Secretion of pigment epithelium-derived factor Mutagenic study Hanshuang Shao, Iris Schvartz and Shmuel Shaltiel* Department of Biological Regulation, Weizmann Institute of Science, Rehovot, Israel Pigment epithelium-derived factor (PEDF), a neurotrophic and antiangiogenic protein, is an extracellular component of the retinal interphotoreceptor matrix which has been shown to be secreted by human fetal retinal pigment epithelial cells. It belongs to the serpin superfamily and contains the typical exposed reactive center loop. The function of this loop is still unknown. In this study we used site-directed mutagenesis of the cDNA encoding PEDF to show that (a) truncation of the C-terminal tail (Pro415–Pro418) of PEDF, (b) deletion of the Pro373–Ala380 segment that resides within the reactive center loop of the protein, and (c) alanine substitution of amino-acid residues Asn391–Thr403 located within its hydrophobic core inhibit PEDF secretion, but not its tran- scription, by cells transfected with the various PEDF cDNAs. On the basis of the crystal structure of PEDF, these mutations are presumed to alter the protein conformation, suggesting that conservation of the 3D structure of PEDF is essential for its secretion. In addition, we show that replacement of Gly376 and Leu377 with alanine prevents PEDF secretion. As these two residues are located within the highly exposed segment of the reactive center loop, we pro- pose a novel function for this loop in PEDF. Our results imply that the reactive center loop, specifically Gly376 and Leu377, is involved in the interaction of PEDF with com- ponents of the quality control system in the endoplasmic reticulum, thus ensuring its efficient secretion. Keywords: crystal structure; neurotrophic activity; pigment epithelium-derived factor (PEDF); secretion; serpin. Pigment epithelium-derived factor (PEDF) was originally identified as an extracellular component of the retinal interphotoreceptor matrix and found to be secreted by human fetal retinal pigment epithelial cells [1,2]. It has been shown to be a neurotrophic factor that induces neurite outgrowth in cultured human retinoblastoma Y-79 cells [1,2]. The neurotrophic activity of PEDF was further demonstrated by its ability to promote neuronal survival of the cerebellar granule cells and of developing spinal motor neurons [3], to protect neurons against rapid glutamate toxicity [4], and to inhibit apoptosis induced by hydrogen peroxide of rat retinal neurons [5]. Recently, PEDF was shown to be a very potent inhibitor of neovascularization in a murine model of ischemia-induced retinopathy [6]. The inhibition of neovascularization was associated with endothelial cell apoptosis [6], probably by increasing Fas ligand (FasL) mRNA and surface FasL in these cells [7]. Sequence analysis of intact human PEDF, a 50-kDa glycoprotein of 418 amino-acid residues, shows a high homology to the serpin (serine protease inhibitors) super- family [8]. NMR measurements and analysis of the X-ray structures revealed that all the known serpins contain an exposed reactive center loop (RCL) [9], which is susceptible to cleavage by specific proteases [10]. As a consequence of this cleavage, the RCL of the inhibitory members of the serpin superfamily is inserted into b-sheet A (as a new strand) leading to a transition from a stressed (S) to a relaxed (R) loop structure. Subsequently, the specific protease, attached to the RCL, moves to the base of the serpin, where it is partially unfolded resulting in inhibition of its activity [11,12]. The serpin superfamily also contains noninhibitory members, among which are ovalbumin, angiotensinogen and PEDF [13]. Similarly to ovalbumin and angiotensino- gen, PEDF does not undergo the S fi R transition upon cleavage of its RCL [13]. The crystal structure of human PEDF was recently solved to 2.85 A ˚ and revealed that the RCL is highly exposed [14]. The function of the RCL of PEDF has not yet been resolved. The C-terminal amino-acid residues have been shown to play an important role in the secretion of many proteins [15–19]. Truncation of the four amino-acid residues (391–394) at the C-terminus of a1-proteinase inhibitor (A1Pi) prevented its secretion by cells transfected with the mutant A1Pi cDNA [16]. Proline at position 391 was found to be important for the A1Pi secretion as replacement of Pro391 by various amino acid residues severely restricted its secretion [17]. A possible mechanism for the impaired secretion of the A1Pi variants was suggested to be associated with loop–sheet polymerization, whereby the reactive center loop of one molecule is inserted into the b-sheet of another molecule [20–22]. Correspondence to I. Schvartz, Department of Biological Regulation, Weizmann Institute of Science, Rehovot 76100, Israel, Fax: + 972 8 934 4116, Tel.: + 972 8 934 2483, E-mail: iris.schvartz@weizmann.ac.il Abbreviations: PEDF, pigment epithelium-derived factor; serpin, serine protease inhibitor; RCL, reactive center loop; A1Pi, a1-proteinase inhibitor; ER, endoplasmic reticulum. Dedication: dedicated to the memory of our mentor and distinguished scientist, Shmuel Shaltiel. *Note: Deceased. (Received 27 August 2002, accepted 18 November 2002) Eur. J. Biochem. 