Department of Biological Sciences, University of Maryland Baltimore County, Catonsville, MD 21228, USA
* Present address: Department of Biological Chemistry, Johns Hopkins University School of Medicine, 725 N. Wolfe Street, Baltimore, MD21205, USA
Summary
The iron-storage protein ferritin was found to beassociated with highly purified coated vesicles (CV)from chicken liver. Chicken liver fer.ritin was mor-phologically similar to ferritin from horse spleenand could be isolated using a specific anti-ferritinmonoclonal antibody. This antibody recognized a240xl03Mr form of chicken ferritin but not the22xlO3Mr ferritin subunit after protein transfer tonitrocellulose. CV purified by controlled-poreglass-bead chromatography also contained ferritinwhen assayed by monoclonal anti-ferritin antibodyusing a sensitive enzyme-linked assay. Ferritinremained associated with CV even after re-chroma-tography. Ferritin particles were observed to be
associated with CV by electron microscopy. CV-associated ferritin could be quantitatively removedfrom CV by treatment of the CV with 0-5M-Tris-HCl + 2M-urea at pH8-5, conditions that also leadto dissociation of the clathrin lattice. Triton X-100detergent treatment did not affect the association offerritin with CV. These results indicate that puri-fied CV from chicken liver contain ferritin inassociation with the clathrin lattice. The possiblefunctional significance of this association is dis-cussed.
Coated vesicles (CV) are transient intracellular inter-mediates in the receptor-mediated endocytosis of a var-iety of nutrients and proteins such as low densitylipoprotein (LDL) (Mello et al. 1980), iron-transferrin(Booth & Wilson, 1980; Omary & Trowbridge, 1981) andasialoorosomucoid (Wall et al. 1980) or growth factorssuch as epidermal growth factor (EGF) (Ushiro &Cohen, 1980; Dunn et al. 1986) and insulin (Czech,1982; Kasuga <?<«/. 1982; Petruzzelli et al. 1982; Pilch etal. 1983). In addition, the biosynthesis of endogenous orviral-encoded proteins may depend on CV for the segre-gation and exocytosis of membrane-bound proteins(Campbell et al. 1983; Rothman et al. 1980), althoughproteins constitutively transported to the cell surface maybe transported in non-clathrin-coated vesicles (Orci et al.1986). Therefore, CV appear to be important mediatorsof both coated pit-dependent endocytotic events andperhaps sorting of ligands and receptors (Brown et al.1983).
CV are identified morphologically by a distinctive coat(Roth & Porter, 1964), which appears to be shed withinminutes of the invagination of a coated pit (Brown et al.1983) or budding of a trans-Golgi vesicle (Balch et al.1984; Rothman, 1981; Rothman et al. 1980; Orci et al.1986). This coat surrounds a lipid bilayer that is derived
from the particular cellular membrane that buds into thecytoplasm to form the coated vesicle. One of the majorproteins comprising the coat is the 180K (K = 103/V/r)clathrin molecule, which exists as a trimer in associationwith three light chains of molecular weight 32-36K(Brodsky et al. 1983; Crowther & Pearse, 1981; Kirchau-sen & Harrison, 1981; Ungewickell & Branton, 1981).This basic unit is the triskelion, which can dissociate andre-associate in vitro (Keen et al. 1979; Woodward &Roth, 1978) and is responsible for the characteristichexagon/ pentagon lattice on the surface of the mem-brane vesicle (Crowther & Pearse, 1981; Kirchausen &Harrison, 1981; Ungewickell & Branton, 1981). Of themany remaining peripheral membrane proteins that com-prise the CV, the 100K and 50K polypeptides arebelieved to act as the triskelion nucleation site on themembrane vesicle (Unanue et al. 1981; Pearse & Robin-son, 1984; Robinson & Pearse, 1986).
In addition to the known CV structural proteins manyother cellular proteins have been found to associate withpurified CV. For example, integral membrane proteinreceptors such as the transferrin receptor (Omary &Trowbridge, 1981), the mannose phosphate receptor(Sahagian et al. 1981; Steiner & Rome, 1982), the LDLreceptor (Schneider et al. 1982), and the insulin receptor(Pilch et al. 1983) have been identified as CV com-ponents. Several molecules believed to be involved in the
187
catalysis of clathrin coat assembly and disassembly havebeen isolated in association with coated vesicles. Theseinclude calmodulin (Linden et al. 1981), Ca2+-ATPase(Blitz et al. 1977), Ca2+-independent protein kinase(Pauloin et al. 1982), uncoating ATPase (Schlossman etal. 1984) and assembly polypeptides (Zaremba & Keen,1983). A proton ATPase that may be responsible for theacidification of CV has also been described (Stone et al.1983; Forgac <?/«/. 1983).
Ferritin is the major iron-storage protein in most cells(Theil, 1987). In liver it is probably involved in theexchange of iron between the cell and transferrin in theblood (Osaki et al. 1971; Trump & Berezesky, 1977).Under conditions where the liver contains adequatereserves of ferric iron, which must be mobilized to meetthe needs of other cells, ferritin loses iron to the blood-stream, perhaps at the plasma membrane through themediation of the serum copper enzyme, ferroxidase I(Osaki et al. 1971). Conversely, if iron must be stored inthe liver or taken up by other cells, the iron carriertransferrin binds a specific receptor, which is endocy-tosed in coated vesicles (Van Renswoude et al. 1982;Dantry-Varsat et al. 1983) and iron is released aftervesicle fusion with an acidic endosomal compartment(Mellman et al. 1986). Although the ultimate fate ofendocytosed iron is storage in ferritin, the mechanismthat sequesters iron after release from transferrin is notknown. In this study we wished to investigate thepossible direct association of ferritin, which is a cytoplas-mic protein, with coated vesicles.
We and others (Kelly et al. 1983; Wiedenmann &Mimms, 1983; Pfeffer et al. 1983) have shown thatanother group of cytoplasmic proteins, cvand /Jtubulin,are components of highly purified coated vesicles frombovine brain. In order to minimize contamination of ourCV with cellular tubulin, we purified CV using gelpermeation chromatography (Pfeffer & Kelly, 1981).Using these techniques to isolate highly purified CV fromliver and a sensitive enzyme-linked immunoassay foranalysis of CV-bound proteins, we report that endogen-ous ferritin, the major iron-storage protein of the cell, isfound associated with coated vesicles from chicken liver.
