(Received for publication, January 23, 1997, and in revised form, March 24, 1997)
From the Laboratory of Cell Biology, National Institute of Mental Health, Bethesda, Maryland 20892-4090
The transmembrane topology of the
Na+- and Cl-dependent
-aminobutyric
acid transporter GAT-1 has been studied using protein chimeras in
Xenopus oocytes. A series of COOH-terminal truncations was
generated to which a prolactin epitope was fused. Following expression
of transporter-prolactin chimeras in Xenopus oocytes, the
transmembrane orientation of each chimera was determined by testing for
protease sensitivity in an oocyte membrane preparation. Data from
protease protection assays with GAT-1-prolactin chimeras has shown that
residues in the loops connecting hydrophobic domain (HD)3 and HD4 and
HD7 and HD8 are accessible to protease in the cytoplasm and suggest the
presence of pore loop structures which extend into the membrane from
the extracellular face. Such pore loop structures may be involved in
the formation of the substrate-binding pocket. Studies presented herein
confirm that the NH2 and COOH termini are cytosolic and
hydrophobic domains span the membrane in a manner consistent with the
predicted hydropathy model for Na+- and
Cl
-dependent transporters. These data also
provide insight into GAT-1 transmembrane assembly and suggest that a
complex series of topogenic sequences directs this process. A potential
pause-transfer sequence has been identified and may be responsible for
the translocational pausing observed in this study.
The Na+- and Cl-dependent
transporters are a family of complex integral membrane proteins
responsible for the reuptake of many neurotransmitters, amino acids,
osmolytes, and a variety of other substrates (1-3). The GAT-1
aminobutyric acid (GABA)1 transporter was the first member of
this family for which a cDNA was isolated and characterized (4).
Subsequent to identification of a cDNA for the norepinephrine
transporter (NET) (5), comparison of the NET amino acid
sequence with that of GAT-1 revealed a significant degree of homology
and indicated that these transporters might be members of a larger
transporter family. The extent and diversity of this transporter family
was revealed by the identification of cDNAs for several members
with the use of homology cloning strategies. Hydropathy analyses of NET
and GAT-1 primary sequences predict that these proteins share many
structural features. These include a motif of 12 hydrophobic regions
connected by hydrophilic loops; a large extracellular loop connecting
hydrophobic domains (HDs) three and four which contains potential sites
for N-linked glycosylation; cytoplasmic localization of the
amino terminus, based on the lack of a recognizable signal sequence;
and cytoplasmic localization of the carboxyl terminus (1-3). In a
small number of orphan members of this transporter family, the
predicted structure differs in the presence of a larger hydrophilic
loop connecting HDs seven and eight with potential sites for
N-linked glycosylation (6-8). Some features of this
predicted model have been confirmed. Immunocytochemical studies with
sequence-directed peptide antibodies have shown that the amino and
carboxyl termini of NET are located in the cytoplasm (9). In the same
study, the large loop connecting HD3 and HD4 and the loop connecting
HD7 and HD8 were shown to be extracellular (9). In addition, several
groups have confirmed that the sites for N-linked
glycosylation in the large loop connecting HD3 and HD4 are utilized in
the norepinephrine, glycine, and serotonin transporters (10-13),
consistent with the extracellular placement of this loop. While some
features of the predicted model for the transmembrane topology of these
transporters have been confirmed, detailed experimental testing of the
model has only recently begun (14, 15).
Knowledge of the transmembrane topology of these important proteins is
necessary for understanding how they function. Complex or polytopic
integral membrane proteins, such as the Na+- and
Cl-dependent transporters, are synthesized
and assembled into their native structure at the endoplasmic reticulum
(ER) (16). Assembly of polytopic proteins is thought to be directed by
a series of topogenic sequences which interact with the ER
translocation machinery (17, 18), resulting in translocation and
integration of the nascent chain across and into the ER membrane. The
sidedness of the hydrophilic loops connecting HDs can be utilized to
infer the cellular location of the loops and the transmembrane
orientation of the HDs. Protease protection assays of nascent chains in
vesicles prepared from ER membranes are one approach that takes
advantage of the sidedness of integral membrane proteins. This method
has been used to study the transmembrane topology of several proteins including the human P-glycoprotein (19-21), the GluR3 glutamate receptor (22), and nicotinic acetylcholine receptor subunits (23).
Study of the transmembrane topology of MDR1 (19-21) and GluR3 (22) has
shown that models based on hydropathy analyses can be misleading and
has resulted in the generation of novel topological profiles for these
proteins.
In this study the transmembrane topology of GAT-1 has been studied as a
representative member of the Na+- and
Cl-dependent transporter family. While the
data suggest that the transmembrane topology of GAT-1 HDs is not
significantly different from that predicted by hydropathy analysis, the
data show that the positioning of some extracellular loops may differ
from the predicted model. Protease protection assay data has shown that residues in the loops connecting HD3 and HD4 and HD7 and HD8 of GAT-1-prolactin fusion proteins are accessible to protease in the
cytoplasm. These residues may extend into a central pore from the
extracellular face forming pore loop structures. These data also
suggest that coordinate actions of several topogenic sequences are
necessary for translocation and membrane integration of GAT-1.
