THE JOURNAL OF CHEMIS~Y Vol. 269, No. 43, Issue of 27115 ... · THE JOURNAL OF BIOWXCAL CHEMIS~Y 0 1994 by The American Society for Biochemistry and Molecular Biology, Inc. Vol. 269,

THE JOURNAL OF BIOWXCAL C H E M I S ~ Y 0 1994 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 269, No. 43, Issue of October . 27115-27124, 1994 Printed in U.S.A.

Synthesis and Targeting of Insulin-like Growth Factor-I to the Hormone Storage Granules in an Endocrine Cell Line*

(Received for publication, April 25, 1994, and in revised form, August 1, 1994)

Walter K. Schmidt$ and Hsiao-Ping H. Moore5 From the Department of Molecular and Cell Biology, University of California at Berkeley, Berkeley, California 94720-3200

Export of growth factors is generally believed to be restricted to the constitutive secretory pathway, whereas peptide hormones are typically secreted in a regulated manner. Here we show that insulin-like growth factor (1GF)-I, a growth factor released constitu- tively from the liver, is synthesized and secreted from the mouse pituitary AtT-20 cell line via the regulated pathway. IGF-I production is 1500-fold less than the pep- tide hormone ACTH. Secretagogue induces IGF-I secre- tion in a manner similar to ACTH. Like ACTH, IGF-I is sorted into the regulated pathway >35-fold more effi- ciently than a constitutively secreted protein. Dense core granules isolated from cells transfected with a hu- man IGF-I cDNA contain both ACTH and human IGF-I. AtT-20 cells also synthesize IGF-binding proteins, and at least one of these is secreted by the regulated pathway. Human IGF-I does not exhibit milieu-induced, concen- tration-dependent aggregation, in contrast to secre- togranin I1 which sorts by a proposed aggregation mech- anism. These data suggest that 1) growth factors are not solely released from tissues via the constitutive path- way, 2) IGF-I may contain information for correct granu- lar targeting, and 3) IGF-I may be sorted by a mechanism distinct from that proposed for the secretogranins.

Proteins destined for secretion are unidirectionally trans- ported via vesicular carriers through either the constitutive or the regulated secretory pathway (Burgess and Kelly, 1987; Miller and Moore, 1990). These pathways begin to diverge at the trans-Golgi network (TGN)’ (Orci et al., 1987; Tooze and Hutt- ner, 1990). Post-TGN transport through the constitutive path- way is mediated by small, clear vesicles that are continually secreted. Proteins secreted via the regulated pathway are tar- geted to storage granules whose release is stimulus-dependent.

While all cells secrete proteins constitutively, neuronal, exo- crine and endocrine cells possess additional regulated secretory

* This work was supported by United States Public Health Service Grant GM 35239 and the American Cancer Society Grant CD-497, CB-89A (to H.-P.M.). The costs of publication of this article were de- frayed in part by the payment of page charges. This article must there- fore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

$ Supported by a predoctoral grant from the Howard Hughes Medical Institute.

and Cell Biology, University of California at Berkeley, 142 Life Sciences 5 To whom correspondence should be addressed: Dept. of Molecular

Addition #3200, Berkeley, CA9472C-3200. %I: 510-643-6528; Fax: 510- 643-8708; Email: Hsiao-Ping [email protected].

The abbreviations used are: TGN, trans-Golgi network; IGF, insulin- like growth factor; hIGF, human insulin-like growth factor; IGFBPs,

VSV, vesicular stomatitis virus; TG, truncated VSV-glycoprotein; insulin-like growth factor-binding proteins; RIA, radioimmunoassay;

POMC, pro-opiomelanocortin; GH, growth hormone; SgII, secretogra- nin 11; PBS, phosphate-buffered saline; MES, 4-morpholineethanesu1- fonic acid; DMEM, Dulbecco’s modified Eagle’s medium; PAGE, polyac- rylamide gel electrophoresis; 8-Br-cAMP, 8-bromo-CAMP.

pathways. In these cells, the sorting of proteins into either the constitutive and the regulated pathways occurs at the level of the TGN (Orci et al . , 1987; Tooze et al., 1987; Sossin et al., 1990) and continues in immature granules (Grimes and Kelly, 1992; Kuliawat and Arvan, 1992). Segregation of soluble peptide hor- mones into dense core granules is thought to require a positive sorting determinant: constitutive proteins can be redirected to the regulated pathway when tagged to granular proteins (Moore and Kelly, 1986; Rosa et al., 1989; Stoller and Shields, 1989). This positive determinant is not encoded by a common peptide sequence as required for nuclear import and endoplas- mic reticulum retention, nor is it due to a common glycosylation signal as in mannose 6-phosphate-dependent lysosomal target- ing. A favored model is that the low pH (6.4) and high calcium (10 mM) milieu of the TGN causes the selective aggregation of granular proteins but not constitutive proteins. This model is supported by the biophysical properties of several granule pro- teins which undergo concentration-dependent and calciudpH- induced aggregation in vitro (secretogranins (Gerdes et al., 1989; Gorr et al., 1989; Yo0 and Albanesi, 1990; Chanat and Huttner, 1991) and pancreatic zymogens (LeBlond et al., 1993)). The aggregation of mature insulin to the exclusion of C peptide has also been proposed to result in the preferential removal of the insulin C peptide from immature granules (Ar- van et al., 1991; Neerman and Halban, 1993). However, several granule proteins have been shown not to require such homo- typic oligomerization or aggregation for sorting (Quinn et al., 1991; Turkewitz et al., 1991). Thus, the exact mechanisms for sorting may differ for different proteins.

We have focused on the structural determinants required for sorting of insulin into the regulated secretory pathway. It has previously been established that the insulin sorting determi- nant resides in the mature protein (Powell et al., 1988) and that sorting is independent of its multimeric state (Quinn et al., 1991). Insulin belongs to a gene family which includes the insulin-like growth factors (IGF-I and IGF-11) and relaxin (Blundell and Humbel, 1980). Members of this family share a common tertiary structure as determined by computer model- ing and epitope mapping (Bedarkar et al . , 1977; Blundell et al . , 1978; Murray-Rust et al., 1992). Moreover, the six cysteine residues involved in disulfide bond linkages are conserved in this family.

Human insulin and IGF-I share 45% amino acid identity (Fig. 1). Insulin is synthesized and stored in the dense core granules of pancreatic B cells and is secreted in a stimulus- dependent fashion. IGF-I is primarily synthesized by hepato- cytes which have no known regulated secretory pathway; se- cretion of proteins examined thus far from this cell type has been constitutive (Salamero et al., 1990; Saucan and Palade, 1992). However, it is unknown whether constitutive secretion of IGF-I from hepatocytes is due to the lack of a sorting domain on IGF-I or the lack of sorting machinery in hepatocytes. IGF-I has been detected in tissues such as heart, kidney, and brain (D’Ercole et al., 1984; Aguado et al., 19921, and also in the

27115

27116 Synthesis and Targeting of IGF-1

I N S U L I N F V N Q H ;H::W:a RREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKR GYGSSSRRAPQT

B P E P T I D E C P E P T I D E A P E P T I D E l L B z U U FIG. 1. Comparison of primary sequences of human insulin and human IGF-I. The amino acid sequences of human insulin (Bell et al.,

1979; Sures et al., 1980) and human IGF-I (Jansen et al., 1983) are aligned. Identical residues are shaded in black.

pituitary and pituitary-derived GH, cell line (Fagin et al., 1987); the exact pathway of IGF-I secretion in these cell lines, however, has not been characterized.

In this study, we have examined the sorting of IGF-I in cells with a regulated secretory pathway. We have found that IGF-I is synthesized and secreted by the mouse anterior pituitary corticotroph-derived cell line AtT-20 which contains well char- acterized constitutive and regulated pathways for secretion. We characterize the secretion pathway of both endogenous and transfected human IGF-I in these cells and show that IGF-I is secreted by the regulated secretory pathway. We have also in- vestigated the possible role of aggregation in the sorting of IGF-I.

MATERIALS AND METHODS Antisera and Hormones-Affinity-purified rabbit anti-porcine ACTH

and rabbit anti-VSV were prepared as described previously (Moore and Kelly, 1985). Rabbit anti-serum hIGF-I (UB3 189) was supplied by Drs. L. Underwood and J. J. Van Wyk, Department of Pediatric Endocrinol- ogy, University of North Carolina at Chapel Hill, through the National Hormone and Pituitary Program (Baltimore, MD). Guinea pig anti- porcine insulin was purchased from Linco Research, Inc. (St. Louis, MO). lZ6I-ACTH was a generous giR of the Metabolic Research Unit, University of California, San Francisco. '261-hIGF-I was kindly supplied by Dr. C. Nicoll, University of California, Berkeley.