Materials and methods
MaterialsHorse spleen ferritin was purchased from Miles Laboratories.Controlled-pore glass beads were obtained from Electro-Nucleonics, Inc. CNBr-activated Sepharose 4B was fromPharmacia Fine Chemicals, Ultrapure urea was from Schwarz-Mann. Acrylamide, SDS, and other electrophoresis reagentswere from Bio-Rad Laboratories. All other chemicals werereagent grade.
Purification of coated vesicles by gel-permeationchromatographyRooster livers, chicken brain, or bovine brain obtained from alocal slaughterhouse, were minced on ice with single-edgedrazor blades. The volumes of buffer given are for startingmaterial consisting of 300g of tissue. When other amounts oftissue were used the volumes were adjusted proportionately.
Tissue was added to a Waring blendor and homogenized bythree 30-s bursts in 800 ml total volume of 0-1 M-Mes at pH6-S,lmM-EGTA, lmM-MgCl2, and 0-02% NaN3 (IB). Aftercentrifuging this homogenate at 20000g'for SOmin, the pelletswere discarded and the supernatants (chicken liver extract)were centrifuged at 100000gfor 1 h. The resultant microsomalpellets were resuspended in 300ml of isolation buffer (IB;0-1 M-Mes, pH6-5, 1 mM-MgCl2, 0-02% NaN3) and processedaccording to the method of Nandi et al. (1982). Briefly,microsomes were centrifuged at 20 000 g1 for 20min to removeaggregates, after which the supernatants were centrifuged at100 000g for 1 h. This low-speed/high-speed centrifugationwas repeated with the resuspended pellets. The high-speedsupernatants were combined and used for purification of ferritin(see below). The final pellets were resuspended in 18 ml of IB,divided into three equal parts and layered onto three BeckmanTi60 centrifuge tubes containing 8 % sucrose in IB prepared inD2O. After centrifugation at 80 000 g for 2h, the pelletscontaining ferritin, coated vesicles and uncoated membranevesicles were resuspended in IB containing 10% sucrose andcentrifuged at 20000^for 20min to remove aggregates. Part ofthe resulting supernatant (impure CV) was applied to a1-5 cm X 100 cm column containing controlled-pore glass beadsof 1902 A average diameter (Pfeffer et al. 1983; Kelly et al.1983). The column was equilibrated and eluted with columnbuffer (CB): 10mM-Hepes, pH7-0, lOmM-EGTA, 0-2M-SU-crose, 0-3M-NaCl, 0 0 2 % NaN3. After several experiments,glass beads were re-generated with 10M-urea at 37°C for 5 minand washed five times with distilled water. The beads were thenretreated with 1 % PEG 20 000 that had been filtered at 24 °Cthrough successive membrane filters of 5-0^m, 1-2fun and0-45 ftm. About 500 ml of PEG was added to 150 ml of gel andstirred by gentle shaking for 20 min. After washing in distilledwater, five times, and degassing, the beads were washed withCB and packed into a glass column by vortexing.
AntibodiesMonoclonal antibody recognizing chicken ferritin was preparedin collaboration with Dr D. Fambrough, using chicken livercoated vesicle preparations as the immunogen, following themethod of Fambrough & Bayne (1983). Anti-ferritin IgG waspurified from the ascites fluid using DEAE chromatography.Rabbit anti-mouse IgG coupled to horseradish peroxidase(RAM-HRP) was obtained from Sigma Chemical Company.
Immunoprecipitation of chicken fenitinMonoclonal anti-ferritin antibody IgG was coupled to CNBr-activated Sepharose 4B using established procedures (Van Eijk& van Noort, 1976). Washed beads (3-5 ml wet vol.) wereresuspended to 14 ml total volume in phosphate-buffered saline(PBS) containing 0 0 2 % NaN3. Three sources of antigen fromthe CV purification were used: (1) microsomal pellets, (2) thehigh-speed supernatant fractions, and (3) impure CV. Themicrosomal pellets were solubilized prior to use by incubatingthe resuspended pellets in 2% TX-100 (Triton X-100) in IB.After incubation at 24°C for 20 min, the mixture was centri-fuged at 20 000 £ for 20 min. One ml of this detergent-solublemicrosomal supernatant, one ml of impure CV or 15 ml of thehigh-speed supernatant fraction, was incubated with 0-2 ml ofIgG-coupled Sepharose at 4°C overnight. The Sepharose beadswere collected by centrifugation on a clinical IEC desk-topcentrifuge for 5 min at a speed setting of 4. The beads were thenwashed sequentially with 15 ml of the following buffers:(1) three washes with 50 mM-Tris-HCl, pH7-5, 0-5% TX-100,1 mM-EDTA; (2) 10mM-Tris-HCl, pH7-5, 0-1 % SDS, 0-1 %TX-100, 0-3 M-NaCl; (3) 50 mM-Tris-HCl, pH 7-5, 0-5 % TX-100, 1 mM-EDTA, 1 M-NaCl; (4) 1 % TX-100 in distilled water.
188 A. Passaniti and T. F. Roth
Antigen was eluted with 0-2 ml of 0-1 M-triethylamine, pH 11-0,with 0-lM-glycine, pH2-0, or with SDS-PAGE buffer (seebelow).
Purification of chicken ferritin by antibody affinitychromatographyAnti-chicken ferritin monoclonal antibody IgG coupled toSepharose was used to prepare an affinity column by packing aplastic 5 ml pipette with 3-Sml of beads containing 2-0 mg ofbound IgG. The column was equilibrated with IB at pH 6-5 at aflow rate of 5 ml h~ . Chicken liver high-speed supernatantfractions (see above) or TX-100-soluble microsomal mem-branes were applied to the affinity column at a flow rate of5 ml h~'. After application, the column was washed with 20 bedvols of IB followed by IB containing 0-1 % TX-100 and 0-5 M-NaCl. Bound protein was eluted with 10 ml of freshly prepared0-1 M-triethylamine, pH 11-0, at a flow rate of lOmlh"1.Samples (0-5 ml) were collected in tubes containing 0-5 ml 1 M-sodium phosphate to neutralize the triethylamine. The columnwas re-equilibrated with IB at pH6-S and kept at 4°C forfurther use. Purified ferritin was concentrated and desalted onan Amicon ultrafiltration apparatus equipped with a PM-10filter using at least three cycles of concentration and dilutionwith low-salt buffer (IB).