[3H]GABA was purchased from DuPont NEN. Restriction enzymes, Vent polymerase, and PNGase F were purchased from New England Biolabs. Taq polymerase was purchased from Promega. SDS-polyacrylamide gels, MultiMark protein standards, and Mark12 protein standards were from Novex.
Plasmid ConstructionsGAT-1 GABA transporter cDNA in
pBluescript plasmid (pBSSKII()), generated as described previously
(24), was digested with XhoI and XbaI and
subcloned into plasmid JG3.6 (25) for the wild-type construct (pGATA).
Polymerase chain reaction was used to generate a construct, pFLAGA-N in
which the 5
-noncoding region of pGATA was replaced with a
KpnI site followed by a consensus Kozak sequence, an ATG,
sequence for the FLAG epitope tag, and nucleotides 4-20 of GAT-1.
Amplification of GAT-1 with the FLAG oligonucleotide paired with an
oligonucleotide directed to nucleotides 559-582 of the coding sequence
yielded a fragment which was digested with KpnI and
EcoRI and ligated into digested pGATA to create pFLAGA-N.
pFLAGA-C, encoding GAT-1 with the FLAG epitope fused at the carboxyl
terminus, was generated by amplification of GAT-1 with an oligo
directed against nucleotides 1641-1671 paired with an oligo directed
against nucleotides 1780-1797, sequence for the FLAG epitope, a TGA,
and an XbaI site. The fragment was digested with
XmaI and XbaI and ligated into digested
pGATA.
Truncated GAT-1 constructs were generated using polymerase chain
reaction with an oligo containing a KpnI site and directed against the 5-noncoding region (nucleotides
146 to
126) paired with oligos directed against defined regions in the GAT-1 coding sequence and containing a BstEII site (putative
extracellular loop 1, EL1, 211-234; putative cytoplasmic loop 1, CL1,
319-339; EL2, 610-627; CL2, 694-711; EL3, 838-855; CL3, 937-954;
EL4, 1094-1110; CL4, 1225-1242; EL5, 1348-1365; CL5, 1450-1467;
EL6, 1579-1596; see Fig. 1, A and B). Polymerase
chain reaction fragments were digested with KpnI and
BstEII and ligated into digested JGPRO. JGPRO was created by
ligating an EcoRI/SphI fragment from
pSPSp+1L.ST.gG.pT, obtained from W. R. Skach
and V. R. Lingappa, into digested JG3.6 creating a vector bearing the
prolactin epitope tag with a BstEII site engineered just
prior to the sequence encoding the tag. PRO, a construct with
full-length prolactin, was generated by ligation of a PstI
and HindIII fragment from a prolactin construct (BPI)
received from W. R. Skach and V. R. Lingappa into pBSSKII(
). Sequence
analysis of constructs confirmed the absence of mutations introduced by
polymerase chain reaction in all but two constructs, EL5 and CL5. In
CL5 a single mutation of C to T (478 in the GAT-1 coding sequence)
located in the large loop connecting HD3 and HD4 results in a Pro to
Ser change. A nonmutant CL5,
CL5, was generated by digesting pGATA
and CL5 with EcoRI and AvrII and replacing the
mutant fragment in CL5 with that from pGATA. Data obtained from
proteinase K experiments with
CL5 were identical to those obtained
with CL5. In EL5 a mutation of T to A (985 in the GAT-1 coding
sequence) is located in HD7 and results in a Cys to Ser change, and a C
to G (1131 in the GAT-1 coding sequence) is located in HD8 and results
in an Ile to Met change. The T to A mutation in EL5 was corrected by
digesting EL5 and CL4 with EcoRI and AvaI and
replacing the mutant fragment in EL5 with that from CL4 generating
EL5. Data obtained from proteinase K experiments with
EL5 were
identical to those obtained with EL5. Correction of the mutations in
CL5 and
EL5 was confirmed by sequencing of the constructs.
Sequencing was performed by NAPCORE (Nucleic Acid/Protein Research Core
Facility, The Children's Hospital of Philadelphia).
RNA Transcription
mRNA was transcribed from the GAT-1 fusion protein constructs with 1 µg of linearized DNA using T7 RNA Polymerase according to the mMessage mMachineTM protocol (Ambion, Inc.).
Xenopus laevis Oocyte ExpressionX. laevis were purchased from Xenopus (Ann Arbor, MI) and oocytes were dissected and prepared as described previously (26). 25 µCi of Trans35S-label (ICN Biomedicals Inc.) (0.25 µl of 10 × concentrated solution) was added to 1 µl of 250 ng/µl transcript and 50 nl of this solution injected per oocyte. Following incubation at 18 °C for 4-6 h, oocytes were pooled and homogenized on ice in homogenization buffer (0.25 M sucrose, 50 mM potassium acetate, 5 mM magnesium acetate, 1.0 mM dithiothreitol, 50 mM Tris, pH 7.5). Following homogenization CaCl2 was added to 10 mM final concentration.
