(Received for publication, February 12, 1997)
From the Departments of Biochemistry and Molecular Pharmacology, and Microbiology and Immunology, Kimmel Cancer Institute, Thomas Jefferson University, Philadelphia, Pennsylvania 19107
Previously we demonstrated that nonvisual
arrestins exhibit a high affinity interaction with clathrin, consistent
with an adaptor function in the internalization of G
protein-coupled receptors (Goodman, O. B., Jr., Krupnick, J. G.,
Santini, F., Gurevich, V. V., Penn, R. B., Gagnon, A. W., Keen,
J. H., and Benovic, J. L. (1996) Nature 383, 447-450).
In this report we show that a short sequence of highly conserved
residues within the globular clathrin terminal domain is responsible
for arrestin binding. Limited proteolysis of clathrin cages results in
the release of terminal domains and concomitant abrogation of arrestin
binding. The nonvisual arrestins, -arrestin and arrestin3, but not
visual arrestin, bind specifically to a glutathione
S-transferase-clathrin terminal domain fusion protein.
Deletion analysis and alanine scanning mutagenesis localize the binding
site to residues 89-100 of the clathrin heavy chain and indicate that
residues 1-100 can function as an independent arrestin binding domain.
Site-directed mutagenesis identifies an invariant glutamine (Glu-89)
and two highly conserved lysines (Lys-96 and Lys-98) as residues
critical for arrestin binding, complementing hydrophobic and acidic
residues in arrestin3 which have been implicated in clathrin binding
(Krupnick, J. G., Goodman, O. B., Jr., Keen, J. H., and Benovic,
J. L. (1997) J. Biol. Chem. 272, 15011-15016).
Despite exhibiting high affinity clathrin binding, arrestins do not
induce coat assembly. The terminal domain is oriented toward the plasma
membrane in coated pits, and its binding of both arrestins and AP-2
suggests that this domain is the anchor responsible for
adaptor-receptor recruitment to the coated pit.
Receptor-mediated endocytosis is a pathway by which plasma
membrane receptors are internalized selectively into cells through clathrin-coated pits (for review, see Refs. 1 and 2). Receptors are
either internalized constitutively (e.g. some nutrient
receptors) or internalized preferentially following ligand binding
(e.g. some hormone receptors). Among those that belong to
the latter category are several G protein-coupled receptors
(GPRs).1 Many GPRs are phosphorylated
rapidly upon agonist activation, and it is this phosphorylation event
that leads to the selective binding of a family of proteins termed
arrestins. To date there are four known arrestins: two visual arrestins
(arrestin and cone arrestin) and two nonvisual arrestins (-arrestin
and arrestin3; for review, see Ref. 3). One role of arrestin binding is
to uncouple the GPR from its cognate heterotrimeric G protein
(desensitization), probably through direct competition with the G
protein for the activated GPR (4). Following desensitization, the
receptor is believed to be translocated into an intracellular
compartment (5), a process commonly referred to as sequestration.
Interestingly, sequestration also appears to be required for
resensitization of receptors by allowing for receptor dephosphorylation
and recycling (6, 7).
Studies of the m2 muscarinic acetylcholine receptor (8) and a
sequestration-deficient mutant form of the 2-adrenergic receptor (9) have implicated GPR kinases and arrestins in GPR sequestration. Recent work from our laboratories has provided a
plausible molecular mechanism for this internalization process, in
which arrestins function as clathrin adaptors in the uptake of
2-adrenergic (10) and other2
GPRs. We found that nonvisual arrestins bind to clathrin with a
Kd of 10-60 nM, comparable to that of
the clathrin associated protein AP-2 (12, 13). Moreover,
2-adrenergic receptor,
-arrestin, and clathrin
colocalize in vivo upon agonist addition in an
arrestin-dependent manner, indicating that the arrestin/clathrin interaction observed in vitro likely
occurs in vivo in the presence of an activated receptor.
These results suggest that the arrestin/clathrin interaction is of
central importance in regulating GPR trafficking.
Clathrin, the major structural component of the coated pit lattice,
consists of three heavy chains (Mr 192,000)
and three light chains (Mr
26,000) (for
review, see Ref. 14). By limited proteolysis of preformed clathrin
cages, the heavy chain has been shown to comprise two distinct domains
(15). An
50-kDa amino-terminal globular region, the terminal domain
(TD), lies at the distal end of each triskelion leg. The remaining
140 kDa of the heavy chain forms the triskelion core, consisting of
the vertex, proximal leg, and a portion of the distal leg of the
clathrin trimer. One clathrin light chain is associated with each
proximal leg. Based on electron microscopic studies and models of the
polygonal lattice, a clathrin hub lies at each vertex, and each lattice
edge is formed by the superposition of two distal and two proximal
clathrin heavy chain legs (16-18).
