(Received for publication, January 24, 1997, and in revised form, February 27, 1997)
From the a Whitehead Institute for Biomedical Research, Cambridge, Massachusetts 02142-1479, the b Department of Medicine, Diabetes Unit, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114, the e Department of Pathology, Harvard Medical School and Beth Israel Hospital, Boston, Massachusetts 02215, and the i Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139
Caveolae are plasmalemmal microdomains that are involved in vesicular trafficking and signal transduction. We have sought to identify novel integral membrane proteins of caveolae. Here we describe the identification and molecular cloning of flotillin. By several independent methods, flotillin behaves as a resident integral membrane protein component of caveolae. Furthermore, we have identified epidermal surface antigen both as a flotillin homologue and as a resident caveolar protein. Significantly, flotillin is a marker for the Triton-insoluble, buoyant membrane fraction in brain, where to date mRNA species for known caveolin gene family members have not been detected.
Within mammalian cells, proteins are segregated into distinct organellar membrane compartments, and this localization influences the function of these proteins. One such class of compartments is the set of plasmalemmal microdomains known as caveolae (1, 2). These "little caves" are 50-100 nm invaginations of the plasma membrane that have distinct morphological and biochemical properties (3, 4). Caveolae are present to some degree in most cell types, but they are particularly abundant in endothelial cells, adipocytes, fibroblasts, smooth muscle cells, and type I pneumocytes (5, 6). Coating the cytoplasmic surface of caveolae are concentric filaments composed at least in part by the family of 20-24-kDa integral membrane caveolin proteins (7). In neurons plasma membrane invaginations that exhibit caveolae-like properties have been observed by electron microscopy (8, 9), but mRNA species for known caveolin gene family members have not been detected.
Caveolae have at least three functions (6). First, in endothelial
cells, caveolae mediate the transcytosis of macromolecules from the
vascular lumen to the sub-endothelial space (10). In accord with this
transport function, caveolae contain proteins that have been implicated
in vesicular trafficking (11). Moreover, in a cell-free system caveolae
bud as vesicles from plasma membranes derived from endothelial cells in
a time- and GTP-dependent manner, and in permeabilized
cells GTP stimulates the endocytosis of cholera toxin B chain via
caveolae (12). Second, caveolae are the sites of potocytosis, whereby
small molecules are concentrated within caveolae by binding to
glycosylphosphatidylinositol (GPI)1-linked
receptors and then traverse the plasma membrane into the cytoplasm via
an unknown transporter (13). Third, caveolae may participate in the
relay of extracellular signals to the cell's interior by organizing
signal transduction molecules ("the caveolae signaling hypothesis")
(5). Caveolin-rich membrane domains purified by either detergent-based
or detergent-free methods are enriched in several distinct classes of
signaling molecules, including G and G
subunits of
heterotrimeric GTP-binding proteins, Src-like tyrosine kinases, protein
kinase C
, and small GTP-binding proteins such as H-Ras and Rap
GTPases (14-16). Some of these signaling molecules physically interact
with caveolin (15).
During their biosynthesis, caveolin-1 monomers associate to form homo-oligomers of ~350 kDa (14-16 monomers per oligomer) (17). These higher order structures of caveolin monomers may form a scaffold on which caveolin-interacting signaling molecules are organized or sequestered. These purified homo-oligomers also can undergo a second stage of oligomerization and assemble in vitro into structures that are similar in size to caveolae. However, it remains unknown whether other novel integral membrane proteins contribute to the structural organization of caveolae membranes in vivo.
To systematically identify novel protein components of caveolae, we have purified membrane domains that are significantly enriched in caveolin from murine lung tissue (18, 19). This method of purification relies upon the characteristic insolubility of caveolae domains in Triton X-100 at 4 °C and their characteristic buoyant density when subjected to sucrose density gradient centrifugation. A ~45-kDa component of these purified caveolin-rich membranes was one of 10 predominant polypeptides easily detected by Ponceau S staining (Fig. 1B in Ref. 19). Microsequencing of this ~45-kDa component has revealed several novel peptide sequences, as well as peptide sequences from epidermal surface antigen (ESA). ESA was identified and cloned as a keratinocyte cell surface protein (20). Due to the emerging importance of caveolae in vesicular trafficking and signal transduction, we proceeded to clone the cDNA corresponding to the novel caveolae protein, which we have named "flotillin."2 Surprisingly, flotillin is a close homologue of ESA, and together they define a new family of integral membrane proteins in caveolae.
Dulbecco's modified Eagle's tissue culture medium lacking methionine, cysteine, and glutamine was purchased from ICN Pharmaceuticals, Inc. The Express Protein Labeling Reagent, a mixture of [35S]methionine and -cysteine was purchased from DuPont NEN.
Cell Culture3T3-L1 mouse fibroblasts (American Type Culture Collection, Rockville, MD) were propagated and differentiated to adipocytes as described (21).
