(Received for publication, October 28, 1996)
From the Whitehead Institute for Biomedical Research,
Cambridge, Massachusetts 02142-1479 and
Shriners Hospitals for
Crippled Children, Massachusetts General Hospital, Department of
Anesthesia, Harvard Medical School, Boston, Massachusetts 02114
Caveolin, a 21-24-kDa integral membrane protein,
is a principal component of caveolae membranes. We have suggested that
caveolin functions as a scaffolding protein to organize and concentrate certain caveolin-interacting proteins within caveolae membranes. In this regard, caveolin co-purifies with a variety of
lipid-modified signaling molecules, including G-proteins, Src-like
kinases, Ha-Ras, and eNOS. Using several independent approaches, it has
been shown that a 20-amino acid membrane proximal region of the
cytosolic amino-terminal domain of caveolin is sufficient to mediate
these interactions. For example, this domain interacts with
G-protein subunits and Src-like kinases and can functionally
suppress their activity. This caveolinderived protein domain has
been termed the caveolin-scaffolding domain. However, it remains
unknown how the caveolin-scaffolding domain recognizes these
molecules.
Here, we have used the caveolin-scaffolding domain as a receptor to
select random peptide ligands from phage display libraries. These
caveolin-selected peptide ligands are rich in aromatic amino acids and
have a characteristic spacing in many cases. A known caveolin-interacting protein, Gi2, was used as a ligand
to further investigate the nature of this interaction.
Gi2
and other G-protein
subunits contain a single
region that generally resembles the sequences derived from phage
display. We show that this short peptide sequence derived from
Gi2
interacts directly with the caveolin-scaffolding
domain and competitively inhibits the interaction of the
caveolin-scaffolding domain with the appropriate region of
Gi2
. This interaction is strictly dependent on the
presence of aromatic residues within the peptide ligand, as replacement
of these residues with alanine or glycine prevents their interaction
with the caveolin-scaffolding domain. In addition, we have used this
interaction to define which residues within the caveolin-scaffolding
domain are critical for recognizing these peptide and protein ligands.
Also, we find that the scaffolding domains of caveolins 1 and 3 both
recognize the same peptide ligands, whereas the corresponding domain
within caveolin-2 fails to recognize these ligands under the same
conditions. These results serve to further demonstrate the specificity
of this interaction. The implications of our current findings are
discussed regarding other caveolin- and caveolae-associated
proteins.
Caveolae are plasma membrane-attached vesicular organelles that have a characteristic diameter in the range of 50-100 nm (1, 2). Caveolae are present in most cell types but are especially abundant in adipocytes, endothelial cells, fibroblasts, and smooth muscle cells (3). In adipocytes and smooth muscle cells, they represent up to 20% of the total plasma membrane surface area. Endothelial cells contain ~5,000-10,000 caveolae/cell. Although they were originally implicated in cellular transport processes (4), recent evidence suggests that they may participate in signal transduction-related events (5-11).
Caveolin, a 21-24-kDa protein, is a principal integral membrane component of caveolae membranes in vivo (12, 13). Using either Triton X-100-based methods or detergent-free methods, caveolin co-purifies with certain lipid-modified signaling molecules (such as G-proteins, Src family tyrosine kinases, and Ha-Ras) (5, 6, 7, 9, 10, 14-16). In addition, caveolin was first identified as a major v-Src substrate that undergoes tyrosine phosphorylation in Rous sarcoma virus-transformed cells (17). Based on these and other observations, we have proposed the "caveolae signaling hypothesis," which states that caveolar localization of certain signaling molecules could provide a compartmental basis for their actions and explain cross-talk between signaling pathways (18-20).