Protease Protection AssayProteinase K (Boehringer Mannheim) was added to aliquots of oocyte homogenates (0.2 mg/ml final) in the presence or absence of 1% Triton X-100 and incubated on ice for 1 h. Protease was inactivated by addition of 10 mM phenylmethylsulfonyl fluoride (Sigma) in Me2SO, dilution with 10 volumes of 1% SDS, 0.1 M Tris, pH 8.0, and boiling for 5-10 min. Samples were diluted with 4 volumes of 1.25 × RIPA buffer (187.5 mM NaCl, 1.25% Triton X-100, 1.25% deoxycholate, 62.5 mM Tris, pH 8.0, 2.5 mM EDTA, 0.125% SDS) and set at 4 °C with rocking ~12-16 h. Samples were centrifuged at 14,000 × g for 15 min and proteins immunoprecipitated from the supernatant as described below.
Carbonate ExtractionOocyte homogenates were extracted with sodium carbonate as described by Fujiki et al. (27) and Skach et al. (21). Briefly, homogenates were diluted in 400 volumes in either 0.1 M sodium carbonate, pH 11.5, or 0.1 M Tris, pH 7.5, and set on ice for 30 min. Membranes were pelleted by centrifugation at 50,000 rpm (230,000 × g) for 30 min using a 70.1 Ti rotor (Beckman). Proteins in the supernatants were precipitated with 15% trichloroacetic acid and pelleted. Trichloroacetic acid pellets and membrane pellets were dissolved in 1% SDS, 0.1 M Tris, pH 8.0, 100 µM 4-(2-aminoethyl)benzenesulfonyl fluoride (ICN Biomedicals, Inc.) and samples were boiled for 5-10 min. Following addition of 4 volumes of 1.25 × RIPA buffer, transporter-prolactin chimeras were immunoprecipitated as described below.
PNGase and Endo H TreatmentOocyte homogenates were denatured and peptide:N-glycosidase F (PNGase, 1000 units) or endoglycosidase H (Endo H, 2500 units) treatment performed at 37 °C for 1 h, according to the New England Biolabs protocol provided with the enzymes. Samples were solubilized in 1 × RIPA buffer, as described above, and subsequently immunoprecipitated.
ImmunoprecipitationFusion proteins were immunoprecipitated with rabbit anti-ovine prolactin polyclonal serum (ICN Biomedicals, Inc.) at 1:500 for 4 h at 4 °C. Protein A Affi-Gel (50 µl) (Bio-Rad) was added and samples were set at 4 °C for 3 h followed by 3 washes with 1 × RIPA buffer and 2 washes with 0.1 M NaCl, 0.1 M Tris, pH 8.0. Proteins were eluted from Protein A Affi-Gel in 2 × Laemmli buffer (28) at 55 °C for 15 min. Samples were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (PAGE) and imaged on the Fujix PhosphorImaging system (Fuji).
Transfection of CV-1 CellsCV-1 cells were plated at a density of ~2-3 × 105 cells per well in 24-well plates directly in the well or on 1.2-cm diameter coverslips. Cells were transfected with 1 µg of DNA/well using LipofectACE (Life Technologies, Inc.) at a ratio of 1:5 DNA/lipid (w/w). Cells were incubated with the DNA/lipid solution at 37 °C for 16-24 h at which time the medium was removed and replaced with complete medium. At 72 h post-transfection, cells were assayed for transport and immunocytochemistry was performed.
Transport AssayTransfected cells were incubated with [3H]GABA (50 nM, 1 µCi/µl) in a modified Krebs-Ringer-HEPES buffer (24) in the presence of 100 µM aminooxyacetic acid, a GABA transaminase inhibitor, for 30 min at 37 °C. Uptake was stopped by placing cells on ice and washing with ice-cold assay buffer. Cells were solubilized in 1% SDS, and the amount of accumulated [3H]GABA was determined by liquid scintillation counting.
ImmunocytochemistryCV-1 cells transfected with pFLAGA-N or pFLAGA-C were rinsed with phosphate-buffered saline and fixed by incubation in 4% formaldehyde in phosphate-buffered saline for 30 min at room temperature. Fixed cells were incubated in blocking solution (1% bovine serum albumin, 3.3% normal goat serum in phosphate-buffered saline) in the presence or absence of 0.1% Triton X-100 for 60 min at room temperature. Cells were incubated in blocking solution for 16 h prior to addition of anti-FLAG M2 monoclonal antibody (Eastman Kodak Co.) at 1:300 in blocking solution. A 1-h incubation with primary antibody was followed by three 5-min washes with blocking solution. Incubation with Cy3-conjugated AffiniPure goat anti-mouse IgG (Jackson Immunoresearch Labs., Inc.), at 1:300-1:500 in blocking solution, was done for 1 h followed by three 5-min washes with blocking solution. Cells were then covered with coverslips after adding SlowfadeTM reagent (Molecular Probes). Observation of the fluorescence staining was performed with a laser scanning microscope (Zeiss, LSM 410, at the Light Imaging Facility, NINDS, Bethesda, MD).
