retinol-binding protein: the transport

Retinol-Binding Protein: the Transport

Protein for Vitamin A in Human Plasma

MAsAmrrsu KANAI, AMIRAMRAZ, and DEWrrr S. GOODMAN

From the Department of Medicine, Columbia University College of Physiciansand Surgeons, NewYork 10032

A BSTRACT Vitamin A circulates in humanplasma as retinol bound to a specific transportprotein. This protein differs from the known lowand high density plasma lipoproteins and has a hy-drated density greater than 1.21. In order to studythis protein, volunteers were injected intravenouslywith retinol-15-14C. Plasma was collected 1-3 dayslater, and the purification of retinol-binding pro-tein (RBP) was monitored by assaying for 14C andalso by following the fluorescence of the protein-bound retinol. Purification of RBPwas effected bythe sequence: Cohn fractionation, chromatographyon columns of Sephadex G-200 and diethylamino-ethyl (DEAE)-Sephadex, preparative polyacryl-amide gel electrophoresis, and finally chromatog-raphy on Sephadex G-100. These procedures re-sulted in a preparation of RBPwhich was at least98%o pure and which had been purified more than1500-fold. Purified RBP has al mobility on elec-trophoresis and has a molecular weight of ap-proximately 21,000-22,000. There appears to beone binding site for retinol per molecule of RBP.Solutions of RBP are fluorescent (characteristicof retinol) and have ultraviolet absorption spectrawith peaks at 330 m~m (resulting from the boundretinol) and at 280 mu. There are no fatty acid or

This paper is No. 8 in the series "Vitamin A andcarotenoids" from the laboratory of DeW. S. Goodman.An abstract of this work was presented to the SeventhInternational Congress of Biochemistry in Tokyo, Japan,in August 1967 (39), and the work was presented, inpart, at the 11th International Conference on the Bio-chemistry of Lipids, Jerusalem, Israel, August 1967, andat the 1968 annual meeting of the American Society forBiological Chemists (40).

Masamitsu Kanai's permanent address is Shinshu Uni-versity Hospital, Matsumoto City, Japan.

Received for publication 6 February 1968.

fatty acyl chains present in purified RBP. Theusual concentration of RBP in plasma is of theorder of 3-4 mg/100 ml. In plasma, RBP ap-parently circulates as a complex, together withanother, larger protein with prealbumin mobilityon electrophoresis. The RBP-prealbumin complexremains intact during Cohn fractionation and dur-ing chromatography on Sephadex and on DEAE-Sephadex columns. The complex dissociates dur-ing gel electrophoresis, permitting the isolationand subsequent purification of each of the com-ponents. The complex is again formed by mixingtogether solutions of the separated RBP and ofprealbumin. Retinol transport in plasma thus ap-pears to involve both a lipid-protein (retinol-RBP) interaction and a protein-protein (RBP-prealbumin) interaction.

INTRODUCTION

It is well established that vitamin A mainly cir-culates in plasma as retinol (vitamin A alcohol),apparently associated with a transport protein ofhydrated density greater than 1.21 (1-3). Thisprotein differs from serum albumin in man (1)and has been reported to have a, mobility (2) ora2 mobility (4) in the rat. Recently, Alvsaker,Haugli, and Laland (5) have suggested that thisprotein is identical with the tryptophane-rich pre-albumin in human serum. Detailed informationabout the transport of retinol in plasma is, how-ever, not available. Wenow report the purificationand partial characterization of retinol-binding pro-tein (RBP), the specific protein to which retinolis bound in human plasma. RBP has-a, mobilityand differs from all previously identified plasmaproteins.

The Journal of Clinical Investigation Volume 47 1968 2025

METHODSPlasma. The studies reported here were carried out

with fresh blood, collected in acid citrate dextrose (ACD)anticoagulant, from healthy young adult donors. The iso-lation of RBP was carried out three times: first with280 ml of plasma from one donor, second with 1700 mlof plasma from six donors, and third with 5 liters ofplasma from 16 donors. The plasma used in the first ofthese preparations, and 300 ml of the plasma used in thesecond preparation, was derived from subjects previouslyinjected with "C-labeled retinol.

Assay. The purification of RBP was quantitativelymonitored by assaying the fractions from each puri-fication procedure for protein-bound retinol. In theinitial stages of this work this assay was carried out bythe extraction of plasma or plasma fractions with CHCI3-CHaOH, 2:1 (v/v), followed by the chromatographyof the total lipid extracts on small columns of deacti-vated alumina (6) in order to isolate retinol. The retinolwas then assayed spectrophotometrically (in our labora-tory Eing = 1625 at 330 mg& for retinol in benzene).During most of this work, however, a simpler assay wasused, in which the protein-bound retinol was first labeledwith '4C, and the purification of RBP was then monitoredby radioassay, for '4C, of the total lipid extracts. Tothis end, normal volunteers were injected intravenouslywith retinol-15-"'C dispersed in their own plasma. Thedispersion was achieved by adding, via a microsyringeand under sterile conditions, 0.5 ml of acetone contain-ing 12 gc of retinol-15-"C (specific radioactivity 29.6Ac/mg) to 20 ml of freshly collected plasma. Plasma wascollected 1-3 days after "C-retinol injection and wasused for RBP purification. The radioactivity ("C) inthese plasma samples was shown by extraction and chro-matography to reside almost entirely in retinol. The "4Cassay procedure for protein-bound retinol was validatedfor each purification method by showing that comparableresults were obtained by "4C assay and by spectrophoto-metric assay for retinol. In addition, RBP purificationwas also qualitatively monitored by observing under ul-traviolet light the visible, green fluorescence of protein-bound retinol.

Chromatography. Gel filtration (7) was carried outwith columns of Sephadex 1 G-100 or G-200, with phos-phate buffer, usually at pH 7.4 or 7.6, and containing 0.2M NaCl. Ion-exchange chromatography was conductedwith columns of DEAE-Sephadex 1 A-50, with phos-phate buffer, pH 7.6, and a gradient of increasing con-centration of NaCl. Specific details for these proceduresare provided for representative examples in the legendsto the figures. All column chromatography was carriedout in a cold room at 4°-5°C.

Electrophoresis. Analytical and preparative poly-acrylamide gel electrophoresis were performed in 7%acrylamide gel without a concentrating gel, with a con-tinuous buffer system of Tris-glycine-HCl, pH 8.1, underconditions similar to those described by Nerenberg (8).Analytical disc gel electrophoresis employed a current

1 Pharmacia Fine Chemicals, Inc., Piscataway, N. J.

of 1 mampper tube at room temperature and was termi-nated just when the tracking dye ran out of the gel.For preparative purposes, the Buchler2 "polyprep" ap-paratus was used in a manner similar to that describedby Jovin, Chrambach, and Naughton (9). The preparedgel column, about 9.5 cm high, was prerun for 1 hr at30 mamps before the sample was applied. In the usualprocedure, carried out at 2°-5°C, a sample of 30-60 mgof protein was dissolved in 4.5 ml of upper buffer, 0.5 mlof 40% sucrose solution was added, and the mixed samplewas then applied to the top of the gel. Electrophoresiswas begun at a current of 15-20 mamps. After 30 minthe current was adjusted to 30-35 mamps and was main-tained at this level until the voltage reached 500 volts,after which time electrophoresis was continued at a con-stant voltage of 500 volts. The elution rate was 30-40ml/hr. The complete procedure usually required 8-10 hrtime. During gel electrophoresis the effluent stream wasmonitored continuously for absorption of light at wave-length 280 m,.3 The effluent stream from most columnchromatographic runs was similarly monitored.

Paper electrophoresis was performed in glycine-ace-tate buffer, pH 8.6, ionic strength 0.15 (10). Zone elec-trophoresis on Pevikon was kindly performed by Dr.Jane H. Morse, with barbital buffer, pH 8.6, ionicstrength 0.1 (11). At the end of electrophoresis, 1 cmstrips of Pevikon were collected and eluted. The eluateswere analyzed for protein and for 'C (after CHC13-CH30Hextraction).

Analytical ultracentrifugation. Sedimentation equi-librium and velocity analyses were kindly carried out byDr. Herbert S. Rosenkranz in a Spinco model E ultra-centrifuge. The molecular weight of RBP was estimatedby the sedimentation equilibrium (Archibald) method asdescribed by Schachman (12). Analyses were carriedout on a solution of 10 mg of RBP per ml in 0.05 M po-tassium phosphate buffer, pH 7.2. Centrifugation was con-ducted at 10.6°C at two speeds (9341 and 15,220 rpm),and photographs were taken with Schlieren optics. Pho-tographs taken at each speed at bar angles of 70 and of80° were measured in a microcomparator, together withphotographs taken from a run in a synthetic boundarycell. Calculations were made only from the measure-ments at the top of the cell, since some denaturation ap-parently occurred at the oil-water interface at the bottomof the cell. The molecular weight was calculated as de-scribed by Schachman (12), using an assumed value of0.75 as an estimate for the term Vp. This value is closeto the value of Vp (0.73) which can be calculated fromthe amino acid composition, the partial molal volume foreach amino acid, and the measured density of the solu-tion. It is hence felt that the assumed value for Vp didnot introduce any major error in the estimation of themolecular weight. It should, however, be noted that thecalculated molecular weight is only an approximate esti-mate, because of the uncertainty in the value of Vp.

