalian cells were grown in Dulbecco's modified Eagle's medium, supplemented with 10% (v/v) fetal bovine serum, in a 37 °C incubator containing 5% (v/v) CO2. For radioisotope labeling, cells were washed twice with Hank's balanced salt solution (8 g/liter NaCl, 0.4 g/liter KCl, 0.2 g/liter MgSO4·7H2O, 60 mg/liter Na2HPO4, 60 mg/liter KH2PO4, 1 g/liter glucose, 140 mg/liter CaCl2), then incubated in cysteine/methionine-free medium containing 10% (v/v) dialyzed fetal calf serum (Life Technologies, Inc.) for 1 h, followed by addition of 50 µCi/ml [35S]cysteine/methionine (NEN Research Products) and incubation at 37 °C for 6 h before harvesting.
Construction of GST Expression VectorsTo construct
pGST-LT-R(CD), a 540-base pair PstI/HindIII
cDNA fragment (nucleotides 767-1305) encoding almost the full
length of the human LT-
R cytoplasmic domain was amplified by reverse transcription-PCR using a HepG2-derived cDNA template and the primers 5
-CGGGATCCATGCTCCTGCCTTGGGCCAC-3
(sense) and
5
-GCGGATCCTGGGGGCAGTGGCCTAATGG-3
(antisense). The PCR product was
end-filled, cloned into SmaI-cut pGEX-2T vector, and
sequenced to confirm that the pGST-LT-
R(CD) construct contained no
deleterious PCR errors. Deletion mutants were generated by restriction
enzyme digestion and subsequent religation of pGST-LT-
R(CD).
The
expression and purification of GST-fusion proteins were carried out as
described by Darnay et al. (21) with the following modifications. JM109 bacteria containing the pGST-LT-R(CD) were grown to A600 = 0.2 at 37 °C before adding
isopopyl-1-thio-
-D-galactopyranoside to 0.1 µM. After 3 h of induction, bacteria were lysed with
buffer A (20 mM Tris, pH 8.0, 200 mM NaCl, 10%
(v/v) glycerol, 1 mM phenylmethylsulfonyl fluoride, 2 µg/ml leupeptin, 2 µg/ml aprotonin, 0.1% (v/v) 2-mercaptoethanol, 0.5% (v/v) Nonidet P-40) containing 100 µg/ml lysozyme. After brief
sonication, the bacterial lysate was centrifuged at 3500 × g, for 5 min at 4 °C to remove insoluble debris. The
supernatant was incubated with 50% (v/v) glutathione-agarose beads
(Sigma) at 4 °C for 30 min on a rotatory shaker. The slurry was then
washed five times with buffer A, and stored as a 50% (v/v) slurry at 4 °C.
The in vitro binding assay was modified from
that described by Darnay et al. (21). Briefly, 1 × 107 [35S]cysteine/methionine-labeled HepG2
cells were resuspended in 1 ml of lysis buffer (20 mM Tris,
pH 7.7, 0.5% (v/v) Nonidet P-40, 200 mM NaCl, 50 mM NaF, 0.2 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 2 µg/ml leupeptin, 2 µg/ml aprotonin, and 0.1% (v/v) 2-mercaptoethanol) for 1 h at
4 °C, followed by centrifugation at 9000 × g for 10 min to remove cell debris. Before incubation with the GST-LT-R(CD)
fusion protein, the supernatant was precleared with 15 µg of GST and
50 µl of 50% (v/v) glutathione-agarose beads for 2 h at
4 °C, followed by centrifugation at 800 × g for 3 min at 4 °C. The supernatant was then transferred to a fresh tube and mixed for 2 h with 10 µg of GST-LT-
R(CD) attached to 50 µl of 50% (v/v) glutathione-agarose. Finally, the
glutathione-agarose beads were washed six times with lysis buffer. The
sample was fractionated by SDS-PAGE (10% (w/v) acrylamide). Gels were
dried and analyzed by autoradiography.
In vitro kinase assays
were carried out as described by Darnay et al. (21).
Following incubation with HepG2 cell lysates GST-LT-R(CD) was
incubated with 50 µl of kinase reaction buffer (20 mM
HEPES, pH 7.4, 10 mM MgCl2, 0.2 mM
NaF, 0.1 mM sodium orthovanadate) and 10 µCi of
[
-32P]ATP for 10 min at 37 °C, in the presence or
absence of Mg2+ or Mn2+. Reactions were stopped
by addition of Laemmli's sample buffer. Samples were fractionated by
SDS-PAGE (10% (w/v) acrylamide), and proteins were visualized by
Coomassie Blue staining. Phosphorylated proteins were identified by
autoradiography.