Enzyme-linked immunoassays (ELISA)Chicken liver ferritin was detected in fractions from the gel-permeation chromatographic procedure by a direct-bindingELISA (Kelly et al. 1983) using purified anti-ferritin mono-clonal antibody IgG. Rabbit anti-mouse IgG conjugated toHRP was used as second antibody and the absorbance of theHRP reaction product was measured at 490 nm using a Dyna-tech ELISA plate reader.
A competitive ELISA was used to quantify the amount ofantigen in CV fractions or tissue extracts essentially as described(Kelly et al. 1983).
Polyacrylamide gel electrophoresis (PAGE)SDS-PAGE 3-75 % to 15 % acrylamide slab gels were preparedaccording to the method of Laemmli (1970). The slab gels werestained with either 0 1 % Coomassie Brilliant Blue in 40%methanol:acetic acid: water (40: 10:50, by vol.) and destainedin 10% acetic acid, or with silver using the procedure of Merrilel al. (1981).
Electrophoretic transfer of protein to nitrocellulose paperFerritin was purified by immunoaffinity chromatography andresolved on 3-75 % to 15 % acrylamide gels (SDS-PAGE) afterheating the sample at 100°C for 5 min or at 24°C for 5 min. Gelswere transferred to nitrocellulose according to the method ofTowbin et al. (1979). Second antibody coupled to HRP wasused to detect specific binding of primary antibody.
Electron microscopyCopper grids (400-mesh) were coated with Formvar dissolvedin ethylene dichloride and air dried. Carbon with vacuum-evaporated onto the Formvar-coated grids and a drop of sampleapplied. Grids of samples containing sucrose from the glass-bead chromatographic step were rinsed with deionized water.Uranyl acetate at 2% was freshly prepared, centrifuged, andapplied to the sample on the grids. Samples were viewed andphotographed on a Hitachi-600 electron microscope at either50kVor75kV.
Routine methodsProtein was quantified by the method of Bradford (1976). An
LKB model 4050 spectrophotometer was used for monitoringabsorbance at 595 nm for the protein assays and absorbance at280nm for gel permeation fractions. ELISA microtitre plateswere placed in a microELISA minireader from DynatechCorporation (MR590) equipped with a 490 nm filter to monitorthe absorbance of HRP-generated product in the microtitrewells. A BCD-HP41C interface was used to record data onto anHP peripheral printer.
Results
Identification of ferritin in coated vesicle preparationsfrom chicken liverCoated vesicles isolated from chicken liver by the low-sucrose procedure of Nandi et al. (1982) have most of thetypical CV proteins seen in bovine brain CV (Fig. IB,lanes a,b). Clathrin at 180K was the predominant pro-tein. Additional proteins at 100K to 110K and 32K to36K were also evident. The tubulin polypeptides at 53Kand 56K were present in chicken liver CV, although inreduced amounts compared to bovine brain CV. Incontrast, CV from chicken liver contained a major bandat molecular weight 22K that is barely detectable in CVfrom bovine brain. When chicken liver CV were furtherpurified by gel-permeation chromatography using con-trolled-pore glass beads, membrane vesicles (Fig. 1A,C,peak I) were separated from coated vesicles (Fig. 1A,D,peak II). In addition, macromolecules smaller than CVwere separated from CV (Fig. 1A, peak III). Clathrin,the major CV polypeptide, eluted predominantly in peakII (Fig. IB) confirming that these fractions contained themajority of coated structures. By contrast, the elutedfractions from peak III contained the 22K polypeptide asthe major protein.
Since the 22K polypeptide migrated with a molecularweight similar to ferritin subunits from horse spleen(Stefanini et al. 1982), we wished to confirm whether the22K polypeptide was ferritin and whether it was alsopresent in the coated vesicle fractions. Using a mono-clonal antibody that recognized ferritin (Figs 2, 3), wefound that it bound to immobilized CV on microtitreplates in a competitive ELISA (Fig. 2). 60 ng of compet-ing CV (impure fraction) was required to yield 50 %inhibition of binding of antibody to the microtitre plates(arrow). In contrast, 240 ng of chicken liver extract wasrequired for 50 % inhibition, indicating that the purity ofthe antigen was increased fourfold during the isolation ofCV from liver extracts. In contrast, 100- to 1000-foldmore bovine brain CV was required to achieve similarinhibition. These results suggest that the ferritin antigenis present in CV and liver extracts from chicken liver butthat bovine brain CV contain very little of this antigen(see below). This correlates well with the obvious pres-ence of the 22K protein on SDS-PAGE of chicken liverCV (Fig. IB, lane a) and its virtual absence in CV frombovine brain (Fig. IB, lane b).
Purification of ferritin and quantification in CVfromvarious tissuesIn order to affinity purify the antigen from TX-100-soluble liver microsomal membranes (Fig. 3, lane a) or
Coated vesicles and ferritin 189
80 KM) 120Fraction number
I III
MW62 66 70 74 78 82 86 90 94 98102106110
200 —
116 —93 •
66 »
45 9
31 -
21 5
200 V
. 180 -
66
45 T26 f
I• • ? • • .
221812
Fig. 1. Coated vesicle purification by gel-permeation chroniatography (A) and analysis of the eluted fractions by SDS-PAGE(B) and electron microscopy (C,D). Coated vesicles were prepared from chicken liver by the method of Nandi el al. (1982).Membrane pellets from the D2O-sucrose step were resuspended by homogenization in isolation buffer (IB) at pH 6-5. A 2-5 mlsample in 10% sucrose as applied to a 1-5 cm X 100cm column of controlled-pore glass beads. In A, 1-ml fractions werecollected at a flow rate of 6mlh~' and the absorbance was measured at 280 nm. In B, column fractions for SDS-PAGE (75 jul)were diluted with a 4X concentrated SDS sample buffer (25//I) and heated at 100°C for 2min before application of 50f(l to a3-75 % to 15 % acrylamide gradient gel. Lanes a and b in B are samples of chicken liver and bovine brain preparations beforeapplication to the gel permeation column. Samples that comprise the lanes in groups I, II and III are from the fractions in A.For electron microscopy, samples from peak I (C) or peak II (D) were adsorbed to carbon-reinforced Formvar-coated grids,negatively stained with 2 % uranyl acetate, and viewed in the Hitachi 600 electron microscope. X75 000.