270, 822–831 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03374.x In this study, we constructed a series of PEDF mutants, expressed them in Chinese hamster ovary (CHO) cells, and detected their secretion by these cells. The results of Western-blot and Northern-blot analysis as well as of immunofluorescence microscopy suggest that conservation of the 3D structure of PEDF is essential for its efficient secretion. In addition, we show that Gly376 and Leu377, which are located within the RCL of PEDF, play an important role in PEDF secretion. This is, to our know- ledge, the first time that a physiological function for the RCL of PEDF has been reported. Materials and methods Materials Full-length human PEDF cDNA was kindly provided by Dr N. P. Bouck (North-western University, Chicago, IL, USA). Transfomer TM Site-Directed Mutagenesis Kit was purchased from Clontech. All oligonucleotide primers were synthesized by the Expedite TM Nucleic Acid Synthesis System (Workstation, Path Framinghan, MA, 1 USA) at the Weizmann Institute of Science, Rehovot, Israel. Restriction enzymes were purchased from Roche. Goat anti-(rabbit IgG) conjugated to horseradish peroxidase were purchased from Sigma. Midi-purification kit was purchased from QIAGEN. DMEM/F-12 medium, MEM, and Lipofect- AMINE reagent were purchased from Life Technologies Inc.ECL,Ni 2+ -chelated Sepharose Fast Flow column and Ready To Go random primer labeling kit were purchased from Amersham Pharmacia Biotech. Bradford reagent was purchased from Bio-Rad. Pfu DNA polymerase and SV total RNA isolation kit were purchased from Promega. Cy TM 2-conjugated AffiniPure Goat anti-rabbit IgG (H + L) was purchased from Jackson Laboratory 2 .All other materials were the best commercially available grade. Construction of the PEDF mutants The full-length PEDF cDNA in pBluescript II SK (+) was used as a template for mutagenesis. Replacement and deletion of amino-acid residues of PEDF were performed using an oligonucleotide site-directed mutagenesis kit. PEDF cDNAs with the correct mutations were amplified by PCR using the sense and antisense primers (Table 1). Truncation of the C-terminus of PEDF was by PCR amplification using the sense and truncation primers listed in Table 1. Pure PCR products digested with HindIII and EcoRI were ligated into the multicloning site of pcDNA3. DNA sequencing analysis (PE-ABI 377 DNA sequencer) confirmed the nucleotide sequence of the PEDF mutants. Table 1. Oligonucleotide primers. Bold letters indicate mutated nucleotides; letters in parentheses indicate original nucleotides; letters with underline are the recognition sequence by HindIII or EcoRI. Oligonucleotide Sequence PT121 5¢- CGGGAATTCCCGTTAGGTACCATGGATGTCTGGGC-3¢ PT250 5¢- CGGGAATTCCCGTTAAACAGCCTTAGGGTCCGACATC-3¢ PT350 5¢- CGGGAATTCCCGTTAGGGTTTGCCTGTGATCTTGC-3¢ PT412 5¢-CGGGAATTCCCGTTAAATCTTGCCAATGAAGAGAAG-3¢ PT414 5¢- CGGGAATTCCCGTTAGTCCAGAATCTTGCCAATGAAG-3¢ PT414-M 5¢- CGGGAATTCCCGTTAATGGTGATGGTGATGGTGGTCCAGAATCTTGCCAATGAAG-3¢ PT415 5¢- CGGGAATTCCCGTTAGGGGTCCAGAATCTTGCCAATG-3¢ PT416 5¢- CGGGAATTCCCGTTACCTGGGGTCCAGAATCTTGCC-3¢ PT417 5¢- CGGGAATTCCCGTTAGCCCCTGGGGTCCAGAATCTTG-3¢ PM44/45 5¢- GTGGAGGAGGAGGC(A)TG(C)CTTTCTTCAAAGTCC-3¢ PM116/117 5¢- CTTGATCAGCAGCG(C)CAGC(A)CATCCATGGTACC -3¢ PM246/247 5¢-CCCATGATGTCGGC(A)CG(C)CTAAGGCTGTTTTAC-3¢ PM372/373 5¢- GGGGCGGGAACCG(A)CCG(C)CCAGCCCAGGGCTG-3¢ PM374/375 5¢- GGAACCACCCCCGC(AG)CG(C)CAGGGCTGCAGCCTG-3¢ PM376/377 5¢- CACCCCCAGCCCAGC(G)GGC(CT)GCAGCCTGCCCAC-3¢ PM378/379 5¢- CAGCCCAGGGCTGGC(CA)GG(C)CTGCCCACCTCACC-3¢ PMD373-380 5¢- GGGGCGGGAACCACCCACCTCACCTTCCCGC-3¢ PM391 5¢ -CTGGACTATCACCTTGC(AA)CCAGCCTTTCATCTTC-3¢ PM392 5¢ -GACTATCACCTTAACGC(CA)GCCTTTCATCTTCG-3¢ PM393 5¢ -CTATCACCTTAACCAGG(C)CTTTCATCTTCGTAC-3¢ PM392/393 5¢- GACTATCACCTTAACGC(CA)GG(C)CTTTCATCTTCGTAC-3¢ PM391/392/393 5¢-CTGGACTATCACCTTGC(AA)CGC(CA)GG(C)CTTTCATCTTCGTACTG-3¢ PM394/395 5¢- CCTTAACCAGCCTGC(TT)CGC(AT)CTTCGTACTGAGGG-3¢ PM396/397 5¢- CCAGCCTTTCATCGC(TT)CGC(GT)ACTGAGGGACACAG-3¢ PM398/399 5¢- GCCTTTCATCTTCGTAGC(CT)GGC(AG)GGACACAGACACAGG-3¢ PM400/401 5¢- CTTCGTACTGAGGGC(A)CG(A)CAGACACAGGGGCCC-3¢ PM402/403 5¢- GTACTGAGGGACACAGC(A)CG(A)CAGGGGCCCTTCTC-3¢ Sense primer 5¢- CCCAAGCTTGGGAGGATGCAGGCCCTGGTGCTACTC-3¢ Antisense primer 5¢- CGGGAATTCCCGTTAATGGTGATGGTGATGGTGGGGGCCCCTGGGGTCCAGAATC-3¢ Ó FEBS 2003 PEDF secretion by mutagenesis (Eur. J. Biochem. 270) 823 Transient expression of the PEDF mutants in CHO cells CHO cells were grown in six-well tissue culture plates (Nunc, Inc.) at 37 °Cina5%CO 2 incubator in DMEM/ F-12 medium containing 10% fetal bovine serum, 2 m M L -glutamine and antibiotics (growth medium) until the cells were 60–80% confluent. pcDNA3 (2 lg), carrying the mutant PEDF cDNA prepared using the midi-purification kit, was transfected into CHO cells using the Lipofect AMINE reagent following the manufacturer’s instructions. The transfected CHO cells were grown for 18 h in growth medium (2 mL per well), and thereafter the medium was replaced by serum-free DMEM/F-12 medium containing 2m ML -glutamine and antibiotics. After three days, the medium was collected and assayed for PEDF immuno- reactivity. SDS/PAGE and immunoblotting Samples (10 lL) of the conditioned medium of CHO cells transfected with the cDNA of either wild-type (wt)PEDF or one of