Purification and Partial Sequencing of Saxiphilin, a …Purification and Partial Sequencing of Saxiphilin, a Saxitoxin-binding Protein from the Bullfrog, Reveals Homology to Transferrin*

THE JOURNAL OF BIOLOGICAL CHEMISTRY [c) 1991 by The American Society for Biochemistry and Molecular Biology, Inc

Vol. 266, No. 23, Issue of August 15, pp. 15481-15487,1991 Printed in lJ S. A.

Purification and Partial Sequencing of Saxiphilin, a Saxitoxin-binding Protein from the Bullfrog, Reveals Homology to Transferrin*

(Received for publication, April 19, 1991)

Yi Li and Edward MoczydlowskiS From the Department of Pharmacology and Department of Cellular and Molecular Physiology, Yale University School of Medicine, New Haven, Connecticut 06510

Plasma from the bullfrog, Rana catesbeiana, con- tains a soluble component of unknown function that specifically binds the neurotoxin, [3H]saxitoxin, with a Kd of -0.2 nM. Saxiphilin, the protein responsible for this activity, was purified -440-fold from bullfrog plasma by column chromatography on heparin-Seph- arose followed by chromatofocusing. The purified sax- iphilin preparation exhibits a binding capacity of 9.6 nmol/mg protein and a Kd of 0.32 nM for [3H]saxitoxin. Analysis of the preparation by sodium dodecyl sulfate- polyacrylamide gel electrophoresis shows a predomi- nant band migrating with an apparent M, of -89,000 which is similar to the expected size of saxiphilin pre- viously estimated by nondenaturing size exclusion chromatography. Amino-terminal sequencing of the -89-kDa protein and sequencing of four different tryptic peptide fragments revealed that each of the partial saxiphilin sequences can be aligned by homol- ogy with members of the transferrin protein family with sequence identity as high as 69%. The available sequence corresponding to conserved residues that comprise part of the two Fe3+ binding sites in lacto- transferrin show several substitutions in saxiphilin, suggesting that saxiphilin is not an Fe3’-binding pro- tein. Saxiphilin appears to be a monomeric -89-kDa protein that is evolutionarily related to transferrin but which binds saxitoxin instead of Fe3+.

Saxitoxin (STX)] is a low molecular weight neurotoxin containing two cyclized guanidinium groups that has been well characterized as a blocker of voltage-dependent sodium channels involved in the generation of action potentials (Hall et al., 1990). A tritiated derivative of STX has been widely used as a radioligand for measuring the density of sodium channels in cell membranes from various excitable tissues (Ritchie and Rogart, 1977). In previous studies of [“HISTX binding in frog tissues, an unusual, high affinity site for [‘HI STX was found in a soluble form in frog heart extracts (Doyle et al., 1982; Tanaka et al., 1984). Our laboratory extended this observation to soluble extracts of bullfrog (Rana catesbeiana)

* This work was supported by Grants AR38796 and HL38156 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore he hereby marked “aduertisement” in accord- ance with 18 U.S.C. Section 1734 solely to indicate this fact.

$ To whom correspondence should be addressed Dept. of Phar- macology, Yale University School of Medicine, 333 Cedar St., New Haven, CT 06510.

The abbreviations used are: STX, saxitoxin; PVDF, polyvinyli- dene difluoride, SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis; Mes, 2-(N-morpholino)ethanesulfonic acid; Mops, 3-(N-morpholino)propanesulfonic acid; NaOAc, sodium ace- tate.

skeletal muscle (Moczydlowski et al., 1988) and further char- acterized the pharmacology and biochemical properties of the soluble STX binding component (Mahar et al., 1991). These latter studies showed that soluble, high affinity ( K d 0.2 nM) binding activity for [3H]STX is present in many nonexcitable bullfrog tissues with the highest concentration found in plasma at 380 2 60 pmol of [3H]STX binding sites per ml. This soluble binding site is extremely specific for STX and its derivatives. For example, another low molecular weight guanidinium toxin, tetrodotoxin, which is a potent competi- tive inhibitor of STX binding to sodium channel proteins (Ritchie and Rogart, 1977) is inactive as an inhibitor of [“HI STX binding to the soluble bullfrog site. Observed structure- activity relationships for binding of various STX derivatives are also unlike that of known sodium channel subtypes (Ma- har et ul., 1991). The soluble [3H]STX binding component elutes with an apparent molecular weight of M, = 74,000 2 8,000 in size exclusion chromatography and exhibits a basic isoelectric point of 10.7 (Mahar et al., 1991). Based on this and other evidence, we have concluded that this soluble [“HI STX binding activity in bullfrogs is due to a protein that is biochemically distinct from sodium channel proteins. The function of this soluble STX-binding protein named saxi- philin is currently unknown, but we have speculated that it may be part of a species-specific detoxification mechanism for STX (Mahar et al., 1991).

In order to elucidate the identity and function of saxiphilin, we have pursued a biochemical approach toward the isolation and structure of this protein. By a combination of chromatog- raphy on heparin-Sepharose and subsequent chromatofocus- ing, we have achieved a -440-fold purification of [‘HISTX binding activity from bullfrog plasma. Electrophoretic analy- sis of such high activity preparations indicates that saxiphilin is a monomeric protein with an apparent molecular mass of -89 kDa. Direct sequence analysis of the amino terminus and four tryptic peptides of saxiphilin reveals that its sequence is homologous to previously described members of the transfer- rin family of proteins (for reviews see Huebers and Finch (1987); Bowman et al. (1988)) which function as the major iron transport protein in higher organisms. The homology of an STX-binding protein to a well known class of iron-binding proteins provides a model for testable hypotheses on the possible function of saxiphilin in frogs. These results also support the possibility that other transferrin-like proteins may exist which have evolved to function as binding proteins of other biologically important toxins or metabolites.