In the coated pit, the TDs are believed to be located beneath each vertex of the surface lattice and to project inward toward the plasma membrane (16). It has been shown that the TD contains a binding site for plasma membrane-localized AP-2 and that this site is required for AP-2-mediated coat assembly (19). Indeed, cryoelectron microscopy analysis suggests that AP-2 lies between the membrane and the shell of clathrin TDs (16), suggesting that this interaction may be important in the AP-2 adaptor function as well as AP-2 assembly activity.
Here we show that the TD also contains a distinct arrestin binding site. By deletion and alanine replacement mutagenesis of recombinant TDs, we identify clathrin heavy chain residues 89-100 as critical for arrestin binding, but not required for AP-2 binding. Site-directed mutagenesis experiments indicate that several highly conserved and invariant residues contribute to this binding. Finally, we show that arrestins are not sufficient to support clathrin assembly, consistent with previous assembly models. These observations suggest that a general function of the clathrin TD is to interact with multiple adaptors, targeting their cognate surface receptors to coated pits.
Glutathione-agarose was purchased from Sigma.
Clathrin and AP-2 were prepared from bovine calf brains as described
previously (20). L-1-Tosylamido-2-phenylethyl chloromethyl
ketone-trypsin was from Worthington Biochemical, Inc., and HEPES and
isopropyl-1-thio--D-galactopyranoside were from
Boehringer Mannheim. All other chemicals were reagent grade or better.
Buffers utilized are as follows. Buffer A is 0.1 M Na-MES,
pH 6.8, 0.1 mM dithiothreitol, 0.1 mg/ml soybean trypsin
inhibitor. Buffer B is 20 mM K-HEPES, pH 7.3, 120 mM potassium acetate, 0.1 mM dithiothreitol,
0.1 mg/ml soybean trypsin inhibitor, 1 µg/ml each leupeptin,
pepstatin, and antipain and 0.1% Triton X-100. Buffer C is 0.1 M Na-MES, pH 6.8, 0.1 mM phenylmethylsulfonyl fluoride.
Glutathione S-transferase (GST)-TD fusion proteins were constructed from polymerase chain reaction fragments derived from a bovine brain cDNA library. Amino acid residues 1-579 were amplified and cloned into the BamHI and SmaI sites of pGEX-2T (GST-TD, Pharmacia). GST-TD(1-389) was generated by subcloning the BamHI-EcoRI fragment of pGST-TD into pGEX-2T. TD deletion and site mutants were subsequently obtained from pGST-TD template by polymerase chain reaction-based mutagenesis (primer sequences available upon request). All constructs were confirmed by DNA sequencing (Nucleic Acids Facility, Kimmel Cancer Institute).
Expression and Purification of Recombinant Proteins in Escherichia coliProteins were expressed in E. coli
BL21 cells. A 5.0-ml overnight culture grown in LB medium supplemented
with 100 µg/ml ampicillin was diluted into 250 ml of medium and grown
2 h at 37 °C (A600 ~ 0.6). Cultures
were then induced with 75 µM
isopropyl-1-thio--D-galactopyranoside for 2 h at
30 °C. Bacterial pellets were resuspended in 5 ml of phosphate-buffered saline, 1 mM phenylmethylsulfonyl
fluoride, 2 mg/ml lysozyme, incubated at room temperature for 15 min,
and then supplemented with Triton X-100 (1% v/v). Lysates were
obtained by two rapid freeze-thaw cycles, treated with 300 units of
DNase for 15 min on ice, and centrifuged at 50,000 rpm for 10 min.
Supernatants were supplemented with 2 mM dithiothreitol and
gently agitated in the presence of 200 µl of glutathione-agarose
beads overnight at 4 °C. Beads were washed three times with 10 ml of
phosphate-buffered saline, 1% Triton X-100, and three times with 1.4 ml of phosphate-buffered saline in the absence of detergent. Protein
content of the beads was determined by densitometric analysis of
Coomassie-stained SDS-PAGE gels (Molecular Dynamics). Yields were
typically 1.0 mg/250 ml of original culture volume. Recombinant
arrestins were expressed and purified as described previously (10).