its mutants were subjected to SDS/PAGE (10% gel) under reducing conditions using the buffer system of Laemmli [23]. To determine the intracellular content of the different PEDF mutants, CHO cells (1 · 10 6 ) transfected with a PEDF cDNA were lysed with 200 lL Laemmli’s sample buffer, boiled for 3 min, and then subjected (5 lg total proteinÆ10 lL )1 ) to SDS/PAGE under reducing con- ditions as above. Subsequently, the proteins were trans- ferred to a nitrocellulose membrane as described previously [24]. Immunoblotting was carried out using rabbit poly- clonal antibodies, raised against the denatured plasma PEDF (1 : 1000 final dilution). We have shown these antibodies to be highly specific for recombinant PEDF (unpublished). The procedure was followed by incubation of the membrane with goat anti-(rabbit IgG) conjugated to horseradish peroxidase. The signal was detected on Fuji X-ray film using the ECL detection reagents. Purification of the PEDF mutants The medium of CHO cells transfected with a secreted form of PEDF mutant was collected (2 mL) and clarified by centrifugation (600 g for 5 min). The supernatant was appliedtoaNi 2+ -chelated Sepharose Fast Flow column (1 mL) equilibrated with 50 m M phosphate buffer, pH 8.0, containing 0.3 M NaCl. The column was washed extensively with 5 m M imidazole, and the bound PEDF was eluted in 200 m M imidazole (0.2 mL). To purify the intracellular truncated mutant PEDF 1–414, six histidine residues were added to the C-terminus of this PEDF mutant by a second PCR using the sense and PT414-M primers (Table 1). CHO cells were transfected with this mutant and grown in growth medium. After 2 days, the cells (4 · 10 7 ) were lysed by ultrasonic treatment in 1 mL RIPA buffer (50 m M Tris/ HCl, pH 8.0, 150 m M NaCl, 1 m M EGTA, 1% Triton X-100, 0.1% SDS and 1% sodium deoxycholate) in the presence of protease inhibitors (0.1 m M phenyl- methanesulfonyl fluoride, 1 l M pepstatin, 1 m M benzami- dine, 10 l M leupetin and 1 lgÆmL )1 aprotinin). The cell lysate was centrifuged (10 000 g for 10 min at 4 °C). The supernatant was collected and purified on a Ni 2+ column as described above. Purified PEDF mutants were dialyzed against NaCl/P i , and the protein concentration was deter- mined using the Bradford reagent. Northern-blot analysis Total RNA of transfected CHO cells was isolated using SV total RNA isolation kit and separated (10 lg) by 1% formaldehyde/agarose gel electrophoresis in a minigel system. The RNA was then transferred to a Hybond-N membrane (Amersham Pharmacia Biotech). Thereafter, the membrane was fixed for 2 h at 80 °C in a vacuum oven and then hybridized with a PEDF cDNA probe labeled with [a- 32 P]dCTP using the Ready To Go random primer labeling kit. The membrane was extensively washed with NaCl/Cit buffer (150 m M NaCl, 15 m M sodium citrate, pH 7.0) containing 1% SDS at 60 °C, and finally exposed to Fuji X-ray film at )80 °Cfor18h. Immunofluorescence microscopy CHOcellsgrownoncoverslipsingrowthmediumwere transfected with the cDNA of either wtPEDF or one of its mutants. After 48 h the cells were rinsed twice with NaCl/P i and fixed (30 min) in NaCl/P i containing 2% paraformal- dehyde. The fixed cells were permeabilized by incubating them in NaCl/P i containing 0.2% Triton X-100 (5 min on ice). The permeabilized cells were washed with NaCl/P i (3 times, 10 min each wash), and incubated (at 22 °C) in NaCl/P i containing 1% BSA. After 30 min, the cells were incubated in NaCl/P i containing anti-PEDF Igs (1 : 500 final dilution; 1 h at 22 °C), and then washed (3 times, 10mineachwash)withNaCl/P i . Subsequently, Cy2- conjugated AffiniPure Goat Anti-Rabbit IgG (H + L) (1 : 250 final dilution in NaCl/P i ) was added and incubated for 45 min at 22 °C. The cells were washed (3 times, 10 min each wash), and the coverslips were then transferred inversely to a glass slide carrying a drop of MowiolÒ 44–88 and photographed (Nikon EFD 3). Neurite outgrowth assay Human Y-79 retinoblastoma cells (obtained from ATCC) were assayed for neurite outgrowth as described previously [25]. Briefly, 2 ml Y-79 cells (1 · 10 5 cellsÆmL )1 )were incubated in the presence of 20 n M either wtPEDF or a PEDF mutant in MEM supplemented with nonessential amino acids, 1 m ML -glutamine, antibiotics, 0.1% insulin– transferrin–selenium (ITS) 3 ,10m M Hepes, pH 7.5, and 1m M sodium pyruvate. After 7 days in culture, the cells were transferred to poly( D -lysine)-coated plates, and their neurite outgrowth was monitored by microscopy at various periods of time. Results Truncation of the C-terminal tail of PEDF prevents its secretion Recent studies showed that PEDF is a potent inhibitor of neovascularization [26]. In an attempt to localize the site in PEDF that is responsible for this biological activity, we 824 H. Shao et al.(Eur. J. Biochem. 