EXPERIMENTAL PROCEDURES

Materiak--[~”H]STX was purchased from Amersham Corp. and was repurified and standardized as described previously (Moczyd- lowski et al., 1988). The repurified [“HISTX had a specific activity of 34 Ci/mmol and a radiochemical purity of 78%. Unlabeled STX was

15481

15482 Purification of Saxiphilin

purchased from Calbiochem. Mops, Mes, Tris, tricaine methanesul- fonate, choline chloride (3 X crystallized) and protein standards for electrophoresis were from Sigma. Heparin-sepharose CL-GB, PBE 94 anion exchanger, and Pharmalyte 8-10.5 ampholytes were from Phar- macia LKB Biotechnology. AG 50W-X2 cation exchange resin was from Bio-Rad. All other chemicals and solvents were reagent or HPLC grade from standard suppliers.

Equilibrium f 'H]STX Binding Assays-For monitoring the activity of column fractions, assay of ['HISTX binding was carried out in the presence of 2.5-5 nM ["HISTX, 0.2 M choline chloride, 0.1 mM EDTA, and 10 mM Mops-NaOH, pH 7.4 (Moczydlowski et al., 1988; Mahar et al., 1991). Fractions from chromatofocusing experiments were assayed in the presence of increased buffer, 80 mM Mops-NaOH, pH 7.4, to ensure constant pH. Aliquots (-5 pl) of various fractions were added to the assay mixture in a final volume of 125 p1 and incubated at 0 "C for 30 min. Free [:'H]STX2+ was separated from bound by passing 100 pl of the reaction mixture through 1-ml columns of cation exchange resin (AG 50W-X2, Tris' form, 100-200 mesh), followed by a rapid wash of 0.5 ml of 20 mM Tris-HC1, pH 7.2, directly into scintillation vials. Preparation of assay columns included a prewash with 1 ml of 10 mg/ml bovine serum albumin in wash buffer. Specific binding is reported as the difference between total and nonspecific binding as determined by duplicate assays in the presence of 10 p~ unlabeled STX. Under standard assay conditions, bound [3H]STX in the assay did not exceed 20% of total ["HISTX to maintain linearity between binding activity and sample protein. Saturation binding curves for Scatchard plot analysis were generated similarly with various total ["HISTX concentrations ranging from 0.1 to 10 nM and duplicate assays of 100-p1 samples. Tritium was quantitated by liquid scintillation counting a t 40% efficiency.

Despite the fact that saxiphilin has a basic isoelectric point (Mahar et al., 1990), the above assay does not underestimate [3H]STX binding activity due to column retention of the saxiphilin protein. This was confirmed by independent determinations of [3H]STX binding to saxiphilin samples by rapid separation of bound and free ["HISTX by gel filtration. The microporous exclusion limit ( M , < 2,700) of the AG 50W-X2 resin prevents retention and adsorption of saxiphilin while effectively binding free ['HIST"'.

Purification of Saniphilin from Bullfrog Plasma-Adult bullfrogs (R. catesbeiana) were purchased from Connecticut Valley Biological Supply (Southhampton, MA). Animals were handled according to guidelines of the Yale Animal Care and Use Committee. Frogs were anesthetized with 30 mg/kg tricaine methanesulfonate by intraperi- toneal injection. Blood was withdrawn from the aortic arch into a syringe containing 0.1 ml of 1 mg/ml heparin. Plasma was prepared by centrifuging whole blood for 3 min a t 1,500 X g in a clincal centrifuge. The yield of plasma was about 10 ml from a large frog. ["HISTX binding activity in samples of plasma that were quick- frozen and stored a t -80 "C was stable for at least 1 year.

The following procedures were carried out a t 4 "C using solutions containing 5 mM EDTA, 1 p~ leupeptin, and 1 p M pepstatin to inhibit proteolysis. A column of heparin-Sepharose CL-GB (1.4 X 20 cm, 30 ml) was pre-equilibrated with 25 mM sodium acetate (NaOAc), 10 mM Mes-NaOH, pH 6.0. A 5-ml sample of bullfrog plasma (21-32 mg protein/ml) was diluted 5-fold with 10 mM Mes-NaOH, pH 6.0, and applied to the column at a flow rate of 1 ml/min. The sample was followed by 200 ml of 150 mM NaOAc, 10 mM Mes-NaOH, pH 6.0, and a linear gradient of 150 ml of 0.3-0.8 M NaOAc in 10 mM Mes- NaOH, pH 6.0. Various fractions (4 ml) were assayed for [3H]STX binding and protein by absorbance a t 280 nm, and the NaOAc gradient was monitored by conductivity.

The pool of ["HISTX binding from the previous step was dialyzed (15-kDa cutoff) against two changes (2.5 h each) of 4 liters of deionized H,O a t 4 "C. The dialyzed sample was applied at 0.4 ml/ min to a 20-ml column (1.4 X 15 cm) of PBE 94 anion exchanger gel that was pre-equilibrated with 25 mM triethanolamine-HCI, pH 10.5. The column was then eluted overnight at 0.4 ml/min with -200 ml of pH 8.0 buffer prepared from 1.8 ml of Pharmalyte 8-10.5 plus 8.9 ml of Polybuffer 96 in a final volume of 250 ml. Various fractions (5 ml) were assayed for ['HH]STX binding activity, protein by absorbance at 280 nm, and pH. The pooled peak of ['HISTX binding activity eluting at pH -8.8 was dialyzed against deionized H20 t o remove Polybuffer and ampholytes before SDS-PAGE and sequencing.

