(Received for publication, April 17, 1997, and in revised form, May 30, 1997)
From the P-glycoprotein, the overexpression of which is a
major cause for the failure of cancer chemotherapy in man, recognizes
and transports a broad range of structurally unrelated amphiphilic compounds. This study reports on the localization of the binding site
of P-glycoprotein for iodomycin, the Bolton-Hunter derivative of the
anthracycline daunomycin. Plasma membrane vesicles isolated from
multidrug-resistant Chinese hamster ovary B30 cells were photolabeled
with [125I]iodomycin. After chemical cleavage
behind the tryptophan residues, 125I-labeled peptides were
separated by electrophoresis and high performance liquid
chromatography. Edman sequencing revealed that [125I]iodomycin had been predominantly incorporated into
the fragment 230-312 of isoform I of hamster P-glycoprotein. According
to models based on hydropathy plots, the amino acid sequence 230-312
forms the distal part of transmembrane segment 4, the second
cytoplasmic loop, and the proximal part of transmembrane segment 5 in
the N-terminal half of P-glycoprotein. The binding site for iodomycin is recognized with high affinity by vinblastine and cyclosporin A.
P-glycoprotein, which belongs to the large family of
ABC1 transporters (1), binds
and transports a broad range of structurally unrelated compounds (2).
Its overexpression may cause the phenomenon of multidrug resistance
(MDR) during cancer chemotherapy, whereby the tumor cells become
resistant to a variety of antineoplastic agents due to a reduced
intracellular accumulation of drugs (2-4). The MDR phenotype can be
overcome by modulators, i.e. substances that are bound by
P-glycoprotein and inhibit its drug excluding function (3). The
medically important substrates of P-glycoprotein comprise anticancer
drugs such as Vinca alkaloids and anthracyclines (2-4) and
approved drugs that turned out to be potent modulators such as
verapamil (5, 6), calcium antagonists (7), and cyclosporins (8).
The major issue of how P-glycoprotein can handle so many substrates has
been mainly approached by photoaffinity labeling and mutagenesis
studies. Mutants that arose spontaneously during drug selection of
cells and thereby changed their resistance profile were detected in the
first cytoplasmic loop of human MDR1 protein (Gly185 Hydropathy plots deduced from the primary sequence predict that
P-glycoprotein consists principally of two symmetrical halves, each of
which contain a membrane domain with six membrane-spanning segments and
a subsequent cytosolic nucleotide binding fold (18-20). The topology
of P-glycoprotein in vivo, however, may be more variable than predicted, e.g. under certain experimental conditions,
TM segments were detected outside the membrane and loops postulated to
face the cytosol were found on the extracellular side (21-25).
Although the three-dimensional structure of the drug binding pocket(s)
is unknown, substantial progress has been made in the localization of
the binding site(s) in the primary sequence by mutagenesis (see above)
and by photoaffinity labeling. Photoreactive derivatives of cyclosporin
(26), daunomycin (27), verapamil (28), prazosin (29), Vinca
alkaloids (30, 31), and calcium channel blockers (32) have been shown
to become photoincorporated into P-glycoprotein. Typically the
sensitivity and specificity of the interaction between drugs and
P-glycoprotein has been substantiated by the inhibition of
photoaffinity labeling by substrates of P-glycoprotein such as
vinblastine or verapamil. The affinity of cytotoxic agents to compete
for the photobinding site decreases for most photoreactive compounds,
in the order vinblastine > daunomycin > colchicine (29). In
the case of the photoreactive 1,4-dihydropyridine derivatives azidopine
(29, 33, 34), iodoarylazidoprazosin (35-37), and AIPP-forskolin (38),
the binding sites were allocated by immunoreactivity of enzymatic
digests to epitope-specific antibodies in or close to TM6 and TM12,
i.e. domains of the molecule which, from structure-function comparisons with other ABC transporters, are assumed to be directly involved in the transport of substrate (12).
Iodomycin is the Bolton-Hunter derivative of the inherently
photoreactive anthracycline daunomycin (27) (Fig.
1). Iodomycin binds to P-glycoprotein in
plasma membrane vesicles and intact cells with an affinity of
107 to 108 M
In this study we describe the localization of the major binding site
for [125I]iodomycin in multidrug-resistant CHO B30 cells.