270) Ó FEBS 2003 constructed three truncated mutants: PEDF 1–121, PEDF 1–250, and PEDF 1–350 (Scheme 1), and expressed them in CHO cells. After 3 days in culture, the conditioned medium of the transfected cells was collected, and the presence of PEDF was evaluated by immunoblotting using antibodies to PEDF. To our surprise, none of the truncated mutants was secreted by the transfected CHO cells as efficiently as wtPEDF (Fig. 1), whereas their intracellular content (Fig. 1) and mRNA level (data not shown) in CHO cells were comparable to those of wtPEDF. These results indicate that truncation of the C-terminal edge of PEDF prevents its secretion. To determine the minimum length required for PEDF secretion, a series of C-terminal truncated mutants differing in length by only one amino acid was constructed and expressed in CHO cells (Scheme 1). As shown in Fig. 1, the truncated mutants PEDF 1–417 and PEDF 1–416 were secreted by the transfected cells to the same extent as wtPEDF. However, PEDF 1–415 was secreted to a lower extent than wtPEDF (Fig. 1), and shorter molecules, such as PEDF 1–414 and PEDF 1–412, were not secreted at all (Fig. 1). The intracellular level of each of these mutants in the transfected CHO cells was similar to that of wtPEDF (Fig. 1), suggesting that the truncated mutants are effect- ively synthesized but are not efficiently secreted by the cells. This was further confirmed by the finding that the mRNA level of these mutants in the transfected CHO cells was comparable to that of wtPEDF (Fig. 2). To localize the intracellular pools of the PEDF mutants, transfected CHO cells were examined by immunofluores- cence microscopy. In cells expressing wtPEDF (Fig. 3B), PEDF 1–417 (not shown), PEDF 1–416 (Fig. 3C), or PEDF 1–415 (not shown), perinuclear spots with prominent brightness and cytoplasmic reticular staining were observed, which correspond to the Golgi apparatus and endoplasmic reticulum (ER), respectively [17,19]. In cells expressing PEDF 1–414 (Fig. 3D) and shorter molecules, such as PEDF 1–412 (Fig. 3E), cytoplasmic reticular staining was observed. It should be noted that no immunofluorescence staining was observed in mock-transfected CHO cells under the same experimental conditions (Fig. 3A). Therefore, these results indicate that Pro415 is important for the transport of PEDF from the ER to the Golgi apparatus, and consequently for its secretion. Scheme 1. Schematic presentation of the PEDF mutants. The reactive center loop is indicated by a black area. The amino-acid substitutions are identified by standard one-letter designations and are positioned along the PEDF amino-acid sequence. Fig. 1. Secretion of wtPEDF and its mutants by transfected CHO cells. Upper panel, samples (10 lL) of conditioned medium of CHO cells transfected with the cDNA of either wtPEDF or one of its mutants were subjected to SDS/PAGE (10% gel) under reducing conditions followed by immuno- blotting with anti-PEDF Ig (1 : 1000 final dilution) as described in Materials and meth- ods. The PEDF band was visualized with ECL reagents. Lower panel, transfected CHO cells were lysed in 200 lL Laemmli’s sample buffer, boiled for 3 min, and samples (5 lgtotal proteinÆ10 lL )1 ) were subjected to SDS/ PAGE and immunoblotting as described in the upper panel. Ó FEBS 2003 PEDF secretion by mutagenesis (Eur. J. Biochem. 270) 825 Mutations within the RCL of PEDF affect its secretion The crystal structure of human PEDF reveals that the central segment of the RCL is highly exposed, presumably to ensure its interaction with potential targets [14]. However, such targets have not yet been discovered. To discover whether the central segment of the RCL (residues 373–380) is important for the secretion of PEDF, a deletion mutant, PEDF D373–380, was expressed in CHO cells, and its secretion by these cells was measured. As shown in Fig. 1, PEDF D373–380 was ineffectively secreted by the transfected cells. The intracellular content of this mutant (Fig. 1) and its mRNA level in the cells (Fig. 2) were similar to those of wtPEDF. The impaired secretion of PEDF D373–380 was shown to be associated with its ineffective transport from the ER to the Golgi apparatus as revealed by the immunofluorescence staining (Fig. 3F). Therefore, these results indicate that the deletion of the central segment of the RCL prevents the PEDF secretion. To specifically identify the amino-acid residues within this segment that play a role in the PEDF secretion, we constructed a series of mutants in which two consecutive residues were replaced by alanine (PEDF T372/A P373/A, PEDF S374/A P375/A, PEDF G376/A L377/A, and PEDF Q378/A P379/A, Scheme 1). Expression of the mutants in CHO cells showed that three of them (PEDF T372/A P373/ A, PEDF S374/A P375/A, and PEDF Q378/A P379/A) were effectively secreted (Fig. 1), transcribed (Fig. 2) and transported to the Golgi apparatus (Fig. 3G,H). In con- trast, the PEDF mutant G376/A L377/A, although suc- cessfully transcribed (Fig. 2), was not secreted (Fig. 1) and was localized primarily in the ER (Fig. 3I), implying that the amino-acid residues Gly376 and Leu377 may play an important role in PEDF secretion. Mutations within the hydrophobic core of PEDF impair its secretion A series of PEDF mutants was constructed in which consecutive amino-acid residues between Asn391 and Thr403 were replaced by Ala. These residues are located within b-sheet B which has been shown to be involved in formation of the hydrophobic core of PEDF [14]. To determine whether this segment is important for the PEDF secretion, we transfected CHO cells with the following constructs: PEDF N391/A, PEDF Q392/A, PEDF L398/A R399/A, PEDF D400/A T401/A, and