impure coated vesicles (Fig. 3, lane b), we coupled theanti-ferritin monoclonal antibody to Sepharose 4B andafter incubation with the liver fractions, eluted thespecifically bound ferritin. A polypeptide migrating at anapparent molecular weight of 240x 103 was present whenimmune-purified protein was not heated in SDS-PAGEsample buffer prior to PAGE (Fig. 3, lane d). Incontrast, a polypeptide having an apparent molecular
weight of 22X 103 was the major band observed whensamples were heated at 100°C (Fig. 3, lane c). Whencommercially available ferritin from horse spleen wasanalysed in the same manner, unheated samples con-tained a prominent band around 200K (Fig. 3, lane f),which was not present when samples were boiled. Thus,the disappearance of the higher molecular weight form ofthe chicken or horse ferritin resulted in a concomitant
190 A. Passaniti and T. F. Roth
103 104
Protein (ng)
F i g . 2 . Detection of ferritin in chicken liver extracts andcoated vesicles by competit ive E L I S A . Anti-ferritinmonoclonal ant ibody was used at 2-5 n g m l ~ to assay for thepresence of ferritin. Impure chicken liver coated vesicles(50fig m l " ' ) were used to coat the wells of a vinyl microti treplate. A mixture of antibody and various amounts of eitherchicken liver coated vesicles ( O ) , chicken liver extract ( • ) , orbovine brain coated vesicles ( A ) were incubated with the CVon the vinyl wells. After incubation at 37°C for 2 h , thosean t ibody-an t igen complexes bound to the microti tre wellswere detected using a second ant ibody coupled to horseradishperoxidase. Activity is expressed as percentage of b indingcompared to 100% binding in the absence of compet ingprotein. T h e amount of protein is plotted on a logarithmicscale.
increase in the amount of the lower molecular weightform. This suggests that the higher molecular weightprotein is composed of complexes of the lower molecularweight polypeptides.
To determine which form of chicken ferritin themonoclonal antibody recognized, purified antigen onSDS-PAGE was transferred to nitrocellulose and as-sayed for anti-ferritin antibody binding. The monoclonalantibody appeared to recognize only the 240K form of theantigen (Fig. 3B). The 22K protein was not recognizedby the antibody even when samples were not heated.That this antigen is ferritin is also evident from theelectron-microscopic examination of the material elutingin the included volume of the controlled-pore glass beadcolumn (Fig. 4C,D), which contained high levels ofantibody binding activity. Affinity-purified chicken ferri-tin was morphologically identical to the chromatographi-cally isolated ferritin (Passaniti & Roth, unpublisheddata).
To quantify the amount of chicken ferritin in our CVpreparations, we isolated CV from chicken liver, chickenbrain and bovine brain. Since the isolation procedures forall tissues were identical, any differences in antibodybinding would reflect differences in levels of ferritin inchicken brain and chicken liver CV and/or differences inantigenic determinants between chicken and bovine brainferritins. We used affinity-purified chicken ferritin to coatmicrotitre wells and generate a standard binding curvefor the monoclonal anti-ferritin antibody. We found thatferritin accounted for 3 % to 7 % of the total protein in theimpure coated vesicle preparations before purification by
gel permeation chromatography (data not shown).Chicken brain CV contained 0-13% ferritin, which isonly 0-05 of that found in chicken liver CV (data notshown). Bovine brain CV did not bind antibody nearly aswell as chicken brain CV. Therefore, since bovine ferritinbinds the anti-chicken ferritin antibody very poorly, wecould not estimate the amount of ferritin in bovine brain.
Specific association of ferritin and CV afterchromatographyWhen we examined the amounts of chicken ferritin in CVpurified by gel-permeation chromatography, we foundthat both peaks II and III (Fig. 4A) contained anti-ferritin antibody binding activity. Of the total proteinrecovered from several pooled chromatographs, 60 %eluted in the CV peak II and 24% eluted in peak III. Theremaining 13% was found in the excluded peak I.Fractions in peak III consisted of uniform, electron-dense particles characteristic of ferritin when viewed byelectron microscopy (Fig. 4C). These particles withelectron-dense centres were surrounded by negativelystained protein typical of the appearance of horse spleenferritin (not shown). That these fractions containedferritin is also evident from SDS-PAGE. Samples thatwere not heated exhibited a major band at 240K (Fig. 4B,lane f) and when heated a band at 22K (Fig. 4B, lane e).In contrast, fractions from peak II contained mostly CVas seen in the electron microscope (Fig. ID), but someferritin was seen associated with the CV (Fig. 4D).SDS-PAGE of unheated column fractions from peak I(lane b) and peak II (lane d) showed the 240K band,indicate of ferritin.
In order to determine how much of the ferritin in peakII from the controlled-pore glass (CPG) columnremained associated with the highly purified coatedvesicle fraction after re-chromatography, we pooled,concentrated, and then re-chromatographed the fractionsin peak II on the same CPG-bead column. After re-chromatography, we found that the majority of theapplied material eluted at the position of peak II (Fig. 5).The shoulder apparent before peak II suggests thepresence of some peak I vesicles. In contrast, very littlematerial eluted from the column at the position of peakIII. Bradford (1976) protein assays of material in frac-tions 95-120 from the peak III region showed undetect-able levels of protein. However, when the fractions fromFig. 5 were assayed for anti-ferritin antibody binding bydirect ELISA, the presence of ferritin was confirmed inboth the peak II and peak III regions (data not shown).In order to determine if the re-chromatographed peak IIIactivity was originally present in peak II coated vesiclesor if it was spill-over from peak III (Fig. 4A), wequantified the amount of ferritin in peak II coatedvesicles by competitive ELISA before and after chroma-tography. Before pooling the peak II fractions (Fig. 4A),fraction 73 contained 2% ferritin (12f«g). When peak IIfractions 70-88 were pooled and concentrated, they alsocontained 2% ferritin (280 fig out of 14 mg). After re-chromatography, the recovery of protein was 50 % (7 mg)and fraction 73 in peak II still contained 2% ferritin.Since there is no change in the amount of ferritin in peak
Coated vesicles and ferritin 191
a b c dmi*
e f a b c d e f
240200 -200
180- -97
-66
- H C
LC
22
-26
-18-12
18
Fig. 3. Detection and identification of ferritin in microsomal membrane and coated vesicle fractions from chicken liver byaffinity chromatography (A) and protein transfer to nitrocellulose (B). A. Anti-ferritin monoclonal antibody coupled toSepharose was incubated with Triton X-100 extracts of liver microsomes (A, lane a) or coated vesicles (A, lane b). Afterwashing with mixed detergents and high salt, the ferritin was eluted with 0 1 M-glycine, pH2-0 (A, lane d), or SDS-PAGEsample buffer (A, lane c). Eluted antigen was neutralized with 1 M-phosphate (1:1, v/v) at pH6-8 and then incubated withSDS-PAGE buffer at 24°C (A, lane d) or 100°C for S min (A, lane c). For comparison, commercial horse spleen ferritin wasdissolved in SDS buffer at 24°C (A, lane f) or at 100°C for 5 min (A, lane e) and electrophoresed on a 3-75 % to 15 %acrylamide gradient gel. Gels were stained with silver. HC, heavy chain IgG; LC, light chain IgG. B. Immune-purifiedmaterial was resolved on SDS-PAGE and transferred to nitrocellulose. The immune-purified ferritin was dissolved in SDSbuffer at 24°C (lanes a, c, e) or at 100°C for 5 min (lanes b, d, f). The nitrocellulose paper was stained with Amido Black todetect protein (lanes a,b) or incubated with anti-ferritin antibody followed by HRP-coupled second antibody to detect antigen(lanes c,d) or incubated with second antibody alone (lanes e,f). Specific antibody binding was detected using 4-chloro-l-naphthol as HRP substrate.