Photolabeled P-glycoprotein was cleaved behind its tryptophan residues
with BNPS-skatol (40), and the resulting peptides were purified by electrophoresis and HPLC. Edman sequencing revealed that the
binding site for iodomycin resides in a region in the N-terminal half of the protein that is distinct from the binding site described for
[125I]iodoarylazidoprazosin (37).
CHO B30 cells were grown in For the isolation
of plasma membranes, culture medium was replaced by 1 mM
EGTA, 2 mM EDTA, 5 mM dithiothreitol, 20 mM Hepes buffer, pH 7.4. Cells from 50 tissue culture
plates (diameter, 9 cm) were scraped off with a rubber policeman and
disrupted by ultrasonication until 80% of the cells were lysed as
checked by light microscopy. After removal of cell debris by
centrifugation (4,000 × g for 10 min), the membraneous
fraction of the supernatant was collected by centrifugation at
40,000 × g for 1 h. The pellet was subsequently
centrifuged for 90 min (40,000 × g) through a 35%
sucrose cushion. After resuspension of the interphase in 5 mM Tris buffer, pH 7.5, the plasma membrane fraction was
collected by a spin of 100,000 × g for 1 h. The
protein pellet was solubilized in 5 mM Tris-HCl, pH 7.5, with 8.6% sucrose (w/v) and stored at Metaphor agarose (FMC) was dissolved in 500 mM
Tris, 160 mM borate, pH 8.5, to give a concentration of 7%
for the resolving gel (2/3) and 3.5% for the stacking gel (1/3). The
boiled agarose was casted vertically in a 12.5 × 8.5 × 0.2-cm large chamber consisting of two prewarmed scratched glass
plates. A maximum of 450 µg of ethanol-precipitated membrane protein
dissolved in 62.5 mM Tris-HCl pH 6.8; 7.5% (w/v) FicollTM
type 400; 0.001% (w/v) bromphenol blue, and 7 mg/ml dithiothreitol was
loaded onto the agarose gel. Electrophoresis was run at 8 V/cm for
1 h and then 12 V/cm for 3 h at 4 °C. The anodal buffer
consisted of 90 mM Tris, pH 8.5, 90 mM borate,
and 0.1% SDS. The cathodal buffer consisted of 9 mM Tris,
pH 8.5, 9 mM borate, and 0.1% (w/v) SDS. The agarose
stripe containing P-glycoprotein was cut out of the gel and chopped
into small pieces, collected in an Eppendorf tube, covered with a
10-fold excess of buffer (50 mM Tris, pH 10.4, 3 mM EDTA, 0.5% (w/v) N-lauroylsarcosinate), and
melted at 80 °C for 15-30 min until the suspension appeared homogeneous. The samples were solidified in an ice bath and stored for
3 h at The presence of P-glycoprotein was
detected on immunoblots of 6% PAGE-separated membrane proteins (42)
with polyclonal rabbit antiserum mdr (AB-1) (Oncogene Science) or
monoclonal antibody C219 (Signet) as the primary antibody and
CPDSTAR chemiluminescence in strict accordance with a
protocol of the manufacturer (Tropix). Purified P-glycoprotein was also
detected on immunoblots with the primary mouse monoclonal antibody
C494 (Signet).
[125I]Iodomycin (2000 Ci/mmol) was prepared from daunomycin by reaction with
125I-labeled Bolton-Hunter reagent (ICN) and subsequently
purified as described previously (27) (Fig.
1). The absence of contamination with
unreacted Bolton-Hunter reagent and daunomycin in the purified product
was controlled by thin layer chromatography (27). For photoaffinity
labeling, maximal 75 µg of plasma membrane protein from CHO cells and
125I-labeled iodomycin (0.075 pmol for the competition
experiment and 0.75 pmol for the cleavage experiments) were suspended
in 150 µl in a quartz cuvette and were illuminated for 15 min with visible light, emitted from a 500-watt xenon lamp that had passed through two 3-cm filters of water and a saturated aqueous
CuSO4 solution. For the competition experiment with
cyclosporin A, variable concentrations of the drug were added prior to
the photoreaction. Protein was precipitated with ice-cold ethanol
(final 70% (v/v)) at
For the detection of
radioactivity the gels were either exposed to Kodak X-Omat AR film or
to a Fuji imaging plate type BAS-IIIs. Liquid samples of
[125I]iodomycin-labeled proteins and peptides were
measured in a Compu Mild tryptic fragmentation (43) cleaves
P-glycoprotein in plasma membranes into two large fragments. Plasma
membrane protein was digested with trypsin in a volume of 100 µl and
an enzyme to a substrate ratio of 1:50 for 30 min at 37 °C.