In-gel Kinase AssayThe in-gel kinase assay was modified
from that developed by Kameshita and Fujisawa (27). Samples were fractionated on SDS-polyacrylamide gel (10% (w/v) acrylamide) containing GST-LT-
R(CD) fusion protein. The gel was then washed with
20% (v/v) propan-2-ol to remove SDS. The gel was denatured with 6 M guanidine HCl, then renatured in 5 mM
2-mercaptoethanol, 50 mM Tris-HCl, pH 8.0, 0.04% (v/v)
Tween 40. This was followed by washing at 4 °C for 30 min with two
changes of buffer (20 mM HEPES, pH 7.4, 0.2 mM
NaF, 0.1 mM sodium orthovanadate), and incubation at
22 °C for 3 h in the kinase reaction buffer containing 50 µCi of [
-32P]ATP. After removal of unreacted
[
-32P]ATP, protein kinases were visualized by
autoradiography of the dried gel. Signals due to protein kinases were
also detected and quantified using a PhosphorImagerTM (Molecular
Dynamics) to allow comparison of signal strengths.
Phosphoamino Acid AnalysisPhosphoamino acid analysis was
carried out as described by Guo et al. (28). Samples were
transferred to ImmobilonTM-P membrane after fractionation by SDS-PAGE
(10% (w/v) acrylamide). A strip of membrane containing
phospho-GST-LT-
R(CD) was hydrolyzed in 6 N HCl for 90 min at 110 °C, then dried by speed vacuum. After resuspension in
H2O, samples were spotted onto Whatman KC2
ethyl reverse-phase thin layer chromatography (TLC) plates and then
analyzed by ascending TLC in buffer containing 40% (v/v) methanol,
1.5% (v/v) acetic acid, and 0.5% (v/v) formic acid. The migration
standards were visualized by ninhydrin staining.
To identify
and characterize the proteins associated with LT-R, we fused the
cytoplasmic domain of LT-
R with glutathione S-transferase
and expressed the GST-LT-
R(CD) fusion protein in Escherichia
coli. After purification with glutathione-conjugated agarose
beads, a protein migrating at 50 kDa was observed by SDS-PAGE. This
corresponds to the predicted molecular mass of the fusion protein.
Since HepG2 cells have been shown to contain abundant LT-
R
receptors, both by reverse transcription-PCR and immunoprecipitation analysis,2 the purified GST-LT-
R(CD)
fusion protein was incubated with [35S]cysteine/methionine-labeled HepG2 cell lysates to
identify proteins that bind to the cytoplasmic domain of LT-
R. After
incubation, several proteins with molecular mass values of 75-80, 61, 50, 43, 34, and 29 kDa were found to be associated specifically with GST-LT-
R(CD) (Fig. 1, lane 3), but not
with GST under the same conditions (Fig. 1, lane 2) or with
GST-LT-
R(CD) in the absence of HepG2 cell lysates (data not
shown).
To test for the presence of kinase activity in the proteins associated
with GST-LT-R(CD), we carried out an in vitro kinase assay. As shown in Fig. 2, a 50-kDa protein
corresponding to the molecular mass of GST-LT-
R(CD) was
phosphorylated after incubation with HepG2 cell lysates in the in
vitro kinase assay (Fig. 2B, lane 1).
However, there was no evidence that GST was phosphorylated under the
same condition (Fig. 2B, lane 3), nor was the
GST-LT-
R(CD) fusion protein phosphorylated in the absence of HepG2
cell lysates (Fig. 2B, lane 2). This result
suggests that the cytoplasmic domain of LT-
R does not have intrinsic
kinase activity, but that it might associate with one or more kinases
found in HepG2 cell lysates. The smaller phosphorylated proteins
(visible in lane 1 of Fig. 2B) may be
phosphoproteins associated with GST-LT-
R(CD), but are more likely to
be due to degradation of GST-LT-
R(CD) since the same species were
also observed on the SDS-polyacrylamide gel following Coomassie Blue
staining (Fig. 2A). Furthermore, an 80-kDa phosphoprotein
(p80) was found to associate with GST-LT-
R(CD) in the in
vitro kinase assay (Fig. 2B, lane 1).
Similar results were obtained using lysates from other cell lines
expressing LT-
R, including K562, U937, and HeLa (data not
shown).
Serine/Threonine Kinase Activity Associated with LT-
The
LT-R-associated kinase activity was found to be optimal in the
presence of Mg2+ or Mn2+ (Fig.
3A) and was inhibited by the serine/threonine
protein kinase inhibitor staurosporine. In contrast, the kinase
activity was not inhibited by calphostin C, which is highly specific
inhibitor of protein kinase C. In addition, neither genistein nor
tyrophostin A25 inhibited the LT-
R(CD)-associated kinase activity,
although both of these reagents have a broad spectrum inhibitory effect on protein tyrosine kinases (Fig. 3B). Phosphoamino acid
analysis showed that serine residues were phosphorylated as were
threonine residues, but to a lesser extent (Fig. 3C).