2 Buchler Instruments, Inc., Fort Lee, N. J.3Uvicord II absorptiometer, LKB Instruments, Inc.,

Rockville, Md.

2026 M. Kanai, A. Raz, and D. S. Goodman

Sedimentation velocity studies were carried out on asolution of 1 mg/ml in the same buffer, using ultra-violet absorption optics. The sedimentation coefficient wascalculated for a solution of this protein concentration inwater at 20'C (Sio,,v).

Amino acid analysis. The amino acid composition ofRBP was kindly determined by Dr. Robert E. Canfield.Equal-sized portions (about 0.5 mg each) of RBP wereeach dissolved in 1 ml of constant boiling HC1 and hy-drolyzed in evacuated sealed tubes kept at 1080 ± 10C for24, 48, or 72 hr. The HC1 was removed by evaporationin vacuo over KOH. Amino acid analyses were per-formed by the method of Spackman, Stein, and Moore(13) on a Spinco Model 120B amino acid analyzer. Noattempts were made to determine the M cystine contentafter oxidation or reduction and alkylation of RBP.

The tryptophane content of RBP was estimated by de-termining the ratio of tyrosine to tryptophane by themethod of Goodwin and Morton (14). Purified apo-RBP(preparation RBP., see Results), 0.87 mg, was dissolved

in 1.0 ml 0.1 N NaOHand the absorption spectrum mea-sured in a Cary model 14 spectrophotometer. A base lineof "irrelevant absorption" was obtained by extrapolatingthe absorption between 360 mu and 340 mAt back to 260myt. The corresponding base line values were subtractedfrom the absorbances observed at 280 mA& and at 294.4m;& and the ratio of tyrosine to tryptophane then cal-culated as described (14).

Other procedures. Double diffusion in gel was per-formed by the method of Ouchterlony (15), with gelsmade of 0.7% agar in Na barbiturate-glycine buffer, pH7.6.

Cohn fractionation was carried out according to method10 of Cohn et al. (16). In early studies, whole plasma(ACD anticoagulant) was used for Cohn fractionation.In later work, fibrinogen (Cohn Fraction I) was firstremoved from the plasma by addition of 3 U of bovinethrombin' per ml of plasma. The plasma was incubatedfor 2 hr at room temperature with stirring and the in-soluble fibrin removed by filtration.

The distribution of retinol among the different plasmalipoproteins was determined on a sample of plasma col-lected from a subject 27 days after the intravenous in-jection of retinol-1'C. It was assumed that considerableequilibration would have occurred by this time betweenthe injected labeled retinol and the body pools of vita-min A. A portion of the plasma was extracted and thelipid extract chromatographed on deactivated alumina(6). Almost all (94%o) of the 14C applied to the col-umn was recovered after chromatography; 88%o of therecovered radioactivity was found in the retinol frac-tion. These results are virtually identical with those ob-tained in our laboratory for the extraction and chromatog-raphy of pure retinol, and establish that the 'C inplasma was present in retinol. Another portion of theplasma was separated into lipoprotein classes of densityless than 1.063, between 1.063 and 1.21, and greater than1.21, by the method of Havel, Eder, and Bragdon (17).

4 Parke Davis & Company, Detroit, Mich.

The lipoprotein fractions were not washed by recentri-fugation. Each of the fractions was extracted, and thelipid extracts were chromatographed on alumina. Theretinol fractions were collected and assayed for radioac-tivity.

Protein concentrations were determined by the methodof Lowry, Rosebrough, Farr, and Randall (18), withcrystalline bovine serum albumin as a standard.

Radioassay was carried out by dissolving samples in 15ml of 0.5% diphenyloxazole in toluene, followed by as-say with a Packard Tri-Carb liquid scintillation counter.Significant quenching was not observed.

Absorption spectra and absorbances were usually mea-sured with a Beckman DB spectrophotometer. In someinstances spectra were recorded with a Cary model 14spectrophotometer.

The content of retinol in purified RBP preparationswas determined by extracting 1 ml of RBP solution,together with 1 ml of freshly drawn plasma, with 20volumes of CHCI3-CHsOH (2: 1) in the usual way.After evaporation of the chloroform, the lipid residuewas chromatographed on 2 g of deactivated alumina (6);the retinol fraction was evaporated to dryness and theresidue dissolved in 1 ml of benzene. The benzene solu-tion was then analyzed spectrophotometrically. Controlsamples of 1 ml of plasma alone were processed in thesame manner and the ultraviolet absorption of the cor-responding column fraction subtracted from that of theRBP sample. The plasma was added before extractionin order to protect the retinol against oxidative lossduring the procedure. For a series of eight standardsamples, in which small measured amounts of retinol(not RBP) were added to the CHCI3-CH8OH extracts,the recovery of retinol (assayed spectrophotometrically)was 89.9 ± 1.7% (mean ± SEM). In the absence of addedplasma, the recovery was less than 80% and was quitevariable.

A purified RBP preparation was analyzed for itspossible content of free fatty acids and fatty acyl chains.Duplicate 1 ml samples, each containing 1.1 mg of theRBP preparation (see Results), were extracted withCHC1-CH30H containing 5 ,ug of heptadecanoic acid(to serve as an internal standard). The lipid extract wasevaporated, the residue methylated, and gas-liquid chro-matography of the fatty acid methyl esters carried out,as described previously (19). Standard samples of dis-tilled water plus 5 or 10 ug of lysolecithin or free fattyacids (FFA) were carried through the same procedureat the same time. With these standard samples the addedlipid was clearly and quantitatively detected.

Materials. Retinol-15-14C was the generous gift ofHoffmann-La Roche, Inc. of Basel, Switzerland. Unla-beled retinol was purchased from Eastman Kodak Chem-icals, Rochester, N. Y. Antisera were obtained fromBehringwerke A.G.5 The purity of the commercial anti-sera was not tested. Proteins of known molecular weightwere obtained as a "molecular weight marker kit." 6

5 Hoechst Pharmaceutical Co., Kansas City, Mo.B Mann Research Labs, Inc., New York.

Vitamin A Transport in Human Plasma 2027

RESULTSLipoprotein distribution. The distribution of

retinol-14C among different plasma lipoproteinswas determined as described above. Of the recov-ered retinol-14C, 90% was found in the proteinfraction with density greater than 1.21, 67o wasfound in the high density (d 1.063-1.21) lipopro-tein fraction, and 4%o in the low density (d lessthan 1.063) lipoprotein fraction. These findingsconfirm previous observations (1) that vitamin Acirculates in human plasma as retinol mainly inassociation with proteins of density greater than1.21.

Cohn fractionation. The results of Cohn frac-tionation are shown in Table I. Three-fourths ofthe plasma retinol, as indicated both by 14C assayand by spectrophotometric assay for retinol, wasrecovered in Cohn Fraction I + III. All of thisretinol was present in Cohn Fraction III, sinceremoval of fibrinogen (fraction I) by addition ofthrombin removed none of the 14C, and none ofthe retinol, from plasma.

Column chromatography. Chromatography ofCohn Fraction I + III (derived from plasma con-taining retinol-14C) on a large column of SephadexG-200 (Fig. 1) resulted in the recovery of morethan 90%o of the applied 14C in the total eluatefrom the column. More than 80%o of the recovered14C was eluted in a peak which appeared after theelution of the major protein peaks from the column(Fig. 1, top). The elution volume of this peak wasclose to that expected for serum albumin. All ofthe visible (green) fluorescence was eluted in thecorresponding portion of the column effluent (Fig.

TABLE IThe Distribution of Retinol in Cohn Fractions

of HumanPlasma*

Per cent distribution of

Retinol(spectro-

Cohn Total Retinol- photometricFraction protein 14C assay)

I +III 22 73 76II 10 3 6

IV 4 8V 64 13 18

VI 1 3

* Plasma (280 ml) was collected 66 hr after the intravenousinjection of retinol-14C and was immediately fractionatedby the method of Cohn et al. (16).

1, bottom). It was estimated that gel filtration re-sulted in an approximately 8-fold purification ofretinol-binding protein, compared to Cohn frac-tion I + III.