RNA Isolation and AnalysisPoly(A)+ RNA was
isolated from mouse tissues and from 3T3-L1 cells at preconfluence and
at progressive stages of differentiation as described (22), with the
exception that RNA used for construction of the adipocyte phage library
was eluted from the oligo(dT)-cellulose column in 10 mM
Tris-HCl, pH 7.4, 1 mM EDTA, without SDS.
Formaldehyde/agarose gel electrophoresis of poly(A)+ RNA,
transfer to nylon membranes, and 32P labeling of DNA were
performed as described (22). Hybridizations were carried out overnight
at 42 °C in 50% formamide, 5 × SSC, 25 mM sodium
phosphate, pH 7.0, 10 × Denhardt's solution, 5 mM EDTA, 1% SDS, and 0.1 mg/ml poly(A). The nylon membranes were washed
sequentially in 2 × SSC, 0.1% SDS, and 0.1 × SSC, 0.1% SDS at 50 °C. Autoradiography was performed at 80 °C with an intensifying screen.
Microsequencing of the ~45-kDa region from mouse lung caveolin-rich membrane domains was performed as described (19). Sense and antisense degenerate primers were designed based upon the novel peptide sequences listed in Table I and were synthesized by Research Genetics, Huntsville, AL. These sense and antisense primer pairs were used for nested PCR with mouse lung first-strand cDNA as the template. Mouse lung first-strand cDNA was prepared from 1 µg of mouse lung polyadenylated RNA, according to the manufacturer's instructions (Life Technologies, Inc.). A discrete 0.3-kilobase pair band was generated by hemi-nested PCR as follows. The first reaction used the sense primer "DLV" (GAYHTNGTNAAYATGGGNAT) and the antisense primer "AYQ" (TGRTANGCNARRTCNGCYTG) (note: where Y = C,T; H = A,C,T; N = G,A,C,T; R = G,A; D = G,A,T) with the following conditions: 1.5 mM MgCl2, 0.2 mM each dNTP, 100 pmol of each primer, and 2.5 units of Amplitaq (Perkins-Elmer) in a 50-µl reaction volume. The cycling parameters were as follows: 95 °C for 3 min, (95 °C for 1 min, 42 °C for 1 min, 72 °C for 2 min) times 5 cycles, (95 °C for 1 min, 50 °C for 1 min, 72 °C for 2 min) times 25 cycles, and 72 °C for 10 min. This first reaction did not produce a discrete band. Of this first reaction 1 µl was used as template for the hemi-nested second reaction with sense primer DLV and antisense primer "EVN" (RTTNACYTCDATRTCRTA) (see note above for explanation). The conditions of the second reaction were identical to the first, except for the amplification cycles (95 °C for 1 min, 55 °C for 1.5 min, 72 °C for 2 min times 30 cycles). The 0.3-kilobase pair PCR product was subcloned into pCR-Script (Stratagene, Inc.) according to the manufacturer's instructions, sequenced, and found to correspond to an open reading frame. Because amino acid residues encoded by DNA sequences internal to the PCR primers matched residues from the microsequenced peptides, we concluded that this cDNA corresponded to the protein originally isolated from caveolin-rich membrane domains. The identical fragment could also be amplified from 3T3-L1 adipocyte first-strand cDNA.
|
Poly(A)+ RNA (5 µg) from 3T3-L1 adipocytes
at day 8 of differentiation was used as the template to construct a
EXloxTM library according to the manufacturer's instructions
(Novagen, Inc., Madison, WI). Oligo(dT) was used to prime the
first-strand cDNA synthesis. Prior to ligation to the phage arms,
the cDNA was size-fractionated (>1.5 kilobase pairs) using a
potassium acetate gradient (5-20%) as described (23). The resulting
library represented ~1 × 106 independent clones.
The library was amplified once, and this singly amplified library was
used for screening.
The
3T3-L1 adipocyte cDNA library described above was screened with a
digoxigenin-labeled cDNA probe that corresponded to the original,
partial clone generated by PCR (nucleotide position 571-843, Fig. 1).
Screening was performed according to the manufacturer's instructions
(Boehringer Mannheim). Several positive clones were obtained, none of
which extended further 5 than base 296. To generate a cDNA probe
that would correspond to the 5
-end of flotillin, the 3T3-L1 adipocyte
cDNA library described above was used as template in nested PCR
with a vector-based primer (T7 Gene-10: TGAGGTTGTAGAAGTTCCG) and two
antisense flotillin primers (TGGTCATCATGAATATCC and
AGCATCTCCTTGTTCTGG). After removal of vector sequence by restriction digestion, the resulting PCR fragment was then digoxigenin-labeled and
used to screen the 3T3-L1 adipocyte library as above. A positive clone
was plaque-purified. The plasmid containing the positive clone,
pFlotillin-1, was excised from the phage arms via Cre-lox recombination
as described in the manufacturer's protocol (Novagen, Inc.). This
plasmid was then fully sequenced on both strands, by manual sequencing
using Sequenase (U. S. Biochemical Corp.) and by automated sequencing
using an Applied Biosystems 373 Sequencer. This clone extended to
nucleotide position 146 in the 5
-untranslated region. Because the 5
PCR-generated fragment extended farther upstream than pFlotillin-1, we
confirmed this upstream sequence by amplifying the 5
-end of flotillin
as above from an independently constructed library (random-primed,
mouse brain
gt10 library from CLONTECH). This
fragment contained an identical sequence to that amplified from the
adipocyte library, and even began with the same base. Accordingly, the
sequence represented in Fig. 1 contains sequence data from these
PCR-generated 5
-ends and pFlotillin-1.