Cell Culture-AtT-20 cells were maintained in DMEM (BioWhit- taker, Walkersville, MD) supplemented with 10% fetal calf serum under 15% CO, at 37 "C. Cells were stably transfected with a plasmid con- taining sequences corresponding to bovine GH signal sequences and the mature form of hIGF-I (Bayne et aE., 1987). Stable lines were main- tained in media containing 0.25 mg/ml (active concentration) Geneticin (Life Technologies, Inc.). Cells were grown to 7545% confluence for all experiments.

PC-12 cells were maintained in DMEM supplemented with 5% horse serum (Gemini Bioproducts, Calabasas, CA) and 5% enriched calf se- rum (Gemini Bioproducts) under 10% CO, at 37 "C. Cells stably trans- fected with the plasmid pRSV-rIns containing the genomic rat proinsu- lin I1 DNA (Orci et al., 1987) were maintained in media containing 0.33 mg/ml (active concentration) Geneticin. Cells were grown to 7545% confluence for all experiments.

ZGF-Z Secretion Studies-Identical cultures of AtT-20 cells grown in lOcm dishes were thoroughly washed to remove hormones contained in the growth media. The cells were rinsed twice with 10-15 ml of PBS, then incubated in 10-15 ml of serum free DMEM for 30 min at 37 "C, and then washed for a third time with PBS. The cells were next incu- bated for 2 h at 37 "C with either DMEM alone, or DMEM containing 100 &ml cycloheximide to prevent new protein synthesis (Brion et al., 1992). After this pretreatment, the cells were washed with PBS and incubated with 3 ml DMEM for 4 h at 37 "C to measure unstimulated secretion. A parallel culture was incubated with DMEM containing 5 m~ 8-Br-cAMP to induce secretion from the regulated pathway. For the cycloheximide experiments, the incubation media also contained 100 &xnl cycloheximide. The media was collected and cleared of cell debris by centrifugation. The cells were washed with DMEM and then scraped into 3 ml of DMEM. These samples were then subjected to RIA analysis.

Extraction of ZGF-I-binding Proteins-Prior to analysis by IGF-I RIA, samples were treated in order to remove IGF-I-binding proteins by an acetondformic acid extraction procedure (Bowsher et al., 1991). Briefly, 0.625 volume of ice-cold 8 M formic acid, 0.5% Tween 20 was added to each sample. After vortexing, ice-cold acetone (4.4 volume) was then added to precipitate high molecular weight proteins including IGF- binding proteins (small peptides including IGF-I remain in the super- natant). Samples were then centrifuged at 3000 x g for 30 min at 4 "C, and 83% of the total supernatant was transferred to a new tube, and dried in an vacuum oven and a Speed Vac. Samples were reconstituted to 0.25-0.50 of the original volume, neutralized with 10 N NaOH, and incubated for 12 h at 4 "C prior to RIA analysis.

RL4-The IGF-I RIA (Bowsher et al., 1991) was carried out on seri- ally diluted, extracted samples. The samples were adjusted to 100 pl

with diluent from a mock media sample prepared identically as above to normalize the salt concentration in all tubes. The standard, human recombinant IGF-I (R&D Systems, Minneapolis, MN), was diluted from 2 ng to 0.975 pg per tube in IGF-I RIA buffer (50 mM NaH,PO,, pH 7.5, 10 m~ EDTA, 0.02% sodium azide, 0.02% protamine sulfate, 0.05% Tween 20), with added mock samples to normalize salts. Approximately 10,000 cpm of lZ6I-hIGF-I and anti-IGF-I (at 1:15000 in the final assay) were added to each sample and the t o t a l volume was adjusted to 0.5 ml with IGF-I RIA buffer. After incubation at 4 "C for 16 h, antibody- antigen complexes were precipitated with the addition of 1 ml of GAW PEG/NRS for 30 min at room temperature. The GAR.PEG/NRS is com- posed of 1:150 goat anti-rabbit (BABCO, Richmond, CA), 5% polyethylene glycol 8000, and 1:3500 normal rabbit serum (BABCO) prepared in water. Supernatants from a 3000 x g, 25-min centrifugation were removed and the pellets counted in a gamma counter. Counts were analyzed by an RIA potency program (F'. Licht, University of California, Berkeley) based on Faden and Rodbard (1975).

The ACTH RIA (charcoal-dextran method) was carried out as de- scribed previously (Moore et al., 1983). Briefly, reconstituted RIA samples were serially diluted in 100 pl ACTH RIA buffer (50 m~ NaH,PO,, pH 7.6, 0.25% bovine serum albumin, 0.5% P-mercaptoetha- nol, 0.02% sodium azide). Porcine ACTH from 36 ng to 4.4 pg per tube in 100 pl of RIA buffer was used as a standard. This was followed by the addition of an equal volume of RIA buffer containing 25 ng/ml affinity- purified rabbit anti-ACTH antibody and 16,000-20,000 cpm of ACTH. All samples were incubated for 16 h at 4 "C, 100 pl of charcoal- dextran solution (33 mM NaH,PO,, pH 7.6, 2% Norit A charcoal, 0.5% dextran 70 (90 kDa), 10% horse serum, 0.02% sodium azide) was added, and the samples were incubated for 15 min before centrifugation at 3000 x g for 5 min. Two hundred-microliter supernatant fractions were counted in a y counter and analyzed by the RIA potency program de- scribed above.

Cell fiansfections-AtT-20 cells were seeded in a 10 cm tissue cul- ture dish (1 x lo6 cells). After 36 h, the cells were washed three times with 10 ml of serum free DMEM and then incubated with transfection media (TM) for 8 h at 37 "C. The AtT-20 TM was composed of DMEM containing 75 pg of Lipofectinm (Life Technologies, Inc.), 30 pg of plas- mid containing hIGF-I, and 6 pg of pSV2-neo. After the transfection period, the TM was removed, and the cells were replated onto new 10-cm culture dishes. Cells were re-seeded 48 h post-transfection and grown an additional 16 h. AtT-20 growth medium containing 0.25 mg/ml active Geneticin was then added (72 h after the beginning of the trans- fection). Individual stable clones were screened for the production of hIGF-I by RIA.

PC-12 cells (obtained from Dr. Stuart Feinstein, University of Cali- fornia, Santa Barbara) were seeded in a well of a six-well tissue culture plate (1.5 x lo5 cells). After 24 h, the cells were washed three times with 2 ml of serum-free DMEM and then incubated with TM for 6 h at 37 "C. The PC-12 TM was composed of DMEM containing 32 pg of Lipo- fectamineTM (Life Technologies, Inc.), 2.5 pg of pRSV-rIns, and 0.6 of pg pSV2-neo. After the transfection period, the TM was supplemented with an equal volume of PC-12 growth medium, and the incubation was continued for an additional 12 h. The cells were then collected, seeded onto a 6-cm tissue culture dish, and grown for an additional 30 h in PC-12 growth medium. PC-12 growth medium containing 0.33 mg/ml active Geneticin was then added (48 h after the beginning of the trans- fection). Individual stable clones were screened for the production of insulin by radiolabeling and immunoprecipitation.

Isolation of Secretory GranulesSecretory granules were isolated from hIGF-I transfected AtT-20 cells using a D,O/Ficoll gradient isola- tion procedure previously described (Chavez et al., 1994). The cell sus- pension was homogenized on ice using an EMBL homogenizer (8.02-mm bore, 7.99-mm or no. 2 ball, eight strokes). A postnuclear supernatant prepared from the homogenate was subsequently centrifuged at 23,000 x g (or 16,000 rpm) in a SS-34 rotor for 40 min. The resulting pellet (P2) was separated on a D,O/Ficoll gradient. A 50-p1 sample of each gradient fraction was solubilized and assayed for ACTH as described above. One and one-half ml of each fraction were diluted to 3 ml with H,O and then acidacetone-extracted prior to assay by IGF-I RIA as described above.