The radioactive, fluorescent fractions from theSephadex G-200 column were pooled and the pro-tein chromatographed on a column of DEAE-Sephadex (see Fig. 2). In the experiment shownin Fig. 2, 98% of the 14C applied to the columnwas recovered in a single, sharp peak which waseluted after the peak of serum albumin and cerulo-plasmin. The column fractions comprising the "C-containing peak were strongly fluorescent. Thesefractions were combined and the protein rechro-matographed on a smaller column of DEAE-Sephadex (see Fig. 3). This resulted in the elu-tion of a single sharp peak of fluorescence, of pro-tein, and of radioactivity. In the experiment shownin Fig. 3, 93 % of the 14C applied to the columnwas recovered from the column, and of this 97%owas recovered in the single peak shown. The frac-tions comprising this peak (fractions 42-52 inFig. 3) were combined, dialyzed against water,and lyophilized. The resulting preparation wascalled the "D" preparation. It was estimated thatthe D preparation contained retinol-binding pro-tein purified approximately 600 to 700-fold, ascompared to whole plasma.

The D preparation. Analysis of the D prepara-tion by polyacrylamide disc gel electrophoresis(Fig. 4) revealed the presence of two major pro-tein components: one with prealbumin mobilityand one with al mobility. In addition, there weretwo minor components (one with a2 mobility,considered to be ceruloplasmin, and one with a,mobility, migrating just anodal to the major a,band), and three trace components. The major a,band was strongly fluorescent; none of the othercomponents was visibly fluorescent. This findingindicated that the protein-bound retinol of theD preparation was present in association with themajor a, component.

Confirmatory evidence that retinol circulates inplasma bound to a protein of a, mobility was ob-tained by paper electrophoresis and by zone elec-trophoresis on Pevikon. A portion of the D prepa-ration from a subject injected with retinol-14Cwas analyzed by paper electrophoresis, togetherwith a sample of whole plasma as a referencestandard on an adjacent paper strip. After elec-


.8 - . FRACTION I + m

S.1. / @v 9P~q/ g 6007

26 1000 <O40020 40 6lO4

35 .

3 lo040I I EcF vG f n 200(l2 o g al

Intheexperient shown in he upper pane, th ape 8 'lvoue e

a

0~ ~ ~A

coum 40 60inenldaee 80 cm0in 140thEltinwa0arid u

35-FRACTION llE

4P0.30-~~ ~ ~ ~ ~ ~ ~~~0~

25-

('Jhr Frcin of 9mleceecletd I hxeietsoni

of20 zc

OD <

colmn -h 2bobacofecwrcina8 m a esrdsa

<l t0

CID5o

Flso 1 Gel fira o Chne Fracen I+r (t), or Fracine for

lyze ovrih gis h aebfe cnann al. oeltr

Inurethneexermntdsowutainletlihe uppwer panel,thbampes87m volumsene, de-

colunt 5 cmtinted X 1 cm in lhu c doFIGUEluIo was 27-30tio mohr aond Fractions of20-2 mlteach wereFrctollected

(bto) ncolumn.TheSeabsorbance0ofheeachnfractiona 8 a measurdissoaneest miimatvoueofth poticntn of tphshefaction thfer soli curve thuws show

Inthe eluinxfperoteint fromnithupepal,thclm.eml portion of7the fraction were

aolsonassye fotr al (uppeer pael brokemn curve), tor were examrined fortfluorscene under ulophtrvoet light (oloweripnel bars; fluoraescenc semi-4

quantitaivl easti730matedro andscaleiofsof200-3+). wrecllctd


1

'K. T. I'I. r\NA ..A

10 20 30 40

FRACTION

5 60

NUMBER

70 80 90 100 110

FIGURE 2 Diethylaminoethyl (DEAE) -Sephadex chromatography after Cohn fractionation and gel filtration.Plasma (1700 ml from six donors, one of whom had been injected with retinol-P'C) was subjected to Cohn frac-tionation and gel filtration on Sephadex G-200. The radioactive, fluorescent fractions from the G-200 column were

pooled and the protein precipitated with ammonium sulfate to 90% saturation. The precipitate was dissolved inNa phosphate buffer, 0.02 mole/liter, pH 7.6, and dialyzed against this buffer for several hours. The entire sam-

ple (volume 110 ml) was then applied to a column of DEAE-Sephadex, 2.7 cm (I.D.) X 58 cm (bed volume 330ml) packed with about 15 g of DEAE-Sephadex which had been extensively washed with the above phosphatebuffer. Elution was carried out with 0.02 M Na phosphate buffer, pH 7.6, with a linear gradient of NaCl from0 to 0.6 mole/liter. Fractions of 20 ml each were collected at a flow rate of 25-40 ml/hr. The fractions were as-

sayed for protein by absorbance at 280 myA (solid curve), and small portions were extracted and assayed for 14C(broken curve). It was estimated that the radioactive peak (fractions 74-81) was eluted at a NaCl concentra-tion of 0.35-0.4 mole/liter. The relatively large protein peak in the center of the figure (fractions 55-60) con-

sisted of serum albumin, some of which had not been removed during Cohn fractionation. Ceruloplasmin was

eluted as a small shoulder (fractions 62-67) at the bottom of the descending limb of the albumin peak.

trophoresis the strips were dried in the dark un-

der warm nitrogen. The strip containing the Dpreparation was cut lengthwise into two sections(one-fourth and three-fourths the width of thestrip), and the smaller section was stained forprotein. Bands of protein were seen at positionscorresponding to prealbumin and to proteins of a,mobility (as determined by reference to the stripcontaining whole serum); a faint protein bandwas also seen between the two major protein

bands. The unstained section of the strip was thenaligned with the stained section and was cut intofour parts, corresponding respectively to prealbu-min, albumin, a-proteins, and the remainder ofthe strip (including the origin). The parts were

extracted with warm CHCl3-CH8OH and the ex-

tracts assayed for 14C. 88% of the 14C applied tothe strip was recovered, and 68%o of this was

found in the a zone. The remaining small amountof 14C was distributed throughout the rest of the


9 r

8 -

0-%

i 7

,1 6E0

005CMj I.

4wz

m

cr0(I)m

5000 -

-1

4000z0

0

3000 crILl

31-

2

wa.

0

2000

E0.

1000 0

E0CMcm 3.0

LLI

z

miom

1.0

0 10 20 30 40 50 60 70

FRACTION NUMBER

paper strip, consistent with the conclusion thatit represented the occurrence of some smearingof protein during paper electrophoresis.

Zone electrophoresis was carried out with 8 mlof whole plasma, drawn from the first subject 35days after he had been injected with retinol-14C.Fig. 5 shows the results of this experiment. 92%oof the 14C applied to the Pevikon block was recov-ered in the albumin-a-protein area. The position ofthe peak of radioactivity was slightly behind(cathodal to) the major protein peak (of serumalbumin), and this position corresponded to thatexpected for a,-proteins in this system.

These studies therefore demonstrated that, inthree different kinds of electrophoretic systems,plasma retinol was found in association with aprotein of a, mobility. This was true both for wholeplasma and for the approximately 650-fold puri-fied D preparation, and was indicated both byfluorescence observations and by 14C assay forretinol.

Fig. 6 shows the ultraviolet absorption spectrumof the D preparation (dashed curve). The spec-trum had a peak at 330 mp, as well as the usualprotein absorption peak near 280 mju. Since retinolitself in solution in organic solvents (e.g., ben-zene, hexane) has its peak of absorbance at or

Z FIGURE 3 Repeat DEAE-Sephadex chromatog-O raphy. The fractions comprising the radioactiveo, (and fluorescent) peak from the first DEAE-< Sephadex column (Fig. 2) were pooled, dialyzedLAE against water, and lyophilized. The dried pro-XBr tein was dissolved in 15 ml of 0.02 M Na phos-w phate buffer, pH 7.6, and applied to a column ofa. DEAE-Sephadex, 1.6 cm X 22 cm (bed volumeE about 60 ml), containing about 2 g of DEAE-O Sephadex. Elution was carried out with the same

Na phosphate buffer, with a convex gradient ofo) NaCl from 0 to 0.44 mole/liter (using a constant

volume mixing chamber of 250 ml). The elutionrate was 20-25 ml/hr, and fractions of 6.3 ml eachwere collected.

wP D R

y

/3a

a b

PA

.... .',~~~~~ ~ ~~~~~~~ ~ ~~~~~~~~~~......

........... ..

I ~ ~~~~~4

FIGuRE 4 Polyacrylamide disc gel electrophoretic analy-sis of samples of (from left to right) (a) whole plasma(WP); (b) the D preparation (D); and (c) the RBPpreparation (R). (See Results for further details.)After electrophoresis, the gels were examined under ul-traviolet light and the positions of fluorescent bands pre-cisely determined. The gel was then stained for protein.In each of the two right-hand gels shown, a stronglyfluorescent band was observed before staining at exactlythe position of the major al band.


- A^

70

0

0u

z0

LA.

wa-0

zGiiI--0it

a.

