THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 265, No. 22, Issue of August 5, pp. 12876-12879,199O 0 1990 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

Purification and Crystallization of a Retinoic Acid-binding Protein from Rat Epididymis IDENTITY WITH THE MAJOR ANDROGEN-DEPENDENT EPIDIDYMAL PROTEINS*

(Received for publication, February 7, 1990)

Marcia E. Newcomer and David E. Ong From the Department of Biochemistry, Vanderbilt University, Nashville, Tennessee 37232

The retinoic acid binding activity in the lumen of the rat epididymis (Ong, D., and Chytil, F. (1988) Arch. Biochem. Biophys. 267, 474-478) has been purified to homogeneity. The protein exists in two forms, one form having an additional three amino acids at the amino terminus. The amino acid sequence of the pro- tein was determined to 20 amino acids and proved to be identical to that of the major androgen-dependent proteins from rat epididymis as deduced from the cDNA sequence. These proteins are thought to play a role in sperm maturation, perhaps, it can be suggested now, by delivering retinoic acid to the sperm. The retinoic acid-binding protein has sequence homology to the serum retinol-binding protein and is predicted to have the same overall fold of the polypeptide chain. The epididymal retinoic acid-binding protein has been crystallized from 39 to 43% saturated ammonion sul- fate, 10 mm Tris, pH 8.0. The crystals are space group P21, with a = 39.4, b = 56.9, c = 65.4 A, B = 105 ’ 16 min.

A number of biological processes require vitamin A, includ- ing vision, the maintenance of properly differentiated epithe- lial tissue, and spermatogenesis. Because male rodents pro- vided with retinoic acid as the sole source of the vitamin are invariably sterile, it has been concluded that only retinol is active in the functions of the vitamin related to spermatogen- esis (1). However, the organ distribution of cellular retinoic acid-binding protein (CRABP)’ suggests that retinoic acid does indeed function in some aspect(s) of spermatogenesis. Both rat testis and epididymis contain CRABP (2). In the testis CRABP is found in the late germinal cells and the maturing sperm, with the sperm retaining some CRABP during transit of the epididymis (the site of sperm maturation) and on into the vas deferens (3). Furthermore, a very specific retinoic acid binding activity has been detected in the lumen of the epididymis (4). The extracellular location of the activity and the apparent molecular mass (18,500 daltons) of the binding activity were not consistent with that of CRABP (15,000 daltons). However, a major androgen-dependent epi- didymal protein has been described which, like the retinoic

* This work was supported by National Institutes of Health Grant HD25206 and BRSG RR05424. Core facilities utilized were supported by NIH Grant HD05797. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be herebv marked “aduertisement” in accordance with 18 USC. Section 1732 solely to indicate this fact.

1 The abbreviations used are: CRABP, cellular retinoic acid-bind- ing protein; CRBP, cellular retinol-binding protein; TTNPB, tetrah- ydrotetramethyl-napthalenyl propenylbenzoic acid; SDS, sodium do- decyl sulfate.

acid binding activity, exists in two forms differing in charge which are located in the lumen (5). The two forms are consid- ered to be the product of a single gene and are referred to as the B and C proteins (6). The amino acid sequence, deduced from the cDNA, places this protein in the cYPr-globulin super family (6, 7). This super family consists of small secretory proteins which bind hydrophobic and/or labile ligands. The x-ray crystallographic structures for three of the members of this family have been reported (8-ll), and despite nominal sequence homology there is striking structural homology among them. The first of the tertiary structures reported was that for the retinol-binding protein from human serum (8). Its structure, which is presumably representative of all the structures for the protein family, is that of an eight-stranded up and down P-barrel which wraps around its ligand. Two other members of the protein super family also bind retinol (P-lactoglobulin (12, 13) and purpurin (14)). Because the major androgen-dependent protein from the lumen of the epididymis is in the same protein family as the retinol-binding protein (with sequence homology less than 20% for this pair), it was suggested that it might be a vitamin A-binding protein (7). It appeared to be a possible candidate for the luminal activity which bound retinoic acid (4). The studies presented here confirm this suggested identity. In addition we report the crystallization of the epididymal protein for an x-ray crystallographic structural determination.