Amino-terminal Sequencing of Saniphilin-A dialyzed sample of purified saxiphilin (-40 pg) from a chromatofocusing experiment was subjected to SDS-PAGE (4 lanes, 10 pg each) and the protein was electroblotted from the slab gel to polyvinylidene difluoride (PVDF) membranes (Immobilon-P, Millipore, Bedford, MA) after the method

of MOOS et al. (1988) as follows. Two PVDF membranes were wet with methanol, rinsed with H20, and soaked in Tris/glycine buffer (25 mM Tris base, 192 m M glycine, pH 8.3, 15% methanol). Directly after running, the SDS-PAGE gel was immersed for a few seconds in Tris/glycine buffer, sandwiched between three layers of Whatman 3 MM paper on one side and two PVDF membranes on the other side and placed in a Bio-Rad Trans-Blot apparatus filled with Tris/glycine buffer, with the PVDF side facing the anode. The gel was electro- blotted for 2.5 h a t a constant voltage of 30 V. The electroblotted PVDF membranes were washed for 3 min in H,O, stained with 0.1% Coomassie Blue in 50% methanol for 3 min and destained in 50% methanol for 3 min, rinsed in H 2 0 for 5 min, air-dried and stored frozen at -20 "C. Approximately 63% of the original 89-kDa band blotted on PVDF was cut out and subjected to gas phase sequencing by Edman degradation using an Applied Biosystems 470A sequenator with on-line detection of phenylthiohydantoin derivatives of amino acids.

Sequencing of Selected Tryptic Peptide Fragments-A purified and dialyzed sample of saxiphilin corresponding to -1 nmol (91 pg by amino acid analysis) was lyophilized to dryness and dissolved in 50 ~1 of 8 M urea, 0.4 M ammonium bicarbonate, pH 8.0. The denatured protein was reduced and alkylated by addition of 5 p1 of 45 mM dithiothreitol, incubation a t 50 'C for 20 min followed by cooling to 22 "C, addition of 5 p1 of 100 mM iodoacetamide, and incubation for 20 min. Trypsin (1.8 pg) was added. The sample was brought to 210 pl with H20 and digested at 37 "C for 24 h. The tryptic digest was then resolved by reverse phase chromatography on a 0.46 X 24-cm Vydac C-18 column (Nest Group, Southboro, MA) a t a flow rate of 0.5 ml/min. The elution gradient consisted of three sequential linear gradients: 2-37.5% B over 0-60 min, 37.5-75% B over 60-90 rnin, and 75-98% B over 90-105 min, where the buffer composition is expressed as the volume percentage of buffer B/(buffer B + buffer A); A = 0.06% trifluoroacetic acid, and B = 0.025% trifluoroacetic acid in 80% acetonitrile. Four of the well resolved peptide peaks were individually subjected to automated gas phase microsequencing as described above.

Other Methods-SDS-PAGE was performed according to the method of Laemmli (1970) using a water-cooled slab gel (0.75-mm thickness) apparatus. Total acrylamide concentration was 7% (%T) with 2.7% bisacrylamide content (%C) in the separating gel. Protein samples for SDS-PAGE were prepared by boiling for 90 s in 62.5 mM Tris-HC1, pH 6.8, 2% SDS, 10% glycerol, and 5% mercaptoethanol. Gels were stained with Coomassie Blue R-250 according to Diezel et al. (1972). RF values and integrated areas of stained bands were measured with the aid of an Isco model 1312 gel scanner. Protein was assayed according to the method of Peterson (1977) or by the bicin- chonic acid method (Smith et al., 1985) after precipitation of protein with deoxycholate and trichloracetic acid as described in Peterson (1977). Bovine serum albumin was used as a standard in the protein assay. Conductivity measurements on column fractions were taken with a Jenway PCM3 conductivity meter after 100-fold dilution of a sample aliquot with deionized water. Water was deionized by a Milli- Q system (Millipore Corp., Bedford, MA).

RESULTS

Plasma prepared from whole blood of adult bullfrogs is a good starting source for the purification of saxiphilin. We have found that plasma concentrations of [3H]STX binding sites in different bullfrogs range from 300 to 400 pmol/ml with specific activities in the range of 9.7-18.5 pmol/mg protein (Mahar et al., 1991). If the protein mass of one STX binding site corresponds to 74 kDa or 13,500 pmol sites/mg protein as suggested by its apparent size on gel filtration columns (Mahar et al., 1991), then a purification factor in the range of -1,000-fold would be required to achieve homoge- neity. Since such a purification factor is within the realm of conventional protein chromatography, we explored various protein separation methods such as gel filtration, ion ex- change, isoelectric focusing, and hydroxyapatite chromatog- raphy. Trial and error evaluation of such techniques led to the following two-step procedure illustrated by the represent- ative experiments shown in Figs. 1 and 2.

Fig. 1 shows the elution profile of a sample of 5 ml of bullfrog plasma fractionated on a 30-ml column of heparin-

Purification of Saxiphilin 15483

froctlon cumber

FIG. 1. Fractionation of bullfrog plasma on heparin-Seph- arose. A 5-ml sample of plasma was chromatographed on a 30-ml column of heparin-Sepharose CL-GB, and 4-ml fractions were assayed for [,'H]STX binding (O), absorbance at 280 nm (0), and conductivity (dashed line) as described under "Experimental Procedures." The large conductivity increase at fraction 105 was due to 2 M NaCl added to wash the column.