PEDF D402/A T403/A. The transfected cells secreted the PEDF mutants to the same extent as cells transfected with wtPEDF cDNA (Figs 1 and 3J,K). However, cells expressing PEDF mutants in which Pro393, Phe394 or Phe396 were replaced by alanine (PEDF Q392/A P393/A, PEDF N391/A Q392/A P393/A, PEDF F394/A I395/A, and PEDF F396/A V397/A) did not secrete them in measurable amounts. The impaired secretion of these mutants was not due to inefficient transcription (Fig. 2), but rather to ineffective transport from the ER to the Golgi apparatus, as revealed by the immunofluo- rescence microscopy (Fig. 3L–O). Note that the single-site mutation of Pro393 to alanine (PEDF P393/A) significantly reduced mutant secretion, but did not abolish it completely (Fig. 1). Replacement of Asp44 Pro45, Pro116 Asp117, and Asp246 Pro247 with alanine did not affect PEDF secretion Analysis of the crystal structure of PEDF [14] also reveals that segments I, II and III (Fig. 4) constitute three distinct exposed loops in the PEDF molecule. To discover whether these specific segments are involved in the PEDF secre- tion, we constructed three mutants in which Asp44 Pro45, Pro115 Asp116, or Asp246 Pro247 of segments I, II, and III, respectively, were replaced by alanine and then determined the secretion of these mutants by transfected CHO cells. The secretion of each of these mutants (Fig. 1), their transcription (Fig. 2), and their intracellular localization (Fig. 3P) were essentially very similar to that of wtPEDF. Effect of PEDF mutants on Y-79 differentiation One of the well-studied biological activities of PEDF is its ability to induce neurite outgrowth in retinoblastoma cells [2]. This activity was recently shown to be located at the exposed parts of helices C and D and at loop 90 [14]. As a few of the mutations described above were found to prevent PEDF secretion, we aimed to test whether these mutations would affect the ability of PEDF to induce retinoblastoma differentiation. As shown in Fig. 5, the PEDF mutants exhibited neurotrophic activity similar to that of wtPEDF, indicating that these mutations do not essentially change the ability of PEDF to induce Y-79 differentiation. Discussion In this study we produced a series of mutations in the PEDF cDNA and determined their effect on the protein Fig. 2. Northern-blot analysis of the transiently transfected CHO cells. Total RNA (10 lg) from each transfected cell line was separated by 1% formaldehyde/agarose gel electrophoresis, transferred to a Hybond-N membrane, and hybridized with 32 P-labeled PEDF cDNA probe as described in Materials and methods. The RNA level in the different transfected cell lines was essentially identical, therefore only repre- sentative samples are presented. 826 H. Shao et al.(Eur. J. Biochem. 270) Ó FEBS 2003 secretion. From our results we can divide the mutations into two categories: (a) those that presumably bring about a change in the 3D structure of PEDF, which leads to inhibition of PEDF secretion; (b) those that do not appear to confer a major change in the PEDF structure, but affect PEDF secretion, probably through a different mechanism. Conservation of the 3D structure of PEDF is important for its secretion We have shown that truncation of the C-terminal tail of PEDF before Pro415 inhibits PEDF secretion. This is in line with the findings that truncation before Pro391 prevented secretion of A1Pi [16]. Analysis of the 3D structure of PEDF Fig. 3. Immunofluorescence microscopic localization of wtPEDF and its mutants. CHO cells transfected with the cDNA of either wtPEDF or one of its mutants were fixed and examined by immunofluorescence microscopy using anti-PEDF Igs (1 : 500 final dilution) and Cy TM 2-conjugated AffiniPure Goat Anti-Rabbit IgG (H + L) (1 : 250 final dilution) as described in Materials and Methods. (A) Mock-transfected CHO cells; (B) wtPEDF; (C) PEDF 1–416; (D) PEDF 1–414; (E) PEDF 1–412; (F) PEDF D373–380; (G) PEDF T372/A P373/A; (H) PEDF Q378/A P379/A; (I) PEDF G376/A L377/A; (J) PEDF N391/A; (K) PEDF L398/A R399/A; (L) PEDF Q392/A P393/A; (M) PEDF N391/A Q392/A P393/A; (N) PEDF F394/A I395/A; (O) PEDF F396/A V397/A; (P) PEDF D44/A P45/A. Ó FEBS 2003 PEDF secretion by mutagenesis (Eur. J. Biochem. 270) 827 reveals that Pro415 is mostly buried and interacts primarily with Phe231 and Leu223 (Fig. 4). Truncation of PEDF at this site probably results in disruption of the hydrophobic interactions imposed by Pro415 and exposure of Asp414 (a negatively charged amino acid) to the negatively charged C-terminus. We therefore assume that this mutation brings about a conformational change in the PEDF molecule that leads to its inefficient secretion. A conformational change in the 3D structure of PEDF is also suggested to occur on replacement of Pro393, Phe394, or Phe396 with alanine. These residues are located within the central b-strand of b-sheet B (red segment in Fig. 4). Analysis of the crystal structure of PEDF reveals that this strand is important for the formation of the hydrophobic core of PEDF. The two Phe residues, i.e. Phe394 and Phe396, appear to interact with the hydrophobic aromatic residues Phe362 and Trp364 located within b-sheet A [14]. Mutations to alanine are assumed to result in loss of