Several independent lines of evidence suggest that caveolin may function as a scaffolding protein within caveolae membranes: (i) both the amino- and carboxyl-terminal domains of caveolin remain entirely cytoplasmic and are therefore accessible for cytoplasmic protein-protein interactions (21); (ii) caveolin forms high molecular mass homo-oligomers of ~350 kDa (22, 23) that have the capacity to interact with specific lipids (cholesterol and glycosphingolipids; Refs. 24 and 25) and lipid-modified signaling molecules (G-proteins, Src family kinases, and Ha-Ras; Refs. 16 and 26-29); (iii) these caveolin homo-oligomers can self-associate to form caveolae-like structures in vitro (22, 25); and (iv) recombinant overexpression of caveolin in caveolin-negative cells (lymphocytes and Sf 21 insect cells) is sufficient to drive the formation of caveolae-sized vesicles (30, 31). Thus, it appears that the caveolin protein has the capability of interacting with itself, specific lipids, and other proteins and can serve to orchestrate caveolae formation.
Using a variety of domain-mapping approaches (deletion mutagenesis,
GST1 fusion proteins, and synthetic
peptides), a region within caveolin has been defined that mediates the
interaction of caveolin with itself and other proteins (16, 26-29).
This cytoplasmic 41-amino acid membrane proximal region of caveolin is
sufficient to mediate the formation of caveolin homo-oligomers (22),
and the carboxyl-terminal half of this region (20 amino acids, residues
82-101) mediates the interaction of caveolin with G-protein subunits and Src family tyrosine kinases (26-29). This interaction is
sufficient to suppress the GTPase activity of G-proteins and inhibits
the autoactivation of Src family tyrosine kinases (29). As this caveolin domain (residues 82-101) is critical for caveolin
homo-oligomerization and the interaction of caveolin with certain
caveolin-associated proteins (G-proteins, Ha-Ras, and Src family
kinases), we have previously termed this protein domain the
caveolin-scaffolding domain (29).
Here, we have used the caveolin-scaffolding domain and one of its
protein ligands (a G-protein subunit, Gi2
) to
explore the possible protein sequence requirements that underlie this molecular recognition event. As a first step, a fusion protein carrying
the caveolin-scaffolding domain was used to select random peptide
ligands from phage display libraries. Many of these peptide ligands
share two properties with a particular region of the G-protein: (i) a
preponderance of aromatic amino acids in a short stretch; and (ii) a
characteristic spacing between these aromatic residues. We show that
this region of Gi2
interacts directly with the caveolin-scaffolding domain, and a peptide encoding this G-protein domain competitively inhibits the binding of the appropriate region of
Gi2
to the caveolin-scaffolding domain.
The 15-mer library and
bacterial strains were from Dr. George Smith (University of Missouri,
Columbia, MO), whereas the decapeptide (10-mer) library was constructed
as described elsewhere (32). The GST-caveolin-1-(61-101) fusion
protein was purified and immobilized on glutathione-agarose beads
(Sigma) as described previously (26). The incubation was performed
using 6 × 1010 transforming units of phage in
presence of 50 µl of beads in TNET buffer (10 mM Tris pH
8.0, 150 mM NaCl, 1 mM EDTA, 0.05% Tween 20)
at 4 °C under agitation for 16 h. Beads were then washed with
20 ml of TNET before acidic elution with 100 µl of 50 mM glycine, pH 2.2. The eluate was neutralized with 0.1 volume of 1 M Tris, pH 10.0, and used to infect Escherichia
coli K91-kan cells for amplification. Purified phage were prepared
as described elsewhere (33). Four rounds of selection were performed.
Colonies from the last titering plate were picked to inoculate 2 ml of Luria broth/tetracycline medium. The culture was grown overnight, and
phage DNA was prepared by polyethylene glycol precipitation of the
culture supernatant, phenol-chloroform extraction and ethanol precipitation (33). The region corresponding to the selected peptide
was sequenced using a Sequenase II kit from U.S. Biochemical Corp. An oligonucleotide specific for the fuse5 vector downstream of the cloning site was used for sequencing
(5-GCCTGTAGCATTCCACAGACAA-3
).