40 0 1 0 8 1

ot- p j o 1 0 10 20 30 40 50

fraction number

FIG. 2. Purification of saxiphilin by chromatofocusing. The pool of ["HISTX binding activity from the heparin-Sepharose column in Fig. 1 was dialyzed and chromatographed on a 20-ml column of PBE 94 with a descending pH gradient, and 4-ml fractions were assayed for ["HISTX binding (O), absorbance at 280 nm (0) and pH (dashed line) as described under "Experimental Procedures." The column was washed with 1 N HC1 at fraction 42 to remove adsorbed protein.

Sepharose. In this procedure, all of the initial [3H]STX bind- ing activity was adsorbed to the column when applied at pH 6.0 and low ionic strength. The column was then washed with buffer containing 0.15 M sodium acetate to remove some of the extraneous protein, and the activity was eluted with good recovery (-60%) in a sharp peak with a linear gradient of sodium acetate up to 0.8 M. This column procedure was not particularly effective in terms of increasing the initial specific activity of ['HISTX binding (1.5-4.2-fold range in six differ- ent experiments); however, this step was essential for the removal of contaminating bands seen in SDS-PAGE of the final preparation.

The pooled peak of activity from the heparin-Sepharose step was dialyzed and subjected to chromatofocusing on PBE 94 anion exhanger with a descending pH gradient of pH 10.5- 8.0 as shown in Fig. 2. This step was found to be very effective with a typical increase in specific activity of [3H]STX binding approaching 200-fold over the previous step. The efficacy of this step can be appreciated by the low protein absorbance of the activity peak at pH 8.8 in comparison with the large peak of extraneous protein washed from the column by final addi- tion of 1 N HC1 (Fig. 2).

Table I is a typical purification chart showing a -440-fold increase in specific activity over the starting plasma. In six different experiments the final purification factor ranged from 170- to 690-fold with this procedure, as judged by a routine assay in the presence of 5 nM [3H]STX to evaluate purifica- tion. The final protein sample from the experiment of Table I was also titrated with ["HISTX to obtain the Scatchard plot

TABLE I Purification of P'HISTX binding activity from bullfrog plasma

A 5-ml sample of plasma was subjected to various purification steps and assayed for ['HISTX binding activity and protein as described under "Experimental Procedures." Data are from a representative experiment.

Fraction binding Protein ~~~~~~ Purification

pmol mg pmollmg -fold

Plasma 1720 138 12.5 1 Heparin-Sepharose 1130 36.3 31.1 2.5 Chromatofocusing 986 0.181 5450 436

1

0 2 4 6 0 1 0 Bound (nrnol/rng protein)

FIG. 3. Scatchard plot for [3H]STX binding to purified sax- iphilin. A high activity sample pooled from a chromatofocusing experiment as in Fig. 2 was titrated with 0.1-10 nM ["HISTX at a protein concentration of 0.12 pg/ml in the presence of 80 mM Mops- NaOH, pH 7.4,0.2 M choline chloride, and 0.1 mM EDTA. Nonspecific binding in the presence of 10 p~ STX was subtracted, and the data were fit to a one-site binding equilibrium with Kd = 0.32 nM and 9.6 nmol/mg protein by linear regression.

shown in Fig. 3. This result shows that ["HISTX binding to this sample is consistent with a single class of sites with a Kd of 0.32 nM for ['HISTX and a maximal binding capacity of 9.6 nmol sites/mg protein. This value is similar to the Kd of 0.16 f 0.03 nM for ['HISTX binding to crude plasma found previously (Mahar et al., 1991), implying that the binding site is not drastically altered in the process of purification. The binding capacity of 9.6 nmol/mg determined from the Scat- chard analysis approached the theoretical maximum value expected for a 74-kDa protein (13.5 nmol/mg) that binds one molecule of STX as discussed earlier. Assuming that all of the purified saxiphilin protein in the sample is completely active in binding ['HISTX, we would expect that the protein composition of such a purified sample would consist of at least 70% saxiphilin. Thus, resolution of such samples by SDS-PAGE after the chromatofocusing step should be highly informative regarding the protein composition of saxiphilin.

Fig. 4 shows SDS-PAGE analysis of various fractions from the purification procedure. After a 2.5-fold purification from the initial plasma by the heparin-Sepharose step there are still many contaminating protein bands. In contrast, the pooled fraction from the chromatofocusing column shows only one major band at -89 kDa, with several faint bands. These faint bands never amounted to more than -7% of the staining intensity in preparations purified by this procedure. The apparent molecular weight of the major band was estimated from a standard curve of log M , versus Rr. fit by linear regression. This method gave an average apparent M , = 89 f 5 kDa (+S.D., n = 7). This nominal molecular weight is within experimental error of the expected size (74 + 8 kDa, +S.D., n = 3) of the [3H]STX binding component as previously esti-

15484

A B C D ”

kDa

200 -

116 - 97 -

66 -

45 -

Purification of Saxiphilin

31 - ”” . 1

FIG. 4. Protein composi t ion of var ious f rac t ions as ana lyzed by SDS-PAGE. Lnnc A shows molecular weight standards: myosin (200 kDa), $-galactosidase (116 kDa), rabbit muscle phosphorylase b (97.4 kDa), hovine serum albumin (66.2 kDa), hen ovalbumin (45 kDa), and bovine carbonic anhydrase (31 kDa). The protein samples and specific activity of [ ‘HISTX binding shown in the other lanes are: lnnr H , 40 pg of bullfrog plasma, 1 2 5 pmol/mg; lone C, 40-pg pool from heparin-Sepharose column, 31 pmol/mg; lane 11, 10-pgpool from chromatofocusing column, 5450 pmol/mg. SDS-PAGE using a Yr; acrylamide gel was performed as descrihed under “Experimental Procedures” and stained for protein with Coomassie Blue. The dark smear at the hottom of lane I ) is due to the presence of residual nmpholytes and Polyhuffer not completely removed hv dialysis.