EXPERIMENTAL PROCEDURES

Materials

Rat Epididymis were purchased from Biotrol (Indianapolis, IN). Retinoic acid, buffers, salts, and Lipidex were purchased from Sigma. Imidazole was recrystallized from benzene before use. All other re- agents were used as supplied. DE52 cellulose was from Whatman. [3H]TTNPB was provided by the NC1 Biological and Chemical Prevention Program. Ammonyx LO (lauryldimethylamine oxide) was purchased from Stepan Co. (IL).

Methods

Preparation of Protein-Eighty g of frozen tissue were homogenized in four volumes of Tris acetate buffer (10 mM, pH 8.3) in a stainless steel Waring blender. The cell debris was removed by centrifugation at 20,000 x g for 15 min. The supernatant liquid was filtered through cheese cloth and the oH adiusted to 5.0 with 1 M HCl. Precipitated material was removedby centrifugation at 20,000 X g for 15 min. The supernatant liquid was again filtered through cheesecloth and the pH was adjusted to 8.0 with 1 N NaOH. The extract was concentrated to 20 ml and dialyzed overnight against 10 mM Tris acetate (pH 8.3). Prior to the application of sample onto a column containing Sephadex G-75 (5 x 80 cm), 5 ~1 of the retinoic acid analog [3H]TTNPB (3.5 X -lo mol, 24.1 Ci/mM) in ethanol was added. The column was eluted with 10 mM Tris acetate (8.3) at 1 ml/min. Fractions (100 X 15 ml) were monitored for 1) cpm, 2) Azso, 3) Aw, and 4) fluorescence (excitation 330 nm, emission 470 nm).

The fractions containing radioactivity were pooled and applied to

12876

Epididymal Retinoic Acid-binding Proteins 12877

a DEAE-cellulose column (DE52, 2.6 x 23 cm) equilibrated in 0.05 M Tris acetate, pH 8.3. The column was developed with a linear gradient from 0.05 to 0.33 M Tris acetate, pH 8.3. Fractions were monitored for cpm, AzM, and Aas,,. Two peaks of radioactivity were recovered, pooled separately, and concentrated by ultrafiltration (Millipore PLGC 10,000 NMWL membrane) to approximately 30 ml. The buffer was exchanged by diafiltration (with a minimum of 10 volumes of buffer) for 0.01 M imidazole acetate, pH 6.4, and 2.5 rmol of retinoic acid in 100 ~1 of dimethyl sulfoxide was added. The samples were separately applied to two DE52 columns (2.6 x 23 cm) equilibrated in the same buffer. The sample which had eluted first from the previous pH 8.3 column was eluted with a gradient of 0.01-0.10 M imidazole acetate, pH 6.4. The second column was developed with a gradient of 0.01-0.13 M imidazole acetate, also at pH 6.4. Fractions were monitored for AT80 and Aa5,,. The fractions which contained retinoic acid-binding proteins were identified and examined by both SDS and native gel electrophoresis (Pharmacia Phast gels, 8-25% gradient, native and SDS buffer strips). On the native gel system the two proteins were clearly resolved but had identical migration in the SDS gel system. The appropriate fractions were pooled.

For all ion-exchange columns the total gradient volumes were 550 ml. One-hundred 5-ml fractions were collected.

Determination of the Extinction Coefficient of Bound Retinoic Acid-Fifty ~1 from each of five fractions from the final column run was extracted as follows. The sample was acidified by the addition of 10 ~1 of 1 M HCI and to the solution was added 120 ~1 of absolute ethanol, 480 ~1 of hexane, and 120 ~1 of water. After each addition the sample was vigorously mixed. After final phase separation the organic layer was removed and its UV absorption spectrum deter- mined. As a control, 50 ~1 of an ethanolic solution of retinoic acid at an AzsO comparable to that for the retinoic acid in the protein samples was extracted in the same fashion. Recovery was 98%. The ratio of the optical density at 350 nm for the bound and free retinoic acid was used to calculate the molar extinction coefficient of bound retinoic acid. The extinction coefficient for retinoic acid in hexane was taken to be 44,400.