these hydrophobic interactions that are essential for the correct folding of PEDF. In addition, our results indicate that deletion of the central segment of the RCL (Pro373 through Ala380) prevents PEDF secretion. As the RCL connects between the b-strand of two b-sheets, A and C [14], our results suggest that deletion of these amino-acid residues, which occupy a gap of 17.6 A ˚ in the polypeptide chain (Fig. 4), prevent the correct folding of PEDF. The importance of correct folding of secretory or membrane proteins for their efficient secretion has been previously documented [27–29]. Hammond & Helenius [28] described a stringent quality control system in the ER and in downstream compartments of the secretory pathway. This system ensures secretion of correctly folded and assembled proteins by promoting their proper folding and by retaining misfolded and incompletely assembled proteins, which are eventually degraded [28]. This quality control system includes lumenal and membrane-bound chaperones that have been shown to play a role in several human diseases. For example, genetic variants of human A1Pi that are unable to fold into the native conformation of the protease were shown to be bound to the membrane-associated chaperone calnexin [30]. Similarly, a cystic fibrosis trans- membrane conductance regulator (CFTR) mutant (DF508) was found to be bound to a cytosolic chaperone, Hsc70, and recently Hsc70 together with a cochaperone molecule (CHIP) were shown to target aberrant forms of the CFTR for proteosomal degradation [31]. In line with these findings, we propose that the misfolded PEDF mutants described in our study may be retained in the ER by chaperones and then destined for degradation. A possible candidate for retaining the PEDF mutants is BiP, a lumenal chaperone shown to bind proteins with exposed hydrophobic domains on their surface [32,33]. Gly376 and Leu377 play an important role in PEDF secretion Substitution of alanine for Thr372 Pro373, Ser374 Pro375, or Gln378 Pro379 did not affect PEDF secretion. How- ever, replacement of Gly376 and Leu377 with alanine prevented it (Fig. 1). As these amino-acid residues are located within the RCL, which has been shown to be very exposed in the 3D structure of PEDF [14], we suggest that these residues are not involved in intramolecular interac- tions, but rather play a role in intermolecular associations, Fig. 4. View ribbon diagram of human PEDF. The 3D structure of PEDF (Protein Data Bank Code: 1IMV) was further analyzed using the Insight II program (MSI/Biosym Technology, San Diego, CA, USA). The blue segment indicates the C-terminal tail of PEDF (Pro415- Pro418). The red segment indicates the hydrophobic core of PEDF. The distance between the Ca atoms at the ends of the unstructured central segmentoftheRCLofPEDFis17.6 A ˚ . The pink segment indicates the neurotrophic activity region of PEDF (residues 78–121). Segment I, II and III designate three exposed loops in PEDF (see Results). 5 Fig. 5. Neurotrophic activity of wtPEDF and its mutants in human retinoblastoma Y-79 cells. Y-79 cells (2 mL; 1 · 10 5 cellsÆmL )1 )were incubated in the presence of either wtPEDF or one of its mutants (final concentration 20 n M ) as described in Materials and methods. Neurite outgrowth was monitored by microscopy at various periods of time, and photographed day 10 after plating. (A) Serum-free medium; (B) wtPEDF; (C) PEDF D44/A P45/A; (D) PEDF P116/A D117/A; (E) PEDF D246/A P247/A; (F) PEDF T372/A P373/A; (G) PEDF T374/ A P375/A; (H) PEDF G376/A L377/A; (I) PEDF Q378/A P379/A; (J) PEDF D373–380; (K) PEDF Q392/A; (L) PEDF P393/A; (M) PEDF Q392/A P393/A; (N) PEDF N391/A Q392/A P393/A; (O) PEDF F394/A I395/A; (P) PEDF F396/A V397/A; (Q) PEDF L398/A R399/ A; (R) PEDF D400/A T401/A; (S) PEDF D402/A T403/A; (T) PEDF 1–414. 828 H. Shao et al.(Eur. J. Biochem. 270) Ó FEBS 2003 for example, with cargo receptors or escort proteins, which are known to escort proteins out of the ER to the Golgi apparatus [27]. Our results imply, for the first time, a novel function for the RCL of PEDF, namely, binding to components of the secretory pathway to ensure efficient secretion of PEDF. Ó FEBS 2003 PEDF secretion by mutagenesis (Eur. J. Biochem. 270) 829 Neurotrophic activity of PEDF The neurotrophic activity of PEDF was originally demon- strated by its ability to induce neurite outgrowth in human retinoblastoma cells [2]. From studies with synthetic peptides, this activity was located at the N-terminal edge of PEDF (residues 78–121) [34]. Recently, Simonovic et al. [14] mapped the neurotrophic activity of PEDF to the exposed parts of helices C and D and to loop 90. We examined whether the mutations that we produced (trun- cation of the C-terminal tail of PEDF, deletion of the central segment of its RCL, or alanine substitution along the PEDF molecule) affect the neurotrophic activity of PEDF. As none of the mutations was shown to inhibit Y-79 differentiation, we suggest that these mutations, although they induce conformational changes in the PEDF molecule as discussed above, do not significantly affect the exposed C and D helices, and hence do not lead to inhibition of the neurotrophic activity. It should be noted that Pro116 and Asp117 are located within the