minutes

FIG. 5. Reverse phase chromatography of a t ryps in d iges t of saxiphi l in . A sample (91 pg of protein) from a chromatofocusing experiment was reduced, alkylated, digested with tyypsin, and suh- jected to reverse phase separation on a C-18 column as described under “Experimental Procedures.” The lnrgc p e a f ~ s eluting a t 5-12 min and at 22 min correspond to reagent peaks observed in a control run minus saxiphilin. The peptide paofis marked 97, 101, 127, and 19.7 were sequenced as described under “Experimental Procedures.”

mated by nondenaturing size exclusion chromatography (Ma- har et ul., 1991). Since there are no other observed protein bands in the preparation that approach an expected compo- sition of -70% based on specific binding activity, we conclude that the -89-kDa band is in fact the saxiphilin protein.

To determine the amino-terminal sequence of the protein, we electroblotted the 89-kDa band from an SDS-PAGE gel onto a PVDF membrane and subjected this sample to auto- mated sequencing. The sequence of the PVDF sample (see Fig. 6) could be interpreted to 25 cycles of Edman degradation with an overall yield of 18-23% based on the starting sample loaded on SDS-PAGE and the measured amounts of the 2nd (Pro) and 4th residues (Ala). We also obtained a tryptic map of peptide fragments of reduced and alkylated saxiphilin as shown in Fig. 5 . Four of the well resolved peptide peaks indicated in Fig. 5 ( p e a k 97, 101, 127, and 133) were also subjected to automated sequencing. Each of these peaks gave recognizable sequence information ranging in length from 17 to 31 residues and initial cycle yields ranging from 124 to 452 pmol compared with a starting sample of -1,000 pmol of saxiphilin.

The amino-terminal sequence and the four tryptic peptide sequences derived from saxiphilin were compared with known protein sequences stored in the Protein Identification Re-

source of the National Biomedical Research Foundation using computer software based on the method of Pearson and Lipman (1988). This search consistently detected significant homology of the individual saxiphilin fragments with various members of the transferrin family. No other protein homology was consistently observed for all of the five sequenced saxi- philin fragments. A summary of the saxiphilin/transferrin homology is presented in Fig. 6, which shows an alignment with four previously sequenced transferrin proteins: STF, human serum transferrin (Yang et al., 1984); LTF, human lactoferrin (Metz-Boutique et al., 1984); MTF, human melan- otransferrin or p97 antigen (Rose et al., 1986) and OTF, hen ovotransferrin (Jeltsch and Chambon, 1982).

The complete primary amino acid sequence of the four

. . A P N A K O R . X X X A I f l D L n O K f l X N D L V G

STF(1.24) V P D . K T V - R W C A V S E H E A T K C O S F R D LTF(1-25) G R R R . R S V - O W C A V S O P E A T K C F O W O R

OTF 11-25) MTF 2041) G . M E V - R W C A T S D P E O H K C G N M S E

A P P - K S V l u R W C T I u u u S S P E E K K C N N L R D

. . I G D V I F I V P H T V V F E N T D G K N P A V W A K

STF(ZW.217) LTF(203.220)

G D V A F V K H S T I F E N L A N K . . . . . . .

G D V A F V K H T T V N E N A P D L - . . . . . . OTF(M3-220) G D V A F V K H S T V L E N T D C K . . . . . . . MTF(222-239)

R E S T V F E D L S D E - . . . . . - G D V A F l

STF(528.552) G D V A F V K H O T V P O N T G G K N P D P W A K LTF(552.576) G D V A F V K D V T V L O N T D G N N N E A W A K MTF(568.592 G D V A F R H T T T V F D N T N G H N S E P W A A OTF1535-5591 G D V A F I O H S T V E E N T G G K N K A D W A K -

FIG. 6. Sequence a l ignment of f ive saxiphi l in f ragments wi th four members o f the t ransfer r in fami ly . The amino te r - minus of saxiphilin (SAX-NH2) and that of four tryptic fra, (SAX-133, SAX-127, SAX-97, and SAX-101) were sequenced as

Uments

described under “Experimental Procedures.” Below each saxiphilin sequence, the first four sequences correspond to alignments ofamino- terminal domains and the next four sequences correspond to align- ments of carboxyl-terminal domains of four different members ofthe transferrin family: S‘TF, human serum transferrin (Yangrct nl., 1984); LTF, human lactotransferrin (Metz-Boutique rt nl., 1984); ”I’F, human melanotransferrin or p9’i antigen (Hose et nl., 1986); and 07%: hen ovotransferrin (Jeltsch and Chambon, 1982). I)n.shes are placed as gaps needed to maximize homology. In the SAX sequences, X refers to a uncertainty in the sequence. Bold lrttrrs refer to positions of identity between saxiphilin and transferrins. Howcs are drawn to indicate absolute identity for each alignment. The two orrows at particular positions nhour the SAX-13.3 and SAX-127 sequences note the position of residues in the Fe“ binding site of lactotransferrin as discussed in the text. The single-letter amino acid code is used.

Purification of Saxiphilin 15485

previously sequenced transferrin proteins ranges from 679 to 719 residues in length and is characterized by a 2-fold internal homology corresponding to two Fe3+ binding sites per trans- ferrin molecule. For example, the amino-terminal half (resi- dues 1-338) of human lactotransferrin can be aligned with 37% overall identity with its own carboxyl-terminal half (res- idues 339-703) as previously demonstrated (Metz-Boutique et al., 1984). With the exception of the amino terminus of saxiphilin, we do not know the relative homologous location of our fragments in the NHz-terminal uersus the COOH- terminal half of the transferrins. Therefore, for the tryptic fragments of saxiphilin, Fig. 6 shows a possible alignment with both NHz- and COOH-terminal domains of the various transferrin proteins.