Protein Assays-Pierce Biochemicals BCA protein assay reagent was used for all protein determinations. Standard curves were calcu- lated with bovine serum albumin, and incubations were for 30 min at 37 “C.

Determination of Stoichiometty of Ligand Binding-Purified pro- tein (2 mgs/ml) was incubated for 30 min on ice with a 20-fold excess of retinoic acid in 0.1% lauryldimethylamine oxide. Subsequently, the sample (100 ~1) was applied to a 2-ml bed volume Lipidex-1000 column. The fractions were monitored for protein and retinoic acid absorbance. The protein eluted in the void volume and free retinoic acid was retained on the gel. A molar extinction coefficient (17,400) was calculated for the protein from the amino acid content, and the ~280 for retinoic acid was determined from the spectrum of retinoic acid in ethanol (c = 5900). The observed AzsO was then corrected for the contribution of the retinoic acid before calculating the ratio of protein to ligand.

Crystallization and Space Group Determination-Conditions for crystallization of the purified apoprotein were scanned in hanging drop-vapor diffusion experiments. As is typically the case, various precipitants, buffers, pH values, temperatures and concentrations of protein and precipitants were tried. Plate-like crystals grew in clusters from ammonium sulfate buffered with either 20 mM Tris (8.0) or 120 mM phosphate (6.6) from 39 to 43% saturated ammonium sulfate.

A single piece of crystal was cut from a cluster and mounted in a glass capillary. The capillary was then sealed with wax and mounted on an Enraf Nonius precession camera (75-mm crystal to film dis- tance). The camera was mounted on a Rigaku rotating anode. Nifil- tered x-ray diffraction photographs were taken at 50 mA 180 kV.

Screened precession photographs were taken for zones hk0 and Okl. The two zones were separated by an angle of 74 ’ 44 min on the camera spindle.

RESULTS

These studies establish the identity of the retinoic acid binding activities in the lumen of rat epididymis with the androgen-dependent epididymal secretory proteins, previ- ously described but with no assigned function. To accomplish this the luminal retinoic acid binding activities were purified to homogeneity. To simplify the preparation whole rat epidid- ymis was used. An initial pH 5 extract of the tissue was trace-

labeled with [3H]TTNPB, as previous observations* indicated that these epididymal retinoic acid-binding proteins were also able to bind this potent retinoic acid analog. Retinoic acid was not added at this point in the hope that the proteins might be isolated with endogenous ligand. The concentrated labeled extract was then applied to a Sephadex G-75 column. One can note from Fig. la. that the radioactivity eluted in an asymmetric peak in the last third of the column elution volume. The trailing shoulder on the peak was due to CRABP, known to be present in rat epididymis (2). This peak also contained cellular retinol binding protein (CRBP), also abun- dant in the epididymis (2).

The fractions containing radioactivity were collected and then applied to a DEAE-cellulose column at pH 8.3. The elution profile from this column (Fig. lb) showed resolution of two major protein peaks with bound tritium-label and a third abundant protein. CRABP, as identified by bound reti- noic acid and by migration position on SDS gels, and CRBP, as identified by its characteristic fluorescence spectrum with bound retinol (15), eluted between the two major retinoic acid-binding peaks.

The two peaks identified by bound radioactivity were pooled separately. The samples were applied then to DEAE-cellulose columns at pH 6.4. Prior to application to the column a calculated nominal excess of retinoic acid was added to each sample since no absorbance at 350 nm from endogenous retinoic acid had been detected in the previous separation step. After gradient elution the column fractions were then monitored for both Azso and A350. Identical elution profiles for Azso and Asso (shown in Fig. lc and d) were observed for both samples, demonstrating that retinoic acid coeluted with the major protein peak in each case. This step completed the purification, removing remaining contaminating proteins. The fractions across the retinoic acid-binding peaks showed single bands on both SDS and native gel electrophoresis (results not shown). From 80 g of rat epididymis 25 mg of the first and 37 mg of the second binding protein were obtained.