Plasma membrane proteins were
cleaved either with N-chlorsuccinimide (44) or BNPS-skatol
(40). For the cleavage procedure with N-chlorsuccinimide, 75 µg of plasma membrane protein in 150 µl were incubated immediately
after the photoreaction for 30 min at 37 °C with 75 µl of
H2O and 150 µl of the following mixture: 1 mg of
N-chlorsuccinimide dissolved in 500 µl of H2O,
500 mg of urea, and 500 µl of acetic acid. The reaction products were precipitated with ice-cold ethanol and analyzed on a 4%/12%
discontinuous Tricine-PAGE (45). For the cleavage procedure with
BNPS-skatol, the precipitated protein was dissolved in 7.5 µl of
H2O and 22.5 µl of the following mixture: 1.3 mg of
BNPS-skatol (Fluka) dissolved in 1 ml of 100% acetic acid. The
reaction was carried out at 47 °C for 18 h. The digest was
stopped by diluting the mixture with 24 µl of H2O. The
cleavage products were concentrated in an evaporator and analyzed on a
4%/12% discontinuous Tricine-PAGE. For a preparative gel, the 15 cm × 17 cm × 0.1-cm gel was loaded with 2 mg of total membrane protein or 200 µg of purified P-glycoprotein.
According to the signals generated by photoimaging, the
125I-labeled peptide of lowest molecular weight was
gel-eluted onto a 3-kDa cut-off membrane in the Centrilutor (Amicon,
catalog no. 1047601) for 12 h at 80 V and 4 °C in a glycine
buffer (42). Next, excess volume was removed by centrifugation of the
Centricon tubes at 4 °C at 7,500 × g. This solution
was washed three times by ultrafiltration with the maximum possible
volume of 5 mM Tris-HCl, pH 7.5, to remove SDS to the
greatest possible extent. Subsequently the concentrated peptide
solution was removed from the filtration unit and supplemented with
guanidinium hydrochloride to a final concentration of 6 M.
The membranes of the Centricon units were washed each with 15 µl of 6 M guanidinium hydrochloride in 5 mM Tris-HCl to
avoid any loss of peptide. The collected fractions were separated on an
Aquapore phenyl 7-µm 30 × 2.1-mm Brownlee column by HPLC
reversed phase chromatography (Applied Biosystems 0711-0068). The
linear discontinuous gradient was generated by mixtures of solution A
(0.1% trifluoroacetic acid) and B (60% acetonitrile, 20% 2-propanol;
0.085% trifluoroacetic acid) in Milli Q water: 1 min 1-10% B; 5 min
10-25% B; 50 min 25-50% B, 65 min 50-65% B; 70 min 65-70% B; 85 min 70-99% B (Applied Biosystems/BAI HPLC system, flow rate 0.1 ml/min, detection 214 nm).
Sequence analysis was carried out
with the Applied Biosystems model 477A apparatus according to
Bökenkamp et al. (46).
This P-glycoprotein-overexpressing cell
line (41) grown in medium supplemented with 30 µg/ml colchicine has a
P-glycoprotein content of 10-20% of total membrane protein (Fig.
3A), which makes B30 cells a
suitable source for the large scale purification and characterization
of P-glycoprotein.
To facilitate the
identification of cleavage products of digested P-glycoprotein (see
below), we developed a simple and fast protocol to purify this very
hydrophobic membrane protein. Electrophoretic separation of membrane
proteins in Metaphor® agarose (47) and recovery of
P-glycoprotein by a freeze-squeeze procedure in the presence of the
detergent N-lauroylsarcosine turned out to be a powerful
tool to isolate P-glycoprotein at a preparative scale. 450 µg of
plasma membrane protein yielded between 20 and 40 µg of 70-99% pure
P-glycoprotein, whereby both yield and purity depended on strict
adherence to the protocol (Fig. 3A, lane 4).
Fig. 3B shows the immunoblot of purified iodomycin-labeled
P-glycoprotein with polyclonal antibody mdr AB-1 (Fig. 3B,
left) and monoclonal antibody C494 (Fig. 3B,
right). The latter antibody only recognizes human and
hamster P-glycoproteins of class I (48).