Western Ligand Blot-Samples were collected as for RIAs. Unex- tracted samples (5% of total) separated on a 12.5% SDS-polyacrylamide

Synthesis and Targeting of IGF-I 27117

gel were transferred in 50 lll~ Tris, 40 mM glycine to a 0.45-pm nitro- cellulose filter (Scleicher & Schuell). Filters were then incubated with blocking solution for 30 min at room temperature. The blocking solution was composed of TBS (20 m Tris, pH 7.4, 0.5 M NaCI), 5% dry milk, 0.05% Tween 20, and 0.025% sodium azide. Filters were washed once with TBS, 0.1% Tween 20, then incubated for 12 h at 4 “C with TBS, 1% bovine serum albumin, 0.1% Tween 20,0.025% sodium azide containing 150,000 cpmiml lZ5I-IGF-I. Filters were washed twice as before, dried, and exposed onto a PhosphorImager cassette. The insulin ligand blot was carried out under identical conditions using 9-insulin as the probe.

TG Secretion StudiesPtable TG transfected AtT-20 cells plated onto a six-well dish (7 x lo5 cells/well) were starved in methionine-free DMEM for 30 min at 37 “C. After starvation, untreated cells were in- cubated for 4 h with methionine-free medium supplemented with 400 pCi/ml [35Slmethionine and 5% DMEM. For secretagogue-treated cells, 5 mM 8-Br-CAMP was added to the labeling media. At the end of the labeling period, precleared media and cell extracts were analyzed by a double immunoprecipitation procedure using an anti-VSV antibody as described previously (Moore and Kelly, 1986). Immunoprecipitates were analyzed by SDS-PAGE and PhosphorImager (Molecular Dynamics, Sunnyvale, CAI.

Quantitation of Sorting and Calculation of Sorting Indices-Images from the PhosphorImager were quantitated using the ImageQuant sys- tem quantitation program (Molecular Dynamics). The percentage of total (cells plus media) immunoreactive materials secreted under non- stimulation (percent released - 8-Br-CAMP) condition was subtracted from the percentage secreted from stimulated cells (percent released + 8-Br-CAMP). This value was then divided by the basal release (percent released in the absence of 8-Br-cAMP) to determine the index value.

Aggregation Assays-Human IGF-I transfected AtT-20 cells grown to confluence in two 15 cm tissue culture dishes (Nunc, Inc., Naperville, IL) were rinsed twice with PBSEGTA (PBS containing 4 mM EGTA), then incubated on ice for 10 min in 10 ml of PBSEGTA. The cells were collected by centrifugation (10 min, 350 x g). The pellet was gently resuspended in 15 ml of homogenization buffer (10 nm Hepes, pH 7.2, 250 mM sucrose, 1 mM Mg(OAc),, 1 rm EDTA, 1.6 mM Na,SO,, 0.5 rm phenylmethylsulfonyl fluoride, 10 pg/ml aprotinin) and recentrifuged. The washed pellet was resuspended in 15 ml of homogenization buffer and then homogenized by passing it through an EMBL homogenizer 8 times (8.02-mm bore, 7.99-mm stainless steel ball; ball no. 2). The homogenate was centrifuged for 4 min at 6,300 x g. The supernatant was transferred to a new tube and centrifuged for 1 h at 23,000 xg. The P2 pellet was resuspended to a final volume of 100 PI with homogeni- zation buffer. Twenty microliters of this suspension was diluted into 1 ml of non-aggregative (10 mM MES, pH 7.4, 30 mM KC1, 1.2 mM leupep- tin) or aggregative (10 mM MES, pH 6.4,lO m CaCl,, 1.2 rm leupeptin) buffer with or without 0.5 mg/ml saponin (Chanat and Huttner, 1991). After a 15-min incubation on ice, the samples were centrifuged in a Beckman TLA 100.3 rotor (Beckman Instruments) at 100,000 x g for 30 min. The supernatants were removed and the pellets were solubilized with 50 pl of 2 x NDET (2% Nonidet P-40, 0.8% deoxycholate, 132 mM EDTA, 20 mM “is, pH 7.4). The solubilized pellets were then diluted with 1 ml of the respective incubation buffer. All samples were then normalized for pH, salt, and detergent prior to analysis by IGF-I RIA.

Vesicles containing radiolabeled proinsulin or TG were isolated from stable transfected AtT-20 cells. Vesicles containing radiolabeled proin- sulin or SgII were isolated from stable transfected PC-12 cells. Briefly, two 15-cm dishes of cells were starved for 30 min in medium lacking methionine, cysteine, or sulfate for TG, proinsulin, SgII, respectively. The cells were then labeled for 20 min with 0.5 mCi/ml [35Slmethionine to label TG, or 10-20 min with 0.5 mCi/ml [35S]cysteine to label proin- sulin, or 5 min with 1 mCi/ml 35S04 to label SgII. A P2 pellet was isolated from labeled cells as described above. The AtT-20 cells contain- ing radiolabeled proinsulin were chased at 19.5 “C for 6 h prior to isolation of the P2. Fractions of the resuspended P2 membranes were then subjected to release assays under aggregative and non-aggregative conditions as described above. The samples were then quantitatively immunoprecipitated (proinsulin and TG) or precipitated with trichloro- acetic acid (SgII) prior to analysis by SDS-PAGE and PhosphorImager.

RESULTS Endogenous Synthesis and Secretion of Immunoreactive

IGF-I from AtT-20 Cells-The AtT-20 mouse pituitary cell line has been previously shown to sort the ACTH precursor pro- opiomelanocortin (POMC) and various exogenous peptide hor- mones into the regulated secretory pathway. This cell line was used in this study to analyze the secretion pathway of IGF-I.

Production of IGF-I in pituitary corticotrophs or AtT-20 cells has not been previously documented. Since many tissues and cell lines are known to produce IGF-I, we first determined whether AtT-20 cells synthesized and secreted IGF-I. IGF-I peptide can be detected by a well established RIA that is spe- cific for IGF-I but not other members of the insulin family (the National Hormone and Pituitary Program; Bowsher et al. (1991)). Using this RIA, we detected IGF-I immunoreactivity in 24 h culture media from AtT-20 cells (not shown).

Since the normal growth media for AtT-20 is supplemented with serum which contains IGF-I, it was important to deter- mine whether the detected immunoreactivity was due to con- tamination from serum in culture medium or secretion from the cells. Cycloheximide was used to block cellular protein synthe- sis, and its effect on the observed IGF-I immunoreactivity in the medium was determined. Cultures of AtT-20 cells were washed extensively in PBS and then pretreated for 2 h with 100 pg/ml cycloheximide in serum-free medium; treating AtT-20 cells with this concentration of cycloheximide for 30 min inhib- its >95% of new protein synthesis without affecting the consti- tutive secretory machinery (Brion et al., 1992). Cells were then incubated in serum-free medium containing cycloheximide for 4 h, and the medium and cells were collected. Aparallel culture was incubated in medium lacking cycloheximide. Since IGF-I- binding proteins are present in these samples (see below) and interfere with IGF-I RIA(Bowsher et al., 19911, media and cells were first extracted to remove IGF-I-binding proteins before the RIA assay (see “Materials and Methods”).

As shown in Fig. 2 A , cells that were not treated with cyclo- heximide secreted 94 2 24 pg of immunoreactive IGF-I over the 4-h collection period. Cycloheximide treatment reduced this number to 48 2 21 pg. The background level, obtained by treat- ing identical dishes that did not contain any cells, was 38 * 9 pg. Thus treating cells with cycloheximide reduced secretion to background levels, suggesting that secreted IGF-I immunore- activity is due to endogenous protein synthesis and not to con- tamination from serum components. In the cell extracts (Fig. 2 B ) , we recovered 922 2 42 pg of immunoreactive IGF-I from untreated cells. Inhibition of protein synthesis with cyclohexi- mide reduced intracellular IGF-I to 574 2 32 pg, or 62% of that found in untreated cells. A dish which was treated identically but lacked seeded cells served as the no cell control. Only 18 * 14 pg IGF-I was recovered for this condition. Thus, treating cells with cycloheximide reduced the amount of immunoreac- tive IGF-I in the media by 82%, but the amount in the cells by only 38% (values corrected for background). A possible expla- nation for this discrepancy is that some IGF-I is not constitu- tively secreted but is targeted to an intracellular compartment which has a life time longer than the duration of cycloheximide treatment (6 h total). It should be noted that the total amount of immunoreactive IGF-I recovered from media and extract of a 10-cm culture dish (5 x lo6 cells) was 1 ng. This is roughly 1500 times less than the endogenous hormone ACTH which was produced at the level of 1.5 pg for the same number of cells.