However, there was no incorporation of phosphate into tyrosine
residues. These observations provide direct evidence that a
serine/threonine kinase activity associates with the cytoplasmic domain
of LT-
R.
Characterization of the Substrate Specificity of the LT-
It has been shown that the
TNFRII-associated kinase can phosphorylate the cytoplasmic domain of
TNFRI, but not vice versa (21). We, therefore, investigated
whether the LT-R-associated kinase could phosphorylate TNFRI, which
is abundantly expressed in HepG2 cells (30). As shown in Fig.
4, A and B, the kinase associated
with the GST-LT-
R(CD) deletion mutant
6 (amino acids 324-377,
Fig. 6) could phosphorylate the 50-kDa wild type GST-LT-
R(CD) (Fig.
4, A and B, lane 3), while GST and the
60-kDa GST-TNFRI(CD) fusion protein could not be phosphorylated under
the same conditions (Fig. 4, A and B, lanes
1 and 2). The kinase activity associated with
GST-LT-
R(CD) was able to phosphorylate other substrates, such as
histone H1 and myelin basic protein (MBP), but not casein and bovine
serum albumin (Fig. 4, C and 4D).
Characterization of the LT-
To further characterize the kinase activity associated with
LT-R, an in-gel kinase assay was performed. With GST-LT-
R(CD) as
substrate, two signals of 50 and 80 kDa (Fig.
5A, lane 4) were observed,
corresponding to proteins specifically associated with LT-
R, while a
third protein of 82 kDa was also found to associate with both GST (Fig.
5A, lane 3) and GST-LT-
R(CD) (Fig.
5A, lane 4). This result demonstrates the
presence of two kinases, p50 and p80, specifically associated with
LT-
R. To determine whether these two kinases are able to
autophosphorylate, the same experiment was carried out in the absence
of substrate in the gel. A similar result was observed, but the signals
detected were less than one-tenth of the strength of those seen when
GST-LT-
R(CD) was present in the gel (Fig. 5B). Based on
the observations above, it is clear that both p50 and p80 are
LT-
R(CD)-associated kinases, which are able to autophosphorylate and
to mediate phosphorylation of LT-
R. An in-gel kinase assay was also
carried out with MBP as substrate, and signals of 50 kDa and 80 kDa
were observed (data not shown), confirming that p50 and p80 can
phosphorylate MBP.
Kinase Activity Associated with Amino Acid Residues 324-377 of LT-
We wished to identify the region(s) of LT-R to which
the novel kinases, p50 and p80, bind. For this purpose, we constructed several deletion mutants of GST-LT-
R(CD) (Fig.
6A) and expressed the resultant fusion
proteins in E. coli (Fig. 6B). No phosphoproteins were detected following in vitro kinase assays using the
deletion mutants
1,
8 (Fig. 6C),
2, and
7 (data
not shown). In contrast, it is clear that kinase activity does
associate with deletion mutants
3,
4,
5, and
6 (Fig.
6C). These results suggest that the minimal region for
kinase association is contained within the 54 amino acids present in
deletion mutant
6. It is possible that other regions of LT-
R
might be able to associate with the putative kinase, but are not
phosphorylated by it. To rule out this possibility, we incubated the
deletion mutants
1,
2,
7, and
8 with HepG2 cell lysates. We
then performed in vitro kinase assays with wild type
GST-LT-
R(CD) as substrate. GST-LT-
R(CD) was not phosphorylated
under these conditions (data not shown). Therefore, we conclude that
the minimal region for kinase association is located between amino
acids 324 and 377 of LT-
R. It is interesting to note that both p50
and p80 were also detected when using deletion mutants
3,
4,
5, and
6 as baits in the in vitro kinase assays, but
were absent when using
1,
2,
7, and
8 in the same
experiment (Fig. 6C). This result supports the co-existence
of kinase activity with p50 and p80.
Using the yeast two-hybrid and GST fusion protein systems, many
proteins have been shown to associate with TNFRs. These include: the
tumor necrosis receptor-associated factors, TRAF1-6 (16, 17, 31-33);
tumor necrosis factor receptor-associated proteins, TRAP-1 and TRAP-2
(34); proteins containing death domains, such as TRADD and RIP (25,
35); p60-TRAK, a TNFRI-associated kinase (21); and p80-TRAK, a
TNFRII-associated kinase (22). Among these TNFR-associated proteins,
however, only TRAF3 and TRAF5 have been shown to associate with LT-R
(16, 17). Therefore, we wished to determine whether there are other
proteins associated with LT-
R. Using GST-LT-
R(CD) as a bait, we
have identified several proteins with molecular mass values of 75-80,
61, 50, 43, 34, and 29 kDa that associate with GST-LT-
R(CD) (Fig.