We used 96-well plates (Nunc Maxisorp)
to perform all ELISAs. Phage from selected clones were prepared as
described elsewhere (33). Purified phage in Tris-buffered saline (TBS)
were used for coating the plate, and the incubation proceeded at
4 °C overnight. Plates were saturated for 1 h at room
temperature with TBS/0.05% Tween 20 (TBST) containing 1% bovine serum
albumin. Purified His-tagged Myc full-length caveolin-1 (1 µg/well)
(25) or GST-caveolin-1-(61-101) fusion protein (1 µg/well) (26) was
then added in TBST to the well for 2 h and washed several times
with TBST. Monoclonal antibodies directed against the Myc epitope
(Harvard Monoclonal Antibody Facility, Cambridge, MA) (1:400) or GST
(Santa Cruz Biotechnology) (1:1000) in TBST were then added for 2 h at room temperature followed, after washes, by horseradish peroxidase
anti-mouse IgG (Amersham Corp.) (1:2000). The reaction was revealed
using 2,2-azino-di-[3-ethylbenzthiazoline sulfonate (6)] (Boehringer
Mannheim), and absorbance was measured on a microplate reader at 410 nm. The opposite assay was also performed, in which purified
full-length caveolin-1-Myc-H7 (1 µg/well) or
GST-caveolin-1-(61-101) fusion proteins (1 µg/well) were coated onto
the wells in 100 mM sodium bicarbonate, pH 8.5, and then
incubated with phage. A biotinylated anti-fd polyclonal antibody
(Sigma) (1:1000) was used in addition to horseradish peroxidase-streptavidin (Zymed) (1:2000) and
2,2
-azino-di-[3-ethylbenzthiazoline sulfonate (6)]. For the ELISA
using peptides, 500 pmol/well was used for coating the wells in
bicarbonate coating buffer. Peptide synthesis was performed by
the Biopolymers Facility at the Massachusetts Institute of Technology
(Cambridge, MA) and Research Genetics (Huntsville, AL).
Using the bovine
Gi2 cDNA (provided by Dr. Nukada) as a template, the
polymerase chain reaction was performed to construct GST-Gi2
-full length (1-355), GST-Gi2
-B (120-238),
and GST-Gi2
-C (239-355). The products were subcloned into
the EcoRI site of the pGEX-1
T vector; resulting
constructions were subjected to restriction analysis and sequencing via
the Sanger method. Purification of GST-Gi2
fusion
proteins was essentially as described elsewhere (26, 34). To assess the
quality of the purification, the fusion proteins were analyzed by
SDS-polyacrylamide gel electrophoresis (10% acrylamide) and
transferred to nitrocellulose. After transfer, nitrocellulose sheets
were stained with Ponceau S to visualize protein bands and subjected to
immunoblotting with anti-GST IgG (1:1000) to visualize GST fusion
proteins. For ELISAs, GST-Gi2
fusion proteins (~200
ng/well) were added to peptide-coated wells, and binding was revealed
using a monoclonal antibody directed against the GST moiety.
To the caveolin-1-(82-101) peptide-coated well, a
mixture of ~120 ng of purified GST-Gi2 B fusion
protein (~3 pmol) and increasing amounts of competing peptide was
added. The mixture was assembled just prior to addition to the well.
The procedure following was as described above. We also used
biotinylated peptides in several experiments, and binding was revealed
using horseradish peroxidase-streptavidin (1:2000).
We used a GST-caveolin fusion protein
that contains the caveolin-scaffolding domain as a receptor to randomly
select peptide ligands from two different phage display libraries (Fig.
1). This GST-caveolin fusion protein contains caveolin
residues 61-101 and has been previously shown to be functionally
sufficient to interact with G-protein subunits (including
Gi2
), c-Src, and Ha-Ras (26, 29).
After four rounds of selection, caveolin-binding phage clones were
subjected to DNA sequence analysis to reveal their peptide sequences,
as shown in Table I. Caveolin binding clones from the
10-mer library were rich in tryptophan, and most exhibited a
characteristic spacing conforming to the sequence
WXWXXXXW. Random sequencing of the 10-mer library
indicated that tryptophan was enriched ~3-fold in caveolin-selected
clones (Fig. 2, A and B). In
contrast, caveolin binding clones from the 15-mer library did not
exhibit any characteristic spacing but were also enriched in aromatic
amino acids. More specifically, tryptophan, phenylalanine, and tyrosine
were enriched 3.8-, 1.5-, and 1.8-fold relative to the unselected
library population (Fig. 2, B and C). Also, a
single 15-mer clone
(RNVPPINDV
IA
) accounted for
~60% of all clones.