10 15 20FRACTION NUMBER

near 330 mju, our initial assumption was that the330 m/A peak seen with the D preparation repre-sented the absorbance of the protein-bound retinol.Evidence for the validity of this assumption wasobtained by the finding that further purification ofretinol-binding protein (see following text) wasassociated with an enrichment of the 330 mu ab-sorbance (relative to 280 mAabsorbance) identi-cal with the extent of RBP purification as moni-

0.8

0.7

0.6

wo 0.5z

m 0.4cr0(0 0.3m

I

/

z00-

E0.0

,o FIGURE 5 Zone . electrophoresis onPevikon of whole plasma from asubject injected intravenously withretinol-'C. (See Results for furtherdetails.)

tored by 14C assay. Proof of this assumption wassubsequently obtained by the development of amethod for the extraction of retinol from RBPinto heptane without denaturation of the protein.During the course of this extraction the 330 m1uabsorption peak steadily diminished, whereas the280 mJU peak remained unchanged. In preliminaryexperiments it has also been possible to add retinolback to the extracted RBP, thus showing that the

I \

FIGURE 6 Ultraviolet absorptionspectra of the D preparation (dashedline) and of purified holo-RBP(RBPh) (solid line). (See Resultsfor further details.)

WAVELENGTH (mP)


//

protein is capable of reassociating with addedretinol. The details of these methods, togetherwith the results of experiments carried out withthese methods, will be reported in a subsequentcommunication.

Preparative polyacrylamide gel electrophoresis.Further purification of RBP was effected by pre-parative polyacrylamide gel electrophoresis (seeFig. 7). In the experiment shown in Fig. 7 theelution of protein, and of protein-bound retinol,were determined, respectively, by monitoring theeluate for ultraviolet absorption at 280 mp and at330 mMA. Two major, and well-separated, peaks ofprotein were obtained after electrophoresis. Thefirst peak consisted of prealbumin (as determinedby disc gel electrophoresis). The second (a,)protein peak contained 44% of the recovered pro-tein (280 mp absorption) and 93%b of the re-covered retinol (330 mp absorption). All of thevisible (green) fluorescence was found in thefractions comprising the second peak. In this ex-periment the total recovery of applied material, inthe entire effluent from the gel, was 98%o for pro-tein (280 mp absorption) and 89% for retinol(330 mp absorption).

The D preparation derived from the experi-ment shown in Fig. 3 was subjected to preparativegel electrophoresis in three batches. The averagerecovery of applied 14C in the second (a,) peak was77%o; virtually all of the 14C recovered in the totaleffluent was found in this peak.

The fractions comprising the second electro-

- 0.7 -

0.-E0

O.-o' 05

04 I20 0.3

02z

0co)co4

0 50

phoretic peak (180-290 ml effluent volume in Fig.7) were pooled and the protein then chromato-graphed on a column of Sephadex G-100 (seeFig. 8). The major peak obtained after chroma-tography contained all of the visible fluorescenceand almost all of the protein-bound retinol (asmonitored by 330 mu absorbance). The fractionscomprising this peak (fractions 16-19 in Fig. 8)were pooled, and the resulting protein preparationwas designated RBP.

The extent of purification of retinol-bindingprotein during preparative gel electrophoresis andsubsequent Sephadex G-100 chromatography wasestimated from the observed ratios of absorbance(330 mu/280 mju) in each preparation. For theexperiments shown in Figs. 7 and 8, this absorb-ance ratio was 0.35 for the D preparation, 0.65 forthe second peak (pooled fractions) after gel elec-trophoresis, and 0.80 for the RBPpreparation. Itwas thus estimated that retinol-binding proteinwas 2.3-fold purified in the RBP preparation, ascompared to the D preparation, and that the RBPpreparation contained retinol-binding protein puri-fied approximately 1500-fold, compared to wholeplasma. These estimates of purification are mini-mal estimates, since retinol is an unstable mole-cule, and since small amounts of protein-boundretinol are lost during the handling involved ineach purification procedure.

RBP: separation of two components. Analysisof the RBPpreparation by disc gel electrophoresisrevealed the presence of two close bands with a1

FIGuRE 7 Preparative polyacrylamide gelelectrophoresis of the D preparation. The gelcolumn was 9.5 cm high (about 120 ml, vol-ume). Fractions of 12 ml (0-180 ml and be-yond 290 ml, effluent volume) or of 7 ml(180-290 ml, effluent volume) were collected.

450EFFLUENT VOLUME(ML)


I 3.5

2 3.00

- 2.50

A 2.0zoE

0'14

wz

0CD)

k~

1.0 _

0.5 _

010

ol

_1~O0

5 10 15FRACTION NUMBER

20

FIGURE 8 Sephadex G-100 chromatography after pre-parative gel electrophoresis. The sample from gel elec-trophoresis was dialyzed against water for several hoursand then lyophilized. The dry sample was dissolved in 5ml of 0.05 M K phosphate buffer, pH 7.4, which was also0.2 M in NaCl, and applied to a column 1.6 cm (I.D.) X80 cm (bed volume 155 ml) in size. Elution was carriedout with the same buffer solution at a rate of 11 ml/hr.Fractions of 7.1 ml were collected and assayed for ab-sorbance at 280 m/A and at 330 miu. The total recovery,after chromatography, of 280 m~uabsorption (sum of 280mju absorbance X volume for all fractions) was 82%o, andof 330 m~u absorption was 76%. Of the recovered ultra-violet absorbing material, the major (RBP) peak con-tained 64% of the recovered 280 m/A absorption and 86%of the recovered 330 m/A absorption.

mobility (Fig. 4, right). The more slowly migrat-ing and major band (the upper band in Fig. 4,right) was strongly fluorescent (before stainingfor protein).

The RBPpreparation was subjected to prepar-ative polyacrylamide gel electrophoresis, in orderto try to separate and isolate each of the twoa,-protein bands. As shown in Fig. 9, this proce-dure substantially separated the two bands, withthe smaller, more anodal band (Fig. 4) beingeluted first, as a small peak, followed by a muchlarger, strongly fluorescent peak, correspondingto the upper band in Fig. 4. The protein-boundretinol (330 m/u absorbance and fluorescence) wasassociated with the major peak (upper band,Fig. 4).

The pooled fractions comprising the second(major and fluorescent) peak from three runssuch as that shown in Fig. 9 were combined andchromatographed on a small column of DEAE-Sephadex. A single major peak was obtained.

The fractions comprising this peak were dialyzedagainst water and then lyophilized. The resultingpreparation was designated RBPh. The fractionscomprising the first peak from the same three runswere similarly treated to produce a preparationdesignated RBPa. The absorbance ratio (330 mu/280 mib) of the RBPh preparation was 0.95. Thehighest absorbance ratio observed on any of thefractions at the center of the RBPh peak, duringthe three electrophoretic runs, was 1.05.

Analysis of the RBPhand RBPapreparations bydisc gel electrophoresis (Fig. 10) confirmed thefact that RBP. consisted almost completely of theprotein corresponding to the lower band of theRBP preparation, whereas RBPh was greatly en-riched with the protein corresponding to theupper band of the RBP preparation.

The amino acid composition of each of the twoa,-protein preparations (RBP. and RBPh) wasdetermined. The compositions of the two prepara-tions were virtually identical (see Table II). Inaddition, the two a,-proteins had identical mo-bilities on columns of Sephadex G-100 and of

4

E 0.30

~ 0.2

E0GoN

wz

m

m

5 10FRACTION NUMBER

FIGURE 9 Preparative polyacrylamide gel electrophore-sis of the RBP preparation. Approximately 9-10 mg ofthe RBP preparation was subjected to electrophoresis inthe usual way. Fractions of about 6 ml each were col-lected and assayed for absorbance at 280 mu and at 330m/A. The total recovery of 280 my absorption was 86%,and of 330 my absorption was 83%. In a second run, withsimilar separation, the recoveries of ultraviolet absorbingmaterial were 85 and 77%, for 280 m/.& and 330 m/& ab-sorbance, respectively.


X X T. 0, fry .4~E-r-zSF

-. .

R....

N...) '- :̂.

(+)

FIGURE 10 Polyacrylamide disc gel elecanalysis of (from left to right): (a) the RBtion (R); (b) RBP. (R.); and (c) RBPhRBPa and RBPh preparations were derivedRBP preparation shown on the left.

DEAE-Sephadex. These data stronglythat the two a,-proteins were in fact iderthat both represented RBP, but that tfluorescent protein (RBPh) consistedRBP (RBP containing bound retinol)the nonfluorescent protein (RBPa) coiapo-RBP (RBP without bound retinol)

RBP: amino acid composition. Theamino acid analysis of RBPh are indicatble II. Histidine was the least abundaacid residue, comprising 1.05% of the tber of amino acids (on a FLmolar basissigning values of 1, 2, or 3 for the numitidine residues per RBP molecule, it vlated that the corresponding molecularRBP would be in the range of about 11000, or 33,000, respectively. Since theweight, as estimated by other methodslow), appeared to be near 21,000, thisthat the RBPmolecule must contain tw4residues, together with the numbers ofor each of the other amino acids as inthe last two columns of Table II.