In order to establish if these proteins were indeed the androgen-dependent epididymal proteins, also called the B and C proteins, material was submitted for blind analysis to an interdepartmental protein sequencing facility. The results of the sequencing of peak 1 were unequivocal up to amino acid 20. As shown in Fig. 2 these first 20 amino acids were identical to those previously established by partial sequencing of the androgen-dependent B protein and consistent with the 166 amino acid sequence predicted by the cDNA sequence, after the putative signal sequence (6). However, the second peak, presumably corresponding to the androgen-dependent C protein, was not blocked at the amino terminus as expected from the previous report (6). Instead, the protein isolated here revealed a sequence identical to the B protein with the excep- tion that it contained an additional three amino acids at its amino terminus. These additional amino acids were identical to those predicted by the cDNA sequence, but previously assigned to the signal sequence. These observations strongly indicate that these proteins are the previously described B and C proteins. They appear to be the product of the same gene but have undergone different processing of the signal sequence or an additional cleavage step to generate the two forms observed here. The established rules for sites of signal sequence cleavage fit both examples here (16). The reason for the C protein being recovered with a blocked N terminus in the previous work is unclear.

In order to determine the stoichiometry of ligand binding, extinction coefficients were determined for the protein moiety

* D. E. Ong, unpublished results.

12878 Epididymal Retinoic Acid-binding Proteins

FIG. 1. Elution profiles for G-75 and DE52 columns. a, the elution of concentrated pH 5.0 epididymal extract on Sepadex G-75, 20 mm Tris acetate, pH 8.3. One-hundred fractions of 15-ml volume were collected. Fluoresence at 470 nm (excitation 330 nm) is indicated by A. Radioactivity (cpm in 100 ~1) is indicated by 0. b, elution of pooled ra- dioactive peak from DE52 at pH 8.3. The column was developed with a linear gra- dient 0.05-0.33 in Tris acetate. Fractions were monitored for AzaO (A), AabO (0) and cpm, 50 ~1 (Cl). c, elution of peak I from DE52 at pH 6.4 with a linear gradient 0.01-0.10 M imidazole acetate. The col- umn was monitored for AZsO (0) and Aah (A). d, elution of peak II from DE52 at pH 6.4 with a linear gradient 0.01-0.13 M imidazole acetate. The column was monitored for AzRO (Cl) and AzsO (A). The gradient volumes for b-d were all 550 ml. One-hundred 5-ml fractions were col- lected.

A"WDF'I'KFLGWB1 - A""KDFDISSFLGFWyBIAFA=K*G -

(C) TEGA"'"KDFDISKFLOWEIAP - (d) . ..C"GLAAGTEGA""KDFDISXFLGF'WEIAFASSMG...

FIG. 2. A comparison of the amino acid sequences of peak I (b) and peak II (c) with the amino-terminal sequence deter- mined for purified androgen dependent B protein (a) from rat epididymis and the deduced amino acid sequence (d) reported by Brooks et al. (6). The dashes indicate the end of the amino- terminal sequencing data. Amino acids which were not determined in the analyses are indicated by *. The gene sequence information extends in both directions.

and bound retinoic acid. Bound retinoic acid was extracted from the complex and a molar extinction coefficient for the bound ligand calculated based on the recovered retinoic acid (determined from its absorbance in hexane). The value ob- tained was 40,000 (compared with c = 44,400 for retinoic acid in hexane). The extinction coefficient for the protein was calculated as 17,400 (El% = 9.4) by summation of the extinc- tion coefficients of the constituent amino acids. The calcu- lated extinction coefficient agrees with that determined em- pirically by comparing the observed Azso with the concentra- tion of the protein as measured by the bicichoninic acid protein assay (Pierce Chemical Co.). The partially saturated protein (100 ~1 of 2 mg/ml) was incubated with excess retinoic acid in 0.1% lauryldimethylamine oxide and allowed to pass through a Lipidex-1000 column (2 ml bed volume, 10 mM Tris 8.0). The Lipidex effectively removes only free retinoic acid from the solution and only protein bound retinoic acid elutes with the protein in the void volume. As the molar excess of retinoic acid was increased from l- to 20-fold (X 1, X 5, X 10, x 20) the amount of retinoic acid eluting with protein in- creased in a hyperbolic manner to a fractional saturation of 0.90. The absorbance at 280 was corrected for the contribution of bound retinoic before calculation of the fractional satura- tion. The UV spectrum for the protein which eluted from the Lipidex is shown in Fig. 3. The absorbance spectra of the two