|
An ELISA was developed to evaluate the interaction of these peptide
ligands with various regions of caveolin. Nine caveolin-derived peptides that correspond to regions of the cytoplasmic amino-terminal domain of caveolin were used to as receptors to capture these ligands
(Fig. 3A). All phage clones tested only
interacted with the peptide that corresponds to the
caveolin-scaffolding domain (residues 82-101; Fig. 3A) and
with recombinant full-length caveolin-1 purified from E. coli (not shown). The clones tested include the three most
abundant clones from the 10- and 15-mer libraries (see Table I) and the
clone containing the sequence
VE
AVSR
N. Identical results were
obtained when a biotinylated peptide corresponding to the sequence of
the most abundant 15-mer phage clone was tested.
When the caveolin-scaffolding domain is divided into two halves (residues 84-92 and 93-101), we have previously shown that this inhibits its functional activity (26). Similarly, each of these halves was unable to interact with the caveolin-selected phage clones and their corresponding biotinylated peptides (Fig. 3B).
Recent studies have shown that caveolin is the first member of a multigene family of related molecules (13, 27, 28, 35, 36); caveolin has been retermed caveolin-1. Thus, we next tested the interaction of these caveolin-selected phage clones with the scaffolding domains of caveolins 1-3. Fig. 3B shows that these peptide ligands only interact with the scaffolding domains of caveolins 1 and 3 but fail to interact with the homologous domain in caveolin-2. This suggests that the scaffolding domain of caveolin-2 has different protein sequence requirements for its interaction with other molecules and further demonstrates the selectivity of these interactions, as these peptide ligands were selected using the scaffolding domain of caveolin-1.
Interaction of Gi2By comparing the
peptide ligands obtained for the caveolin-scaffolding domain with the
protein sequences of a known class of caveolin-interacting proteins
(G-protein subunits), we deduced two possible caveolin-binding
motifs:
X
XXXX
and
XXXX
XX
, where
is an aromatic
residue (Trp, Phe, or Tyr). These two motifs correspond to the
sequences of most 10-mer phage clones and the most abundant 15-mer
clone, respectively. Fig. 4A shows that
these two motifs are present in most G-protein
subunits in a
composite manner (
X
XXXX
+
XXXX
XX
X
XXXX
XX
).
To evaluate whether this region of G-protein subunits can serve as
a ligand for the caveolin-scaffolding domain, we constructed: (i) a
synthetic peptide containing this G-protein region
(TH
T
KLDLH
KM
DV), termed GP; and (ii) a variety of GST-Gi2
fusion
proteins, including full-length (FL) Gi2
and two
deletion mutants, termed B and C (Fig. 4B). Note that the GP
region is contained within two of these GST-Gi2
fusion
proteins (FL and B).
Fig. 5A shows that both the FL and B
Gi2 fusion proteins interacted preferentially with the
caveolin-scaffolding domain, but little or no interaction was observed
with an adjacent region of caveolin (residues 53-81). Also, the B
region of Gi2
was only recognized by the scaffolding
domains of caveolins 1 and 3 but not by the scaffolding domain of
caveolin-2 (Fig. 5B). This binding profile is exactly what
we observed previously for peptide ligands selected using the
scaffolding domain of caveolin-1 (See Fig. 3B), suggesting a
similar mode of interaction.
Fig. 6A shows that the GP peptide interacts
with the scaffolding domain of caveolins 1 and 3 and that this
interaction is strictly dependent on the presence of aromatic residues
within the GP peptide. When the phenylalanine residues of the GP
peptide (THT
KLDLH
KM
DV) were
changed to either alanine or glycine
(TH
T
KLDLH
KM
DV and
TH
T
KLDLH
KM
DV),
binding of these peptides to the caveolin-scaffolding domain was almost
completely abolished. This is consistent with the idea that aromatic
residues play a key role in recognition by the caveolin-scaffolding
domain. Also, it is important to note that the GP peptide was only
recognized by the scaffolding domain of caveolins 1 and 3 but not by
the scaffolding domain of caveolin-2 (Fig. 6A).