The NHz terminus of bullfrog saxiphilin exhibits maximum homology with the NH2 terminus of hen ovotransferrin, which amounts to 52% identity if we consider only the 21 known residues of the saxiphilin sequence and omit alignment gaps in the homology comparison. The other transferrin members show weak homology at the amino terminus that ranges from 16 to 22% identity. Nevertheless, there are at least 4 residues that are absolutely conserved for all five of the compared proteins within the first 25 NHz-terminal residues as indi- cated by the boxed residues.

The SAX-133 tryptic fragment exhibits an overall homol- ogy preference within the COOH-terminal domain of trans- ferrin that ranges from 35% identity with serum transferrin (391-419) to 41% identity with lactotransferrin (406-440). The SAX-127 fragment exhibits a slight COOH-terminal homology preference, but it could have originated from either an NHz- or COOH-terminal domain of a transferrin-like protein. The highest homology of SAX-127 is 65% identity with ovotransferrin (453-472). The SAX-97 fragment shows an NHz-terminal homology preference at 69% identity for both lactotransferrin (151-170) and ovotransferrin (155-172). The SAX-101 fragment shows a COOH-terminal homology preference at 68% identity with both serum transferrin (528- 552) and ovotransferrin (535-559). In addition to the five saxiphilin sequences shown in Fig. 6, sequencing of the SAX- 97 tryptic fragment also yielded a secondary sequence (not shown) that was followed for 13 cycles of Edman degradation with the sequence Cys-Leu-Lys-Glu-Asp-Met-Gly-Asp-Val- Xaa-Phe-Leu-Met. This partial sequence can also be unam- bigously aligned with the transferrin sequences, e.g. 50% identity with lactotransferrin residues 197-209 and 58% iden- tity with lactotransferrin residues 546-558.

DISCUSSION

This paper describes a relatively simple and rapid method for the purification of saxiphilin, a soluble STX-binding pro- tein from bullfrog plasma. This has allowed us to identify saxiphilin as an -89-kDa monomeric protein that is evolu- tionarily related to members of the transferrin family of Fe3+- binding proteins.

Our measured binding capacity of 9.6 nmol/mg protein is close to the expected value of 11.2 nmol/mg for an -89-kDa protein that binds one molecule of [3H]STX. Assuming that saxiphilin does possess one STX binding site, this agreement is consistent with the apparent purity of the preparation as visualized by SDS-PAGE. However, a variety of errors could affect our quantitation of the binding stoichiometry of STX to the purified protein. These factors may include: 1) the presence of undetected protein impurities including de- natured saxiphilin that has lost the ability to bind [3H]STX, 2) error in absolute protein content of the sample due to relative measurement against a standard of bovine serum

albumin, 3) error in the specific activity of [3H]STX, 4) an incorrect value of the protein molecular weight of saxiphilin as estimated by SDS-PAGE. As possible evidence of one of these problems, we have observed that our present saxiphilin preparation loses [3H]STX binding activity slowly at a rate of 25% loss after incubation at 5 days at 4 “C. This suggests that some of the activity could be lost due to denaturation during the course of purification which may lower the ex- pected binding capacity. Future work will be aimed at stabi- lization of the STX binding activity, more reliable values of protein content by amino acid composition, and an exact molecular weight given by a complete sequence of the protein. This will allow us to rigorously confirm the present finding of 1 STX binding site per molecule that is implied by our present stoichiometry estimate of 0.85 mol of [3H]STX bound per mol of saxiphilin based on the Scatchard plot B,,, of Fig. 3 and an estimated molecular mass of 89,000 daltons.

The surprising homology of the -89-kDa band to members of the transferrin family leads to the question of whether we may have purified bullfrog serum transferrin or a transferrin- like Fe3+-binding protein that fortuitously binds STX. There are several lines of evidence that can be cited to discount these possibilities. Serum transferrin is a major protein con- stituent in the blood of vertebrates. The measured concentra- tion of transferrin in adult bullfrog plasma is 2.6 mg/ml or 34 FM (Valaitis and Theil, 1984). In contrast, the plasma concen- tration of saxiphilin is only 0.38 PM as measured by [3H]STX binding sites or 0.028-0.034 mg/ml, assuming a molecular mass in the range of 74-89 kDa. Thus, the saxiphilin content of bullfrog plasma is 100-fold lower than that of transferrin. Also, the reported isoelectric point of transferrin purified from bullfrog plasma is in the range of 6.3-6.6 (Valaitis and Theil, 1984), whereas saxiphilin has a basic PI in the range of 10.3- 10.7 (Mahar et al., 1991). To our knowledge, transferrin from the particular bullfrog species that we have studied has not yet been cloned or sequenced; however, we can compare our saxiphilin sequence data with that of a full length transferrin clone from another amphibian, the African clawed toad, Xen- opus lueuis (Moskaitis et al., 1990). A similar homology alig- ment comparison with Xenopus transferrin (not shown) as that shown in Fig. 6 for the human and chicken transferrin homologs yields the following homology data for saxiphilin/ Xenopus transferrin residues (SAX/XTF); SAX-NH2 ex- hibits 42% identity with the corresponding Xenopus amino terminus, SAX-133/XTF(403-437) shows 42% identity, SAX-127/XTF(465-485) shows 60% identity, SAX-97/ XTF(l74-193) shows 56% identity, and SAX-l01/XTF(550- 574) shows 76% identity. Surprisingly perhaps, the homology of bullfrog saxiphilin to Xenopus transferrin is no greater than that found for the various human and chicken transfer- rins in Fig. 6. This rather high divergence of two protein family members from related amphibians may also suggest that saxiphilin does not function as an Fe3+-binding protein in the manner of transferrin.