Preparation of intact and clipped clathrin cages reconstituted with light chains has been described elsewhere (21). Arrestin binding to cages was performed as described previously (10). Briefly, arrestins (300 nM) were incubated in the presence of clathrin cage preparations (100 nM trimers) in buffer A for 10 min at room temperature. Complexes were centrifuged through a 75-µl 0.2 M sucrose cushion prepared in buffer A in a TLA100 rotor (Beckman) at 75,000 rpm for 5 min at 4 °C and analyzed by 7.5% SDS-PAGE. Coomassie-stained arrestin bands were quantified by densitometry.
GST-TD Binding AssaysFusion proteins (5 µg in a
5-µl bed volume of glutathione-agarose) and arrestins (
0.4 µg)
were incubated in 100 µl of buffer B for 1 h at 4 °C on a
rotator. Deletion mutant fusion protein additions were adjusted as
necessary to attain equimolar inputs. The beads were centrifuged and
the unbound fraction aspirated. The beads were washed subsequently with
500 µl of buffer B and eluted either with buffer B: 2 M
Tris-HCl, pH 7 (1:1, v/v), for 10 min or with SDS-PAGE sample buffer.
Binding was quantified by immunoblotting 5% of each fraction with the
monoclonal anti-arrestin antibody F4C1 (L. Donoso, Thomas Jefferson
University) and comparison with a 10-fold range of arrestin
standards.
Clathrin assembly by AP-2 has been
detailed elsewhere (20). Briefly, 100 µg of clathrin was dialyzed
overnight against buffer C in the presence of 40 µg AP-2 and/or 8 µg of arrestin. Retentates were microcentrifuged at 10,000 × g for 2 min (generating a low speed pellet fraction) and
then at 75,000 rpm for 5 min in a TLA100 rotor (generating high speed
pellet and supernatant fractions). Five percent of each fraction was
analyzed by 7.5% SDS-PAGE, and the clathrin heavy chains were
quantified by densitometry of Coomassie Blue staining.
To determine which region of the clathrin triskelion contains
a binding site for nonvisual arrestins, we subjected clathrin cages to
limited tryptic digestion. This treatment hydrolyzes clathrin light
chains and cleaves the 192-kDa heavy chains, resulting in release of
the globular 50-kDa TD (15). The remainder of the heavy chains,
comprising the hub, proximal leg, and a portion of the distal leg,
remains assembled and readily sedimentable (20), and we refer to these
structures as truncated cages. As shown in Fig. 1,
binding of
-arrestin and arrestin3 to truncated cages (lanes
3 and 7) was greatly reduced relative to that observed for intact cages (lanes 2 and 6). Reconstitution
of truncated cages with intact clathrin light chains (23) did not
restore arrestin binding (lanes 4 and 8).
Although we cannot rule out the possibility that other portions of the
clathrin heavy chain may also contribute to arrestin binding, these
data indicate that the clathrin TD, corresponding approximately to
residues 1-479 (24), contains an arrestin binding site.
To assess directly the interaction between TD and arrestin, we
expressed and affinity purified recombinant bovine clathrin TD as a
GST-fusion protein (GST-TD). Binding assays were carried out by
incubating arrestins with GST-TD beads or, as a negative control, GST
beads. After washing, bound arrestins were eluted completely with
either boiling SDS-PAGE sample buffer or high concentrations of
Tris-HCl (0.5 M).
In previous studies (10), we demonstrated that although visual arrestin
fails to interact appreciably with clathrin, the nonvisual arrestins,
-arrestin and arrestin3, bind to clathrin cages with high affinity
(Kd
10-60 nM). This arrestin selectivity is recapitulated in the GST-TD binding assay (Fig. 2). As expected, none of the arrestins bound to GST
beads (lanes 1, 4, and 7). Although
visual arrestin also failed to interact appreciably with GST-TD
(lane 8), both arrestin3 and
-arrestin bound GST-TD beads
(lanes 2 and 5, respectively). In general, arrestin3 bound more extensively (60-90% of input) than
-arrestin (30-50% of input).
Localization of the Arrestin Binding Site within the Clathrin TD
To localize more precisely the arrestin binding site within the clathrin TD, we initially employed an in vitro transcription/translation system and a GST-fusion protein containing the COOH-terminal half of arrestin3, the region of nonvisual arrestins known to be involved in clathrin binding (10, 25). In vitro translated clathrin heavy chain residues 1-389 effectively bound to this fusion protein (data not shown), localizing the binding region to the NH2-terminal moiety of the clathrin TD. Unfortunately, TD constructs containing greater COOH-terminal deletions were not expressed stably in the in vitro translation system.