The molar ratio of tyrosine to trypt(determined spectrophotometrically on a

R h the RBPa preparation, was 1.95. This indicated

7777 that RBPmust contain four tryptophane residuesper molecule, since the amino acid analysis indi-cated the presence of eight tyrosine residues permolecule. On a weight basis, it could then be cal-culated that tyrosine makes up approximately 6.1%oand tryptophane approximately 3.5% of theweight of the RBP molecule. These values indi-cate that RBP contains an unusually high con-tent of aromatic amino acids.

RBP: immunodiffusion. The RBP preparation(tested as a solution of 0.34 mg/ml) did not re-act with commercial antisera against whole humanserum, prealbumin, albumin, a,-acid glycoprotein,or ceruloplasmin. In contrast, the D preparationreacted strongly with anti-prealbumin antiserumand also reacted with antisera against albuminand ceruloplasmin. Three distinct lines of im-munoprecipitation were seen when the D prepa-ration was tested against anti-whole human

trophoretc serum.P prepara-

T(Rk). The The sensitivity of the commercial antiserumfrom the against prealbumin was estimated by testing this

antiserum opposite serial dilutions of purified pre-albumin prepared from the D preparation by pre-

suggested parative polyacrylamide gel electrophoresis (as initical, and Fig. 7). The antiserum reacted strongly with thethe major purified prealbumin. A line of immunoprecipitation

of holo- was obtained with a prealbumin concentration ofwhereas 0.0064 mg/ml, but not with a prealbumin concen-

nsisted of tration of half this much. Since the RBP solu-tion of 0.34 mg/ml did not react with anti-pre-

results of albumin antiserum, this indicated that the RBPted in Ta- preparation could not contain as much as 1.9%oint amino (by weight) of prealbumin.total num- RBP: molecular weight and sedimentation ve-). By as- locity. Sedimentation velocity studies demon-ber of his- strated that RBPh migrated as a single, homo-vas calcu- geneous component, with a sedimentation con-

weight of stant (520,w) of 2.26S (for a 1 mg/ml solution).1,000, 22,- The molecular weight was estimated, by sedimen-molecular tation equilibrium analysis, to be 20,800 + 2000

(see be- (mean ± SD). The molecular weight of RBP was

indicated also estimated by chromatography of a portion ofo histidine the RBPpreparation on a standardized column off residues Sephadex G-100 (Fig. 11, solid line). This studydicated in indicated that the molecular weight of RBP was

approximately 20,400 (assuming ideal behavior)phane, as in this chromatographic system). Finally, fromportion of the amino acid composition shown in the last


-i

.1.1

'4"W 0

I -

TABLE I IThe Amino Acid Compositon of RBP*

%distribution of Corrected jumoles of each Estimated No.observed umoles %distribu- amino acid, of residues

tion of relative to per RBPAmino acid RBPa RBPh Observed emoles X 102 jsmolest histidine§ molecule

24 hr 48 hr 72 hr

Lysine 5.73 5.75 8.95 9.17 9.08 5.51 10.50 10-11Histidine 1.06 1.10 1.70 1.75 1.74 1.05 2.00 2Arginine 8.01 8.04 12.70 12.40 12.40 7.59 14.46 14-15Aspartic acid 15.80 15.70 25.10 25.10 25.60 15.35 29.24 29Threonine 5.07 5.07 8.04 7.79 7.89 5.01 9.54 9-10Serine 6.17 5.88 9.36 9.09 8.91 5.93 11.30 11Glutamic acid 10.70 10.60 17.90 18.20 18.10 10.97 20.90 21Proline 3.21 3.05 4.73 4.77 5.12 2.96 5.64 5-6Glycine 6.53 6.41 10.30 10.50 11.10 6.46 12.30 12Alanine 7.65 7.63 12.30 12.60 12.60 7.59 14.46 14-15Half-cystine 2.67 2.63 4.47 5.16 4.76 2.92 5.56 5-6Valine 6.41 6.20 11.10 11.70 12.40 7.45 14.19 14Methionine 1.85 2.08 2.96 2.91 3.09 1.81 3.45 3-4Isoleucine 1.40 1.36 2.80 3.30 3.60 2.17 4.13 4Leucine 7.79 7.85 12.40 12.10 12.40 7.47 14.23 14Tyrosine 4.16 4.50 7.22 6.79 6.86 4.22 8.04 8Phenylalanine 5.78 6.13 9.39 8.96 9.04 5.55 10.57 10-11

100.00 100.00 161.40 162.30 164.70 100.00 190.50 185-193

* Two studies were carried out at an interval of 5 months. In the first study (first 2 columns of numbers) samples ofRBPaand RBPhwere hydrolyzed at the same time, for 22 hr, for the purpose of comparing the amino acid composition ofthe two preparations. In the second study (remainder of columns), portions of RBPhwere hydrolyzed for 24, 48, or 72 hr,in order to more fully define the amino acid composition of RBP.$ The corrected percentage distribution lists the average of the three values observed for each amino acid (for 24, 48, and72 hr hydrolysis) except for: (a) threonine and serine, in which the observed values were extrapolated back to zero time,and this value was used; (b) valine and isoleucine, in which the 72-yr values were used. The final values were adjustedso that their sum equals 100%.§ Relative to histidine which was assigned the value of 2.00 (see Results).

column of Table II (and including the four tryp-tophane residues estimated to be present per mole-cule) it was calculated that the molecular weightshould be in the range of 21,600-22,600. Sincethe results of the different methods for esti-mating the molecular weight of RBP agreedquite well, it seems fairly secure that RBP has amolecular weight of approximately 21,000-22,000.

RBP: spectral studies; retinol content. Theabsorption spectrum of the RBPh preparation isshown in Fig. 6, solid curve. The spectrum hadtwo peaks, of nearly equal height.

The retinol content of the RBPpreparation wasdetermined by extraction and chromatography of1 ml of a solution of RBPwith a peak absorbanceat 330 mu of 0.415. The benzene solution of theextracted retinol was adjusted to the same volumeas that of the RBP solution which had been ex-

tracted (1 ml). This benzene solution had a typi-cal retinol absorption peak at 330 mpe, with an ab-sorbance at 330 m/u of 0.372 (corrected for con-trol values as indicated under Methods). Whencorrected for the 89.9% recovery of retinol ob-tained with standard samples in this procedure,the absorbance of the benzene solution of retinolwas calculated to be 0.414 at 330 mpt. The absorb-ance at 330 miu of the retinol extracted from RBPand in solution in benzene was thus identical withthe absorbance of the aqueous solution of RBPfrom which it had been derived. Since retinol isknown to be quantitatively extracted from plasmaby the method used, this indicates that, at the ab-sorption peak of approximately 330 miu, the molarextinction of retinol bound to RBP (in aqueoussolution) is identical with that of retinol in solu-tion in benzene.


O-O 0-100A---A 0-200

" 6

5 \-m_-

45 5.0LOG MOLECULAR WEIGHT

FIGURE 11 Estimation of molecular weight by gel filtration. A portionof the RBP preparation was dissolved in 0.05 M K phosphate buffer,pH 7.5, also 0.2 M in NaCl, and chromatographed on a column ofSephadex G-100, 1.6 cm (I.D.) X 90 cm in size. Elution was carriedout with the same buffer at a flow rate of 12 ml/hr. The absorbanceof the effluent stream at 280 mg was monitored continuously. A portionof the D preparation was similarly chromatographed on a column ofSephadex G-200. The columns were standardized by chromatographyon them of small samples (2-4 mg) of proteins of known molecularweight. A small amount of blue dextran polymer (mol wt 2 X 10')was added to each sample before chromatography, in order to deter-mine the void volume (Vo). After chromatography, the effluent volume(V.) corresponding to the center of the peak of eluted protein wasmeasured, and the values of V./Vo were then plotted against the logof the molecular weight. The observed values of V./Vo for the RBPand the D preparations are indicated above by the left-hand and theright-hand arrows, respectively; the molecular weights of these prepa-rations were then estimated as shown (20). The standard proteinswere: No. 1, ribonuclease; No. 2, myoglobin; No. 3, chymotrypsin;No. 4, trypsin; No. 5, human serum albumin; No. 6, bovine serum

albumin; No. 7, gammaglobulin.

The E'm at 280 miA was determined by mea-

suring the protein content (18) of solutions ofRBP which had been assayed in the spectropho-tometer. In two determinations, the identical valueof 16.5 was obtained. This value is unusually highand is consistent with the very high content oftyrosine and tryptophane found in RBP. Further-more, it is almost certain that the true value forthe Emat 280 mp is higher than this, since itis known that the extinction coefficient of the

Lowry color reaction is increased by aromaticamino acids (18, 21). The protein content wouldtherefore have been overestimated by the methodas used, i.e., when standardized against bovineserum albumin. In studies with thyroxine-bindingprealbumin, a protein fairly rich in aromatic aminoacids, Oppenheimer, Surks, Smith, and Squef de-termined that the protein level measured by the

Lowry color reaction (standardized against bo-vine serum albumin) should be multiplied by 0.89


'2e2

2.0VeVo

1.51-

1.04.0 5.5

25 r

in order to obtain the true protein concentration(21). Since RBP has an even higher content ofaromatic amino acids than does prealbumin, weestimated that the measured protein levels shouldhave been multiplied by a factor of 0.85 in orderto obtain the correct protein concentration. Thecorrected E'm at 280 mu was thus estimated tobe approximately 19.4.