As preliminary evidence of an actual functional difference, we have tested commercially available human serum transfer- rin, human milk (a source of lactotransferrin), and rat plasma (a source of rat transferrin) for the ability to bind [3H]STX under our standard assay conditions and observed no hint of specific [3H]STX binding.2 Likewise, our efforts to demon- strate direct binding of Fe3+ to saxiphilin by spectrophoto- metric assays used for transferrin or by competition between [3H]STX and Fe3+ have also been negative.

The available preliminary sequence comparison between saxiphilin and structurally characterized transferrins also

Y. Li and E. Moczydlowski, unpublished observations.

15486 Purification of Saxiphilin

provides important information on the possible functional relationships of these proteins. Crystallographic analysis of Fe3+-saturated human lactoferrin has revealed that there is one bound Fe3+ ion in equivalent locations in each of the two homologous domains comprised by residues 1-338 ofthe NH2- domain and residues 339-703 of the COOH domain (Anderson et al., 1987). Fe3+ is coordinated by Asp", Tyrg3, Tyr'", and

His6'' in the COOH domain. It has been noted that lactotrans- ferrin, serum transferrin, and ovotransferrin exhibit absolute conservation of these latter residues that comprise the Fe3+ binding site (Anderson et al., 1987). However, examination of the SAX-133 alignment in Fig. 6 shows that a critical aspar- tate residue in one of these sites ( A ~ p ~ ' / A s p ~ ' ~ of lactotrans- ferrin) appears to be replaced by a glutamate in saxiphilin (see arrow at 2nd residue of SAX-133 in Fig. 6). The crystal- lographic analysis of lactotransferrin has also implicated res- idues Arg"' in the NH2 domain and the equivalent Arg477 in the COOH domain as the groups responsible for coordinating two bicarbonate anions. Bicarbonate is known to be essential for the binding of Fe3+ to transferrin and is believed to form a bridge between Fe3+ and arginine in these proteins (Ander- son et al., 1987; Baker et al., 1987). Examination of the SAX- 127 alignment shows that this arginine group is absolutely conserved in lactotransferrin, serum transferrin, and ovo- transferrin but is replaced by an aspartate residue in saxi- philin. Other important residues in lactotransferrin are the hydroxyl groups of Thr"8(Thr473) and Thr'22(Thr478) which appear to contact the bicarbonate anion density observed in the crystal structure (Baker et al., 1987). These groups are conserved as either threonine or serine in lactotransferrin, serum transferrin, and ovotransferrin, but one of them is replaced by an isoleucine residue in saxiphilin at the residue equivalent to Thr'22(Thr478) in lactotransferrin (see the SAX- 127 alignment in Fig. 6). Although the sequence information that is presently available does not allow us to fully compare all of the homologous Fe3+-coordinating residues of saxiphilin, the structural differences and preliminary biochemical evi- dence noted above suggest that saxiphilin does not function in the binding of Fe3+.

Assuming that saxiphilin is a distinct STX-binding protein that is evolutionarily related to the transferrin family, what implications does this structural homology have for the pos- sible function of saxiphilin? Transferrin serves as the major iron transport protein in higher organisms and has been found to be essential for cell growth (Huebers and Finch, 1987; Thorstensen and Romslo, 1990). Iron is delivered to cells by binding of the Fe3+-transferrin complex to a specific transfer- rin receptor protein on the cell surface. The receptor-trans- ferrin complex is then internalized and recycled in a process that has been classically described as receptor-mediated en- docytosis (Dautry-Varsat, 1986; Ward, 1987). Transferrin synthesis has been detected in many different tissues, but the liver has been identified as the major source of plasma trans- ferrin (Idzerda et al., 1986). Although saxiphilin appears to be much less abundant than transferrin in the bullfrog, it also exhibits a widespread pattern of tissue distribution (Mahar et al., 1991). If saxiphilin functions in a detoxification pathway for STX as previously suggested (Mahar et al., 1991), such a system may operate in the fashion of transferrin with the STX-saxiphilin complex binding to a cell surface receptor and being internalized by endocytosis. Once sequestered in- ternally, STX may then be neutralized by some of the known drug conjugation and oxidation pathways. Although such a pathway for STX detoxification is plausible, one might alter- natively speculate that STX or an STX-like compound may

His2S2 in the NH2 domain and by Asp407, Tyr447, T Y ~ ~ ~ ' , and

be an endogenous neuromodulator or a defensive toxin in the skin secretions of certain amphibian species. In this case saxiphilin might function in the delivery and/or sequestration of such a potent and dangerous metabolite.

Additional evidence for the wider biological role of trans- ferrin-like proteins comes from crystallographic studies (An- derson et al., 1987; Baker et al., 1987) which have noted a remarkable structural relationship between human lacto- transferrin and certain bacterial periplasmic binding proteins (Quiocho et al., 1987) which function in the membrane trans- port of sugars, amino acids, and inorganic ions such as sulfate. For example, the tertiary structure of the 310-residue sulfate- binding protein of Salmonella typhimurium was found to have the same overall folding topology as either one of the two homologous domains of lactotransferrin despite a low level of primary sequence homology. This observation and the present case of saxiphilin suggests that transferrin-like proteins may serve as a generalized class of binding proteins that have evolved to bind a variety of small molecules including essential growth factors and toxins. The existence of other transferrin homologs with unique functions is also suggested by the case of melanotransferrin, or p97, a membrane bound protein that is expressed at high levels on cultured melanoma cells (Rose et al., 1986). While the NHz-terminal domain of melanotrans- ferrin appears to contain a functional Fe3+ binding site (Brown et ab, 1982), Baker et al. (1987) have noted that there are several nonconservative substitutions of critical residues predicted to comprise the Fe3+ binding site in the COOH domain. This suggests that the COOH domain of melano- transferrin has lost the ability to bind Fe3+ or has evolved to perform a different function.