The GP peptide also competitively inhibits the binding of the B region
of Gi2 to the caveolin-scaffolding domain, and this
occurs in a dose-dependent manner (Fig. 6B). In
this regard, it is important to note that the B region contains the
sequence that corresponds to the GP peptide. Virtually identical
results were obtained using the peptide ligand selected from the 15-mer phage display library
(RNVPPI
NDVY
IA
), indicating that
this caveolin-selected peptide, the GP peptide, and Gi2
are all recognized by the caveolin-scaffolding domain in a similar
fashion. Importantly, mutated GP peptides lacking phenylalanine
failed to show any competition for binding.
Fig. 7 shows that the GP peptide also recognizes the
full-length intact caveolin-1 molecule (caveolin-1-Myc-H7).
However, as predicted, the GP peptide was not recognized by
full-length caveolin-2 (caveolin-2-Myc-H7) under
identical conditions. These results provide an extraordinary
demonstration of the specificity of this interaction given the close
protein sequence homology between caveolins 1 and 2. Caveolin-2 is 58%
similar and 38% identical to caveolin-1 (27).
As the protein sequence encoded by the GP peptide appears to serve as a
ligand for the caveolin-scaffolding domain, we identified the location
of this sequence within the known three-dimensional structure of
G-protein subunits. The GP region lies directly between switch I
and switch II regions and precisely defines the space between where
switch I ends and switch II begins (Fig. 8). Interestingly, we have previously shown that a Gln
Leu mutation (located 5 amino acid residues downstream from the end of the GP
region) prevents the interaction of Gs
with caveolin-1 (26). This mutation also locks the G-protein
subunit in the GTP-liganded and -activated conformation.
Mutational Analysis of the Caveolin-1-scaffolding Domain
Two
different approaches were used to define critical residues within the
caveolin-1-scaffolding domain that are required for binding peptide
ligands. First, deletion mutagenesis of the 82-101 region indicated
that a minimal length of 16 amino acids is required (residues 86-101)
(Fig. 9A). Second, alanine-scanning mutagenesis was then performed using this minimal
caveolin-1-scaffolding domain. Our results indicate that a central core
of four amino acids (92FTVT95) is strictly
required for interaction of the caveolin-1-scaffolding domain with both
the GP peptide and the corresponding region of Gi2 (Fig.
9B). This region is F
S in caveolin-3 and
F
S in caveolin-2. Thus, this small difference may explain
why these peptide ligands (library peptides and GP peptide) and protein ligands (GST-Gi2
fusion proteins) are recognized by the
scaffolding domains of caveolins 1 and 3, but not by the scaffolding
domain of caveolin-2 (as shown in Figs. 3B,
5B, and 6A).
Here, we have identified peptide and protein ligands for the scaffolding domain of caveolin and characterized the sequence requirements of this reciprocal interaction. Several independent lines of evidence indicate that this interaction is extremely specific: (i) the identified peptide and protein ligands interacted only with the caveolin-scaffolding domain, but not with other regions of the caveolin protein; (ii) the interaction also occurred with the purified full-length caveolin molecule expressed as a polyhistidine-tagged protein; (iii) the interaction was sequence-specific; mutation of critical phenylalanine residues within the ligand prevented the binding of these ligands to the caveolin-scaffolding domain; also, mutation of certain critical residues within the caveolin-scaffolding domain abrogated the binding of peptide and protein ligands; and (iv) these peptide and protein ligands bound selectively to the scaffolding domains of caveolins 1 and 3 but not to the scaffolding domain of caveolin-2. This is despite the fact that these domains within caveolins 1-3 are extremely homologous. Also, it is important to note that these ligands were identified using the scaffolding domain of caveolin-1, suggesting that other potential ligands may exist that selectively recognize the scaffolding domain of caveolin-2. These results are consistent with the previous observations that the scaffolding domains of caveolins 1 and 3 can both act as GDP dissociation inhibitors for heterotrimeric G-proteins, and both inhibit the autoactivation of Src family kinases (16, 26-29). In contrast, the scaffolding domain of caveolin-2 exhibits GTPase-activating activity toward heterotrimeric G-proteins and fails to affect the activity of Src family kinases (27, 29).