COS-7 Cell Transfections, in Vivo Labeling, and Immunoprecipitation
cDNAs were transiently transfected into
COS-7 cells (10-cm dishes) by the DEAE-dextran method (24). 48 h
after transfection, cells were labeled for 10 min in 3 ml of
Dulbecco's modified Eagle's medium lacking methionine and cysteine
and supplemented with 0.5 mCi (1000 Ci/mmol) of Express Protein
Labeling Reagent. Thereafter, cells were washed 3 times with chase
medium (Dulbecco's modified Eagle's medium containing unlabeled
methionine and cysteine at 1 mM and cycloheximide at 300 µM). Cells were then incubated for 5, 10, 20, 30, and 60 min with 2 ml of chase medium. The cells were washed twice with cold
PBS and then scraped into TNET/OG lysis buffer (1% Triton X-100, 60 mM octyl glucoside, 5 mM EDTA, 20 mM Tris, pH 8.0, 150 mM NaCl, 1 mM
phenylmethylsulfonyl fluoride, and 1 µg/ml leupeptin). Insoluble
debris were removed by centrifugation for 10 min at 15,000 × g. The postnuclear cell lysates were then incubated for 30 min at 4 °C with Protein A-Sepharose (Pharmacia Biotech Inc.). The
Protein A-Sepharose was removed by centrifugation, and fresh Protein
A-Sepharose was added along with the corresponding antibody.
Immunoprecipitation was performed for 3 h at 4 °C.
Immunoprecipitates were then washed 5 times with lysis buffer (lacking
octyl glucoside), released from the Protein A-Sepharose by boiling in
Laemmli sample buffer, and analyzed by SDS-PAGE (25). The gel was
prepared for fluorography by incubation with 1 M sodium
salicylate for 20 min. The dried gel was exposed to Kodak X-OMAT AR
film at 80 °C.
Antibodies against a flotillin peptide were generated in rabbits by Research Genetics, Huntsville, AL. This peptide, KAQRDYELKKATYD (residues 211-224, indicated within Fig. 1), was coupled to keyhole limpet hemocyanin and injected. The resulting antisera were affinity-purified over a CNBr-activated Sepharose column (Pharmacia) to which the immune peptide had been coupled according to standard methods (26). Antibodies against a carboxyl-terminal mouse GDI-1 peptide (FENMRKKQNDVFGEADQ) were generated in rabbits by Research Genetics as described above. Anti-Myc epitope IgG (mAb 9E10) was obtained from Santa Cruz Biotech. Anti-caveolin-1 IgG (mAb 2297) and anti-ESA monoclonal antibody were obtained from Transduction Laboratories. Polyclonal anti-Acrp30 antiserum was prepared as described (27).
Extraction of Cells with Sodium CarbonateExtraction of 3T3-L1 adipocytes with sodium carbonate was performed according to a previously described protocol (28). Adipocytes were washed 3 times in PBS and once in a solution containing 150 mM NaCl. The cells were then scraped into 1 ml of 100 mM sodium bicarbonate buffer, pH 11.5, and homogenized by 5 strokes in a 2-ml Dounce homogenizer. After 30 min incubation at 4 °C, the homogenates were centrifuged at 100,000 rpm at 4 °C in a TLA 100.2 rotor (Beckman). The pellets were resuspended in 0.5 ml of 100 mM sodium bicarbonate buffer, pH 11.5, and pellet and supernatant fractions were immediately mixed with an equal volume of SDS-PAGE (Laemmli) sample buffer. Any DNA in the pellets was sheared by several passages through a 26-gauge needle to decrease the viscosity; these samples were then used directly for SDS-PAGE and Western blot analysis.
Preparation of Cell Lysates from 3T3-L1 Cells During Adipocyte DifferentiationPlates of 3T3-L1 cells at interval stages of adipogenesis were washed twice with cold PBS, lysed in TNET/OG lysis buffer, incubated on ice for 15 min, and then centrifuged for 10 min at 4 °C in a Beckman Microfuge at 14,000 rpm. Protein determinations were made on the post-nuclear supernatants using the BCA assay (Pierce).