Iodomycin, the
Bolton-Hunter derivative of the anthracycline daunomycin (Fig. 1),
belongs to the compounds to which multidrug-resistant CHO cells are
cross-resistant (27). The inherently photoreactive [125I]iodomycin is incorporated with high specificity
into P-glycoprotein in plasma membranes of CHO B30 cells (Fig. 2,
lane 1). Photolabeling was competitively inhibited by
typical substrates of P-glycoprotein with differential efficacy; for
example, more than 10 µM colchicine, but less than 0.5 µM vinblastine or cyclosporin (Fig. 2) were necessary to
cause 50% reduction of [125I]iodomycin photobinding (17,
27, 39, 49). According to these data, the photobinding site of
P-glycoprotein for iodomycin represents a high affinity site for
cyclosporin and vinblastine, but a low affinity site for
colchicine.
Our next step was to investigate in
which part of the protein the photoreaction takes place. According to
the protocol published by Georges et al. (43), it is
possible to cleave P-glycoprotein in plasma membrane vesicles by mild
tryptic digestion into a ~110-kDa N-terminal fragment and a ~65-kDa
C-terminal fragment. Both fragments were detected in immunoblots by
monoclonal antibody C219, which recognizes a common epitope in the two
nucleotide binding folds (data not shown).
[125I]Iodomycin predominantly labeled the 110-kDa
N-terminal fragment and only a minor portion of radioactive drug was
photoincorporated into the C-terminal fragment (Fig.
4).
After the major photobinding site had been
assigned to the N-terminal half of P-glycoprotein, its localization was
refined by internal cleavage at the tryptophan residues with
N-chlorsuccinimide (NCS) or BNPS-skatol. Since the amino
acid sequence of hamster P-glycoprotein contains only 11 tryptophan
residues (Table I), the chemical cleavage
of Trp-X peptide bonds was expected to result in fewer fragments than
the enzymatic digestion with one of the commonly employed
endoproteinases.
Table I.
The tryptophan cleavage pattern of P-glycoprotein (pgp1)
Klinische Forschergruppe,
Niedersächsisches
Institut für Peptidforschung,
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
FOOTNOTES
ACKNOWLEDGEMENTS
REFERENCES
Val) (9) and in transmembrane segment (TM) 6 of hamster P-glycoprotein
(Gly338
Ala/Ala339
Pro) (10). Active
mutagenesis identified further motifs and positions in the N- and
C-terminal half of P-glycoprotein which influence substrate specificity
(11, 12). For example, the exchange of amino acids in the first
cytoplasmic loop (11), TM5, TM6, TM10, TM12 (13-16), and the cytosolic
linker peptide (17) resulted in all in an altered multidrug resistance
phenotype, suggesting that during binding and transport the substrate
is recognized by multiple residues located either in the cytosolic, membraneous, or ectoplasmic domains.
1 (39)
and belongs to the compounds to which multidrug-resistant cells are
cross-resistant, i.e. the accumulation of iodomycin is
reduced in P-glycoprotein-overexpressing cells compared with their
drug-sensitive parents (27). The anthracycline recombines with its
binding sites in a diffusion-controlled reaction with kon rates of (2-4) × 109
M
1 s
1. P-glycoprotein is
specifically photolabeled by [125I]iodomycin (27). Since
the photoreaction is several orders of magnitude slower than the
formation of the non-covalent drug-protein complex, the thermodynamic
affinity constant determines the concentration-dependent yield of photoincorporation of iodomycin into protein and the competitive reduction of photobinding of radioactive
[125I]iodomycin by increasing concentrations of
non-radioactive anthracycline (39). Vinblastine and cyclosporin compete
strongly with iodomycin for its photobinding site in P-glycoprotein,
whereas colchicine and nifedipine are poor inhibitors of the
photobinding (27). Photolabeling of protein is optimal at 488 nm, which
coincides with the absorption maximum of anthracene derivatives (39), implying that the photoreaction of iodomycin with amino acids is
different from that of azido compounds (28-32).
Fig. 1.
The structure of
[125I]iodomycin. Daunomycin reacts with
125I-Bolton-Hunter reagent to
[125I]iodomycin.
[View Larger Version of this Image (16K GIF file)]
Cell Culture
-minimum
essential medium supplemented with glutamine, nucleosides, 10% calf
serum, and 30 µg/ml colchicine (41).
70 °C. The yield varied
between 1 and 2 mg of plasma membrane protein. During the whole
procedure, the following antiproteases were added to avoid any protein
degradation: leupeptin (final concentration 4 µM)
(Sigma), N-p-tosyl-L-arginine-methyl
ester (final concentration 26 µM) (Sigma), and Pefabloc
(final concentration 0.4 mM) (Merck).