What are the possible implications of these interactions? Other modular protein domains (such as Src homology 2 and 3, WW, and PID) have been previously defined and their corresponding peptide and protein ligands identified using techniques that are similar or identical to the ones used here for the caveolin-scaffolding domain. Src homology 2 domains and PID domains both recognize phosphotyrosine and 3 or 4 surrounding residues; Src homology 3 domains recognize proline-rich sequences, especially with the consensus PXXP; and the WW domain recognizes proline-rich sequences such as PPPY (see Prosite data base for specific references). Protein kinases also recognize short peptide sequences to direct phosphorylation of specific peptide and protein substrates: protein kinase A (R-R-X-[ST]); protein kinase C ([ST]-X-[RK]); casein kinase II ([ST]-X-(2)-[DE]); and tyrosine kinases ([RK]-X-(2,3)-[DE]-X-(2,3)-Y). This is also the case for other posttranslational modifications, including: N-glycosylation (N-X-[ST]); N-myristoylation (MG); dual acylation (MGC); and prenylation (C-aliphatic-aliphatic-X). Similarly, antibody binding depends on the recognition of specific epitopes that can be as small as 5-10 amino acids in length. Thus, many diverse cellular processes rely heavily on the recognition of short consensus peptide motifs. However, not all of these motifs are recognized, as this depends on whether a given motif is cytoplasmic or extracellular or is exposed on the surface of the protein; exposure on the surface of the protein may even be conformation-specific. Also, recognition may depend on the subcellular localization of the protein and the potential interacting partner, e.g. nuclear, cytoplasmic, plasma membrane, or Golgi-associated. These added criteria for interaction thus strictly modulate whether the interaction may or may not take place and greatly increase the specificity of these interactions. Of course, these same constraints would apply to the recognition of peptide and protein ligands by the caveolin-scaffolding domain.
An increasing number of reports suggest that many different classes of
molecules are concentrated within caveolin-rich membrane domains. Thus,
we searched the protein sequences of known caveolin- or
caveolae-associated proteins for aromatic-rich sequences that contain a
specific spacing (X
XXXX
,
XXXX
XX
, or
X
XXXX
XX
, where
= Trp,
Phe, or Tyr), as defined here using the caveolin-scaffolding domain.
Most molecules reported to be associated with caveolin or caveolae
contain cytoplasmically accessible sequences that resemble those that
we have defined here as peptide and protein ligands for the
caveolin-scaffolding domain (summarized in Table II).
Although it is not yet known whether all these molecules interact
directly with caveolin, our studies provide a rational and systematic
basis for investigating whether these protein sequences are indeed
recognized as cytoplasmic ligands by the caveolin-scaffolding domain.
This type of interaction could provide a simple mechanism for
sequestration of a diverse group of molecules within caveolin-rich regions of the plasma membrane.
|
The caveolin-binding motif is reminiscent of another motif identified using the chaperone BiP for selection: Hy(W/X)HyXHyXHyXHy, where Hy is a large hydrophobic amino acid (most frequently Trp, Leu, or Phe) and X is any amino acid (37). In fact, the best peptides tested for binding to BiP contain one Phe and three Trp, which suggests that they could also bind to caveolin. However, for BiP, no single peptide sequence was enriched over others. This is in contrast to the present situation, in which strong enrichments were observed for particular caveolin-selected peptide sequences. The apparent similarity between BiP and caveolin in this matter may suggest a potential role for caveolin as a chaperone. Although to our knowledge no chaperone described thus far is an integral membrane protein, caveolin could fulfill the role of a "membrane-bound chaperone" by interacting with signaling molecules and maintaining them in an inactive conformation.
We thank Drs. Gerald R. Fink, Peter S. Kim, and Harvey F. Lodish for their enthusiasm and encouragement; Dr. Ton Schumacher (Kim Laboratory) for help with the phage display experiments and for critical discussions; Dr. Philipp Scherer for critical discussions; Dr. John R. Glenney for monoclonal antibodies (2297 and 2234) directed against caveolin; and members of the Kim and the Lisanti laboratories for insightful discussions during the course of this study.