To examine the transmembrane topology of
the GAT-1 GABA transporter a series of constructs were generated in
which the transporter sequence was truncated at the COOH-terminal
region of each of the putative hydrophilic loops connecting the
putative HDs (see Fig. 1A). Each of the
truncated transporter fragments were ligated into a vector (JGPRO)
in-frame with the sequence for the carboxyl-terminal fragment of the
secretory protein prolactin (codons 56-199). This prolactin fragment
lacks intrinsic translocation activity and has previously been shown to
serve as a faithful reporter for translocation when following topogenic
sequences in chimeric proteins (17, 20, 22, 23). GAT-1-prolactin
chimeras are diagrammed in Fig. 1B. Transcripts were made
from the constructs and injected into X. laevis
oocytes. Following incubation for 4-6 h, oocytes were homogenized.
Endo H sensitivity, a reliable marker of protein residence in the ER
membrane, was used to confirm chimera residence in the ER, and absence
from the plasma membrane, at 4-6 h post-injection (data not shown).
Because the chimeras have not exited the ER, it is evident that the
membrane vesicles containing the chimeras display a single orientation.
Furthermore, subcellular localization of GAT-1 expressed in oocytes as
late as 2-5 days following transcript injection revealed that the vast
majority of the transporter remained in intracellular compartments
(29). Protease protection assays were used to determine the
transmembrane spanning abilities of hydrophobic domains and the
cellular location of regions connecting these domains. Nascent chains
in the ER are oriented such that extracellular domains are located in
the ER lumen and cytoplasmic domains are on the cytosolic face of the
ER membrane. Exposure of nascent chains in vesicles prepared from ER
membranes to proteinase K, a protease that cleaves nonspecifically,
results in cleavage of cytoplasmically exposed domains while domains in
the ER lumen are protected. The data for study of the amino terminus
and HD1 to HD6 are presented in Fig. 2, A-G.
Proteinase K treatment of homogenates from oocytes expressing
full-length prolactin shows that membrane vesicles in the homogenate
are intact as prolactin is fully protected from the protease (Fig.
2A). The intensity of the protected prolactin band is nearly
identical to the intensity of the prolactin control band (Fig.
2A), indicating that the majority of vesicles in the
Xenopus oocyte membrane homogenate are in the correct
orientation. Prolactin is digested in the presence of proteinase K and
a nonionic detergent (Fig. 2A) indicating that prolactin is
not protease-insensitive. A protease-resistant band of 14-15 kDa was
detected for the full-length prolactin as well as each of the
GAT-1-prolactin fusions tested. This fragment corresponds to the
predicted size of the prolactin tag and is not indicative of a
particular membrane orientation. The GAT-1-prolactin chimera extracellular loop 1, EL1, was protected from proteinase K digestion in
the absence but not the presence of a nonionic detergent (Fig. 2B). These data strongly suggest that the first hydrophobic
domain (HD1) has been translocated into the ER lumen but that HD1 has not yet integrated into the membrane (schematized in Fig.
2B) since the amino terminus also has been protected from
digestion. Cytoplasmic loop 1, CL1, was partially digested by
proteinase K as indicated by the reduced size of the protected fragment
in Fig. 2C (open arrow, lane 2). Comparison of
the size of the untreated chimera with the protected fragment shows a
difference of approximately 6 kDa. Partial digestion of CL1 indicates
that HD1 has become associated with the membrane, and that a portion of
CL1 slightly larger than the amino terminus has been digested. These
data suggest that the amino terminus of CL1 is in the cytosol. No
protected fragments were detected with EL2 (Fig. 2D), the
chimera with the prolactin tag fused at the COOH terminus of the large
loop connecting HD3 and HD4 and containing three canonical sites for
N-linked glycosylation (Fig. 1A). This finding
was surprising as several groups have shown that the sites for
N-linked glycosylation in this loop are utilized, and
therefore that this loop is extracellular (10-13). An 18-kDa fragment
(approximately 2 kDa without the prolactin tag) was detected following
protease treatment of CL2 (Fig. 2E, open arrow, lane 2). The
approximate point of digestion of CL2 based on gel mobility is the
NH2-terminal side of HD4 suggesting that the COOH-terminal
part of the large loop connecting HD3 and HD4 extends into the membrane
to the extent that it is accessible to protease. Protected fragments
were detected following proteinase K treatment of EL3, whereas no
protected fragments were detected following treatment of CL3 (Fig. 2,
F and G). These data confirm that the loop
connecting HD5 and HD6 is extracellular and that connecting HD6 and HD7
is intracellular as predicted by hydropathy analysis (Fig.