II before or after re-chromatography, these resultssuggest that: (1) the ferritin present at the peak IIIposition after re-chromatography of peak II probably waspeak III material that was not originally separated frompeak II (it is not dissociated ferritin from peak II); and(2) the ferritin in peak II is tightly associated with thecoated vesicles.
Topological distribution of CV-associated ferritinWe next investigated the nature of the association be-tween ferritin and CV by two approaches: (1) permeabil-ization of CV with 1 % saponin to release ferritin from theinterior of the CV; and (2) perturbation of the CVsurface with agents that disrupt CV structure.
Coated vesicles were obtained from the CPG columnand incubated in IB in the presence or absence of 1 %saponin for 15 min at 24°C. Competitive ELISA wasthen used to quantify the amount of ferritin present inpermeabilized and non-permeabilized CV. No increase inantibody binding was observed when CV were assayed inthe presence of saponin (data not shown). In order topreclude the possibility that the clathrin coat prevented
saponin from solubilizing the CV membrane as has beenreported for filipin-CV interactions (Steer et al. 1984),we treated CV with 1 M-KC1 or 0-5 M-Tris-HCl at pH 8-5to extract the clathrin coat (see below) before permeabil-ization with saponin. Even though 30%—50% of theclathrin could be removed by KC1 or Tris treatment, weobserved no increase or decrease in anti-ferritin antibodybinding in the presence of saponin (data not shown).
We next wished to determine if disruption of thecytoplasmic coat on the CV also removed ferritin. It hasbeen shown that treatment of CV with Tris or ureapreferentially removes the clathrin coat (Pearse &Bretscher, 1981; Woodward & Roth, 1978; Steer et al.1984) and that KC1 treatment removes clathrin andperipheral proteins (Kelly et al. 1983; Wiedenmann &Mimms, 1983). However, TX-100 has been shown to beineffective in disrupting the clathrin lattice (Woodward &Roth, 1978; Pearse, 1982) but may perturb the lipidbilayer of the CV. Purified CV from the CPG columnwere treated separately with IB containing the followingcomponents: 1 M - K C 1 ; 1% TX-100; 1% TX-100 + 0-2% SDS; 60 mM-dithiothreitol (DTT); 0-5 M-
192 A. Passaniti and T. F. Roth
20 40 60 80 100Fraction number
120 140
Ba b
IIc d e f
-240
— -180
-22
Fig. 4. Gel-permeation chromatography of chicken liver coated vesicles: elution of anti-ferritin binding activity. Chicken liverCV were applied to a column of controlled-pore glass beads as in Fig. 1. A. Fractions were collected and assayed for protein bythe Bradford method (absorbance at 595 nm) and for ferritin using monoclonal anti-ferritin antibody in the direct-bindingELISA (absorbance at 490nm). Samples corresponding to the peak regions I (lanes a,b), II (lanes c,d) and III (lanes e,f) inFig. 1 were analysed by SDS-PAGE (B). Samples were dissolved in SDS buffer at 24°C (b,d,f) or at 100°C for 5 min (a,c,e)before application to the gel. Electron micrographs of the eluted fractions from peak III (C) or a higher magnification of theCV-containing peak II (D) were obtained as described in Materials and methods. Arrowheads indicate ferritin particlesassociated with CV. C, X100000; D, x 137 500.
Tris-HCl, pH8-5; or0-5 M-Tris-HCl, pH8-5 + 2\i-urea.After incubation, samples were centrifuged on a 10% to30 % sucrose gradient and the CV pellets were analysedfor the presence of protein, ferritin and clathrin. Inpreliminary experiments, free ferritin did not sediment
beyond the 20%-30% sucrose interface in the 10% to30% sucrose velocity gradients, whereas CV sedimentedto the bottom of the tube. Sedimentation of purifiedchicken ferritin was not affected by the CV-perturbingagents. We found (Table 1) that 1 M-KC1 and 0-5 M-Tris-
Coated vesicles and fenitin 193
U-i
0-4
0-3
0-2
0-1
I II
A
in
* a b
• I
Table 1. Extraction of CPG-purified coated vesicles
0 20 40 60 80 100 120Fraction number
Fig. 5. Re-chromatography of coated vesicles from acontrolled-pore glass bead column. Fractions 70-88 of peakII from several chromatograms (inset) were pooled,concentrated, and 14 mg of protein was re-applied to theglass-bead column. Eluted fractions were assayed forabsorbance at 280 nm. The inset shows an SDS-PAGE gel ofthe pooled CV peak II applied to the column (lane a). PeakIII is shown for comparison (lane b). Note the presence ofthe 22K polypeptide in peak II.
HC1 were quite effective in extracting non-clathrin pro-teins and ferritin from the CV but not very effective inextracting clathrin. This suggests that the associationbetween CV and ferritin may depend on electrostaticinteractions. A combination of 0-5 M-Tris-HCl and 2 M-urea was more effective in extracting clathrin but notmore effective than 1 M-KC1. As expected, 1 % TX-100did not dissociate the clathrin lattice structure, sinceclathrin still sedimented through the sucrose gradientafter 1 % TX-100 treatment. Similarly, ferritin was notextracted from the CV with TX-100 and co-sedimentedwith clathrin through the sucrose gradient, suggestingthat it is associated with the coat lattice.