70 °C. The agarose was thawed on ice and centrifuged for 30 min at 4 °C at 13,000 × g. The agarose was
pelleted, whereas P-glycoprotein remained dissolved in the supernatant
(freeze-squeeze procedure). To remove any residual agarose and
detergent and to increase protein concentration, the solution was
centrifuged through a 100,000-Da cut-off filter in a Centricon tube
(Amicon).
20 °C for 4 h and analyzed by gel
electrophoresis (42) or stored at
70 °C until further
processing.
Fig. 2.
Photoaffinity labeling of P-glycoprotein in
plasma membranes of CHO B30 cells. 50 µg of plasma membrane
protein were labeled with 0.05 pmol of [125I]iodomycin in
the presence of various concentrations of cyclosporin A (lanes
1-5 (from left to right): 0, 3 nM, 30 nM, 300 nM, and 3 µM). The binding of iodomycin was documented by 1-week
exposure of the dried 6% PAGE gel to a Kodak X-Omat film. 300 nM cyclosporin A or more displaces iodomycin out of its
binding site.
[View Larger Version of this Image (67K GIF file)]
counter.
The Cell Line CHO B30
Fig. 3.
Purification of P-glycoprotein with an
agarose gel system. A, lanes 1-3, 6% PAGE of 50 µg of plasma membrane protein of CHO B30 cells (lane 3)
preincubated with endoglycosidase F (lanes 1 and
2). Note the shift of the P-glycoprotein band in the
Coomassie stain by deglycosylation. Lane 4, silver stain of 1 µg of P-glycoprotein purified with the agarose gel system.
B, immunoblot of purified P-glycoprotein. Lanes 1 and 5, 2.5 µg; lanes 2 and 6, 5 µg; lanes 3 and 7, 3 µg of purified protein. C, 6 µg of plasma membrane protein of CHO B30 cells.
Lanes 1-3, immunoreactivity of P-glycoprotein with
the polyclonal antibody mdr AB-1; lanes 5-7,
immunoreactivity of P-glycoprotein with the monoclonal antibody
C494, which only recognizes human and hamster P-glycoproteins of class
I. Occasionally an extra 110-kDa band was detected by the polyclonal
antibody (left panel), but not by C494 (right
panel).
[View Larger Version of this Image (75K GIF file)]
Fig. 4.
Mild tryptic digest. Autoradiogram of a
dried 6% PAGE gel is shown. Iodomycin-labeled P-glycoprotein in plasma
membrane vesicles was tryptically cleaved into two large fragments
according to the protocol by Georges et al. (43). Lane
1 (50 µg) and lane 3 (25 µg), iodomycin-labeled
plasma membrane protein (not digested with trypsin); lane 2,
50 µg of plasma membrane protein digested with trypsin. Iodomycin is
predominantly photoincorporated into the ~100-kDa N-terminal fragment
(2-week exposure to a Kodak X-Omat film).
[View Larger Version of this Image (46K GIF file)]
No.
N-terminal amino acid
C-terminal amino
acid
Molecular weight
1
1
44
5343
2
45
133
9635
3
134
159
3093
4
160
209
5566
5
210
229
2062
6
230
312
9045
7
313
695
42,214
8
696
705
1244
9
706
800
10,824
10
801
852
5366
11
853
1105
27,797
12
1106
1276
18,812
Treatment of [125I]iodomycin-photolabeled plasma
membranes (Fig. 5, A and
B) or purified P-glycoprotein (Fig. 5C) with NCS
(Fig. 5A) or BNPS-skatol (Fig. 5, B and
C) yielded a pattern of photolabeled peptides that was more
complex than predicted by the primary sequence of P-glycoprotein for a
total digest at the tryptophan residues. This finding could result from
(a) nonspecific cleavage, (b) partial digestion,
(c) more than one photobinding site in P-glycoprotein, and/or (d) labeling of further proteins. Explanation
(a) is unlikely to apply because despite their distinct
mechanisms of cleavage both NCS and BNPS-skatol gave the same pattern
of radiolabeled peptides (cf. Fig. 5, A and
B). Partial digestion of protein, however, had to be
seriously considered because the reaction was found never to proceed to
completion. To differentiate between partial digestion and/or multiple
photobinding sites, BNPS-skatol digests of total plasma membrane
protein (Fig. 5B) and purified P-glycoprotein (Fig.