Accordingly, we next generated a series of COOH-terminal-deleted GST-TD
constructs and assessed their ability to bind arrestins (Fig.
3). Interestingly, the first 100 residues of TD
supported arrestin binding to an extent comparable to that observed for the full-length TD (1-579). Deletion of an additional fifteen residues
(1-85) resulted in the complete abrogation of binding of both
arrestin3 and -arrestin, suggesting that a crucial arrestin binding
determinant is contained within residues 86-100 of the clathrin heavy
chain (Fig. 3).
To identify residues in this region involved in arrestin binding, we
initially performed triplet alanine replacement mutagenesis of residues
86-100, expressed in the 1-579 context (Fig. 4). For both arrestins, the most potent mutant was the KMK triplet (K98A, M99A,
K100A; Table I). With the exception of the KTL
construct, all alanine triplet replacements exhibited reduced arrestin
binding, with the effects on arrestin3 (Fig. 4A) generally
being somewhat greater than on -arrestin (Fig. 4B). These
findings focused attention on residues 89-100 as contributing to
arrestin binding.
To assess the contribution of individual clathrin residues to arrestin binding, we produced fusion proteins containing the complete clathrin TD (residues 1-579) with site-specific mutations in the 89-100 sequence. This region contains a number of invariant and highly conserved residues, and these were the targets for alteration (Table I). Charged residues were systematically inverted, whereas Ile-90 and Phe-91 were each converted individually to alanine to assess the contribution of hydrophobic interactions. The invariant glutamine residue at position 89, which may participate in hydrogen bonding, was converted to either methionine (approximately isosteric, non-hydrogen bonder) or alanine.
In general, both arrestin3 and -arrestin exhibited the same trends
in binding to the TD site-specific mutants, indicating that both
arrestins interact similarly with residues in this region. However, the
inhibition of arrestin3 binding was, in almost all cases, more complete
than that for
-arrestin (Fig. 5, A and
B). Of the charge inversion mutants, K98E and K96E were most
deficient with respect to arrestin3 and
-arrestin binding. The K100E
mutant exhibited only a modest loss of binding, whereas E94K showed no loss whatsoever. Of the hydrophobic residue mutants, F91A but not I90A
was significantly defective in binding to both arrestins, suggesting
that an aromatic moiety may be important. Both alterations of the
invariant glutamine residue, Q89A and Q89M, were also significantly defective.
To show that the loss of arrestin binding was not due to a global
disruption of TD structure, we subjected the wild-type and mutant
GST-TD fusion proteins most defective in arrestin binding to limited
tryptic proteolysis. In all cases, a major 50-kDa band was obtained
whose electrophoretic mobility was identical to TD derived from bovine
brain preparations (data not shown), confirming that the TD was a
folded structure much more resistant to protease than the linker
peptide connecting it to the GST moiety. Furthermore, we tested binding
of all site mutants to AP-2, which has been shown previously to bind
clathrin TD (19). In contrast to the arrestin binding results, all
mutants retained substantial AP-2 binding. Although two of the mutants
(F91A and K98E) exhibited some reduction, this decrease was
significantly less than that observed for arrestin binding.
Arrestins bind
clathrin with affinity similar to AP-2, a protein originally identified
based on its ability to promote clathrin lattice formation under
physiological solution conditions (22, 26). Like the arrestins, AP-2
binds to clathrin TDs (Ref. 19 and Fig. 5C), but AP-2 also
binds stably to the truncated light chain-free hubs of clathrin
generated either proteolytically (21) or by recombinant methods (27).
It was therefore important to determine whether arrestins can support
clathrin coat assembly. Whereas an equimolar input of AP-2 promoted
clathrin assembly, arrestins did not (Fig. 6): even a
10-fold molar excess of arrestin to clathrin did not yield assembled
lattices (data not shown). Furthermore, although both proteins bind to
clathrin TDs, the presence of arrestins in the assembly reaction did
not alter the ability of AP-2 to stimulate coat formation.
Fusion proteins containing progressive COOH-terminal truncations within the TD along with alanine scanning mutagenesis demonstrated that clathrin heavy chain residues 89-100 within the TD are an important determinant of arrestin binding. Examination of clathrin heavy chain sequences (Table I) indicates that these residues are highly conserved among diverse species, suggesting that this region is involved in a critical function.