1A). The digested fragments from EL3 (Fig. 2F, open
arrows, lane 2) are 7 and 17 kDa smaller than the full-length chimera indicating that the approximate points of digestion were HD1
and the NH2-terminal side of the loop connecting HD3 and
HD4, respectively. The inaccessibility of the loop connecting HD4 and HD5 to protease in chimera EL3 suggests that this chimera has assumed a
conformation where this loop is buried and thus inaccessible to
protease.
The protease protection assay data showed that each of the
GAT-1-prolactin chimeras was translocated across the ER membrane, but
did not address the membrane integration abilities of these proteins.
Polypeptides were extracted from membranes in 0.1 M sodium carbonate, pH 11.5. Under these conditions peripheral
membrane proteins and luminal polypeptides are extracted from the
membranes while integral membrane proteins remain associated and
readily sediment with the lipid bilayer (27). The secretory protein prolactin as well as EL1 are found in both the supernatant and pellet
of Tris and carbonate-treated membranes (Fig. 3). In
contrast, CL1 and CL2 are found in the Tris and carbonate pellets only
(Fig. 3). EL2 and each of the remaining chimeras, EL3 through EL6 and CL3 through CL5, were found in the Tris and carbonate pellets following
extraction (data not shown). The appearance of EL1 in the supernatant
of the carbonate-extracted membranes confirms that HD1 has not
completed integration into the membrane, while all of the chimeras with
the exception of EL1 have integrated. The presence of prolactin and EL1
in the Tris supernatant is most likely due to lysis of some of the
membrane vesicles in the procedure. The presence of some prolactin and
EL1 in the carbonate pellet could be due to attachment of a portion of
the nascent chains to ribosomes or to the presence of a fraction of
sealed vesicles.
The large loop between HD3 and HD4, containing sites for
N-linked glycosylation, appears to be located in the cytosol
based on data obtained in protease protection assays with EL2. If this loop is cytosolic then EL2 should not be glycosylated. To determine the
glycosylation state of chimeras, homogenates were treated with PNGase F
an enzyme that cleaves oligosaccharides between the innermost
N-acetyl moiety and asparagine residues. Neither CL1, which
has no potential sites for N-linked glycosylation, nor EL2
were affected by treatment with PNGase F (Fig. 4). In contrast, CL2 and EL3 (Fig. 4), and all remaining chimeras (data not
shown), showed increased mobility on SDS-PAGE after treatment with
PNGase F indicating that sugar moieties have been removed from these
chimeras. These data indicate that a topogenic sequence in HD4, or in
the loop connecting HD4 and HD5, is directing the extracellular
orientation of the large loop connecting HD3 and HD4 such that it
undergoes glycosylation. Taken together the PNGase F and protease
protection data show that HD1 through HD6 attain a transmembrane
orientation similar to that predicted by hydropathy analysis. However,
these data point to potential differences in the membrane association
of HD1 and the membrane association of the COOH-terminal region of the
loop connecting HD3 and HD4.
Transmembrane Topology of GAT-1 HD7 through HD12 Resembles the Model Predicted by Hydropathy Analysis
With the exception of EL4
and CL4, protease protection assays with chimeras designed to test
loops connecting HD9 through HD12 yielded data consistent with the
predicted topology for GAT-1. There were no detectable fragments on
SDS-PAGE following proteinase K digestion of EL4 (Fig.
5A), suggesting a cytosolic placement for
this loop. However, protease treatment of CL4 resulted in the
protection of a fragment of approximately 20 kDa (4 kDa without the
prolactin tag) (Fig. 5B) indicating that HD7 has integrated into the membrane and that the COOH-terminal portion of the loop connecting HD7 and HD8 is accessible to protease. Proteinase K treatment of CL5 resulted in no detectable fragments (Fig.
5D) which suggests a cytosolic placement of the loop
connecting HD10 and HD11. Several protected fragments were detected for
both EL5 and EL6 (Fig. 5, C and E) suggesting
that the nascent chain goes through various conformational states in
acquiring the appropriate topology and as a result not all cytosolic
loops are accessible to the protease. The sizes of the protected
fragments reveal sites in the chimera which are accessible to protease.
Protected fragments of 29, 27, and 21 kDa were detected for EL5 (Fig.
5C). The approximate sites of cleavage for the 29 and 27 kDa
(13 kDa and 11 kDa without the prolactin tag) are in the loop
connecting HD7 and HD8, while the approximate site of cleavage for the
21-kDa fragment (5 kDa without the prolactin tag) is in the loop
connecting HD8 and HD9. These data confirm the protease accessibility
of the loop connecting HD7 and HD8 that was detected with the CL4
chimera. In addition, these data place the loop connecting HD8 and HD9
in the cytosol, consistent with the predicted transmembrane topology of
GAT-1. Despite the presence of a single mutation in HD8 (Ile to Met, as
described under "Experimental Procedures") the protease protection data obtained with EL5 is confirmed by data obtained with EL6 suggesting that the mutation in EL5 has not had an effect on
transmembrane topology. Four fragments were detected following
treatment of EL6 with protease (34, 31, 28, and 19 kDa) (Fig.