5C) were compared in their profile of photolabeled fragments. Two bands equivalent to fragments of about 50-60 kDa and
smaller than 10 kDa were seen in autoradiograms from both the plasma
membrane preparation and purified P-glycoprotein (cf. Fig.
5, B and C), indicating that they represent
iodomycin-labeled peptides of P-glycoprotein. The <10-kDa radiolabeled
fragment was observed in all cleavage experiments. The hamster pgp1
cDNA sequence (10) predicts that such a small skatol digestion
fragment exists only in the N-terminal half of P-glycoprotein (Table
I), which is in accordance with the outcome of the mild tryptic digest whereby most [125I]iodomycin was photoincorporated into
the N-terminal fragment (see above). When the cleavage with BNPS-skatol
was extended from 4 h to 18 h, the signal of the 50-60-kDa
peptide decreased and that of the small <10-kDa peptide increased
(Fig. 5B). The mutual dependence of the yield of the two
radiolabeled P-glycoprotein subfragments on reaction time indicates
that the 50-60-kDa peptide is a partial digestion fragment which
contains the <10-kDa peptide as the internal
[125I]iodomycin-labeled complete digestion fragment. This
interpretation is consistent with the molecular weights of complete and
partial digestion fragments predicted by the primary amino acid
sequence of hamster P-glycoprotein (fragment 6 versus
fragment [5]-6-7, Table I). The nature of the other radiolabeled
fragments (Fig. 5, A and B) could also be
interpreted as partial digestion fragments from the N-terminal half of
pgp, whereby fragment 6 was an obligatory constituent. Since partial
digestion dissipated the 125I-labeled photoproducts on
multiple peptides, any minor photobinding site, for example that
assigned by mild tryptic digestion to the C-terminal half of
P-glycoprotein (see above), was not detectable by radioimaging after
cleavage with NCS or BNPS-skatol.
The Purification of the Smallest Labeled Peptide
Finally we
wanted to identify the fragment in the N-terminal half of the protein
where the photoreaction takes place.
[125I]Iodomycin-photolabeled substrate was digested with
BNPS-skatol, because the cleavage with NCS produced N-terminal blocked
fragments not suitable for Edman sequence analysis. The smallest
labeled peptide was purified by the combination of gel electrophoresis, electroelution and reversed phase chromatography as described under
"Materials and Methods." The chromatography step was necessary to
remove all peptides that comigrated with the labeled peptide in the
one-dimensional gel. The photocross-linked peptide was eluted in
fractions containing more than 60% solvent B. Fig.
6 shows the elution profiles for solvent,
absorbance, and radioactivity (marker for
[125I]iodomycin). To differentiate between free and
covalently bound iodomycin, all 125I-positive fractions
were tested by gel electrophoresis and subsequent radioimaging.
Radioactivity was associated with the peptide that eluted within the
fraction of 60% solvent B. The sample with highest radioactivity was
subjected to Laemmli gel electrophoresis, blotted onto a PVDF membrane,
stained with Coomassie Blue, and finally subjected to imaging plate
analysis (Fig. 7).
Edman Sequence Analysis
The HPLC fraction with the highest
radioactivity was given either immediately to Edman sequence analysis
or first blotted on the PVDF membrane (v.s.) and then sequenced. The
Edman reaction was carried out on purified peptides from four
independent membrane preparations with identical outcome. Only one
peptide was identified in the reaction cycles independently, whether it
was sequenced directly from the HPLC fraction or after the transfer to
the PVDF membrane. The reaction enabled the reading of the first 17 amino acids of the 125I-photolabeled peptide. The
identified amino acid sequence starts in P-glycoprotein (pgp1) behind
the tryptophan residue number 229, i.e.
125I-photolabel was incorporated into fragment 6 of
pgp1 (amino acids 230-312, Mr 9045, Table I).
According to the topological model established from hydrophobicity
plots (19) this amino acid sequence starts in the fourth transmembrane
domain, encompasses the second cytoplasmic loop, and stops in the fifth
transmembrane domain of P-glycoprotein (Fig.
8).
This study reports on the localization of the binding site for iodomycin in hamster P-glycoprotein. Chemical cleavage of iodomycin-photolabeled P-glycoprotein behind the tryptophan residues revealed that the drug had been predominantly incorporated into the pgp1 fragment 230-312. According to the topology predicted by hydropathy plots (18-20) this amino acid sequence 230-312 forms the distal part of TM4, the second cytoplasmic loop, and the proximal part of TM5 in the N-terminal half of P-glycoprotein (Fig. 8).