Discussion
Isolated coated vesicles are composed of many polypep-tides in addition to the structural clathrin and light-chainproteins. Numerous investigators have reported on theassociation of CV with these other cellular components.However, many of these observations were not madeusing CV that were very highly purified. Separation ofCV from other components by controlled-pore glass-bead(gel-permeation) chromatography is a molecular sizingmethod that greatly extends the established differentialand isopycnic gradient centrifugation methods. Thetraditional methods used to purify CV often do notdistinguish between coated and uncoated vesicles, orbetween coated vesicles and smaller components likeferritin that are of similar density. The further purifi-cation provided by gel-permeation chromatography isnecessary in order to ensure that the CV being analysedare of the highest purity. In this study, as in our previousstudy on the identification of tubulin as a coated vesicle
Peak II fractions from CPG chromatography were pooled andconcentrated by centrifugation. Samples of O'l ml were incubatedwith 0-1 ml of 2x concentrated treatment buffer for IS min at roomtemperature and then placed on top of a 10% to 30% sucrosegradient containing the appropriate treatment buffer. Samples werecentrifuged at 220000gfh~1 at 4°C. Pellets containing CV wereresuspended in 10ml of IB and centrifuged at lOOOOOgh"'. Thewashed pellets were then resuspended to a total volume of 0*2ml.Protein remaining in the pellets after extraction was measured by theBradford method. Ferritin was quantified by competitive ELISA.Clathrin levels in the extracted pellets were quantified bydensitometry of SDS-PAGE gels. IB represents recovered materialfrom untreated samples set arbitrarily at 100%. n.d., not detectable.
component (Kelly et al. 1983), we use highly purified CVobtained by gel permeation to show that ferritin is alsoassociated with CV.
We measured the extent and nature of ferritin's associ-ation with CV by employing a highly specific monoclonalantibody that recognizes chicken liver ferritin. Enzyme-linked immunoassays were used to quantify ferritin levelsin isolated CV and CV that were treated with clathrin ormembrane-perturbing agents. We find that the majorityof ferritin is not bound to CV but that ferritin comprises2% of the total CV protein. In estimating the number offerritin particles per CV, we calculated that if a 'small' CVcontained 100 clathrin molecules (Pearse & Bretscher,1981) then a large CV of about 200 nm diameter wouldcontain 1000 clathrin molecules. Since the majority of theCV we isolate are included in the CPG column, theirdiameter is less than 190nm. Therefore, if we assumethat an average CV contains 400 clathrin molecules(about l-2x 10~16g) and if the majority of the protein inthe highly purified CV preparation is clathrin, 2 % of thetotal protein represents on average about three ferritinmolecules per CV. This estimate is close to the number offerritin molecules (1-4 per CV) estimated by Pearse(1982) from electron micrographs of human placental CVand ferritin.
We found that Triton X-100 did not release ferritinfrom CV. However, if we dissociated the CV coatstructure with Tris-urea, most of the CV-associatedferritin was released. Taken together, our data areconsistent with ferritin being associated with the clathrincoat of the CV and/or with the external face of the lipidbilayer. Using CV purified by conventional, but less-effective gradient methods, Pearse (1982) reported thatferritin, transferrin, and IgG can be co-isolated withplacental coated vesicles in the presence of Triton X-100.Using electron-microscopic localization, Pearse showedthat some ferritin was found inside the isolated CV. In
194 A. Passaniti and T. F. Roth
agreement with the data reported by Pearse, we find thatTriton X-100 does not release ferritin from CV. How-ever, we were not able to detect ferritin inside CV, usingeither our sensitive ELISA methods or electron mi-croscopy. These findings may be due to the differentfunctions of liver and placental membranes or the differ-ent experimental techniques used.
The separation of coated from smooth, uncoatedvesicles by gel permeation depends on the size of thecoated vesicles. In contrast to our previous observationsof CV from bovine brain (Kelly et al. 1983), where theCV fractions were clearly separated from the smoothvesicles, it was clear that the chicken liver CV fractions(Fig. 4, peak II) are not completely resolved from theuncoated membrane fractions. This may be due to thelarger size of CV from chicken liver. Bovine brain CV arealmost exclusively 50-80 nm in diameter, whereas CVfrom chicken liver are typically larger (Pearse, 1982;Woods et al. 1978; Wall et al. 1980).
Endogenous ferritin or exogenously added native ferri-tin has been shown to bind to the cytoplasmic surfaces ofrough endoplasmic reticulum and plasma membrane byelectron-microscopic techniques (Parr, 1979). Since CVare formed from the invagination and budding of plasmamembrane coated pits, it is possible that bound ferritinwould remain or become associated with isolated CV. Inthis report we have demonstrated that ferritin is associ-ated with CV. A possible function for this associationmay be the exchange of iron between cellular storage sitesand transferrin in the circulation. Depending on thedirection of exchange, this process may occur at theplasma membrane (Osaki et al. 1971) or in the endosomalcompartments where ferritin and iron-rich transferrinmight exchange iron across the membrane (Trump &Berezesky, 1977). In the first model (Osaki et al. 1971),ferritin would lose iron into the bloodstream by reductionfrom ferric to ferrous iron. In the second model (Trump& Berezesky, 1977), iron is obtained from transferrinafter endocytosis and release in an intracellular compart-ment. In either case, close association of ferritin withcellular membranes would facilitate iron transfer.Clearly, further studies are needed in order to elucidatethe possible role of ferritin and define the function of CVin this process.
We thank Katherine Rosewell and John Lewis for their experttechnical assistance, Tim Ford, Fran Baldwin and Audrey Ellisfor their assistance in the preparation of the manuscript, andJulie Donaldson for her critical reading and advice.
References
BALCH, W. E., GLICK, B. S. & ROTHMAN, J. E. (1984). Sequential
intermediates in the pathway of intercompartmental transport in acell-free system. Cell 39, 525-536.
BLITZ, A. L., FINE, R. E. & TOSELU, P. A. (1977). Evidence thatcoated vesicles isolated from brain are calcium-sequesteringorganelles resembling sarcoplasmic reticulum. J'. Cell Biol. 75,135-147.
BOOTH, A. G. & WILSON, M. J. (1980). Human placental coatedvesicles contain receptor-bound transferrin. Biochem.J. 196,355-362.