5E). Approximate sites of cleavage for the 34- and 31-kDa
fragments (18 and 16 kDa without the prolactin tag) are in the loop
connecting HD7 and HD8 and in the NH2 terminus of HD8,
respectively, confirming that at least part of this loop is accessible
to protease. The 28- and 19-kDa fragments (12 and 4 kDa without the
prolactin tag) are generated by cleavage of the loop connecting HD8 and
HD9, and cleavage of the NH2 terminus of HD11,
respectively. These data confirm the placement of the loops connecting
HD8 and HD9 and HD10 and HD11 in the cytosol. Together these data
confirm the predicted transmembrane orientation of GAT-1 HD7 through
HD12 and reveal that part of the loop connecting HD7 and HD8 extends into the membrane such that it is accessible to protease.
Amino and Carboxyl Termini of GAT-1 Are Located in the Cytosol
The transmembrane location of GAT-1 amino and carboxyl
termini was studied using immunocytochemistry of GAT-1 with the FLAG epitope fused at the amino (FLAGA-N) and carboxyl (FLAGA-C) termini. Transport of GABA by the FLAG-GAT-1 chimeras transiently expressed in
CV-1 cells was equivalent to that of wild-type GAT-1 (data not shown).
Fig. 6 shows the results of immunocytochemistry of the
FLAG-GAT-1 chimeras with the FLAG M2 monoclonal antibody in the absence
and presence of detergent. There is no immunofluorescence in the
absence of detergent (Fig. 6, A and C) showing
that the epitopes are not accessible to the antibody. In contrast,
permeabilization of the cells makes the epitopes readily accessible to
the antibody and results in robust immunofluorescence (Fig. 6,
B and D). These data are consistent with a
cytosolic localization of both the amino and carboxyl termini of
GAT-1.
Until recently, studies of Na+- and
Cl-dependent transporter structure and
function have been based on the transporter model predicted by
hydropathy analyses (Fig. 1A). GAT-1 was chosen as a
representative of this transporter family in the study described here,
which is among the first attempts to examine experimentally the
transmembrane topology of these proteins in detail. Using protease
protection of defined GAT-1 reporter epitopes our data reveal novel
differences in the topology of some of the loops connecting HDs, while
the topology of GAT-1 HDs is not significantly different from the
predicted model. Specifically, the loops connecting HD3 and HD4, and
HD7 and HD8 are accessible to protease in the cytosol suggesting that
regions of these loops extend into the membrane as depicted in Fig.
7. In addition to elucidating the transmembrane topology
of GAT-1, these methods reveal processes governing the
transmembrane assembly of this transporter. Although the presence of
topogenic sequences throughout GAT-1 makes interpretation of these data
difficult and limits the utility of this methodology, many approaches
have been taken to study the topology of integral membrane proteins,
and each method displays certain limitations. Immunocytochemical
localization of peptide epitopes is a preferred method for studying
transmembrane topology because it is possible to study the full-length
unmodified, and functional protein. Yet it is not always possible to
generate antibodies to epitopes of interest and it is difficult to
distinguish between buried extracellular epitopes and intracellular
ones. Another approach involves the detection of post-translational
modification of residues as an indication of membrane sidedness.
Several integral membrane proteins such as the cystic fibrosis
transmembrane regulator (30), GluR1 glutamate receptor (31), Glut 1 glucose transporter (32), the SGLT1 glucose transporter (33), and most
recently GAT-1 and the GLYT1 glycine transporter (14, 15) have been
studied using glycosylation site insertion. The insertion of such sites throughout a protein has the potential of altering the native topology.
In addition, the insertion of glycosylation sites into the loops of
some proteins can be detrimental to function as has been shown for the
Glut 1 glucose transporter (32), GAT-1 (14), and GLYT1 (15).
The approach taken in this study to examine the transmembrane topology
of GAT-1 has been used by several groups to study other integral
membrane proteins including the human P-glycoprotein (19-21), the
nicotinic acetylcholine receptor subunits (23), and the GluR3 glutamate
receptor (22). In this method the COOH-terminal 144 amino acid residues
of prolactin are fused at defined locations to the GAT-1 transporter.
In a membrane vesicle preparation from X. laevis oocytes
expressing GAT-1-prolactin chimeras, the protease sensitivity of the
epitope reveals the cellular location of the hydrophilic loop being
examined. Protection of the epitope from protease digestion indicates
that the loop is located in the ER lumen and therefore becomes
ultimately extracellular, whereas, digestion of the epitope would
indicate that the loop is cytosolic. In using this approach the
assumption has been made that truncation of GAT-1 and fusion of the
prolactin epitope does not affect the native transmembrane topology.