Response to Stimulation by 5 mM 8-Br-CAMP-Proteins tar- geted to dense-core granules have a half-life of 7-10 h in AtT-20 cells (Moore and Kelly, 1985); this could be the intracellular storage site for the immunoreactive IGF-I observed above. If IGF-I is stored in dense core granules, it is expected that treat- ment with 8-Br-CAMP will increase its rate of release in a fash- ion similar to ACTH. As shown in Fig. 3 A , addition of 5 II~M 8-Br-CAMP increased the rate of IGF-I release from 56 2 24 pg to 289 2 22 pg (corrected for background) over a 4 h period. The cell extracts recovered from treated cultures contained corre- spondingly less IGF-I than the untreated control (compare 904 * 42 pg in unstimulated and 710 -c 55 pg in stimulated cells). Thus

Synthesis and Targeting of IGF-I 27118

0.1 2

0 . 1 a

n w e 0.08

U - 0.06 I ri s2

0.04

0.02

0.00 cycloheximide:

cells:

A

”-

-I-

” 1 .c

T I n w r= U - LL c3

0.8

0.6

0 .4

0.2

0.0 - I - I + I cycloheximide:

- l + I + l cells:

ofAtT-20 cells grown in 10-cm dishes at a density of 6 x lo6 cells per plate were thoroughly washed free of growth medium and incubated in DMEM FIG. 2. Detection of immunoreactive IGF-I in the cell extract and medium of a mouse pituitary cell line, AtT-20. Identical cultures

for 4 h. The medium was collected, and cells were scraped from the dish. Both cell and medium samples were extracted with acid to remove any IGF-I-binding proteins, and the amount of immunoreactive IGF-I was determined by a radioimmunoassay (see “Materials and Methods”). A set of control dishes were treated identically to the above experimental conditions, except that no cells were seeded into plates containing the growth medium. In a second control, cultures were pretreated with serum-free DMEM containing 100 pg/ml cycloheximide for 2 h to inhibit new protein synthesis. The preincubation media was removed and the cells were further incubated for 4 h with fresh medium containing cycloheximide. (A) Immunoreactive IGF-I recovered from the 4 h incubation media and (B ) from the cells collected after the incubation period. Data points represent duplicate samples and are plotted as nanograms of immunoreactive IGF-I recovered from a 10-cm dish under each condition.

n 0) c W

- 4 12

1 .c

0.8

0.6

0 . 4

0.2

0

‘1A T

- CAMP: - + - +

Media Ex t rac t

n P) c W

E Q

zooc

160C

120a

B o a

4 0 0

0

B

T

I I Media Extract

washed and incubated for 4 h in DMEM in the presence or absence of a secretagogue, 8-Br-CAMP (5 mM). The medium and cell samples collected FIG. 3. (A) Stimulated release of immunoreactive IGF-I from AtT-20 cells. Identical cultures of AtT-20 cells in 10-cm dishes were thoroughly

were assayed by an IGF-I RIA as described in Fig. 2. Numbers shown were corrected for background, no cell control. (B) For comparison, fractions of the media and extract samples were assayed forACTH using an ACTH-specific RIA(Moore et al., 1983). 8-Br-CAMP is abbreviated here as CAMP.

the elevated level of IGF-I immunoreactivity in the medium is 15 pg of IGF-I in the media from stimulated cultures, and <lo not due to increased cellular synthesis (and therefore secretion) pg from unstimulated cultures (data not shown). This indicates of IGF-I in 8-Br-CAMP-treated cells. Stimulated release was also that the increased release originates from an intracellular stor- observed in cells treated with cycloheximide; we detected 214 3 age pool rather than new protein synthesis.

Synthesis and Targeting of IGF-I 27119

CAMP:

66 kD-

45 kD-

Media I Extract 1

FIG. 4. Secretion of a constitutive marker from transfected AtT-20 cells under steady-state labeling conditions. Cultures of an AtT-20 cell line stably transfected with a soluble, truncated form ofVSV glycoprotein (TG) grown in six-well dishes were incubated continuously in labeling medium containing [""Slmethionine for 4 h in the presence (+) or absence of (-) 5 mM 8-Br-CAMP. At the end of the incubation, the medium was collected and the cells were extracted with NDET buffer. The samples were immunoprecipitated with an antiserum against VSV and subjected to 12.5% SDS-PAGE and PhosphorImagerlImageQuant analysis.

For comparison, we measured the amounts of ACTH in the same samples using an ACTH RIA. Fig. 3B shows that 8-Br- CAMP increased secretion of ACTH during the 4-h period from 118 2 20 ng to 1220 2 55 ng and caused a corresponding decrease in intracellular ACTH from 1419 2 292 ng to 584 57 ng. From these data it is evident that the amounts of immunoreactive IGF-I are vastly different from ACTH (see the differences in scale between Fig. 3, A and B) . Thus, both IGF-I and ACTH appear to be targeted to the regulated secretory pathway despite such drastic differences in the amount of protein produced.

The above studies show that some endogenous immunoreac- tive IGF-I is secreted via the regulated pathway. The extent of secretagogue-induced secretion of IGF-I, however, is less than that of ACTH (Fig. 3, A and B) . To determine if this degree of sorting is significant, we wished to compare it with a protein that is known to exit AtT-20 cells by the constitutive pathway. Previously we have used a pulse-chase protocol to show that TG, a truncated form of VSV membrane glycoprotein lacking the transmembrane and cytoplasmic domains, is secreted pri- marily via the constitutive pathway in transfected AtT-20 cells (Moore and Kelly, 1985). Since our anti-IGF-I antibodies do not work in immunoprecipitation experiments, we could not per- form similar pulse-chase experiments to directly compare the IGF-I sorting efficiency with that of TG. The RIA used for measuring IGF-I secretion differs from the pulse-chase protocol in that it follows steady-state levels of secretion rather than a pre-labeled pool. We therefore modified our labeling protocol to measure steady-state secretion of TG for direct comparison to IGF-I and ACTH. AtT-20 cells stably transfected with TG were labeled to steady state with [35S]methionine, and materials released into the medium during a continuous 4-h labeling period were collected either in the presence or absence of 5 mM 8-Br-CAMP. The amounts of labeled TG in the cells and the medium were then quantitated by immunoprecipitation. Fig. 4 shows the results of such an experiment for TG. The slightly slower mobility of secreted TG as compared to the intracellular form is due to differences in glycosylation (Moore and Kelly, 1985). As can be seen, secretion of TG is not significantly af- fected by the secretagogue.

To quantitate the efficiency at which the immunoreactive

TABLE I Comparison of sorting indices of IGF-I with other secretory proteins in

AtT-20 cells The sorting index was defined as secretagogue-induced secretion rela-

tive to total secretion in the absence of stimulation. For each protein, the amount of the protein secreted during the 4-h incubation period was divided by the total amount of the protein recovered from both medium and cell extract at the end of incubation to give percent secretion. The sorting index was then calculated from unstimulated and stimulated cultures as follows. Sorting index = increase in secretory rate by stimulationbasal constitutive secretory rate = % secretion in the pres- ence of CAMP - % secretion in the absence of CAMP + % secretion in the absence of CAMP Values for ACTH and IGF-I were taken from Fig. 3 and are expressed in nanogram; those for IGF-I binding proteins (BP) and TG were taken from Figs. 7 and 4, respectively, and are expressed in arbitrary scanning units as determined from PhosphorImager/ ImageQuant analysis.

Amount Amount Sorting secreted stored index

ACTH - 118220 14192292 1537 7.8

Mouse IGF-I - 0.056 2 0.024 0.904 2 0.042 0.960 4.0 + 1220 2 55 584 2 57 1804

(endogenous) + 0.289 f 0.022 0.710 2 0.055 0.999

Human IGF-I - 1.34 2 0.02 4.32 2 0.15 5.66 1.5 (transfected)

24-kDa BP + 4.08 2 0.13 2.72 2 0.46 6.80 - 46.72 0.9 220.0 2 16.5 266.7 1.7 + 129.2 2 1.5 145.0 2 18.2 274.2

29-kDa BP - 32.9 2 0.6 81.7 2 0.2 114.6 0.24 + 39.5 2 0.8

TG 71.1 2 2.6 110.6

- 25.82 7.4 115.3 2 25.2 141.1 0.10 + 34.1 -c 5.3 134.4 2 18.4 168.5

IGF-I is sorted, we determined a "sorting index" which meas- ures the fractional increase in the rate of secretion induced by a secretagogue compared to basal unstimulated levels. The data in Figs. 3 and 4 were used to calculate the sorting indices shown in Table I. Mouse IGF-I exhibited a sorting index of 4.0, which is within a factor of two of that for ACTH (7.8). The sorting index of TG under these conditions was 0.10, or 40 times less than that for IGF-I. These data support the conclu- sion that the intracellular trafficking of IGF-I more closely resembles that of ACTH and not of TG.