Purification of Caveolin-rich Membrane DomainsCaveolin-rich membrane domains were purified according to an established method (29-31) with minor modifications. Briefly, 3T3-L1 adipocytes were washed 3 times with ice-cold PBS, scraped into 2 ml of Mes-buffered saline (25 mM Mes, pH 6.5, 150 mM NaCl; MBS) with 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, and 1 µg/ml leupeptin and homogenized by 10 strokes in a glass Dounce homogenizer. The extract was then adjusted to 40% sucrose by addition of 2 ml of an 80% sucrose solution in MBS, transferred to an ultracentrifuge tube, and overlaid with 8 ml of a linear 5-30% sucrose gradient in MBS containing phenylmethylsulfonyl fluoride and leupeptin at the above concentrations but lacking Triton X-100. The gradients were centrifuged for 16 h at 4 °C in a Beckman SW41 rotor at 39,000 rpm. Caveolin-enriched membranes fractionate as a sharp, light-scattering band at a density of ~15-20% sucrose (18). Equal volumes of each fraction were analyzed by SDS-PAGE and Western blotting. The pellet was brought to a volume equal to that of the fractions in Laemmli sample buffer and solubilized in a Dounce homogenizer prior to boiling.
Detergent-free Purification of Caveolin-rich Membrane DomainsCaveolin-rich membrane domains from 3T3-L1 adipocytes were purified by means of a previously reported protocol (14) with minor modifications. 3T3-L1 adipocytes were washed 3 times in PBS and then scraped into 2 ml of 500 mM sodium carbonate, pH 11.0. Cells were then homogenized on ice with (i) a Polytron tissue grinder on output setting 10 for 30 s (Kinematica GmbH, Brinkmann Instruments, Westbury, NY) and (ii) a sonicator on output setting 2 for 30 s and then setting 5 for 5 bursts (Branson Sonifier 250, Branson Ultrasonic Corp., Danbury, CT). The homogenate was then adjusted to 45% sucrose by addition of 2 ml of 90% sucrose prepared in MBS and placed at the bottom of an ultracentrifuge tube. A 5-35% discontinuous sucrose gradient that contained MBS and 250 mM sodium carbonate was formed above and centrifuged at 39,000 rpm for 16 h in an SW41 rotor (Beckman Instruments, Palo Alto, CA).
Purification of Endothelial Plasma Membranes and Caveolae Using Cationic Colloidal SilicaSilica-based purification of plasma membranes and caveolae was performed as described (32).
ImmunoblottingAfter SDS-PAGE, proteins were transferred to BA83 nitrocellulose (Schleicher & Schuell) using a Hoefer Semi-Phor apparatus. Nitrocellulose membranes were blocked in PBS or Tris-buffered saline with 0.1% Tween 20 and 5% non-fat dry milk. Primary and secondary antibodies were diluted in PBS or Tris-buffered saline with 0.1% Tween 20 and 1% bovine serum albumin. Bound antibodies were detected by enhanced chemiluminescence according to the manufacturer's instructions (DuPont NEN).
Other MethodsSeparation of proteins by SDS-PAGE (Laemmli) was performed according to standard protocols (25). Computer analysis of DNA and protein sequences was performed with the DNASTAR software package, BLAST search software (33), and MacPattern software (34). Densitometry was performed with a Bio-Rad model GS-700 Imaging Densitometer and Molecular Analyst Version 2.1.1 software (Bio-Rad).
Microsequence analysis of a ~45-kDa component of
purified caveolin-rich membrane domains revealed several novel peptide
sequences (Table I). To clone the cDNA for this
novel protein, we generated a probe by reverse transcription-PCR with
degenerate primers that were designed in accordance with the
microsequence data. With the resulting probe we then screened an
appropriate full-length cDNA library. A ~1.7-kilobase clone
was isolated that contained an open reading frame that predicted a
polypeptide of 428 amino acids with a molecular mass of 47 kDa. The
most 5-encoded methionine is indicated in Fig. 1 as the
initiator methionine. However, the next methionine at amino acid
position 11 has a more classic Kozak consensus sequence (G at position
+4 and A or G at position
3) (35). We have not yet defined
experimentally which of these two ATG triplets serves as the start
codon for the flotillin polypeptide.
To show that the isolated cDNA clone could encode a protein, a PCR-generated construct starting with the latter methionine and tagged at the carboxyl terminus with the Myc epitope (pMVVmyc) was expressed in COS-7 cells. In a metabolic pulse-chase experiment with these transfected cells, a ~50-kDa protein was immunoprecipitated with 9E10, a monoclonal antibody that recognizes the Myc epitope (data not shown).