BRADFORD, M. (1976). A rapid and sensitive method for the
quantitation of microgram quantities of protein utilizing theprinciple of protein-dye binding. Analyt Biochem. 72, 248-254.
BRODSKY, F. M., HOLMES, N. J. & PARHAM, P. (1983).
Tropomyosin-Hke properties of clathrin light chains allow a rapid,high-yield purification. J. Cell Biol. 96, 911-914.
BROWN, M. S., ANDERSON, R. G. W. & GOLDSTEIN, J. L. (1983).
Recycling receptors: the round trip itinerary of migrant membraneproteins. Cell 32, 663-667.
CAMPBELL, C. H., FINE, R. E., SQUICCIARINI, J. & ROME, L. H.
(1983). Coated vesicles from rat liver and calf brain contain crypticmannose 6-phosphate receptors. J . biol. Chew. 258, 2628-2633.
CROWTHER, R. A. & PEARSE, B. M. F. (1981). Assembly and packingof clathrin into coats. J. Cell Biol 91, 790-797.
CZECH, M. P. (1982). Structural and functional homologies in thereceptors for insulin and the insulin-like growth factors. Cell 31,8-10.
DANTRY-VARSAT, A., CIECHANOVER, A. & LODISH, H. F. (1983). pH
and the recycling of transferrin during receptor-mediatedendocytosis. Proc. natn. Acad. Sci. U.S.A. 80, 2258-2262.
DUNN, W. A., CONNOLLY, T. P. & HUBBARD, A. L. (1986).
Receptor mediated endocytosis of epidermal growth factor by rathepatocytes: receptor pathway. J. Cell Biol. 102, 24-36.
FAMBROUGH, D. M. & BAYNE, E. K. (1983). Multiple forms of(Na+ + K+)-ATPase in the chicken. Selective detection of themajor nerve, skeletal muscle, and kidnev form by a monoclonalantibody. J. biol. Chem. 258, 3926-3935'.
FORGAC, M., CANTLEY, L., WIEDENMANN, B., ALTSTIEL, L. &
BRANTON, D. (1983). Clathrin-coated vesicles contain an ATP-dependent protein pump. Proc. natn. Acad. Sci. U.S.A. 80,1300-1303.
KASUGA, M., ZICK, Y., BLITH, L., KARLSSON, F. A., HARING, H. V.
& KAHN, C. R. (1982). Insulin stimulation of phosphorylation ofthe /3 subunit of the insulin receptor. J. biol. Chem. 257,9891-9894.
KEEN, J. H., WILLINGHAM, M. C. & PASTAN, 1. H. (1979). Clathrin-coated vesicles: isolation, dissociation and factor-dependentreassociation of clathrin baskets. Cell 16, 303-312.
KELLY, W. G., PASSANITI, A., WOODS, J. W., DAISS, J. L. & ROTH,
T. F. (1983). Tubulin as a molecular component of coatedvesicles. J. Cell Biol. 97, 1191-1199.
KIRCHHAUSEN, T. & HARRISON, S. C. (1981). Protein organization inclathrin trimers. Cell 23, 755-761.
LAEMMLI, U. K. (1970). Cleavage of structural proteins during theassembly of the head of the bacteriophage T4. Xalnre, Loud. 227,680-685.
LINDEN, C. D., DEDMAN, J. R., CHAFOULEAS, J. G., MEANS, A. R.
& ROTH, T. F. (1981). Interactions of calmodulin with coatedvesicles from brain. Pmc. natn. Acad. Sci. U.S.A. 78, 308-312.
MELLMAN, I., FUCHS, R. & HELENIUS, A. (1986). Acidification of
the endocytic and exocytic pathways. A. Rev. Biochem. 55,663-700.
MELLO, R. J., BROWN, M. S., GOLDSTEIN, J. L. & ANDERSON, R.G. W. (1980). LDL receptors in coated vesicles isolated frombovine adrenal cortex: binding sites unmasked by detergenttreatment. Cell 20, 829-837.
MERRIL, C. R., GOLDMAN, D., SEDMAN, S. A. & EBERT, M. H.
(1981). Ultrasensitive stain for proteins in polyacrylamide gelsshows regional variation in cerebrospinal fluid proteins. Science211, 1437-1438.
NANDI, P. K., IRACE, G., VAN JAARSVELD, P. P., LIPPOLDT, R. E. &
EDELHOCH, H. (1982). Instability of coated vesicles inconcentrated sucrose solutions. Proc. natn. Acad. Sci. U.S.A. 79,5881-5885.
OMARY, M. B. & TROWBRIDGE, J. S. (1981). Biosynthesis of thehuman transferrin receptor in cultured cells, j ' . biol. Chem. 256,12888-12892.
ORCI, E., GLICK, B. S. & ROTHMAN, J. E. (1986). A new type of
coated vesicular carrier that appear not to contain clathrin: itspossible role in protein transport with the Golgi stack. Cell 46,171-184.
OSAKI, S., JOHNSON, D. A. & FRIEDEN, E. (1971). The mobilizationof iron from the perfused mammalian liver by a serum copperenzyme, ferroxidase I. J . biol. Chem. 246, 3018-3023.
PARR, E. L. (1979). Intracellular labeling with ferritin conjugates. A
Coated vesicles and fern tin 195
specificity problem due to the affinity of unconjugated ferritin forselected intracellular sites. J. Histochem. Cytochem. 27, 1095-1102.
PAULOIN, A., BERNIER, I. & JOLLES, P. (1982). Presence of cyclicnucleotide-Ca2+-independent protein kinase in bovine brain coatedvesicles. Nature, ljoud. 298, 574-576.
PEARSE, B. M. F. (1982). Coated vesicles from human placenta carryferritin, transferrin, and immunoglobulin G. Proc. natit. Acad. Sci.U.S.A. 79, 451-455.
PEARSE, B. M. F. & BRETSCHER, M. S. (1981). Membrane recyclingby coated vesicles. A. Rev. Biochem. 50, 85-101.
PEARSE, B. M.F. & ROBINSON, M. S. (1984). Purification andproperties of 100-Kd proteins from coated vesicles and theirreconstitution with clathrin. EMBOJ. 3, 1951-1957.
PETRUZZELLI, L. M., GANGULY, S., SMITH, C. J., COBB, M. H.,
RUBIN, C. S. & ROSEN, O. M. (1982). Insulin activates a tyrosine-specific protein kinase in extracts of 3T3-L1 adipocytes and humanplacenta. Proc. naln. Acad. Sci. U.S.A. 79, 6792-6796.