Expression of Human IGF-I in AtT-20 Cells by DNA Dansfection-Since the production level of endogenous IGF-I was low, we generated a high expressing cell line to verify the above results. AcDNAencoding the B-C-A-D peptides of human IGF-I fused to a bovine growth hormone signal sequence was transfected into AtT-20 cells, and stable transformants were isolated. We selected one of the highest expressing clones for subsequent studies. When normalized for total protein, the amount of IGF-I recovered from this transfected cell line was 4.58 2 0.09 ng/mg cellular protein as compared to 0.13 2 0.04 ng/mg in untransfected cells, or a 35-fold increase (Fig. 5 4 ) . The amount secreted into the media during a 4-h incubation was 1.42 2 0.05 ng/mg, which is two orders of magnitude higher than from untransfected cells (Fig. 5A). For subsequent IGF-I secretion studies using this transfected cell line, the amount of endogenous IGF-I was ignored since it did not complicate in- terpretation of the results.

The addition of 5 mM 8-Br-CAMP to the hIGF-I transfected cell line increased the rate of IGF-I release from 1.34 2 0.02 to 4.08 2 0.13 ng (Fig. 5B, Table I). Again, the stimulated cells retained less IGF-I than the control (compare 4.32 2 0.15 ng in unstimulated and 2.72 2 0.46 ng in stimulated cells, Fig. 5B). Although the sorting index for the transfected hIGF-I (1.5) was lower than that of endogenous IGF-I (4.0), it is still 15 fold higher than the constitutive protein TG (0.10).

Co-fractionation of IGF-I and ACTH in Isolated Dense Secre- tory Granules-The above results suggest that both endoge-

0

cells:

A

Synthesis and Targeting of IGF-I

AtT20 1 - I G F I 1 1 - I G F I 1 AtTEO A t T 2 0

I I I

Media Ext rac t

n CJ) C v - L L t3

CAMP:

Media Ext rac t

FIG. 5. Stable expression of human IGF-I inAtT-20 cells by DNA transfection and studies of its secretion characteristics. AtT-20 cells were transfected by the Lipofectinm method with an expression plasmid, carrying a human IGF-I cDNA, pCMV-hGH-hIGF-I. Stable cell lines were isolated, and one of the resulting cell lines (designated as AtT20-IGFI) is characterized here. ( A ) Transfected cells or untransfected AtT-20 cells grown in 10-cm dishes were washed and incubated in DMEM for 4 h as described in Fig. 2. Immunoreactive IGF-I in the medium and the cells was quantitated by RIA, and normalized to the amount of total cellular proteins. ( B ) Identical 10-cm dishes of the transfected cells (approximately 1 x lo6 cells/dish) were washed and incubated for 4 h in DMEM containing or lacking 5 m~ 8-Br-CAMP. The amounts of hIGF-I released into the medium were then determined by RIA.

nous and exogenous IGF-I are sorted into the regulated path- way of AtT-20 cells. To further confirm that IGF-I is present in dense-core granules, we measured the IGF-I content of gran- ules isolated by a previously established procedure (Gumbiner and Kelly, 1981; Powell et al., 1988). Cells transfected with hIGF-I were used for these studies since the isolation proce- dure did not yield sufficient granules for the detection of en- dogenous IGF-I.

Transfected cells were homogenized, and a high speed mem- brane pellet (P2) was prepared. Further fractionation of the P2 pellet on a D,O/Ficoll gradient yielded a secretory granule peak in the dense part of the gradient. Fig. 6 shows the IGF-I, ACTH, and total protein profile of the gradient. The dense core gran- ules, recovered in fractions 3-5, contained both ACTH and IGF-I as detected by RIAs, which were well separated from the bulk of proteins and other cellular membranes (fractions 11- 13). A summary of the recoveries of IGF-I, ACTH, and proteins during the purification procedure is shown in Table 11. The typical yield of granules and thus granular proteins from these preparations is between 1 and 5% of total granules in the crude homogenate (Gumbiner and Kelly, 1981). In this experiment, 2.8% of total immunoreactive ACTH was recovered in the final granule fractions. The amount of hIGF-I in the final granule fractions was 3.3% of total immunoreactivity, similar to ACTH. As expected, there was an enrichment of IGF-I and ACTH over total proteins (recovery of total proteins in the granule frac- tions was 0.18%). The fractionation data thus lend further sup- port to the notion that IGF-I and ACTH are both targeted to dense granule compartments.

Expression and Secretion of Endogenous IGF-binding Pro- teins in AtT-20 Cells-A Western ligand blot was used to deter- mine whether IGF-binding proteins were also synthesized by the AtT-20 cell line. Conditioned media and cell extracts were prepared as described for the detection of IGF-I. Unextracted samples were then subjected to non-reducing SDS-PAGE, transferred onto nitrocellulose, probed with 1251-hIGF-I, and

analyzed by PhosphorImager. Two binding proteins with mo- bilities of 24 and 29 kDa were detected in the media and ex- tracts (Fig. 7). These mobilities are consistent with that of IGFBP-4 and IGFBP-5, respectively (Fielder et al., 1992, 1993).

As in the case of IGF-I, cycloheximide treatment reduced but did not abolish the 24-kDa binding protein from cell extracts, suggesting that this protein was stored intracellularly (data not shown). As with IGF-I and ACTH, the rate of secretion of the 24-kDa binding protein was increased in secretagogue- treated cells (Fig. 7, Table I). Quantitation of the Western li- gand blots showed that the 24-kDa binding protein has a sort- ing index (1.7) which is similar to hIGF-I (1.5). In contrast, the extent of stimulation of the 29-kDa binding protein was much less (Fig. 7); it has a sorting index of 0.24 which was only slightly higher than TG (Table I). The data suggest that IGF-I and at least one of its binding proteins are coordinately pack- aged and secreted via the regulated pathway in AtT-20 cells.

IGF-I is Soluble in a High Calcium and Low pH Enuironment-Chanat and Huttner (1991) have shown that secretogranin I1 forms large aggregates in a high calcium and low pH environment and fails to be released from permeabi- lized TGN vesicles under these conditions. We isolated mem- brane vesicles from hIGF-I transfected cells to address whether IGF-I would also be retained in permeabilized vesicles. These vesicles were incubated under either non-aggregative (pH 7.4, 30 m~ KC1) or aggregative (pH 6.4, 10 m~ CaC1,) conditions in the presence or absence of 0.5 mg/ml of saponin. This amount of saponin has been shown to permeabilize isolated vesicles with- out complete membrane solubilization (Chanat and Huttner, 1991). After centrifugation to separate released materials (the supernatant) from the vesicle pellet, both fractions were ana- lyzed by IGF-I RIA. The aggregation model for protein sorting predicts that IGF-I and other regulated proteins will form ag- gregates under low pHhigh calcium conditions; these aggre- gates will not be released into the supernatant when vesicles are permeabilized. Constitutive proteins, on the other hand,

Synthesis and Targeting of IGF-I 27121

7 5

n c 0 W 6 0 0 .-

E % P) c 4 5

I + 0 3 0

U

a 0

1 5

0 3 5 7 9 1 1 1 3 1 5 1 7

80

6 0

4 0

20

0

bottom FRACTIONS t o P FIG. 6. Targeting of transfected human IGF-I to dense-core granules. A stable AtT-20 cell line transfected with hIGF-I was used to isolate

dense-core granules. Two 15-cm dishes of cells were homogenized, and a postnuclear supernatant fraction was prepared. Membranes were pelleted from this fraction by centrifugation, and the resuspended pellets were subjected to equilibrium density centrifugation. The fractions were assayed for total protein and ACTH and IGF-I by RIAs. The peak centering around fraction 4 represents peptide-hormone containing dense-core granules.

TABLE I1 Recoveries of hIGF-I and ACTH during secretory granule purification

AtT-20 cells stably transfected with human IGF-I were homogenized

trifugation. Membranes from the PNS were pelleted by centrifugation and a post-nuclear supernatant (PNS) was prepared by low speed cen-

at 20,000 x g for 40 min. The resultant P2 pellets were resuspended and layered over a D,O-Ficoll gradient as described under “Materials and Methods.” Dense granule fractions were recovered in fractions 2-6 from the gradient (see Fig. 6). Aliquots from each fraction were assayed for total protein using a Bradford assay and IGF-I and ACTH by RIAs.