Structural and Motif Analysis of Flotillin: Homology with ESA and Two ORFs in the Cyanobacterium SynechococcusThe cDNA for
flotillin encodes a protein of 428 amino acids with a predicted
molecular mass of 47 kDa, which closely matches the size of the
~45-kDa protein that was microsequenced from caveolin-rich membrane
domains. Flotillin contains two hydrophobic domains of 27 and 18 amino
acids, respectively (boxed regions in Fig. 1). The first of
these domains (from position 10 to 36) may represent an atypical signal
peptide or perhaps a transmembrane domain. The second hydrophobic
region (from position 134 to 151) is another potential transmembrane
domain. Computer-assisted motif analysis (34) reveals two potential
tyrosine phosphorylation sites at positions 160 and 292. In addition,
flotillin contains predicted sites for phosphorylation by protein
kinase C (at positions 52, 150, 229, and 315) and protein kinase A (at
position 222). However, there are no predicted N-linked
glycosylation sites. The region of flotillin from 328 to 355 is
predicted to form an helix that may form a triple coiled coil with
other flotillin monomers.3
Beginning at position 47, flotillin demonstrates significant homology
with ESA, for which both human and mouse cDNA clones have been
reported (20, 36) (Fig. 2A). The cDNA for
ESA was cloned from a human foreskin keratinocyte gt11 cDNA
library that was screened with a monoclonal antibody. This monoclonal
antibody, ECS-1, had been raised against cultured human keratinocytes
and was found to stain nucleated epidermal cells in an intercellular pattern. Moreover, the addition of ECS-1 to cultured mouse
keratinocytes led to cell detachment, and this detachment was enhanced
by complement. Based upon these data, ESA has been implicated in
epidermal cell adhesion. Over the region of homology, there is ~47%
identity between flotillin and ESA at the amino acid level. Of the
potential phosphorylation sites in flotillin, only the protein kinase C phosphorylation site at residue 150 and the tyrosine phosphorylation site at residue 152 are conserved in ESA. Flotillin also shares with
ESA a repeat of A/G, E, A/G, E that recurs 6 times in the carboxyl half
of the polypeptide (Fig. 2A, shaded boxes). The significance
of this repeat remains to be determined. The two potential
membrane-spanning domains indicated in the flotillin sequence are not
conserved in ESA. The first such domain (flotillin residues 10-36) is
missing entirely from ESA because the initiator methionine for ESA
corresponds to flotillin residue 46. The second such domain (flotillin
residues 134-151 and ESA residues 89-106) contains two additional
charged amino acids in ESA that make this region unlikely to be
membrane spanning.
Homology searches also reveal a surprising similarity between flotillin and two conceptual proteins from the cyanobacterium Synechococcus PCC7942 (Fig. 2B). The homology extends for nearly the entire lengths of flotillin and the ORF2 protein. Two A/G, E, A/G, E motifs are partially conserved between flotillin and the Synechococcus ORFs (shaded boxes). 24% of the residues in ORF2 protein are identical to their corresponding residues in flotillin. In the region of highest homology (residues 84-240 of flotillin) the amino acid sequence of flotillin is 30% identical to that of ORF2 and 58% identical or conservatively substituted. However, no such homologies are detected between flotillin and the 6-frame translation of the completely sequenced yeast genome.
Tissue-specific Expression of Flotillin Messenger RNABecause
flotillin was initially identified by its co-purification with
caveolin-1, flotillin mRNA should be expressed in tissues that also
express caveolins. Northern blot analysis reveals that flotillin is
expressed at the mRNA level in white adipose tissue, heart muscle,
skeletal muscle (diaphragm), and lung (Fig. 3). As
predicted, these results are similar to the expression pattern of
caveolin-1 (37); however, flotillin is expressed at higher levels in
heart relative to lung than is caveolin-1. Longer exposures of the
autoradiogram also show low levels of expression in spleen, liver, and
testis (data not shown). Of particular note is that flotillin mRNA
is easily detectable in brain. In contrast, there is no detectable
mRNA for any known caveolin family member expressed in brain
(37-39). ESA mRNA has been detected in tissues that contain epithelia, such as pancreas, stomach, and lung, in cultured
keratinocytes and melanocytes, and at low levels in heart (20).
Characterization of the Flotillin and ESA Proteins
To
characterize flotillin at the protein level, we generated an
anti-peptide polyclonal antiserum in rabbits. The epitope within
flotillin was chosen based upon its predicted antigenicity, surface
probability, and the specificity of its sequence. Affinity-purified immune antibodies recognize a 47-kDa protein in Western blots of
caveolin-rich membrane domains isolated from human lung (data not
shown) and from 3T3-L1 adipocytes (Fig. 4). Preimmune
sera do not recognize this band. Furthermore, the immune antibodies do
not recognize the 47-kDa protein in the presence of the peptide that
was used for immunization but do recognize this protein in the presence
of irrelevant peptide. We have not yet found conditions under which our
polyclonal flotillin antibody is able either to immunoprecipitate
flotillin or to demonstrate specific, reproducible staining by
immunofluorescence.
Based upon the two potential transmembrane domains in the predicted
primary amino acid sequence (Fig. 1), we predicted that flotillin would
associate tightly with membranes. Peripheral membrane proteins
dissociate from membranes at high pH, but integral membrane proteins
remain attached (28). When 3T3-L1 adipocytes were lysed at high pH in
carbonate buffer and then subjected to centrifugation, immunoreactive
flotillin was present in the membrane pellet, as was another integral
membrane protein, caveolin-1 (Fig. 5). GDI-1 (a
guanine-nucleotide dissociation inhibitor for Rab family proteins) is a
soluble marker and was present predominantly in the supernatant. Interestingly, despite the absence of a predicted transmembrane domain
in ESA, immunoreactive ESA was detected exclusively in the pellet with
flotillin and caveolin-1. These data collectively indicate that
flotillin and ESA behave as integral membrane proteins.