PFEFFER, S. R., DRUBIN, D. G. & KELLY, R. B. (1983).
Identification of three coated vesicle components as a-and /3tubulin linked to a phosphorylated 50,000-mol-wt polypeptide. J.Cell Biol. 97, 40-47.
PFEFFER, S. R. & KELLY, R. B. (1981). Identification of minorcomponents of coated vesicles by use of permeationchromatography.J. Cell Biol. 91, 385-391.
PILCH, P. F., SHIA, M. A., BENSON, R. J. J. & FINE, R. E. (1983).
Coated vesicles participate in the receptor-mediated endocytosis ofinsulin. J . Cell Biol. 96, 133-138.
ROBINSON, M. S. & PEARSE, B. M. F (1986). Immunofluorescentlocalization of 100K coated vesicle proteins, j ' . Cell Biol. 102,48-54.
ROTH, T. F. & PORTER, K. R. (1964). Yolk protein uptake in theoocyte of the mosquito Aedes aegypti L. J. Cell Biol. 20, 313-332.
ROTHMAN, J. E. (1981). The Golgi apparatus: two organelles intandem. Science 213, 1212-1219.
ROTHMAN, J. E., BURSTYN-PETTIGREW & FINE, R. E. (1980).
Transport of the membrane glycoprotein of vesicular stomatitisvirus to the cell surface in two stages by clathrin-coated vesicles. J.Cell Biol. 86, 162-171.
ROTHMAN, J. E., URBANI, L. J. & BRANDS, R. (1984). Transport of
protein between cytoplasmic membranes of fused cells:correspondence to processes reconstituted in a cell-free system. J.Cell Biol. 99, 248-259.
SAHAGIAN, G. G., DISTLER, J. & JOURDIAN, G. W. (1981).
Characterization of a membrane-associated receptor from bovineliver that binds phosphomannosyl residues of bovine testicular /3-galactosidase. Proc natn. Acad. Sci. U.S.A. 78, 4289-4293.
SCHLOSSMAN, D. M., SCHMID, S. L. & BRAELL, W. A. & ROTHMAN,
J. E. (1984). An enzyme that removes clathrin coats: of uncoatingATPase.J. Cell Biol. 99, 723-733.
SCHNEIDER, W. J., BEISIEGEL, V., GOLDSTEIN, J. L. & BROWN, M.
S. (1982). Purification of the low density lipoprotein receptor, anacidic glycoprotein of 164,000 molecular weight. J. biol. Chem.257, 2664-2673.
STEER, C. J., BISHER, M., BLUMENTHAI., R. & STEVEN, A. C. (1984).Detection of membrane chloresterol by filipin in isolated rat livercoated vesicles is dependent upon removal of the clathrin coat. J.
Cell Biol. 99, 315-319.STEFANINI, S., CHIANCONE, E., AROSIO, P., FINAZZI-AGRO, A. &
ANTONINI, E. (1982). Structural heterogeneity and subunitcomposition of horse ferritins. Biochemistry 21, 2293-2299.
STEINER, A. W. & ROME, L. H. (1982). Assay and purification of asolubilized membrane receptor that binds the lysosomal enzyme a-L-iduronidase. Archs Biochem. Biophvs. 214, 681-687.
STONE, D. K., XIE, X. S. & RACKER, E. (1983). An ATP-drivenprotein pump in clathrin-coated vesicles. J'. biol. Chem. 258,4059-4062.
THEIL, E. C. (1987). Ferritin: structure, gene regulation, andcellular function in animals, plants, and microorganisms. A. Rev.Biochem. 56, 289-315.
TOWBIN, H., STAEHELIN, T. & GORDON, J. (1979). Electrophoretic
transfer of proteins from polyacrylamide gels to nitrocellulosesheets: procedure and some applications. Pmc. natn. Acad. Sci.U.S.A. 76, 4350-4354.
TRUMP, B. F. & BEREZESKY, I. K. (1977). A general model ofintracellular iron metabolism. In Proteins of hvn Metabolism (ed.E. B. Brown, P. Aisen, J. Fielding & R. R. Crichton), pp.359-364. New York: Grune and Stratton.
UNANUE, E. R., UNGEWICKELL, E. & BRANTON, D. (1981). The
binding of clathrin triskelions to membranes from coated vesicles.Cell 26, 439-446.
UNGEWICKELL, E. & BRANTON, D. (1981). Assembly units of clathrincoats. Nature, hand. 289, 420-422.
USHIRO, H. & COHEN, S. (1980). Identification of phosphotyrosine asa product of epidermal growth factor-activated protein kinase in A-431 cell membranes. J. biol. Chem. 255, 8363-8365.
VAN EIJK, H. G. & VAN NOORT, W. L. (1976). Isolation of rattransferrin using CNBr-activated Sepharose 4B. .7. din. Chem. din.Biochem. 14, 475-478.
VAN RENSWOUDE, J., BRIDGES, K. R., HARFORD, J. B. & KLAUSNER,
R. D. (1982). Receptor-mediated endocytosis of transferrin and theuptake of Fe in K562 cells: Identification of nonlysosomal acidiccompartment. Proc. natn. Acad. Sci. U.S.A. 79, 6186-6190.
WALL, D. A., WILSON, G. & HUBBARD, A. L. (1980). The galactose-
specific recognition system of mammalian liver: the route of ligandinternalization in rat hepatocytes. Cell 21, 79-93.
WIEDENMANN, B. & MIMMS, L. T. (1983). Tubulin is a majorprotein constituent of bovine coated vesicles. Biochem. biophvs.Res. Gominun. 115, 303-311.
WOODS, J. W., WOODWARD, M. P. & ROTH, T. F. (1978). Common
features of coated vesicles from dissimilar tissue: composition andstructure. J. Cell Sci. 30, 87-97.
WOODWARD, M. P. & ROTH, T. F. (1978). Coated vesicles:characterization, selective dissociation and reassembly. Proc. natn.Acad. Sci. U.S.A. 75, 4394-4398.
ZAREMBA, S. & KEEN, J. H. (1983). Assembly polypeptides fromcoated vesicles mediate reassembly of unique clathrin coats, jf. CellBiol. 97, 1339-1347.
(Received 26Julv 1988 -Accepted, in revised form, 3 November1988)