Protein ACTH IGF-I

mg % recouery pg %recovery ng % recouery Crude homogenate 160.40 100.00 3.23 100.0 14.30 100.0 P2 17.40 10.85 1.35 41.8 5.50 38.5 Granule fractions 0.30 0.18 0.09 2.8 0.48 3.3

should not aggregate and will be released from permeabilized vesicles under either condition. As expected, the majority of IGF-I was found in the mem-

brane pellet when the incubation was carried out under unper- meabilized, non-aggregative conditions; only 33.6% of the IGF-I was released into the supernatant (Table 111, non-aggregative - saponin). Upon permeabilization, most of the IGF-I immuno- reactivity (87.8%) was released from the vesicles and recovered in the supernatant (Table 111, non-aggregative + saponin). Sur- prisingly, the majority of the IGF-I immunoreactivity (93.9%) was still recovered in the supernatant when the membranes were permeabilized under aggregative conditions (Table 111, aggregative + saponin). Thus IGF-I appears to be soluble under both non-aggregative and aggregative conditions.

We next examined the behavior of another secretory protein, TG. Cells stably transfected with TG were labeled for 20 min with [35Slmethionine, and membrane vesicles were isolated as before. Like IGF-I, the majority of labeled TG was pelleted in unpermeabilized preparations with only 3.6% released (Table 111, non-aggregative - saponin). Unlike IGF-I, permeabilization under either non-aggregative or aggregative conditions did not result in the release of the majority of TG into the supernatant;

only 28.4 and 24.7% was released under these conditions re- spectively (Table 111, non-aggregative + saponin and aggrega- tive + saponin). The behavior of TG thus suggests that the release of IGF-I from permeabilized vesicles was not due to overpermeabilization of membranes.

To confirm that our conditions were indeed suitable to pro- mote aggregation, we repeated this experiment for secretogra- nin 11, the archetypal regulated protein of the aggregation model. Vesicles isolated from PC-12 cells were used since SgII is not a major sulfated species in AtT-20 cells. In complete agreement with studies by Chanat and Huttner (19911, we found that permeabilization under non-aggregative conditions resulted in the release of 69.3% SgII while only 25.1% was released under aggregative conditions (Table 111).

The above results suggest that the regulated protein IGF-I has distinct aggregative properties from the acidic regulated protein SgII. Although we cannot exclude the possibility that these differences are due to utilizing different cell types, this is unlikely. We have found that rat proinsulin I1 exhibits similar properties irrespective of cell types: like IGF-I, the majority of proinsulin is released from permeabilized vesicles under aggre- gative conditions (69.6 and 79.8% from AtT-20 and PC-12 de- rived vesicles, respectively). These results raise the possibility that the sorting of IGF-I and proinsulin may involve a mecha- nism(s) different from the aggregation process proposed to gov- ern the sorting of SgII.

DISCUSSION

Several models have been proposed to explain the selective sorting of soluble regulated secretory proteins into dense-core granules. These models include 1) aggregation of regulated secretory products to the exclusion of constitutive proteins in the TGN (for a review, see Bauerfeind and Huttner, (1993)), 2) removal of non-regulated secretory proteins from immature granules during the condensation of mature secretory products (Kuliawat and Arvan, 1992; Neerman and Halban, 1993), and 31 targeting to dense-core granules by specific sorting domains

27122 Synthesis and Targeting of IGF-I

'c1 Media Extract Serum

-43 kD- 1 40-43 kD BP

I ] 28-30 kD BP

24 kD BP-1 U - 1 1-1- 24 kD BP

FIG. 7. Synthesis and secretion of IGF-I-binding proteins (BP) by the regulated pathway in AtT-20 cells. Identical cultures ofAtT-20 cells grown in 10-cm dishes were washed and incubated for 4 h in the presence or absence of 5 mM 8-Br-CAMP as described in Fig. 3. Media and cell extracts were collected and analyzed by ligand blotting. Samples (5% of total) were separated by 12.5% SDS-PAGE, transferred onto nitrocellulose filters, and probed with '251-hIGF-I. The radiolabeled bands were visualized by exposing the filters to a PhosphorImager cassette. A typical gel of media and cell extracts prepared a t the end of the incubation period is shown. (+) indicates media and extracts prepared from cultures that were stimulated with 5 mM 8-Br-cAMP, and (-1 represents samples from control unstimulated cells. The bands were compared with IGF-I-binding proteins in rat ( R ) or mouse ( M ) serum revealed by ligand blotting of crude serum (2 pl/lane) by the same method. The position of two molecular mass markers, 43 and 29 kDa, are shown.

TABLE I11 Release of secretory proteins from isolated vesicles under various

conditions AtT-20 cells expressing hIGF-I or TG, or PC-12 cells were radiola-

beled and homogenized as described under "Materials and Methods." Fractions of the high speed membrane pellets (P2) were then perme- abilized with 0.5 mg/ml saponin to release soluble materials. The per- meabilization was carried out in either non-aggregative (neutral pH and no Ca") or aggregative (pH 6.4 and 10 mhf Ca") conditions. Re- leased materials were separated from membrane-associated materials by ultracentrifugation. Immunoreactive hIGF-I in the supernatant and the pellets were quantitated by RIA. Immunoprecipitates of radiola- beled TG and trichloroacetic acid precipitates of radiolabeled SgII were quantitated by PhosphorImager after separation by SDS-PAGE. The amount released was calculated as the percent of secretory protein recovered in the supernatant relative to the total (supernatant + pellet) for each condition. Data for IGF-I and TG represent results from du- plicate experiments.

Total released (%)

Non-aggregative - Non-aggregative + Aggregative + saponln saponm saponin

TG IGF-I 33.6 f 1.4 87.8 k 1.1 93.9 k 0.3

SgII 3.6 f 0.2 28.4 -c 7.1

2.5 69.3 24.7 f 4.1

25.1

on regulated secretory proteins (Moore and Kelly, 1986; Chanat et al., 1993). These models all predict the existence of a special structural feature on individual granular proteins that deter- mines sorting. Attempts to identify these structural features, however, have not revealed a common motif among soluble regulated secretory proteins (Powell et al., 1988; Stoller and Shields, 1989; Roy et al., 1991; Castle et al., 1992; Chanat et al., 1993; Tam et al., 1993). An amphipathic CY helix has been sug- gested by comparative computational analysis as a common feature in sorted peptides, but its role in sorting has not be tested (Kizer and Tropsha, 1991). It is possible, though, that the sorting of different secretory proteins may involve distinct mechanisms.

Of the insulin family of peptides, insulin is the only member that has been clearly shown to be packaged in dense-core gran- ules and secreted in a regulated fashion. IGF-I and IGF-I1 are primarily secreted by the liver (Zapf and Froesch, 1986; Le- Roith et al., 19921, whereas relaxin is secreted by the ovarian tissue during pregnancy (Bagnell, 1991). Although the produc- tion of IGFs and relaxin are both regulated by other hormones (IGFs by growth hormone and relaxin by reproductive hor- mones), their regulation in these tissues is at the level of tran- scription (Bichell et al., 1992). Consequently, IGFs and relaxin have been considered to be constitutive secretory proteins that lack granular localization signals. However, IGFs have been shown to be expressed in extra-hepatic tissues, and their modes

of regulation in non-hepatic cells have not been previously ex- amined. Our data show that IGF-I is secreted by a pituitary corticotroph-derived cell line, in which it is targeted to dense granules for regulated secretion. To our knowledge, this is the first time that secretion of IGF-I is shown to be regulated at the exocytic level. Many other growth factors (such as epidermal growth factor, fibroblast growth factor, and interleukins) are also produced by cells lacking classical secretory granules. It remains to be determined whether they also contain informa- tion for granular localization.

IGF-I is known to exert a wide range of effects in diverse tissues, including stimulation of adult skeletal muscle growth, promotion of embryonic neural development, neurite out- growth and neural regeneration, insulin-like glucose regula- tion, and maintenance of neurotransmitter and peptide secre- tion (LeRoith and Roberts, 1993). Although widely expressed in all tissues, IGF-I is mainly produced by cells lacking classical secretory granules; storage in granules is thus unlikely to be a major mechanism regulating its secretion in all tissues. The finding that IGF-I is secreted by the regulated pathway in AtT-20 cells, therefore, is unexpected. Since recent studies sug- gest that AtT-20 cells may also express IGF-I receptors (Fielder et al., 1993), packaging of IGF-I into granules could provide a local regulated response in an autocrine fashion. Future studies may reveal the physiological significance of these observations.