The structural relationship of saxiphilin to the transferrin protein family is evident from the strong homology discussed above and the similarity of their respective molecular masses as judged by migration in SDS-PAGE: 89 kDa for saxiphilin and 80-86 kDa for various vertebrate transferrins (Palmour and Sutton, 1971). In light of this finding, our work provides the first evidence for a new member of the transferrin family with two internally homologous domains that binds a different ligand instead of Fe"+. Future studies aimed at the cloning of bullfrog saxiphilin will provide molecular tools to elucidate its function and may also serve as an inroad to further evolutionary relationships within the transferrin superfamily of proteins.

in collaboration with Dr. Ken Williams and Kathy Stone of the Yale Acknowledgments-Amino acid sequence analyses were performed

Protein Chemistry Facility.

REFERENCES

Anderson, B. F., Baker, H. M., Dodson, E. J., Norris, G. E., Rumball, S. V., Waters, J. M., and Baker, E. N. (1987) Proc. Nutl. Acad. Sei. u. s. A. 84, im-1773

Baker. E. N.. Rumball. S. V.. and Anderson. B. F. (1987) Trends . . Biochem. Sci. 12,350-353 '

Bowman, B. H., Yang, F., and Adrian, G. S. (1988) Ada Genet. 25,

Brown, J. P., Hewick, R. M., Hellstrom, I., Hellstrom, K. E., Doolittle,

Dautry-Varsat, A. (1986) Biochimie 68 , 375-381 Diezel, W. G., Kopperschlager, W. G., and Hofmann, E. (1972) Anal.

Doyle, D., Wong, D., Tanaka, J., and Barr, L. (1982) Science 215,

Hall, S., Strichartz, G., Moczydlowski, E., Ravindran, A., and Rei- chardt, P. B. (1990) in Marine Toxins: Origin, Structure and Mo- lecular Pharmacology (Hall, S., ed) pp. 29-65, American Chemical Society, Washington, D. C.

1-38

R. F., and Dreyer, W. J. (1982) Nature 296, 171-173

Biochem. 48,617-662

1117-1119

Huebers, H. A., and Finch, C. A. (1987) Physiol. Reu. 67,520-582 Idzerda, R. L., Huebers, H., Finch, C. A., and McKnight, G. S . (1986)

Purification of Saxiphilin 15487

Proc. Natl. Acad. Sci. U. S. A. 8 3 , 3723-3727 Peterson, G. L. (1977) Anal. Biochem. 83, 346-356

295 Spring Harbor Symp. Qunnt. Biol. 52,453-463 Jeltsch, J.-M., and Chambon, P. (1982) Eur. J. Biochern. 122 , 291- Quiocho, F. A,, Vyas, N. K., Sack, J . S., and Vyas, M. N. (1987) Cold

Laemmli, U. K. (1970) Nature 227,680-685 Ritchie, J. M., and Rogart, R. B. (1977) Reu. Physiol. Biochem. Mahar, J., Lukacs, G. L., Li, Y., Hall, S., and Moczydlowski, E. (1991) Pharmacol. 7 9 , 1-50

Toxicon 2 9 , 53-71 Rose, T. M., Plowman, G. D., Teplow, D. B., Dreyer, W. J., Hellstrom, Metz-Boutigue, M.-H., Jollbs, J., Mazurier, J., Schoentgen, F., Le- K. E., and Brown, J. P. (1986) Proc. Natl. Acad. Sci. U. S. A. 83,

grand, D., Spik, G., Montreuil, J., and Joll&, P. (1984) Eur. J . 1261-1265 Biochem. 145,659-676 Smith, P. K., Krohn, R. I., Hermanson, G . T., Mallia, A. K., Gartner,

Moczydlowski, E., Mahar, J., and Ravindran, A. (1988) Mol. Phar- F. H., Provenzano, M. D., Fujimoto, E. K., Goeke, N. M., Olson, macol. 33 , 202-211 B. J., and Klenk, D. C. (1985) Anal. Biochem. 150,76-85

Moos, M., Nguyen, N. Y., and Liu, T.-Y. (1988) J. Biol. Chem. 2 6 3 , Tanaka, J. C., Doyle, D. D., and Barr, L. (1984) Biochim. Biophys.

Moskaitis, J. E., Pastori, R. L., and Schoenberg, D. R. (1990) Nucleic Thorstensen, K., and Romslo, I. (1990) Biochem. J . 271 , 1-10 Acids Res. 18 , 6135 Valaitis, A. P., and Theil, E. C. (1984) J. Biol. Chem. 259 , 779-784

Palmour, R. M., and Sutton, H. E. (1971) Biochemistry 10, 4026- Ward, J. H. (1987) Inuest. Radiol. 2 2 , 74-83 4032 Yang, F., Lum, J. B., McGill, J. R., Moore, C. M., Naylor, S. L., van

Pearson, W. R., and Lipman, D. J. (1988) Proc. Natl. Acad. Sci. U. Bragt, P. H., Baldwin, W. D., and Bowman, B. H. (1984) Proc.

6005-6008 Acta 775,203-214

S. A. 85,2444-2448 Natl. Acad. Sci. U. S. A. 81, 2752-2756