segment found to be responsible for the neurotrophic activity of PEDF [34]. However, replacement of these two amino-acid residues with alanine was not shown to affect this activity, suggesting that they do not play a major role in the neurotrophic activity of PEDF. Possible clinical implications of the inefficient secretion of PEDF variants PEDF has been shown to be a very potent inhibitor of angiogenesis in a murine model of ischemia-induced retinopathy [6,35]. It has also been suggested to play a role in the control of tumor growth [36]. Recently, we have shown that PEDF is present in the blood of various mammals and exhibits antiangiogenic activity similar to that observed with the PEDF isolated from bovine eyes (unpublished). On the basis of our present results, we hypothesize that similar mutations along the PEDF molecule may occur in nature and may have clinical implications. This possibility is supported by the findings that individuals homozygous for the production of an A1Pi variant (A1PiZ) have 15–20% of the normal serum levels of the protease [37]. The A1PiZ variant contains an amino-acid substitution at position Glu342, which leads to limited transport of the protease from the ER to the Golgi apparatus and thereby to its inefficient secretion by the liver [38]. Consequently, the proteolytic activity of elastase released from activated neutrophils is not adequately inhibited, leading to severe lung damage culminating in pulmonary emphysema [37]. The reduced secretion of PEDF mutants as observed in the present study may affect its antiangiogenic activity and may thus have pathological consequences. Acknowledgements We thank Dr Rony Seger for critical reading of the manuscript and for his suggestions and comments. We also thank Dr Miriam Eisenstein for her help in analyzing the 3D structure of PEDF, and Dr Dalia Seger for her careful reading of the manuscript. This research was supported in part by the Israel Science Foundation, Award no. 415/99. References 1. Tombran-Tink, J. & Johnson, L.V. (1989) Neuronal differentia- tion of retinoblastoma cells induced by medium conditioned by human RPE cells. Invest. Ophthamol. Vis. Sci. 30, 1700–1707. 2. Tombran-Tink, J., Chader, G.G. & Johnson, L.V. (1991) PEDF: a pigment epithelium-derived factor with potent neuronal differ- entiative activity. Exp. Eye Res. 53, 411–414. 3. Houenou, L.J., D’Costa, A.P., Li, L., Turgeon, V.L., Enyadike, C.,Alberdi,E.&Becerra,S.P.(1999)Pigmentepithelium-derived factor promotes the survival and differentiation of developing spinal motor neurons. J. Comp. Neurol. 412, 506–514. 4. Taniwaki, T., Hirashima, N., Becerra, S.P., Chader, G.J., Etche- berrigaray, R. & Schwartz, J.P. (1997) Pigment epithelium-derived factor protects cultured cerebellar granule cells against glutamate- induced neurotoxicity. J. Neurochem. 68, 26–32. 5. Cao, W., Tombran-Tink, J., Chen, W., Mrazek, D., Elias, R. & McGinnis, J.F. (1999) Pigment epithelium-derived factor protects cultured retinal neurons against hydrogen peroxide-induced cell death. J. Neurosci. Res. 57, 789–800. 6. Stellmach, V.V., Crawford, S.E., Zhou, W. & Bouck, N. (2001) Prevention of ischemia-induced retinopathy by the natural ocular antiangiogenic agent pigment epithelium-derived factor. Proc. NatlAcad.Sci.USA98, 2593–2597. 7. Volpert,O.V.,Zaichuk,T.,Zhou,W.,Reiher,F.,Ferguson,T.A., Stuart, P.M., Amin, M. & Bouck, N.P. (2002) Inducer-stimulated Fas targets activated endothelium for destruction by anti-angio- genic thrombospondin-1 and pigment epithelium-derived factor. Nat. Med. 8, 349–357. 8. Steele, F.R., Chader, G.J., Johnson, L.V. & Tombran-Tink, J. (1993) Pigment epithelium-derived factor: neurotrophic activity and identification as a member of the serineprotease inhibitor gene family. Proc. Natl Acad. Sci. USA 90, 1526–1530. 9. Whisstock,J.,Skinner,R.&Lesk,A.M.(1998)Anatlasofserpin conformations. Trends Biochem. Sci. 23, 63–67. 10. Lawrence, D.A., Ginsburg, D., Day, D.E., Berkenpas, M.B., Verhamme, I.M., Kvassman, J.O. & Shore, J.D. (1995) Serpin- protease complexes are trapped as stable acyl-enzyme inter- mediates. J. Biol. Chem. 270, 25309–25312. 11. Huntington, J.A., Read, R.J. & Carrell, R.W. (2000) Structure of a serpin-protease complex shows inhibition by deformation. Nature (London) 407, 923–926. 12. Egelund, R., Petersen, T.E. & Andreasen, P.A. (2001) A serpin- induced extensive proteolytic susceptibility of urokinase-type plasminogen activator implicates distortion of the proteinase substrate-binding pocket and oxyanion hole in the serpin inhibitory mechanism. Eur. J. Biochem. 268, 673–685. 13. Becerra, S.P., Sagasti, A., Spinella, P. & Notario, V. (1995) Pig- ment epithelium-derived factor behaves like a noninhibitory serpin. Neurotrophic activity does not require the serpin reactive loop. J. Biol. Chem. 270, 25992–25999. 14. Simonovic, M., Gettins, P.G. & Volz, K. (2001) Crystal structure of human PEDF, a potent anti-angiogenic and neurite growth-promoting factor. Proc. Natl Acad. Sci. USA 98, 11131– 11135. 15. Chen, H.W., Chan, W.Y., Chen, C.L., Mansfield, B.C. & Chou, J.Y. (1993) The carboxyl-terminal domain of the human preg- nancy-sepcific glycoprotein specifics intracellular retention and stability. J. Biol. Chem. 268, 22066–22075. 