0.16

0.14

0.12

8 6 0.1

/ A /‘\ I + 0.08 :: II 0.06 0 u 0.04 v a 1

0.02 t u I t I 1 I I

250 300 350 400 wavelength (nm)

FIG. 3. The UV absorbance spectrum of retinoic acid-bind- ing protein (B form) with retinoic acid bound.

proteins with bound retinoic acid were essentially identical. The spectrum is similar to that for serum retinol-binding protein with retinol bound and to that for CRABP with retinoic acid bound in that it is a simple summation of the spectra of the individual components. This is in contrast to what is observed in the spectra of retinol bound to CRBP or CRBP(I1). In both these cases the portions of the spectra which correspond to the ligand are red-shifted with the intro- duction of fine structure (15, 17).

Initial crystallization experiments produced apoprotein (B form) crystals from 39 to 43% saturated ammonium sulfate with 120 mm phosphate buffer at pH 6.6 or 10 mm Tris at pH 8.0. The original phosphate-buffered crystals grew at 4 “C and the Tris-buffered crystals at room temperature. The space group and cell dimensions were determined from precession photographs of the hk0 and Okl zones. The unit cell was monoclinic, @ = 105 o 16 min with a = 39.4, b = 58.9, c = 65.4 A. The corresponding unit cell volume was 146,000 A3. Sys- tematic extinction along the B axis (K # 2n + 1) indicated the space group was P21. The value of V,,, of the crystal with two molecules/asymmetric unit was 1.97 A3/dalton. This value was within the range determined by Matthews (18). Diffraction was observed to 2.1 A resolution. A screened precession (p = 12 “) photograph of Ok1 is shown in Fig. 4.

Epididymal Retinoic Acid-binding Proteins

FIG. 4. A 12 ’ precession photograph (Ok]) of crystals of the epididymal retinoic acid-binding protein (B form).

DISCUSSION

Results here firmly establish that the retinoic acid-binding proteins previously described by Ong and Chytil are identical to the androgen-dependent epididymal proteins B and C. The retinoic acid binding activity had previously been determined to be quite specific as compounds such as retinol and retinal are unable to compete for binding. In addition, retinoic acid analogs which contain altered ring structures, such as aro- matic six membered rings, or five member rings, do not compete, although some of these compounds are effective competitors for the binding of retinoic acid by CRABP (4). Such striking specificity, apparently more stringent than that described for any other known retinoid-binding protein, strongly suggests that the retinoic acid binding activity ob- served in vitro is indeed related to the physiological function of the protein. The failure to observe endogenous retinoic acid bound to the purified protein may simply be due to loss and separation during purification. Both gel filtration and ion- exchange chromatography of purified binding protein com- plexed with retinoic acid led to considerable or complete loss of bound ligand (results not shown). Providing excess retinoic acid in detergent solution to the apoprotein followed by rapid isolation of the protein-ligand complex by filtration on small Lipidex-1000 columns did result in the recovery of a 1:l complex. In contrast the slower and more rigorous steps of gel filtration and ion-exchange chromatography employed in the purification might well allow for dissociation and separa- tion of endogenous liquid from the protein.

Because the B and C proteins are synthesized in response to androgens, they are thought to play a significant role in the maturation of the sperm as it passes through the epidid- ymis. It is interesting to note that these proteins are synthe- sized and secreted by the principal cells of the first portion of the epididymis, the initial segment and proximal caput (5). The principal cells of this region contain the highest levels of CRBP yet observed (19, 20), suggesting considerable vitamin A movement or metabolism occurs in these cells. This could

be related to our postulated function of the B and C proteins: the delivery of retinoic acid to the sperm during epididymal transit.