We have shown previously that caveolin-1 mRNA and protein levels,
as well as the number of caveolae per cell, increase during the
differentiation of 3T3-L1 fibroblasts into adipocytes (37). At the
mRNA level, induction of both caveolin-1 and caveolin-2 occurs
between Day 2 and Day 4 of differentiation (37, 38). In contrast, the
pattern of induction of flotillin mRNA is strikingly different
(Fig. 6A). By Day 0, at which time the 3T3-L1
cells have been confluent for 2 days, flotillin mRNA expression
already has increased to near-maximal levels. Flotillin mRNA levels
then fall to a nadir by Day 2, only to increase again to maximal levels by Day 4. RNA levels for flotillin then remain relatively constant through Day 8. We have noted a similar decrease in mRNA levels for
2-microglobulin on Day 2 of 3T3-L1 adipocyte
differentiation,4 which may reflect a
transcriptional response to one or more components of the media used
for differentiation from Day 0 to Day 2.
At the protein level, flotillin more closely follows the induction pattern of caveolin-1 (filled arrowhead, Fig. 6B). Between Day 2 and Day 4 flotillin protein levels per mg of total protein in post-nuclear cell lysates increase by ~3-fold. There is a further ~2-fold increase from Day 4 to Day 6 and another 2-fold increase from Day 6 to Day 10. A parallel induction pattern is observed for a ~27-kDa immunoreactive protein. We do not know whether this is a cross-reacting protein, a degradation fragment of flotillin, or a physiologically cleaved fragment of flotillin. Hence, we have termed this ~27-kDa protein FCRD (flotillin cross-reacting determinant). There is another protein of ~45 kDa (open arrowhead, Fig. 6B) that is recognized by the polyclonal flotillin antibody on a Western blot of 3T3-L1 adipocytes during differentiation. In contrast to flotillin and FCRD this ~45-kDa protein reaches peak levels on Day 4. Because of its different pattern of expression during 3T3-L1 adipocyte differentiation, the ~45-kDa protein is likely a cross-reacting protein that is recognized by the anti-flotillin polyclonal antiserum but is unrelated to either flotillin or ESA.
Whereas flotillin protein is induced during 3T3-L1 adipocyte differentiation, ESA protein remains at constant levels per mg of protein in post-nuclear cell lysates (Fig. 6B).
Flotillin Is an Integral Membrane Component of Caveolae MembranesTo evaluate the extent of specific localization of flotillin and ESA in caveolae, several complementary methods were used to purify caveolae membranes from 3T3-L1 adipocytes and endothelial cells. In all three methods, immunoblotting with anti-caveolin-1 IgG was used to track the position of caveolae-derived membrane domains.
First, proteins from 3T3-L1 adipocytes were fractionated using an established protocol that separates caveolin-1 and other caveolae-associated proteins from the bulk of cellular proteins and membranes. This procedure is based upon (i) the insolubility of caveolae membranes in the detergent Triton X-100 at 4 °C and (ii) their specific buoyant density in equilibrium sucrose gradients. With this fractionation method, we have previously shown that endogenous caveolin-1 (fractions 5-6) is purified ~2000-fold relative to total cell lysates. In addition, caveolin-1 is separated from greater than 99.95% total cellular proteins and compartment-specific markers for endoplasmic reticulum, Golgi, lysosomes, mitochondria, and non-caveolar plasma membrane, which remain in fractions 9-13 of these bottom-loaded sucrose density gradients (19, 30, 31).
Fractions from these gradients were harvested from the top, and the
proteins of each fraction were resolved by SDS-PAGE, transferred to
nitrocellulose, and blotted with the flotillin antibody and monoclonal
antibodies against ESA and caveolin-1. The top panel of Fig.
7A shows the distribution of total cellular
protein across the gradient by Ponceau S staining. Immunoreactive
flotillin, ESA, and caveolin-1 co-fractionate and are specifically
concentrated within fraction 5, which corresponds to caveolin-rich
membrane domains that can be visualized as a light-scattering band in
the gradient (Fig. 7, A and B). Significantly,
flotillin, ESA, and caveolin-1 are clearly separated from the bulk of
cellular proteins, which remain within fractions 9-12 and the pellet.
FCRD remains in fractions 10, 11, and 12 and does not co-fractionate
with caveolin-1.
Next, we also used a detergent-free protocol to purify caveolin-rich membrane domains as we have reported previously (14). This protocol substitutes the nonionic detergent Triton X-100 with sodium carbonate to disrupt cellular membranes. Using this method, caveolin is highly enriched in a buoyant, light-scattering band just as with the Triton-based method, but in contrast to the Triton-based method, the glycosylphosphatidylinositol (GPI)-anchored protein carbonic anhydrase IV is detected in less buoyant fractions. We applied this carbonate-based method to the fractionation of 3T3-L1 adipocytes, and flotillin, ESA, and caveolin-1 again co-purified (Fig. 7C) in the buoyant, light-scattering fraction that corresponds to caveolae-enriched membranes.