To identify a drug photobinding site in P-glycoprotein, the tools and methodology had to be carefully selected and optimized, i.e. the requirements were a photoreactive substrate for P-glycoprotein of high specific activity, a cellular source of high P-glycoprotein contents, a site-specific reagent that reliably cleaves membrane proteins without blockage of the N terminus, and a purification procedure that separates hydrophobic peptides and has been minimized for loss of photoproduct and contamination by compounds that could affect protein sequencing. Our current protocol is considered to fulfill these criteria. Iodomycin photolabels P-glycoprotein in cells and plasma membrane vesicles with high sensitivity and specificity. The overexpression of P-glycoprotein in CHO B30 cells was instrumental to photolabeling P-glycoprotein in sufficient amounts so that 125I-photoproducts were detectable by radioimaging throughout the whole purification procedure. Enzymatic and chemical means were tested for cleavage of purified P-glycoprotein and crude membrane vesicles. Enzymatic proteolysis was found to be disadvantageous because the self-digestion of enzyme contaminated the sample and impeded purification and sequencing of target peptide. In contrast, site-specific cleavage behind tryptophan with BNPS-skatol enabled low background Edman sequencing of the purified 125I-labeled skatol fragment. Its N-terminal sequence enabled us to discriminate between hamster pgp isoforms and to identify pgp1 as the photobinding acceptor of iodomycin. In summary, peptide identification by Edman analysis was chosen because the method yields unequivocal data on the localization of the binding site in the primary sequence.
Apart from iodomycin, some other photoreactive compounds have already been used for the photoaffinity labeling of drug binding sites in P-glycoprotein. Azidopine and the photoreactive derivatives of prazosin and forskolin label both halves of P-glycoprotein (34, 36, 38). [3H]Azidopine, [125I]iodoarylazidoprazosin, and probably also 125I-AIPP-forskolin bind to a common domain in P-glycoprotein (29). Bruggemann et al. (34) recovered 60% of azidopine binding from the CNBr fragment (amino acids 198-440) in the N-terminal half of human MDR1. Site-specific antibodies recognized the photobinding sites for prazosin and forskolin (35, 36, 38) in or close to TM 6 and TM 12; unfortunately, however, epitope specificity of some antibodies was dependent on the detergent that was used for the solubilization of enzymatic digests of photolabeled protein (36). The role of the C-terminal half of P-glycoprotein could be more clearly discerned in studies on MDR1 (class I)/MDR3 (class III) chimera (37). TM 12 and the loop connecting TM 11 and TM 12 were found to contain determinants for substrate specificity and photobinding of [125I]iodoarylazidoprazosin. The distinct localization of photobinding sites for prazosin and iodomycin may reflect different reaction pathways of photolabeling (50) and/or the presence of more than one binding site for the interaction of P-glycoprotein with its substrate.
Tamai and Safa (51) showed that the interaction of vinblastine with the azidopine binding site was noncompetitive. Cyclosporin A competitively inhibited Vinca alkaloid binding, but inhibited azidopine binding in a noncompetitive manner (52). These data suggested the existence of two binding sites on P-glycoprotein, one specific for Vinca alkaloids, verapamil, and cyclosporins and one binding site specific for azidopine. Ayesh et al. (53, 54) extended the model whereby they proposed two different binding sites. One is recognized by substances such as vinblastine, mefloquine, and tamoxifen and the other one by substances such as verapamil. Cyclosporin A may interact with both sites (53). An even more complex view arose from photolabeling studies with a novel 1,4-dihydropyridine derivative (55). Multiple chemosensitizer domains were identified in P-glycoprotein, which were distinct from the Vinca alkaloid binding site. In contrast to these data, Bruggemann et al. (34) found that vinblastine and azidopine bind to a single site in human MDR1. Vinblastine was only able to prevent azidopine binding if a valine was at position 185 instead of a glycine, which occurred during selection of cells with colchicine (9). These data indicate the direct contact of photobinding sites with substrate specificity influencing amino acids. Friche et al. (56) examined the effect of anthracycline analogues on photolabeling of P-glycoprotein by iodomycin and azidopine. Their data suggest that anthracyclines and Vinca alkaloids may bind to a common site in P-glycoprotein and that azidopine and iodomycin label P-glycoprotein with different binding affinities and/or at different sites. The latter view is supported by binding studies with prazosin. This drug inhibited photoaffinity labeling of P-glycoprotein by azidopine (29), but did not displace iodomycin out of its binding pocket (29, 36).2 The major photobinding site of arylazidoprazosin was localized in and C-terminal to TM6 and TM12 and a second minor site between TM4 and up to TM6 (36), of which the latter corresponds with the iodomycin binding site described in this report (Fig. 8).