16. Brodbeck, R.M. & Brown, J.L. (1992) Secretion of alpha- 1-proteinase inhibitor requires an almost full length molecule. J. Biol. Chem. 267, 294–297. 17. Brodbeck, R.M. & Brown, J.L. (1994) Study of the roles of proline 391 and a highly conserved sequence in the carboxyl-terminal region of members of the serpin family in the secretion of alpha 1-proteinase inhibitor. J. Biol. Chem. 269, 17252–17256. 830 H. Shao et al.(Eur. J. Biochem. 270) Ó FEBS 2003 18. Ghigo, J.M. & Wandersman, C. (1994) A carboxyl-terminal four- amino acid motif is required for secretion of the metalloprotease PrtG through the Erwinia chrysanthemi protease secretion path- way. J. Biol. Chem. 269, 8979–8985. 19. Katsumi, A., Kojima, T., Senda, T., Yamazaki, T., Tsukamoto, H., Sugiura, I., Kobayashi, S., Miyata, T., Umeyama, H. & Saito, H. (1998) The carboxyl-terminal region of protein C is essential for its secretion. Blood 91, 3784–3791. 20. Lomas, D.A., Evans, D.L., Finch, J.T. & Carrell, R.W. (1992) The mechanism of Z alpha 1-antitrypsin accumulation in the liver. Nature (London) 357, 605–607. 21. Lomas, D.A., Evans, D.L., Stone, S.R., Chang, W.S. & Carrell, R.W. (1993) Effect of the Z mutation on the physical and inhibitory properties of alpha 1-antitrypsin. Biochemistry 32, 500–508. 22. Lomas, D.A., Elliott, P.R., Sidhar, S.K., Foreman, R.C., Finch, J.T., Cox, D.W., Whisstock, J.C. & Carrell, R.W. (1995) alpha 1- Antitrypsin Mmalton (Phe52-deleted) forms loop-sheet polymers in vivo. Evidence for the C sheet mechanism of polymerization. J. Biol. Chem. 270, 16864–16870. 23. Laemmli, U. (1970) Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature (London) 227, 680–685. 24. Gechtman, Z., Belleli, A., Lechpammer, S. & Shaltiel, S. (1997) The cluster of basic amino acids in vitronectin contributes to its binding of plasminogen activator inhibitor-1: evidence from thrombin-, elastase- and plasmin-cleaved vitronectins and anti- peptide antibodies. Biochem. J. 325, 339–349. 25. Becerra, S.P., Palmer, I., Kumar, A., Steele, F., Shiloach, J., Notario, V. & Chader, G.J. (1993) Overexpression of fetal human pigment epithelium-derived factor in Escherichia coli. A function- ally active neurotrophic factor. J. Biol. Chem. 268, 23148–23156. 26. Dawson, D.W., Volpert, O.V., Gillis, P., Crawford, S.E., Xu, H., Benedict, W. & Bouck, N.P. (1999) Pigment epithelium-derived factor: a potent inhibitor of angiogenesis. Science 285, 245–248. 27. Ellgaard, L., Molinari, M. & Helenius, A. (1999) Setting the standards: quality control in the secretory pathway. Science 286, 1882–1888. 28. Hammond, C. & Helenius, A. (1995) Quality control in the secretory pathway. Curr.Opin.CellBiol.7, 523–529. 29. Lodish, H., Berk, A., Zopursky, S.L., Matsudaira, P., Baltimore, D. & Darnell, J. (2000) Protein sorting: organelle biogenesis and protein secretion. In Molecular Cell Biology (Tenney, S., ed), 4 pp. 675–750. W.H. Freeman, New York. 30. Le, A., Steiner, J.L., Ferrell, G.A., Shaker, J.C. & Sifers, R.N. (1994) Association between calnexin and a secretion-incompetent variant of human alpha 1-antitrypsin. J. Biol. Chem. 269, 7514– 7519. 31. Meacham, G.C., Patterson, C., Zhang, W., Younger, J.M. & Cyr, D.M. (2001) The Hsc70 co-chaperone CHIP targets immature CFTR for proteasomal degradation. Nat. Cell Biol. 3, 100–105. 32. Blond-Elguindi, S., Cwirla, E., Dower, W.J., Lipshutz, R.J., Sprang, S.R., Sambrook, J.F. & Gething, M.J. (1993) Affinity panning of a library of peptides displayed on bacteriophages reveals the binding specificity of BiP. Cell 75, 717–728. 33. Flynn, G.C.P.J., Flocco. M.T. & Rothman, J.E. (1991) Peptide- binding specificity of the molecular chaperone BiP. Nature (London) 353, 726–730. 34. Alberdi, E., Aymerich, M.S. & Becerra, S.P. (1999) Binding of pigment epithelium-derived factor (PEDF) to retinoblastoma cells and cerebellar granule neurons. J. Biol. Chem. 274, 31605– 31612. 35. Duh, E.J., Yang, H.S., Suzuma, I., Miyagi, M., Youngman, E., Mori, K., Katai, M., Yan, L., Suzuma, K., West, K., Davarya, S., Tong, P., Gehlbach, P., Pearlman, J., Crabb, J.W., Aiello, L.P., Campochiaro, P.A. & Zack, D.J. (2002) Pigment epithelium- derived factor suppresses ischemia-induced retinal neovascular- ization and VEGF-induced migration and growth. Invest. Ophthalmol. Vis. Sci. 43, 821–829. 36. Chader, G.J. (2001) PEDF: raising both hopes and questions in controlling angiogenesis. Proc. Natl Acad. Sci. USA 98, 2122– 2124. 37. Fagerhol, M.K. & Cox, D.W. (1981) The Pi polymorphism: genetic, biochemical, and clinical aspects of human alpha 1-anti- trypsin. Adv. Hum. Genet. 11, 1–62. 38. McCracken, A.A., Kruse, K.B. & Brown, J.L. (1989) Molecular basis for defective secretion of the Z variant of human alpha-1- proteinase inhibitor: secretion of variants having altered potential for salt bridge formation between amino acids 290 and 342. Mol. Cell Biol. 9, 1406–1414. Ó FEBS 2003 PEDF secretion by mutagenesis (Eur. J. Biochem. 270) 831 . structure; neurotrophic activity; pigment epithelium-derived factor (PEDF); secretion; serpin. Pigment epithelium-derived factor (PEDF) was originally identified. Secretion of pigment epithelium-derived factor Mutagenic study Hanshuang Shao, Iris Schvartz and Shmuel Shaltiel* Department of Biological