Human and mouse IGF-I exist in several isoforms which are derived from alternative mRNA splicing (Jansen et al., 1983; Bell et al., 1986). These forms share similar NH,-terminal se- quences, but differ in the COOH-terminal extensions which comprise the E peptides. We do not know which form(s) of IGF-I is produced in AtT-20 cells. Since the human synthetic IGF-I used in the transfection studies encodes the mature form of IGF which lacks the E peptide, the E peptide is not necessary for granular localization. Presently we cannot exclude the possi- bility that the E peptide may encode a redundant sorting sig- nal. Since both insulin and IGF-I are sorted into granules, the targeting information is likely to reside within structures that are similar between these two proteins. The short 8-amino acid D peptide of IGF-I, as well as the 12-amino acid connecting peptide between B and A peptides, are completely different in the two proteins and thus probably do not confer proper local- ization (Fig. 1). This conclusion is supported by our previous studies demonstrating that the C peptide of proinsulin (which corresponds to the connecting peptide of IGF-I) is not necessary for granule targeting (Powell et al., 1988). The sorting informa- tion is therefore likely to be contained within the B and A chains which are 45% identical between IGF-I and insulin (shaded residues in Fig. 1).

Synthesis and Targeting of IGF-1 27123

we have shown that immunoreactive IGF-I is secreted from AtT-20 cells. Since the available antibodies do not work in immunoprecipitation experiments, all of these experiments re- lied on a sensitive RIA to detect IGF-I. One potential problem of the RIA is that it does not distinguish between de novo synthesis from cultured cells and IGF-I contamination from growth media. This latter possibility was ruled out by the ob- servations that 1) control dishes treated with growth media and washed identically did not yield comparable signals (Fig. 2), 2) cycloheximide treatment of cells abolished the signals from the media and reduced the signals from the cells (Fig. 2), and 3) cyclic AMP analogues enhanced the signals detected in the media and caused a corresponding decrease in the signals recovered from the cell extract (Fig. 3). The signals are also unlikely to be due to interference from IGF-I-binding proteins, since the samples were treated to remove IGFBPs. In control experiments we determined that this procedure removes >99% of the binding proteins as assayed by Western ligand blot (data not shown).

IGF-I was shown to be targeted to the regulated secretory pathway in AtT-20 cells by three criteria. Cycloheximide treat- ment revealed a storage pool within the cells (Fig. 21; secretion of both IGF-I and ACTH is stimulated by secretagogue (Fig. 3 and Fig. 5); and subcellular fractionation of IGF-I transfected AtT-20 cells showed that IGF-I and ACTH co-purify in the dense fractions of equilibrium density gradients where secre- tory granules migrate (Fig. 6, Table I). In this regard, IGF-I behaves very similarly to transfected insulin in AtT-20 cells; in the case of insulin, we have also shown that it is packaged in the same granule as ACTH (Orci et al., 1987). ACTH and IGF-I have sorting indices of 7.8 and 4.0, respectively, indicating that they are sorted to the regulated secretory pathway at least 40-fold more efficiently than TG (sorting index of 0.1, see Table I). The ACTH sorting index of 7.8 is higher than the 3.0 value that we have previously obtained using a pulse-chase labeling protocol.2 This is due to the fact that the ACTH antibodies used for the RIA have higher affinities for the mature forms of ACTH than the precursor POMC. Since constitutive secretion of POMC-related peptides is mostly in the form of the precursor, its contribution is underestimated. Taking this consideration into account, endogenous IGF-I is sorted at an efficiency com- parable to ACTH. Although transfected hIGF-I has a lower sorting index (1.51, it is still sorted significantly better than TG.

The liver secretes the majority of IGF-binding proteins (IG- FBPs) which are complexed with IGF-I in the serum (Baxter and Martin, 1989; Sara and Hall, 1990). These IGFBPs exert both negative and positive regulatory effects that include con- trol of interactions with cell surface receptors and prolonging the serum half-life of IGF-I (Clemmons et al., 1993). Much like the distribution of IGF-I, the majority of IGFBPs are synthe- sized by the liver, but other tissues are also known to synthe- size and secrete these binding proteins (Clemmons et al., 1993). Our Western ligand blot analysis detected two major IGFBPs (24 kDa and 29 kDa) in AtT-20 cell extracts and conditioned media. These binding proteins have been recently purified and identified as IGFBP-4 and IGFBP-5, respectively (Fielder et al., 1993).

We have found that IGFBP-4 is secreted coordinately with IGF-I from AtT-20 cells upon stimulation; its sorting efficiency is similar to transfected IGF-I (Fig. 7, Table I). An attractive mechanism for this coordinate sorting is that IGF-I and IG- FBP-4 form a complex in the secretory pathway and this com- plex is sorted as a whole into dense-core granules. This scenario requires that only one protein carries the necessary sorting information; the other protein may be passively targeted to the

H.-P. H. Moore, unpublished results.

granules by binding to the sorted protein. However, there is as yet no evidence for the association of IGF-I with any of its binding proteins during intracellular transport. Furthermore, insulin, which does not bind to IGFBPs (data not shown) and has no known serum-binding proteins, is packaged efficiently into dense-core granules (Orci et al., 1987). Thus, complex for- mation with serum-binding proteins cannot be a prerequisite for the granular targeting of insulin-related peptides.

Comparison of sequences of IGFBP-4 and IGFBP-5 showed that they share a high degree of homology in a cysteine-rich domain, but are quite divergent outside this region. In particu- lar, IGFBP-5, but not IGFBP-4, contains a highly basic COOH- terminal domain which is thought to mediate high affinity binding to heparan sulfate-containing extracellular matrix molecules (Clemmons et al., 1993). Heparan sulfate proteogly- cans are known to exit cells by the constitutive pathway (Tooze and Huttner, 1990). This interaction may explain the preferen- tial entry of IGFBP-5 into the constitutive secretory pathway.

The prevailing model for sorting of proteins into the regu- lated secretory pathway invokes a concentration-dependent, pWCa2+-induced, aggregation of secretory products at the TGN (Bauerfeind and Huttner, 1993). An important feature of the aggregation model is that sorting should be concentration-de- pendent: secretogranin I1 aggregates under low pH and high calcium (aggregative) conditions at concentrations above 1 mg/ ml, but fails to do so at 70 pg/ml (Chanat and Huttner, 1991). However, it is difficult to explain the sorting of IGF-I on the basis of this model. The amount of IGF-I produced endog- enously by AtT-20 cells is extremely low. Based on RIA results, IGF-I is produced at -1500 fold lower levels than ACTH. The estimated concentration of POMC-related peptides in mature granules is 160 mg/ml (assuming vesicle radius of 1000 A and 0.66 fg peptides per vesicle, Gumbiner and Kelly, 1981). Assum- ing that regulated secretory proteins are concentrated 3 to 10 fold during granule maturation, one expects approximately 15-50 mg/ml POMC-related products in the TGN. Therefore, the calculated concentration of IGF-I in the TGN is probably less than 30 pglml and thus not high enough to result in ag- gregation. Furthermore, the transfected hIGF-I would be ex- pected to aggregate and sort more efficiently than endogenous IGF-I since it is expressed at 35 fold higher levels (-1 mg/ml in the TGN). However, this is not the case (Table I).

To directly address whether IGF-I aggregates in within the lumen of secretory compartments, we have performed assays under conditions known to induce the aggregation of SgII. We have found that unlike SgII, IGF-I is released from permeabi- lized vesicles under aggregative conditions. This was not due to cell type differences since proinsulin behaved similarly in both AtT-20 and PC-12 cells. These results suggest that neither IGF-I nor proinsulin form large aggregates in the proposed TGN milieu as has been shown for SgII, and raise the possibil- ity that these peptides may sort by a different mechanism.

An earlier electron microscopic study showed that proinsulin is mostly associated with membranes of the Golgi stacks in pancreatic islet cells, but this membrane association is no long- er observed in the TGN where most of the peptides are found in the lumen of budding granules (Orci et al., 1984). Our biochemi- cal studies of proinsulin correlate well with this morphological behavior, provided that the milieu of the TGN indeed mimics that of our aggregative conditions. Surprisingly, we found that the constitutively secreted protein TG was not released from permeabilized vesicles under either non-aggregative or aggre- gative conditions. TG lacks a transmembrane domain and should behave as a soluble protein. The reason for this obser- vation is unknown, although it is interesting to note that Chanat and Huttner (1991) reported a similar behavior for

27124 Synthesis and Targeting of IGF-I another constitutively secreted protein heparan sulfate proteo- glycan (hsPG). The molecular mechanisms for this membrane attachment and its possible relationship with sorting between constitutive and regulated pathways deserve close future examination.