Finally, from the values of the molar extinctioncoefficient of retinol attached to RBP (the sameas that of retinol in benzene), and the E10%m forRBP (19.4), the weight of protein associated with1 pmole of retinol in RBP could be calculated.In making this calculation we assumed that theabsorbance ratio (330 mu/280 mu) for holo-RBPwas 1.05 (the highest ratio observed in any of theelectrophoretic fractions during preparation ofRBPh). This calculation indicated that in holo-RBP1 jumole of retinol is associated with approxi-mately 22,600 mg of protein, i.e., with approxi-mately 1.0 umole of RBP. This finding suggeststhat the RBP molecule contains one binding sitefor retinol, occupied by a single retinol moleculein holo-RBP. It is, of course, also possible thatRBP contains binding sites for more than onemolecule of retinol, but that these sites were notsaturated in our purified preparation.

RBP: lipid analysis. Analysis of the RBPpreparation for fatty acid and fatty acyl chainsindicated that no fatty acid chains (or FFA)were present, at a level of sensitivity of the methodused which would have detected easily as little asone fatty acyl chain per molecule of RBP. Analysesfor other lipid moieties (e.g., cholesterol) havenot as yet been carried out.

RBP: size reduction during purification. Itwas noted during the course of the purification se-quence described above that the molecular size ofthe retinol-containing protein apparently decreasedduring polyacrylamide gel electrophoresis. Thisfinding is illustrated by the experiment describedin Fig. 12. In this experiment, portions of the Dpreparation, and of the prealbumin and RBP-con-taining peaks obtained after gel electrophoresis ofthe D preparation (as in Fig. 7), were seriallychromatographed under identical conditions onthe same column of Sephadex G-100. Before gelelectrophoresis, the protein and the protein-boundretinol (fluorescence and 330 mu absorption) ofthe D preparation were eluted together in a single

major peak of effluent volume of 48 ml. The posi-tion of this peak was very close to that of humanserum albumin (effluent volume of 51 ml). Aftergel electrophoresis prealbumin was eluted at asimilar effluent volume (51 ml). The RBP ob-tained after gel electrophoresis was, however,eluted much later, at an effluent volume of 72 ml,suggesting that it had decreased considerably insize during the procedure of gel electrophoresis.

In order to quantitate the change in size oc-curring during gel electrophoresis, a portion ofthe D preparation was chromatographed on astandardized column of Sephadex G-200 (see Fig.11). The protein and the protein-bound retinolwere eluted together at an effluent volume cor-responding to a molecular weight of about 83,000.In contrast, as already mentioned the RBP prepa-ration was eluted from Sephadex columns at efflu-ent volumes corresponding to 20,000-21,000.These estimates of size assume, of course, idealbehavior during gel filtration and might be incor-rect if major conformational changes occurredduring the preparation of RBP from the D prepa-ration.

RBP: interaction with prealbumin. Thesefindings suggested that RBP normally circulatesin plasma in the form of a complex which dis-sociates into subunits during gel electrophoresis.Experiments were therefore undertaken to definethe nature of the complex present in the D prepa-ration. In the initial experiments, preparative poly-acrylamide gel electrophoresis was carried out un-der a variety of different conditions, in the hopesof obtaining a protein complex with a, mobility,consisting of more than one RBP subunit. Theseexperiments were completely unsuccessful; underall conditions tested RBP preparations with simi-lar properties were obtained.

The nature of the complex present in the Dpreparation was finally defined by the experimentillustrated in Fig. 13. Purified prealbumin andRBP, both obtained by gel electrophoresis of theD preparation, were mixed together to form threedifferent solutions, with different ratios of pre-albumin to RBP.7 The three solutions were thenserially chromatographed on the same column ofSephadex G-100. As shown in the top panel of

7 Separate studies of the interaction of prealbumin withRBPh and RBP. were not carried out because of thelimited amount of protein available in these preparations.


Fig. 13, when the ratio of prealbumin to RBPwasslightly higher than that present in the D prepara-tion, the two proteins chromatographed togetheron the Sephadex column and were eluted at aneffluent volume of about 96 ml, corresponding toa molecular weight in the range of 85,000 (similarto that seen with the D preparation). In contrast,if RBP had been chromatographed alone on thiscolumn it would have been eluted much later, atan effluent volume of about 140 ml (correspondingto its molecular weight of about 21,000). This ex-periment therefore demonstrated the formation ofa complex between prealbumin and RBP.

E0CoCMj

z0CO

C,)-I.-z

IL

w0wa.

BEFORE

(HSA)

The capacity of prealbumin for complex forma-tion with RBP seems to be quite limited. Thus,when the relative amount of RBP in the mixturewas increased progressively (middle and bottompanels of Fig. 13), the prealbumin apparentlybecame saturated with RBP, and the excess, un-complexed RBP was then eluted at an effluentvolume characteristic of purified RBPalone. Sincethe ratio of prealbumin to RBPin the D prepara-tion lies between the ratios of the solutions shownin the top and middle panels of Fig. 13, it wouldappear that the prealbumin of the D preparationis saturated with respect to RBP. Furthermore,

AFTER

20 40 60 80 100 20 40 60 80 0ooEFFLUENT VOLUME (ML)

FIGURE 12 Gel filtration before and after preparative polyacrylamide gel electrophoresis.A single column of Sephadex G-100, 1.6 cm X 52 cm (bed volume 100 ml) in size, was used forthe four chromatographic analyses shown. Samples of 5-10 mg of protein were each dissolvedin a small volume of 0.02 M K phosphate buffer, pH 7.6, also 0.2 M in NaCl, and then appliedto the column. Elution was carried out with the same buffer at a rate of 7-8 ml/hr. Fractions of3.8 ml were collected. The effluent stream was continuously monitored for absorption of lightof wavelength 280 miu. Panel A, human serum albumin (HSA); panel B, the D preparation;panel C, the prealbumin peak (PA) obtained after electrophoresis; panel D, the RBP peak ob-tained after gel electrophoresis. The major peaks in panel B (at 48 ml) and in panel D (at 72ml) were strongly fluorescent.


00 120 140EFFLUENT VOLUME (ML)

A280 RATIO

(PA/ RBP)

*2.20

1.10FIGURE 13 Gel filtration of prealbumin-RBP mixtures. Three mixtures of identicalvolumes but with different ratios of prealbu-min to RBP were chromatographed seriallyon a single column of Sephadex G-100. Therelative amounts of the two proteins were

estimated from the ratio of the absorbanceat 280 mA& due to prealbumin compared withthat due to RBP, in the final mixture (asindicated on the right). The conditions ofchromatography were similar. to those indi-cated in the legend to Fig. 11. (See Resultsfor further details.)

0.63

since the concentration (in milligrams per milli-liter) of the prealbumin-RBP complex seen in thefractions eluted from the Sephadex column (Fig.13) was only slightly greater than the estimatedlevels of these proteins in whole plasma, it seems

likely that RBP normally circulates in plasma inthe form of a complex with prealbumin.

As a control to the experiment shown in Fig.13, human serum albumin and RBP were mixedtogether in a ratio of approximately twice that ofthe prealbumin-RBP mixture shown in the toppanel of Fig. 13. Chromatography of this mixtureon a column of Sephadex G-100 resulted in theelution of two completely separated peaks, the firstconsisting solely of serum albumin and the sec-

ond of RBP. Hence, no complex formation oc-

curred between serum albumin and RBP. In a

later experiment, a similar mixture of purified a,-acid glycoprotein (22)" and RBP was chromato-graphed on Sephadex G-100; as with serum al-

8 This plasma protein migrates slightly in front ofserum albumin (between albumin and prealbumin) on

disc gel electrophoresis and was purified in our labora-tory by column chromatography and electrophoresis.

bumin, absolutely no complex formation was

discerned between the two proteins. These experi-ments suggest that the formation of a complex be-tween prealbumin and RBP is highly specific forthese two particular proteins.

DISCUSSION

The experiments reported here demonstrate thatretinol circulates in plasma bound to a specificplasma protein, retinol-binding protein. This pro-

tein has al mobility, a molecular weight of ap-

proximately 21,000-22,000, and differs from allpreviously identified plasma proteins. RBP has a

sedimentation constant (s2o,t0) of 2.26S, a hy-drated density greater than 1.21, and contains no

free fatty acid or fatty acyl chains. As isolated,purified holo-RBP contained approximately one

retinol molecule per molecule of RBP, a findingwhich suggests that RBPpossesses a single bind-ing site for one molecule of retinol. From the lev-els of retinol normally circulating in human plasma,we can estimate that the usual level of RBP inplasma is of the order of 3-4 mg/100 ml. Thisestimate indicates that our final RBP preparation


E

0

II

r46

0

Ca,

N)

z

ren

m,

0OD)ai

must have been purified about 1500 to 2000-fold,and confirms our estimate of purification as de-rived from the recovery of protein and protein-bound retinol during each of the purificationprocedures.