Although photoaffinity labeling studies provided strong evidence that P-glycoprotein has more than one binding site for drugs and modulators, it is not definite proof because of the inherent limitations of the methodology. Photoreactions with arylazido groups under UV irradiation and iodomycin with visible light share no common pathways, and correspondingly other reaction centers in the protein are selected, which may be part of the same binding site or another one (50, 57). On the other hand, this differential photoreactivity has the advantage of localizing various epitopes of the three-dimensional binding pocket(s), albeit the number of sites remains elusive. The latter issue is more appropriately addressed by analysis of substrate interaction at equilibrium (39, 58).
Since covalent photolabels just indicate the binding domain by primary sequence, supplementary information about the topology of the protein and its functional states is instrumental for the characterization of the binding site(s). According to the topological model of P-glycoprotein predicted from the hydropathy plot (18-20), the binding site for iodomycin is localized in a sequence that starts at amino acid 230 in TM 4, comprises the second cytoplasmatic loop, and stops at amino acid 312 in TM 5 (Fig. 8).
Recent experimental work (59-61) applying up-to-date methodology to full-length P-glycoprotein confirms the structure of P-glycoprotein established by hydropathy plot analysis. Topologies of P-glycoprotein different than predicted have been observed in cell-free (21, 22, 62), Xenopus oocyte (22), and bacteria (23, 24) expression systems. Despite the fact that all these topologies were found in truncated molecules or fusion proteins, Zhang et al. (25) found a structure of full-length P-glycoprotein different from that predicted by hydropathy plot analysis in CHO B30 cells, which express P-glycoprotein naturally by selection with 30 µg/ml colchicine. They probed the structure of P-glycoprotein with site-specific antibodies and found TM 5 and the loop connecting TM 4 and 5 located extracellularly.
The inconsistent topology may be not only a consequence of different experimental set-up, but could also reflect a dynamic structure of P-glycoprotein associated with different functional states. A paradigm is the colicin Ia channel (63). A mobile segment crosses the membrane from one side to the other as the channel opens. These data support the hypothesis that reversible posttranslational changes of the topology may be an essential mechanism of a channel or transporter like P-glycoprotein.
The loop connecting TM 4 and TM 5 contains a sequence motif that is highly conserved among ABC transporters. The EAA or EAA-like motif (EXAXXXG) predicts substrate specificity of prokaryotic ABC transporters (64-66). Substitution of the central glycine causes transporter dysfunction in bacteria (67) and the exchange of the conserved Glu for Asp and Gly for Pro resulted in a complete loss-of-function of PXA1, a gene product of Saccharomyces cerevisiae homologous to the adrenoleukodystrophy protein, an ABC transporter in the membrane of human peroxisomes (66, 68, 69). The conserved motif is also found in P-glycoproteins and is localized in the second cytoplasmatic loop (Fig. 8) connecting TM 4 and 5 at positions Glu279, Ala281, and Gly285 of hamster pgp1 (10). This region resides within the iodomycin-labeled skatol fragment.
In conclusion, this study demonstrates that the binding site for iodomycin is localized in a domain of P-glycoprotein that recognizes vinblastine and cyclosporin with high affinity. The data about the topology of and the conserved sequence motif within the iodomycin-labeled domain suggest a crucial role for this region in the transport function of ABC transporters in general. With respect to P-glycoprotein, the localization of binding site(s) in the primary sequence supplemented with mutagenesis studies and functional studies will identify the constituents of the binding pocket(s) and hence provide a clue as to how P-glycoprotein can recognize and handle such a broad range of structurally unrelated substrates.
We thank our colleagues at the Institut für Peptidforschung (Dr. Forssmann) and at the Max-Planck-Institut für experimentelle Endokrinologie (Dr. Jungblut) for access to their technical facilities, which was fundamental for the success of the project. We are grateful to Manfred Raida's co-workers for their help and Dr. Weißer for his kind advice.