This assumption is supported by several lines of evidence. First, the
prolactin epitope has no topogenic sequences itself and has been shown
to be permissive to the actions of defined polytopic integral
transmembrane proteins (17). Second, the transmembrane topologies of
some proteins studied with prolactin fusions have been confirmed
through the use of epitope-directed antibodies (20, 23). Finally,
fusion of the prolactin epitope to the COOH terminus of GAT-1 had no effect on transport function (data not shown). The effect of truncation and prolactin fusion on the native transmembrane topology of EL1 through EL6 and CL1 through CL5 could not be tested, however. It is
evident from the limitations described for each of the methods that
convergent data obtained using different methods will prove most
convincing in determining the transmembrane topology of the Na+- and Cl-dependent
transporters.
Due to the presence of topogenic sequences throughout GAT-1 it was not possible to infer the transmembrane topology of a loop based on a single GAT-1-prolactin fusion protein. Protease protection studies to determine the topology of both prokaryotic and eukaryotic integral membrane proteins have shown that data on individual fusion proteins may be misleading (21, 34). Transmembrane assembly of GAT-1 must require the cooperative actions of several topogenic sequences. Therefore, consideration of data collected from studying all the constructs was necessary to provide a two-dimensional picture of this polytopic protein. Cooperative actions were necessary for the assembly of HD1 through HD4 as membrane integration for each of these HDs did not occur until the next downstream HD was present. HD1 did not become membrane integrated until HD2 was present, as evidenced by extraction of EL1 from the membranes at high pH. In addition, the digestion of the amino terminus when CL1 was treated with protease suggests that HD1 has become integrated. As revealed by the difference in the glycosylation states of EL2 and CL2, HD3 did not become inserted in the membrane until HD4 was added. The extracellular placement of the loop connecting HD3 and HD4 is consistent with data from many studies which have shown that the sites for N-linked glycosylation in this loop are utilized (10-13). Immunocytochemical data reveal that this loop is extracellular in the norepinephrine transporter (9). However, EL3 data showed that in this fusion protein the NH2-terminal portion of the large loop connecting HD3 and HD4 is accessible to protease. These data may be unique to the conformation of this particular construct or may reveal that this region of the transporter is topologically different from the predicted model. A different approach will be necessary to confirm that this change exists in the full-length functional transporter. Furthermore, in EL4 and CL4, HD7 and HD8 did not insert in the membrane in the predicted manner, yet they do become integrated as indicated by EL5 and EL6 chimera data. Protected digestion products from both EL5 and EL6 show that the loop connecting HD8 and HD9 is accessible to protease and therefore located in the cytoplasm, consistent with the predicted model for GAT-1. The loop connecting HD7 and HD8 in NET has been shown to be extracellular by immunocytochemistry with epitope-specific peptide antibodies (9). Due to the significant homology between GAT-1 and NET sequences and therefore the expected similarity in transmembrane topology of the proteins, it is assumed that the majority of the loop connecting HD7 and HD8 in GAT-1 is also extracellular. Our immunocytochemical analysis of the amino and carboxyl termini of GAT-1 indicates that both are located in the cytoplasm. This is consistent with data published for the norepinephrine and glycine transporters (9, 11, 35, 36).
While the transmembrane orientation of the HDs is consistent with that determined by hydropathy analysis, accessibility of residues in the loops connecting HD3 and HD4 and HD7 and HD8 was not predicted. Determination of cleavage sites in GAT-1 chimeras was based on gel mobility of cleaved fragments. Since hydrophobic proteins such as transporters can migrate anomolously on SDS gels, these sites of cleavage must be considered approximate until confirmed by mapping. While the specific sites of protease cleavage remain to be defined, the data indicate that regions within the loops connecting HD3 and HD4 and HD7 and HD8, loops which are thought to be extracellular, were accessible to protease in the cytosol. Data from the assay of CL2 indicate that the COOH-terminal portion of the large loop connecting HD3 and HD4 is accessible to protease, and therefore must extend to some degree into the membrane (Fig. 7). Similarly, data from assay of CL4, EL5, and EL6 show that a portion of the loop connecting HD7 and HD8 is accessible to protease (Fig. 7). Prolines and short stretches of hydrophobic residues in the COOH-terminal regions of both of these loops may participate in the formation of loop-type structures that extend into the membrane, or into a channel-like opening, that are accessible from the cytoplasm (Fig. 7). Such pore loop structures have been identified in a number of ion channels (37). Pore loops in the voltage-gated potassium channel are thought to extend into the central pore from the extracellular side of the membrane forming an ion selectivity filter (37). Putative pore loops in GAT-1 may function in substrate selectivity. In fact, Tamura et al. (38) have suggested that the loop connecting HD7 and HD8 in the GABA transporter is involved in forming a pocket to which substrate binds. These authors provide evidence that the loop connecting HD7 and HD8 is involved in determining the GABA binding affinities for different GABA transporters.