Retinol -is a 20-carbon alcohol which resemblescholesterol in many of its physicochemical proper-ties. Cholesterol and other lipids (except for FFA[23] and. lysolecithin [24]) circulate in plasmamainly as part of plasma lipoprotein moleculeswith hydrated densities of less than 1.21 (25).Small amounts of plasma cholesterol (and otherlipids) are also found in association with veryhigh density lipoproteins, of density greater than1.21 (26). On a priori grounds, one might ex-pect the lipid alcohol retinol also to circulate inplasma in association with one or more of theknown plasma lipoproteins. The fact that retinolcirculates instead bound to a different protein,RBP, indicates that the interaction between retinoland RBP must be highly specific. Further char-acterization of this specific interaction may providesome useful information about lipid-protein in-teractions in general. By definition, holo-RBP is alipoprotein, since it consists of a complex betweena lipid alcohol (retinol) and a protein (RBP).

Some information is available about the inter-action between retinol and RBP. In the first place,it is clear that no covalent linkages are involvedin this interaction, since retinol can readily be ex-tracted from RBP by organic solvents. Further-more, it seems likely that the molecular configura-tion of retinol is not markedly different when itis bound to RBP, or when it is in solution in or-ganic solvents. This conclusion derives from thespectral observations that: (a) the position of theabsorption peak of retinol bound to RBP and inaqueous solution (330 mp) was the same (ornearly the same) as that of retinol in solution inorganic solvents; and (b) the molar extinctionat 330 mjA of retinol bound to RBP (in aqueoussolution) was identical with that of free retinol insolution in benzene. In contrast, a fine dispersionof retinol alone in water displays a broad band ofultraviolet absorption below 320 mp with a lowermolar extinction.

RBP isolated from plasma consisted of a mix-ture of two components, both with a, mobility,but which separated from each other on polyacryl-amide gel electrophoresis. The two components had

the same amino acid composition, the same molec-ular size, and identical chromatographic proper-ties. Since the major of these two componentscontained bound retinol, whereas the other com-ponent did not, we tentatively identified thesecomponents as representing, respectively, holo-RBP and apo-RBP. The ratio of holo-RBP toapo-RBP in purified RBP preparations was usu-ally about 4 to 1. Since small amounts of protein-bound retinol were lost during the purificationprocedures, it is likely that much of the apo-RBPfound in the purified product arose during thecourse of the purification. It is, however, alsopossible that a small amount of apo-RBP is nor-mally present circulating in plasma. Further workwill be required in order to answer this question.

The amino acid composition of RBP differsfrom that of any previously reported plasma pro-tein, including the protein moieties of both thehigh density, a-lipoprotein, and the low density,,8-lipoprotein (27-33). The ratio of polar to non-polar amino acids in RBP is not particularly un-usual for a globular protein. RBP is unusual, how-ever, in containing a very high content of aromaticamino acids, with estimated tyrosine and trypto-phane contents of 8 residues/molecule (6.1% byweight), and 4 residues/molecule (3.5% byweight), respectively. The high tryptophane andtyrosine content of RBP is responsible for the es-timated very high value of the extinction coeffi-cient at 280 mu (E1'm = 19.4). The tryptophanecontent and extinction coefficient of RBP areeven higher than the corresponding values for theso-called "tryptophane-rich" prealbumin of plasma,which has been reported to have a tryptophanecontent of 2.5%o (34) or 3.15%o (21), and anE'm at 280 m/A of 14.4 (34) or 13.6 (21). RBPalso contains 10-11 residues of phenylalanine permolecule.

Some inferences can be drawn about the pos-sible structure of the binding site for retinol inRBP. First of all, the retinol molecule contains fiveconjugated double bonds, which together impartsome aromatic character to the molecule. In addi-tion, from crystallographic studies retinol isknown to be a relatively flat, planar molecule, inwhich the trans, conjugated double bonds, and the,8-ionone ring tend to lie within a single plane.Secondly, recent x-ray crystallographic studies ofthe structure of proteins have shown the presence


of clefts in the three-dimensional structure of pro-teins, at their functional centers. Thus, myoglo-bin contains a cleft in its three-dimensional struc-ture, and the heme prosthetic group resides withinthis cleft (35). Lysozyme also contains a cleftin its structure, and the active center of this en-zyme lies within this cleft (36); similar findingswere also recently reported for ribonuclease (37).On the basis of this information about the struc-ture of proteins and of retinol, we suggest thatRBPcontains a thin cleft in its tertiary structure,and that the flat retinol molecule resides withinthe hydrophobic interior of this cleft, with its hy-droxyl group at or near the surface of the pro-tein. Furthermore, since retinol has certain "aro-matic" characteristics, and since RBP contains anunusually high content of aromatic amino acids,it is possible that the aromatic amino acid residuesplay an important role in the structure of theretinol-binding cleft, and in the interaction be-tween retinol and RBP.

It was recently reported by Alvsaker, Haugli,and Laland (5) that vitamin A is present in plasmaattached to the "tryptophane-rich" prealbumin. Intheir studies, these workers chromatographed se-rum on columns of Sephadex G-200 and DEAE-Sephadex and isolated a protein fraction whichwas fluorescent and which contained both vitaminA and prealbumin. This preparation was thusequivalent to our D preparation, which does in-deed contain both prealbumin and RBP. Un-fortunately, the conditions of electrophoresiswhich these workers then employed (pH 4.0 forpaper electrophoresis) were unsatisfactory for theseparation of prealbumin and RBP, and this ledthem to the conclusion that the vitamin A in thepreparation must be bound to the prealbumin.

The preliminary immunodiffusion experimentsreported here demonstrated that RBP did not re-act with commercial antisera against whole humanserum, human serum albumin, prealbumin, ceru-loplasmin, or a,-acid glycoprotein. Studies of theimmunologic properties of RBPand of the RBP-prealbumin complex are in progress.

In plasma, RBP apparently circulates in theform of a complex, together with another, largerprotein with prealbumin mobility. In our studies,the protein-protein complex remained intact dur-ing Cohn fractionation and chromatography on

Sephadex or DEAE-Sephadex columns. The com-plex dissociated into its two different components,however, during electrophoresis. The dissociationof the complex into separated RBP and prealbu-min seemed to be a function of the electric fieldand not of the supporting medium used for elec-trophoresis, since the separation of prealbuminand RBPwas observed during electrophoresis onpolyacrylamide gel, paper, and Pevikon. The com-plex was again formed, however, by mixing to-gether separated prealbumin and RBP. The sizeof the complex was estimated to be in the molecu-lar weight range of approximately 80,000-90,000.The ability of prealbumin to complex with RBPwas quite limited, since the prealbumin was read-ily saturated with RBPby increasing the relativeamount of RBP in the mixture (Fig. 13).

From its electrophoretic and immunologic prop-erties, the prealbumin isolated by us, which in-teracts with RBP, appears to be identical with theso-called "thyroxine-binding" (or "tryptophane-rich") prealbumin in human plasma. This pro-tein has been studied by several groups of in-vestigators (21, 34, 38), particularly with re-gard to its role as one of the plasma transportproteins for thyroid hormone. The molecularweight of prealbumin has been reported to be61,000 (34) and 73,000 (21). This molecularweight range is similar to the molecular weightof the prealbumin isolated by us (as estimated bygel filtration). From the estimated molecularweights of the RBP-prealbumin complex, and ofeach of its components, it appears that one mole-cule of prealbumin can interact with one moleculeof RBP to form a stable complex. The effect ofthe RBP-prealbumin interaction on the bindingand transport of thyroid hormone by prealbuminremains to be determined. It is, of course, also pos-sible that the prealbumin isolated here which in-teracts with RBP is very similar to, but not iden-tical with, the prealbumin involved in thyroxinetransport in plasma.

Retinol transport in plasma thus involves botha lipid-protein (retinol-RBP) interaction and aprotein-protein (RBP-prealbumin) interaction.Each of these interactions appears to serve an im-portant physiological role. Thus, the interaction ofretinol with RBP serves to solubilize the water-insoluble retinol molecule and to protect the un-


stable retinol molecule against chemical degrada-tion. The magnitude of the protective effect is in-dicated by the fact that retinol (bound to RBP)remains intact in whole plasma kept for severalweeks at 4VC, whereas retinol itself is highly un-stable and decomposes rapidly in solution in or-ganic solvents when exposed to traces of oxygenor to light. Similarly, the protein-protein interac-tion clearly serves to protect the RBP molecule,by preventing the glomerular filtration of the rela-tively small RBP molecule, and hence the loss ofRBP in the urine. In addition, it is possible thatthe interaction of RBPwith prealbumin may serveto further stabilize and protect the retinol mole-cule. Further studies are in progress to explorethese questions.