As mentioned above, these epididymal retinoic acid-binding proteins have nominal sequence homology with the serum transport protein for retinol but are presumed to have similar tertiary structures: the motif of an eight-stranded p-barrel wrapped around the ligand-binding site. Retinol in the bind- ing site of retinol-binding protein is encapsulated by a p- barrel, but the terminal hydroxyl group of the retinol is accessible to solvent and apparently does not participate in critical protein-ligand interactions, perhaps because the tail of the bound ligand is at the surface of the protein. This binding site can also accommodate retinal and retinoic acid, when provided in vitro, with affinities comparable to that for the physiological ligand retinol. In contrast, the epididymal retinoic acid-binding proteins described here do not bind either retinol or retinal. Although these two proteins are likely to have similar folds of the polypeptide chain, they must differ with respect to critical structural determinants for ligand specificity. Exactly what these determinants are will become evident once the x-ray crystallographic analysis of the epidi- dymal retinoic acid-binding protein is complete.

REFERENCES

1. Howell, J. MC. C., Thompson, J. N., and Pitt, G. A. J. (1983) J. Reprod. Fertil. 5, 159-167

2. Ong, D. E., Crow, J. A., and Chytil, F. (1982) J. Biol. Chem. 257, 1338513389

3. Porter, S. B., Ong, D. E., Chytil, F., and Orgebin-Crist, M.-C. (1985) J. Androl. 6, 197-212

4. Ong, D. E., and Chytil, F. (1988) Arch. Biochem. Biophys. 267 (2) 474-478

5.

6.

7. 8.

9.

Brooks, D. E., and Higgins, S. J. (1980) J. Reprod. Fertil. 59, -- 363-375

10.

11.

12.

13.

14.

Brooks, D. E., Means, A. R., Wright, E. J., Singh, S. P., and Tiver, K. K. (1986) J. Biol. Chem. 261, 4956-4961

Brooks, D. E. (1987) Biochem. Int. 14 (2), 235-240 Newcomer. M. E.. Jones. T. A.. Aavist. J.. Eriksson. U.. Rask. L..

and Peterson, P. A. (1984) EMbO i. 7; 1597-1664 Papiz, M. Z., Sawyer, L., Eliopoulos, E. E., North, A. C. T.,

Findlay, J. B. C., Sivaprasadarao, R., Jones, T. A., Newcomer, M. E., and Kraulis, P. J. (1986) Nature 324, 383-385

Holden, H. M., Rypniewski, W. R., Law, J. H., and Rayment, I. (1987) EMBO J. 6, 1565-1570

Huber, R., Schneider, M., Epp, O., Mayr, I., Messerschmidt, A., Pflugrath, J., and Kayser, H. (1987) J. Mol. Biol. 195, 423- 434

Futterman, S., and Heller, J. (1972) J. Biol. Chem. 247, 516% 5172

Monaco, H. L., Zanotti, G., Spadon, P., Bolognesi, M., Sawyer, L.. and Elioooulos. E. (1987) J. Mol. B~ol. 197.695-706

Berman, P., dray, Pl, Chen, E., Keyser, L., Ehrlich, D., Karten, H., LaCorbiere, M., Esch, F., and Shubert, D. (1987) Cell 51, 135-142

15. Ong, D. E., and Chytil, F. (1978) J. Biol. Chem. 253, 828-832 16. von Heijne, G. (1984) J. Mol. Biol. 173,243-251 17. Ona, D. E. (1984) J. Biol. Chem. 259, 1476-1482 18. Matthews, B. W., (1968) J. Mol. Biol. 33,491-497 19. Porter. S. B.. Fraker. L. D.. Chvtil. F.. and One. D. E. (1983)

20. Proc; Natl. Acad. Sci. U. 5’: A. 80,‘65i6-6590 I’

Kato, M., Sung, W. K., Kato, K., and Goodman, D. S. (1985) Biol. of Reprod. 32, 173-189