Finally, we used a third independent method for the purification of
caveolae in a different cell type to confirm the membrane compartment
in which flotillin and ESA reside. In this method, the rat lung
vascular system is perfused with cationic colloidal silica particles
(32). These particles adhere to the luminal plasma membranes of
endothelial cells and increase their density such that they can be
highly purified by homogenization and centrifugation. The resulting
membrane pellet retains caveolae on the side of the membrane opposite
the silica. These caveolae are then sheared off of the plasma
membrane-silica complex (P) by homogenization at 4 °C in the
presence of the detergent Triton X-100. The caveolae (V) are then
purified away from the stripped plasma membrane-silica complex (P V)
by sucrose gradient centrifugation. With this method >95% of the
caveolin signal is present in caveolae, whereas >95% of the signals
for the GPI-anchored proteins 5
-nucleotidase and urokinase-plasminogen
activator receptor are present in P
V. Thus, this method can be used
to distinguish between two Triton-insoluble membrane domains as
follows: (i) caveolae, which are enriched in the marker protein
caveolin, and (ii) plasma membrane domains that are enriched in
GPI-anchored proteins. Western blot analysis of these three fractions
(P, V, and P
V) demonstrated that flotillin, ESA, and caveolin-1 are
highly enriched in V, the purified caveolae (Fig. 8,
A and B). Only at nonlinear
exposures of V was any signal for flotillin also detectable in P
V
(middle panel of Fig. 8A). Notably, FCRD is
detected only in the P
V fraction and is excluded from the V
fraction.
Taken together, these data indicate that flotillin and its homologue ESA are resident integral membrane protein components of caveolae.
Flotillin Is a Marker for the Buoyant, Triton-Insoluble Membrane Fraction in BrainAs noted above, flotillin mRNA is easily detectable in mouse brain by Northern blot analysis (Fig. 3), whereas mRNA species for known caveolin family members are undetectable. This observation led to the question of whether flotillin is expressed at the protein level in brain and, if so, whether it sediments in the buoyant, Triton-insoluble membrane fractions. To address these issues, bovine brain tissue (cortical gray matter) was subjected to the same Triton-based purification protocol that we used for 3T3-L1 adipocytes. Indeed, just as is true in adipocytes, flotillin is present in the buoyant, Triton-insoluble fractions in bovine brain tissue (Fig. 7D). Current data do not specify in what cell type or types flotillin is expressed in bovine brain. However, flotillin is detectable in the buoyant, Triton-insoluble membrane fractions of PC12 cells (data not shown), which are derived from a rat pheochromocytoma (a neuroendocrine tumor). This latter finding is suggestive, although not conclusive, that neurons express the flotillin protein.
In this report we have described a new gene family of caveolae-associated integral membrane proteins that so far includes flotillin and ESA. Multiple lines of evidence support the conclusion that flotillin and ESA are resident components of caveolae. First, peptides that are contained within the predicted amino acid sequences of flotillin and ESA have been identified by microsequence analysis of caveolin-rich membrane domains purified from mouse lung tissue. Second, an affinity-purified, anti-peptide antibody raised against an epitope of flotillin recognizes a 47-kDa protein that co-purifies with caveolin-1 in 3T3-L1 adipocytes using either detergent-based or detergent-free purification protocols. Likewise, ESA co-purifies with caveolin-1 in the same assays. Third, caveolae vesicles that have been purified from rat lung endothelial cell plasma membranes following in situ coating with colloidal silica contain almost all of the plasma membrane flotillin and ESA proteins.
Caveolin and flotillin have similar tissue distributions at the mRNA level. However, one important exception to this observation is brain, where flotillin is easily detectable but caveolin family members are absent. We have demonstrated that flotillin protein is expressed in brain and that it fractionates in the Triton-insoluble, buoyant membrane fractions. Several recent reports suggest that caveolae or caveolae-like structures are present in neural tissues (8, 9, 40). Thus, flotillin may represent a new marker for these structures in brain. Whether flotillin-rich membrane domains in neural tissues are functionally related to caveolin-rich membrane domains in adipocytes and other caveolin-expressing tissues will be addressed in future experiments.
The anti-peptide polyclonal antisera we have raised against flotillin
also recognizes a polypeptide of approximately 27 kDa that we have
termed FCRD. Current data do not permit elucidation of the exact
origins of FCRD. One possibility is that FCRD is a degradation product
of flotillin that is produced during sample preparation. Another
possibility is that it is a physiologically relevant proteolytic
fragment of flotillin. Finally, FCRD may be an intact flotillin
homologue or a completely unrelated protein that simply shares a
conserved epitope. However, our current data highlight several
important features of FCRD. First, FCRD is present in the plasma
membrane fractions of rat lung endothelial cells (P and P V), but it
is absent from the caveolar fraction (V). Second, FCRD is soluble in
Triton X-100 in adipocytes and, accordingly, does not co-purify with
caveolin-1. Finally, FCRD polypeptide levels increase during 3T3-L1
adipocyte differentiation in parallel to those of flotillin. These data
are all consistent with a model in which caveolar flotillin is
proteolytically cleaved into FCRD and other fragment(s), with the
result that FCRD leaves the caveola. If this model is correct, then the
full description of FCRD at the amino acid level may lead to the
identification of motifs important for caveolar localization and
retention.