1). Based on molecular mass, the 61-kDa protein might correspond to TRAF3 (62 kDa) and/or TRAF5 (64 kDa). In addition, the results of both
in-gel kinase assays and in vitro kinase assays show the presence of two kinases, p80 and p50, associated with the cytoplasmic domain of LT-
R (Figs. 5 and 6). It is likely that the 80- and 50-kDa
proteins from [35S]cysteine/methionine-labeled HepG2 cell
lysates, that bound to GST-LT-
R(CD) (Fig. 1), correspond to the two
LT-
R-associated kinases, p80 and p50, respectively.
We are confident that the LT-R-associated kinases are distinct from
the serine/threonine kinases that associate with TNFRI and TNFRII for
the following reasons. (i) The LT-
R-associated kinases could only
use LT-
R and not TNFRI as substrate in the in vitro
kinase assay; (ii) the TNFRII-associated kinase is known to
phosphorylate TNFRI, but this phenomenon was not observed for the
LT-
R-associated kinases; (iii) the activity of the
LT-
R-associated kinases could be inhibited by staurosporine, while
the TNFRII-associated kinase activity is only inhibited by CKI-7 and
not by staurosporine (26); (iv) the specificities of these
receptor-associated kinases for other standard kinase substrates, such
as histone H1, casein, and MBP, are all distinct from each other. The
LT-
R-associated kinases can phosphorylate histone H1 and MBP, but
not casein. By contrast, the TNFRI-associated kinase can phosphorylate
both casein and histone H1, but not MBP (21), while the
TNFRII-associated kinase can use MBP as substrate, but not casein or
histone H1 (22).
Receptor clustering is crucial in the activation of the TNFRs. Using
the yeast two-hybrid system, TNFRI (but not TNFRII) has been shown to
self-associate (36, 37), and it has been suggested that this is a
consequence of receptor phosphorylation. Darnay et al. (24)
used GST-TNFRI as bait to identify two phosphoproteins, p55 and p58,
associated with TNFRI. The molecular masses of these species suggested
that they may correspond to endogenous TNFRI (60 kDa). However, this
could not be proved because the anti-TNFRI antibody fails to detect
either p55 or p58 on an immunoblot (24). Using a similar approach, we
found an 80-kDa phosphoprotein associated with GST-LT-R(CD) in the
in vitro kinase assay. Since the self-association of the
intracellular domain of LT-
R has been observed in the yeast two
hybrid system,3 it was possible that p80
corresponded to the endogenous LT-
R, which has a molecular mass of
75-80 kDa (38). To test this, we used anti-LT-
R antiserum to
preclear endogenous LT-
R from HepG2 cell lysates before incubation
with GST-LT-
R(CD). In subsequent in vitro kinase assays,
the intensity of the p80 phosphoprotein was unchanged despite the
complete removal of endogenous LT-
R from the HepG2 cells (data not
shown). Therefore, we conclude that the p80 phosphoprotein is not the
endogenous LT-
R.
The kinase activities associated with TNFRI and TNFRII have been shown
to be up-regulated following TNF stimulation (21-23). Therefore, the
effect of ligand-receptor interaction on the LT-R-associated kinase
activity was examined by treating HepG2 cells with either immobilized
anti-LT-
R antibodies or activated RAMOS cells. The latter have been
shown to express the membrane-bound LT
1/
2 heterotrimer following
phorbol 12-myristate 13-acetate stimulation (6). However, the
LT-
R-associated kinase activities were not affected by stimulation
with either anti-LT-
R or phorbol 12-myristate 13-acetate-activated
RAMOS cells (data not shown).
When using an anti-LT-R monoclonal antibody to precipitate
endogenous LT-
R from HepG2 cells for the in vitro kinase
assay, we observed that the endogenous LT-
R could only be
phosphorylated after treatment of the immunoprecipitates with alkaline
phosphatase.4 This suggests that LT-
R is
constitutively phosphorylated in the intact cells and supports the
argument that the kinase activity could be co-precipitated with
endogenous LT-
R, which is the primary target for its associated
kinases. Similarly, Beyaert et al. (26) and Pennica et
al. (29) immunoprecipitated TNFRII and observed no increase in
phosphorylation of this receptor following TNF stimulation. This led
them to propose that TNFRII is constitutively phosphorylated.
This paper represents the first report of two serine/threonine kinases,
p80 and p50, that associate with LT-R and mediate its
phosphorylation. To further understand the roles of these novel kinases
in LT-
R signaling pathway(s), a yeast two-hybrid system will be
employed to identify the genes encoding p50 and p80. The mechanism by
which LT-
R transduces signals after interaction with membrane-bound
lymphotoxin also remains to be elucidated.
We thank Caroline Milner for critical review of the manuscript. Special thanks also go to Nien-Jung Chen, Shen-Chih Mai, and Hsu-Chin Chen for technical assistance.