ACKNOWLEDGMENTS

We thank Mr. T. Shiratori, Mrs. J. Vormbaum, andMiss E. Miller for expert assistance in this work. Weare most grateful to Dr. R. Canfield for amino acidanalyses, to Dr. H. Rosenkranz for analytical ultracen-trifugal analyses, and to Dr. J. H. Morse for zonalelectrophoresis and for assistance with immunologicstudies. Hoffmann-La Roche, Inc., Basel, Switzerland,generously provided the retinol-15-1'C.

This work was supported by grant No. AM-05968from the National Institutes of Health. Masamitsu Kanaiand Amiram Raz were trainees under grant No. T1AM-5397 from the National Institutes of Health. De-Witt S. Goodman is a Career Scientist of the HealthResearch Council of the City of New York under con-tract No. I-399.

REFERENCES1. Krinsky, N. I., D. G. Cornwell, and J. L. Oncley.

1958. The transport of vitamin A and carotenoids inhuman plasma. Arch. Biochem. Biophys. 73: 233.

2. Garbers, C. F., J. Gillman, and M. Peisach. 1960. Thetransport of vitamin A in rat serum with special refer-ence to the occurrence of unidentified metabolites ofvitamin A in the rat. Biochem. J. 75: 124.

3. Goodman, DeW. S., H. S. Huang, and T. Shiratori.1965. Tissue distribution and metabolism of newlyabsorbed vitamin A in the rat J. Lipid Res. 6: 390.

4. Glover, J., and R. J. Walker. 1964. Absorption andtransport of vitamin A. Exptl. Eye Res. 3: 374.

5. Alvsaker, J. O., F. B. Haugli, and S. G. Laland. 1967.The presence of vitamin A in human tryptophane-richprealbumin. Biochem. J. 102: 362.

6. Huang, H. S., and DeW. S. Goodman. 1965. VitaminA and carotenoids. I. Intestinal absorption and me-tabolism of 'C-labeled vitamin A alcohol and,8-carotene in the rat. J. Biol. Chem. 240: 2839.

7. Gelotte, B. 1964. Fractionation of proteins peptidesand amino acids by gel filtration. In New Biochemical

Separations. A. T. James and L. J. Morris, editors.D. Van Nostrand Co., Inc., London. 93.

8. Nerenberg, S. T. 1966. Electrophoresis. A PracticalLaboratory Manual. F. A. Davis Co., Philadelphia.

9. Jovin, T., A. Chrambach, and M. A. Naughton. 1964.An apparatus for preparative temperature-regulatedpolyacrylamide gel electrophoresis. Anal. Biochem.9: 351.

10. Oppenheimer, J. H., and R. R. Tavernetti. 1962. Dis-placement of thyroxine from human thyroxine-bindingglobulin by analogues of hydantoin. Steric aspects ofthe thyroxine-binding site. J. Clin. Invest. 41: 2213.

11. Mfiller-Eberhard, H. J. 1960. A new supportingmedium for preparative electrophoresis. Scand. J. Clin.Lab. Invest. 12: 33.

12. Schachman, H. K. 1957. Ultracentrifugation, diffusion,and viscometry. In Methods in Enzymology. S. P.Colowick and N. 0. Kaplan, editors. Academic PressInc., NewYork. 4: 32.

13. Spackman, D. IL, W. H. Stein, and S. Moore. 1958.Automatic recording apparatus for use in the chroma-tography of amino acids. Anal. Chem. 30: 1190.

14. Goodwin, T. W., and R. A. Morton. 1946. Thespectrophotometric determination of tyrosine and tryp-tophane in proteins. Biochem. J. 40: 628.

15. Ouchterlony, 0. 1962. Diffusion-in-gel methods forimmunological analysis. In Progress in Allergy. P.Kallos and B. H. Waksman, editors. S. Karger, Basel.6: 30.

16. Cohn, E. J., F. R N. Gurd, D. M. Surgenor, B. A.Barnes, R. K. Brown, G. Derouaux, J. M. Gillespie,F. W. Kahnt, W. F. Lever, C. H. Liu, D. Mittelman,R. F. Mouton, K. Schmid, and E. Uroma. 1950. Asystem for the separation of the components ofhuman blood: quantitative procedures for the separa-tion of the protein components of human plasma. J.Am. Chem. Soc. 72: 465.

17. Havel, R. J., H. A. Eder, and J. H. Bragdon. 1955.The distribution and chemical composition of ultra-centrifugally separated lipoproteins in human serum.J. Clin. Invest. 34: 1345.

18. Lowry, 0. H., N. J. Rosebrough, A. L. Farr, andR. J. Randall 1951. Protein measurement with theFolin phenol reagent. J. Biol. Chem. 193: 265.

19. Goodman, DeW. S., and T. Shiratori. 1964. Fattyacid composition of human plasma lipoprotein frac-tions. J. Lipid Res. 5: 307.

20. Whitaker, J. R. 1963. Determination of molecularweights of proteins by gel filtration on Sephadex.Anal. Chem. 35: 1950.

21. Oppenheimer, J. H., M. I. Surks, J. C. Smith, andR. Squef. 1965. Isolation and characterization ofhuman thyroxine-binding prealbumin. J. Biol. Chem.240: 173.

22. Schmid, K, J. P. Binette, K. Tokita, L. Moroz, andH. Yoshizake. 1964. The polymorphic forms ofa,-acid glycoprotein of normal Caucasian individuals.J. Clin. Invest. 43: 2347.


23. Fredrickson, D. S., and R. S. Gordon, Jr. 1958.Transport of fatty acids. Physiol. Rev. 38: 585.

24. Switzer, S., and H. A. Eder. 1965. Transport oflysolecithin by albumin in human and rat plasma.J. Lipid Res. 6: 506.

25. Fredrickson, D. S., R. I. Levy, and R. S. Lees. 1967.Fat transport in lipoproteins-An integrated approachto mechanisms and disorders. NewEngl. J. Med. 276:34.

26. Alaupovic, P., S. S. Sanbar, R. H. Furman, M. L.Sullivan, and S. L. Walraven. 1966. Studies of thecomposition and structure of serum lipoproteins. Iso-lation and characterization of very high density lipo-proteins of human serum. Biochemistry. 5: 4044.

27. Scanu, A., and W. L. Hughes. 1962. Further charac-terization of the human serum D 1.063-1.21, al-lipo-protein. J. Clin. Invest. 41: 1681.

28. Shore, V., and B. Shore. 1962. The protein subunitof human serum lipoproteins of density 1.125-1.200gram/ml. Biochem. Biophys. Res. Comm. 9: 455.

29. Levy, R. I., and D. S. Fredrickson. 1965. Hetero-geneity of plasma high density lipoproteins. J. Clin.Invest. 44: 426.

30. Margolis, S., and R. G. Langdon. 1966. Studies onhuman serum fi1-lipoprotein. I. Amino acid composi-tion. J. Biol. Chem. 241: 469.

31. Scanu, A., and J. L. Granda. 1966. Effects of ultra-centrifugation on the human serum high-density(1.063 < p < 1.21 g/ml) lipoprotein. Biochemistry. 5:446.

32. Granda, J. L., and A. Scanu. 1966. Solubilization andproperties of the apoproteins of the very low- andlow-density lipoproteins of human serum. Biochemis-try. 5: 3301.

33. Shore, V., and B. Shore. 1967. Some physical andchemical studies on the protein moiety of a highdensity (1.126-1.195 g/ml) lipoprotein fraction ofhuman serum. Biochemistry. 6: 1962.

34. Schultze, H. E., M. Sch6nenberger, and G. Schwick.1956. Uber ein Praalbumin des menschlichen Serums.Biochem. Z. 328: 267.

35. Kendrew, J. C. 1963. Myoglobin and the structure ofproteins. Science. 139: 1259.

36. Blake, C. C. F., L. N. Johnson, G. A. Mair, A. C. T.North, D. C. Phillips, and V. R. Sarma. 1967. Crystal-lographic studies of the activity of hen egg-whitelysozyme. Proc. Roy. Soc. London Ser. B. 167: 378.

37. Kartha, G., J. Bello, and D. Harker. 1967. Tertiarystructure of ribonuclease. Nature. 213: 862.

38. Purdy, R. H., K. A. Woeber, M. T. Holloway, andS. H. Ingbar. 1965. Preparation of crystalline thy-roxine-binding prealbumin from human plasma. Bio-chemistry. 4: 1888.

39. Kanai, M., and DeW. S. Goodman. 1967. Retinoltransport protein: a "new" plasma lipoprotein. Ab-stracts of the Seventh International Congress ofBiochemistry, Tokyo, Japan. 1056.

40. Raz, A., M. Kanai, and DeW. S. Goodman. 1968.Retinol binding protein (RBP): a newly-recognizedplasma lipoprotein. Federation Proc. 27: 594. (Abstr.)

2044 M. Kanai, A. Raz. and D. S. Goodman

retinol-binding protein: the transport

Documents