The close homology between flotillin and ESA and their co-purification with caveolin-1 in 3T3-L1 adipocytes suggest that flotillin and ESA are closely related members of the same gene family. In light of these findings ESA now seems to be an inappropriate appellation because ESA is expressed in cells as divergent as keratinocytes, adipocytes, and endothelial cells. To emphasize that ESA and flotillin constitute a newly recognized gene family, we propose that flotillin henceforth be known as flotillin-1 and ESA be known as flotillin-2. Whether additional members of the flotillin gene family exist is unknown.
A surprising result of these investigations is the identification of two hypothetical proteins of the cyanobacterium Synechococcus that share remarkable homology to flotillin. These homologies point to a very ancient and fundamental function for the flotillin gene family. Although Triton X-100 insoluble, buoyant membrane domains have been described in the yeast Saccharomyces cerevisiae (41), no convincing flotillin homologues have been identified in the S. cerevisiae Genome Data Base.
An important difference between flotillin and ESA is their patterns of protein expression during 3T3-L1 adipocyte differentiation. Whereas the ESA protein is expressed at constant levels as 3T3-L1 cells differentiate into adipocytes, the flotillin protein is induced at least 10-fold. Caveolin-1 mRNA and protein levels and caveolin-2 mRNA levels, as well as morphologically distinct caveolae, also increase significantly during adipogenesis (37, 38). The invariant levels of ESA protein during 3T3-L1 adipocyte differentiation suggest two possibilities as follows: (i) ESA may reside in a non-caveolar pool in preadipocytes and be recruited to caveolae during differentiation, or (ii) the concentration of ESA per caveola may decrease during differentiation.
Of note is the discordance between flotillin mRNA levels and protein levels during 3T3-L1 adipocyte differentiation as detected by Northern and Western blotting, respectively. Flotillin mRNA levels are roughly equal on days 0 and 8 of differentiation, but flotillin protein levels increase by an order of magnitude over the same period. Some other process during 3T3-L1 adipocyte differentiation has the effect of increasing the efficiency of flotillin mRNA translation or the stability of the translated flotillin protein or both. One attractive model is that flotillin protein is stabilized by the increased expression of caveolin protein during 3T3-L1 adipocyte differentiation, and/or by the assembly of caveolae, the number of which also increase during this differentiation program.
What is the function or functions of the flotillin gene family? Their
localization to caveolae suggests some possibilities. First, like
caveolin, flotillin and ESA may contribute to the structure of
caveolae. One model for caveolin function holds that caveolin monomers
assemble into homo-oligomers that pack side-by-side to form a scaffold
onto which signaling molecules can attach (14, 17). The caveolin
scaffold would then provide a mechanism for organizing these signaling
molecules and sequestering them in a given functional state. Flotillin
and ESA may be structural components of this scaffold or may form the
backbone of a distinct part of the caveolae architecture,
e.g. the caveolar neck. Another possible function for
flotillin and ESA is that they participate in signaling, either as
receptors or as membrane-bound effectors. Flotillin and ESA do share a
conserved tyrosine phosphorylation site, but whether they are indeed
phosphorylated at this site remains to be determined experimentally.
Further investigations into the membrane topology of these proteins
will be important to determine whether they are a potentially novel
class of receptors. ESA, in particular, was cloned on the basis of its
being a keratinocyte cell surface antigen. Thus far, the only proposed
function for ESA is a role in cell adhesion. This hypothesis is based
upon the ability of the ECS-1 monoclonal antibody to cause cell
detachment when it is added to intact cultured keratinocytes. That
flotillin also plays a role in cell adhesion is an intriguing
possibility. Other proteins known to mediate or modulate cell-matrix
interactions have been shown to localize in or near caveolae
(urokinase-plasminogen-activator receptor) (42) or to interact with
caveolin directly (1 integrins) (43). Thus, there is
precedence for the involvement of caveolae in cell adhesion.
Whatever the functions of flotillin and ESA, it will be interesting to determine why two such similar proteins are present in the same organelle. One possibility is that flotillin and ESA are functionally redundant. Alternatively, flotillin and ESA might perform similar functions in different subcellular locations. They could define, for instance, two distinct subpopulations of caveolae. Finally, flotillin and ESA may cooperate to execute a coordinated function.
In conclusion, (i) the molecular cloning of flotillin-1 and (ii) the identification of flotillin-1 and flotillin-2 (ESA) as components of caveolae provide important new avenues by which to explore the structure and function of caveolae organelles.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) U90435[GenBank].
We thank Dr. Stefan Constantinescu and Ralph Lin for critical review of the manuscript, Dr. Andreas Stahl and Ethan Wolf for useful discussions, and Dr. Giulia Baldini for assistance in Northern blot analysis.