Recent studies of GAT-1 and GLYT1 (glycine transporter) transmembrane topology using glycosylation site insertion, cysteine mutagenesis, and gene fusion methods (14, 15) suggest that significant revisions of the amino-terminal portion of the transporter model are needed. In these authors' revised model, the first HD extends into the membrane from the cytosolic face as a loop; HD2 is the first true HD; CL1 is located on the extracellular face of the membrane; and an additional HD is found in the amino-terminal portion of EL2. The transmembrane topology of HD4 through HD12 in these authors' model is in agreement with the predicted hydropathy model and the model presented in the current study. Due to the incremental mode of insertion and integration of the transporter prolactin chimeras, our study does not confirm or disprove the revisions suggested by these authors. However, data obtained with EL3 did indicate that the amino-terminal portion of EL2 may be accessible to protease, which would support the presence of an additional HD in this region.
The particular approach chosen here to study GAT-1 transmembrane topology has yielded data that begins to elucidate the unusual transmembrane assembly of this protein. Translocation of a nascent polypeptide across the ER membrane depends upon the presence of a signal sequence. More specifically, signal anchor sequences are responsible for membrane targeting, translocation, and membrane integration of nascent chains. Partial translocation of a nascent chain across a membrane requires the presence of a second type of topogenic sequence, a stop-transfer sequence. It has been proposed that proteins which span the membrane several times acquire their unique topology by a series of alternating signal and stop-transfer sequences (39, 40). Study of the membrane translocation of bovine opsin suggests that alternating signal and stop-transfer sequences are responsible for directing the assembly of this protein (41). Assembly of the carboxyl-terminal portion of GAT-1, including HD5, HD6, and HD9-HD12 appears to occur through an alternating pattern of signal and stop-transfer sequences similar to that observed for bovine opsin (41). My data are confirmed by Olivares et al. (15) studying individual and paired GLYT1 HDs. However, amino-terminal GAT-1 assembly of HD1-HD4 as well as HD7 and HD8 seems complex. Based on the predicted hydropathy model for GAT-1, HD1 should possess signal anchor sequence properties which would direct translocation of the prolactin epitope into the ER lumen and integrate the nascent polypeptide into the ER membrane such that the amino terminus extends into the cytoplasm. However, protease protection data reveal that HD1 does not become integrated into the membrane until HD2 is also present, and HD2, which is expected to have stop-transfer ability based on the predicted topology of GAT-1, does not exhibit this property in the CL1 fusion protein. The membrane integration of HD3 and HD4, as well as that of HD7 and HD8, also reveal the coordinate action of topogenic sequences other than a series of alternating signal and stop-transfer sequences. Similarly, the topogenic properties of GLYT1 individual and paired HD1 and HD2 and individual HD4 do not traverse the membrane in the predicted manner (15).
During synthesis of apolipoprotein B, translocation is paused resulting
in nonintegrated, transmembrane intermediates which become fully
translocated over time (42, 43). The pause in translocation is thought
to be triggered by a pause-transfer sequence that halts the nascent
protein in an aqueous channel until translocation and integration is
resumed. Pause-transfer sequences have also been identified in the
prion protein, and are thought to mediate the unusual translocation of
this protein (43). Pause-transfer sequences from these two proteins
have been identified and are strikingly similar (KPKTNMKHMA,
apolipoprotein B B; KKTKNSEEFA, prion protein STE) (43). Analysis of
the GAT-1 amino acid sequence reveals the presence of a stretch of
residues (KPKTLVVKVQKK) in the amino terminus with remarkable
similarity to the pause-transfer sequences identified in apolipoprotein
B and the prion protein. Pause-transfer sequences and other
unidentified unconventional topogenic sequences may mediate the unusual
translocation and integration of GAT-1. More detailed analyses are
necessary to elucidate the topogenic sequences responsible for GAT-1
transmembrane assembly and may aid in understanding the transmembrane
assembly of other complex polytopic integral membrane proteins.
In the absence of x-ray crystallography data of complex integral
membrane proteins such as GAT-1, one must infer the transmembrane topology of such proteins using methods such as those described above.
While each of these approaches has limitations, valuable information
may be obtained as long as results are interpreted with caution. Given
the significant sequence homology between the members of this
transporter family, it is expected that the model of GAT-1 proposed
here is representative of all Na+- and
Cl-dependent transporters. These results
provide a foundation for future topology studies of this transporter
family using alternative methodologies.
I thank M. Brownstein for generous support and encouragement, R. Seal and T. Usdin for helpful discussions, J. Northup and W. Clark for helpful discussions and critical review of the manuscript, Z. Hall for critical review of the manuscript, and C. Smith for assistance with the laser scanning microscope. V. Lingappa and W. Skach generously provided pSPSp+1L.ST.gG.pT, BPI, and many helpful suggestions.