GST-TD fusion proteins containing either Q89M or K96E mutations had
almost undetectable arrestin3 binding and greatly reduced -arrestin
binding. Interestingly, these are invariant residues in all known
clathrin heavy chains (Table I), suggesting that arrestins or
arrestin-like proteins are ubiquitous and play important roles in
cellular function. The Gln-89 residue is likely to be exposed on the
surface of the TD and participate in hydrogen bonding as replacement by
isosteric non-hydrogen-bonding methionine more extensively diminished
arrestin binding than did replacement by the smaller alanine
residue.
The F91A and K98E mutants had greatly diminished arrestin binding, but both also exhibited some loss of AP-2 binding. This may reflect some alteration in global structure not revealed by the protease sensitivity assay. As the charge of Glu-94 can be inverted without noticeably affecting arrestin binding, this residue likely lies on the surface of the TD but is not involved directly in arrestin binding. However, its conservation throughout all species for which sequence data are available suggests that it is required for other important interactions.
TD as an Arrestin and Adaptor Binding DomainThe globular
50-kDa region at the end of each extended clathrin leg has been
visualized as a distinct globular region and has accordingly been
designated the terminal domain. There is increasing evidence for
specific structural and functional attributes within this region.
Electron microscopy studies suggest that the morphology of the TD has a
variable shape capable of being extended to a greater or lesser extent
in a scroll-like manner (18). Based on proteolytic studies, the TD has
been considered to be composed of residues 1-479, but sequences
further into the linker region, at least 480-497, appear to be
associated with the NH2-terminal region as they remain
physically bound following cleavage (24). Finally, our data indicate
that residues 1-100 comprise a distinct structural domain that
exhibits functional arrestin binding essentially indistinguishable from
the intact TD.
Based on its spatial orientation within the coated pit, the clathrin TD is an apt binding site for arrestin and other candidate adaptor molecules. In the assembled clathrin coat structure, the hubs and distal legs form the cytoplasmic surface of the lattice, a 10-nm thick structure that lies about 15 nm from the membrane surface (16). The TD is believed to be joined to the distal leg by a short, flexible linker region (15, 18). Cryoelectron microscopic images suggest that the TDs project inward at varied angles, forming an inner shell concentric with the surface lattice (16). Therefore, TDs would be in an appropriate position to contact the exposed tails of receptors and the cytoplasmic proteins that bind to them, such as arrestins.
Arrestins as Adaptors for GPR-mediated EndocytosisThe nonvisual arrestins, by virtue of their demonstrated ability to bind tightly to both clathrin and agonist-activated receptor, and their involvement in receptor sequestration at the plasma membrane, constitute a new class of adaptors distinct from the heterotetrameric APs (10, 28). Although both bind clathrin with similar affinities, arrestins and the plasma membrane-associated AP-2 differ in several important respects. Arrestins bind tightly only to the TD region of clathrin. In contrast, AP-2 binds stably to both the TD and the clathrin triskelion core, with each of its large subunits independently recognizing clathrin (29, 30). Arrestin binding to clathrin does not support coat assembly (Fig. 6); interestingly, binding of AP-2 to recombinant clathrin hubs also does not induce lattice formation, although the hubs independently can undergo a polymerization reaction to generate a lattice (27). We note that these observations support the proposal that the AP-2-mediated coat assembly reaction requires bivalent interaction of AP-2 with two clathrin triskelia (20).
In intact cells, AP-2 is a structural component of clathrin-coated pits
and vesicles: it is colocalized with clathrin and can be recovered as a
stoichiometric component of coated vesicles isolated from tissues (14).
In contrast, arrestins seem to be recruited to coated pits only in
response to agonist activation of GPRs (10) and are detected only at
trace levels (1 arrestin:50 triskelia) in purified bovine brain
coated vesicles (data not shown). These observations are consistent
with the model that preformed coated pits bind receptors, or in this
case arrestin-receptor complexes, rather than being formed de
novo in a receptor-induced assembly reaction (31). In contrast,
the properties of AP-2 indicate that it participates both structurally
and functionally in the lattice assembly reaction.
Finally, as potential adaptors, the nonvisual arrestins and AP-2 have remarkably different receptor binding affinities despite possessing similar affinities for clathrin. Whereas arrestins bind cognate GPRs with low or subnanomolar affinity (32), corresponding affinities of APs for cytoplasmic receptor tails appear several orders of magnitude weaker (33). Perhaps the arrestins are sufficient as GPR adaptors, whereas AP-2 works in concert with other potential receptor-binding proteins, such as shc (34) and eps15 (11).
We thank C. Carman and Drs. Z. Huang, A. Gagnon, and L. Kallal for helpful discussions.