Acknowledgments-The authors wish to thank Drs. €3. Denver, K. Chan and other members of the Nicoll lab for their help with the IGF-I RIA, Dr. P. Licht for the RIA potency program, and the Moore laboratory and Dr. E. Hildebrandt for critical discussion.

REFERENCES

Aguado, F., Sanchez-Franco, F., Cacidedo, L., Fernandez, T., Rodrigo, J., and Mar-

Arvan, P., Kuliawat, R., Prabakaran, D., Zavacki, A. M., Elahi, D., Wang, S., and

BagneI1, C. A. (1991) Steroids 56,242-246 Bauerfeind, R., and Huttner, W. B. (1993) Curr Opin. Cell Biol. 5,628-635 Baxter, R. C., and Martin, J. L. (1989) Prog. Growth Factor Res. 1 , 4 9 4 8 Bayne, M. L., Cascieri, M. A,, Kelder, B., Applebaum, J., Chicchi, G., Shapiro, J. A.,

Pasleau, E, and Kopchick, J. J. (1987) Proc. Natl. Acad. Sci. U. S. A. 8 4 , 2 6 3 s 2642

Bedarkar, S., firnell, W. G., Blundell, T. L., and Schwabe, C. (1977) Nature 270, 449451

Bell, G. I., Swain, W. F., Pictet, R., Cordell, B., Goodman, H. M., and Rutter, W. J.

tinez-Murillo R. (1992) Neurosci. Lett. 135, 171-174

Pilkey, D. (1991) J. Biol. Chem. 286, 14171-14174

Bell, G. I., Stempien, M. M., Fong, N. M., and Rall, L. B. (1986) Nucleic Acids Res. (1979) Nature 282,525-527

Bichell, D. P., Kikuchi, IC, and Rotwein, P. (1992) Mol. Endocrinol. 6, 1899-1908 Blundell, T. L., Bedarkar, S., Rinderknecht, E., and Humbel, R. E. (1978) Proc.

Blundell, T. L., and Humbel, R. E. (1980) Nature 287, 781-787 Bowsher, R. R., Lee, W.-H., Apathy, J. M., O'Brien, P. J., Ferguson, A. L., and Henry,

14,7873-7882

Natl. Acad. Sci. U. S. A. 75, 180-184

Brion, C., Miller, S. and Moore, H.-P. H. (1992) J. B i d . Chem. 267, 1477-1483 Bureess. T. L.. and Kellv. R. B. (1987)Annu. Reu. Cell Biol. 3. 243-293

D. P. (1991) Endocrinology 128,8054314

Castle, A. M., Stahl, L. E; and Castle,J. D. (1992) J. Biol. Chem. 267,13093-13100 Chanat, E., and Huttner, W. B. (1991) J. Cell B i d . 11, 1505-1519 Chanat. E.. Weiss, U., Huttner, W. B., and Tooze, S. A. (1993) EMBO J. 12,

~~~~, ~ ~,

2159-2168

Methods Cell Biol. 43,263-288 Chavez, R. A,, Chen, Y.-T., Schmidt, W. K., Carnell, L., and Moore, H:P. H. (1994)

Clemmons, D. R., Jones, J. I., Busby, W. H., and Wright, G. (1993) Ann. N . L Acad.

DErcole, A. J., Stiles, A. D., and Underwood, L. E. (1984) Proc. Natl. Acad. Sci. Sci. 692,lO-21

Faden, V. B., and Rodbard, D. (1975) Radioimmunoassay Data Processing: The U. 5. A. 81,935-939

'Logit-Log'Method and Scatchard Plot, National Technical Information Service, US. Department of Commerce, Washington, DC.

Fagin, J. A., F'ixley, S., Slanina, S., Ong, J., and Melmed, S. (1987) Endocrinology 120,2037-2043

Fielder, P J., Thordarson, G., English, A,, Rosenfeld, R. G., and Talamantes, F. (1992) Endocrinology 131,261-267

Fielder, P. J., Tauber, J. P., Wilson, K. F., and Pham, H. M. (1993) Growth Regul. 77, 1683-1689

Gerdes, H. H., Rosa, P., Phillips, E., Baeuerle, P. A,, Frank, R., Argos, P, and Huttner, W. B. (1989) J. B i d . Chem. 264, 12009-12015

Gorr, S.-U., Shioi, J., and Cohn, D. V. (1989)Am. J. Physiol. 257, E247-EZ54 Grimes, M., and Kelly, R. B. (1992) J. Cell B i d . 117, 539-549 Gumbiner, B., and Kelly, R. B. (1981) Proc. Natl. Acad. Sei. U. S. A. 78, 318322 Jansen, M., van Schaik, F. M., Ricker, A. T., Bullock, B., Woods, D. E., Gabbay, K.

H., Nussbaum,A. L., Sussenbach, J. S., andVan den Brande, J. L. (1983)Nature 306, 609411

Kizer, J. S., and Tropsha, A. (1991) Biochem. Biophys. Res. Commun. 174,586-592 Kuliawat, R., and Arvan, P. (1992) J. Cell Biol. 118, 521-529 LeBlond, F. A., Viau, G., Laine, J., and LeBel, D. (1993) Biochem. J. 291,289-296 LeRoith, D., McGuinness, M., Shemer, J., Stannard, B., Lanau, F., Faria, T. N.,

Kato, H., Werner, H., Adamo, M., and Roberts, C. T. J. (1992) B i d . Signals 1, 173-181

LeRoith, D., and Roberts, C. T. J. (1993) Ann. N . I: Acad. Sci. 692, 1-9 Miller, S. G., and Moore, H.-P. H. (1990) Curr. Opin. Cell Biol. 2, 642-647 Moore, H.-P. H., and Kelly, R. B. (1985) J. Cell B i d . 101, 1773-1781 Moore, H.-P. H., and Kelly, R. B. (1986) Nature 321, 443446 Moore, H.-P. H., Walker, M., Lee, F., and Kelly, R. B. (1983) Cell 35, 531-538 Murray-Rust, J., McLeod, A. N., Blundell, T. L., and Wood, S. P. (1992) BioEssuys

Neerman-Arbez, M., and Halban, P. A. (1993) J. Biol. Chem. 268, 16248-16252 Orci, L., Ravazzola, M., Amherdt, M., Perrelet, A,, Powell, S., Qu~M, D., and

Orci, L., Ravazzola, M., and Perrelet, A. (1984) Proc. Natl. Acad. Sci. U. 5. A. 81,

Powell, S. K., Orci, L., Craik, C . S., and Moore, H.-P. H. (1988) J. Cell Biol. 106,

Quinn, D., Orci, L., Ravazzola, M., and Moore, H.-P H. (1991) J. Cell B i d . 113,

Rosa, P., Weiss, U., Pepperkok, R., Ansorge, W., Niehrs, C., Stelzer, E. H., and

Roy, P., Chewier, D., Fournier, H., Racine, C., Zollinger, M., Crine, P., and Boileau,

Salamero, J., Sztul, E. S., and Howell, K. E. (1990) Proc. Natl. Acad. Sci. U. 5. A.

Sara, V. R., and Hall, K (1990) Physiol. Reu. 70, 591414 Saucan, L., and Palade, G. E. (1992) Hepatology 15, 714-721 Sossin, W. S., Fisher, J. M., and Scheller, R. H. (1990) J. Cell Biol. 110, 1-12 Stoller, T. J., and Shields, D. (1989) J. Cell Biol. 108, 1647-1655 Sures, I., Goeddel, D. V., and Ullrich, A. (1980) Science 208,57-59 Tam, W. W. H.,Andreasson, K. I., and Loh, Y. P. (1993) Eur J. Cell Biol. 62,294-306 Tooze, J., Tooze, S. A,, and Fuller, S. D. (1987) J. Cell Biol. 105, 1215-1226

Turkewitz, A,, Madeddu, L., and Kelly, R. (1991) EMBO J. 10, 1979-1987 Tooze, S. A,, and Huttner, W. B. (1990) Cell 60, 837-847

Yoo, S. H., and Albanesi, J. P. (1990) J. Biol. Chem. 266, 14414-14421 Zapf, J., and Froesch, E. R. (1986) Horm. Res. 24, 121-130

14,325331

Moore, H.-P. (1987) Cell 61, 1039-1051

6743-6746

1843-1851

987-996

Huttner, W. B. (1989) J. Cell Biol. 109, 17-34

G. (1991) Mol. Cell. Endocrinol. 82, 237-250

87,7717-7721