Differential Biosynthesis and Intracellular Transport of Follistatin Isoforms and Follistatin-Like-3
http://www.100md.com
《内分泌学杂志》
Reproductive Endocrine Unit, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts 02114
Abstract
Follistatin (FST) and FST-like-3 (FSTL3) are structurally related proteins that bind and neutralize activin and closely related members of the TGF superfamily. Three FST isoforms (FST288, FST303, and FST315) are produced from the Fst gene that are primarily secreted proteins. FSTL3 is secreted, but is also observed within the nucleus of most cells. We used pulse-chase 35S labeling to examine the biosynthetic and intracellular transport patterns that lead to differential secretion and intracellular retention of these proteins. Among the FST isoforms, FST315 was secreted fastest and FST288 was secreted more slowly, with some remaining intracellular. In contrast, FSTL3 was secreted the slowest, with newly synthesized proteins being both secreted and trafficked to the nucleus. This nuclear FSTL3 was N-glycosylated, although not to the same degree as secreted FSTL3. Both FST and FSTL3 have two Mets in their signal sequence. Mutation of the first Met in FST288 eliminated protein translation, whereas FSTL3 could be translated from either Met. However, although FSTL3 translated from the second Met, which had no signal sequence, was confined to the nucleus, it was not glycosylated. Interestingly, this FSTL3 retained activin-antagonizing activity. Thus, although bioactive, nuclear FSTL3 can be translated from the second Met when the first Met is mutated, the glycosylated nuclear FSTL3 produced endogenously indicates that a different mechanism must be used under natural conditions that apparently includes N-glycosylation. Moreover, the differential biosynthetic and intracellular transport patterns for FST288 and FSTL3 suggest that these two activin-binding proteins may have distinct intracellular roles.
Introduction
LIKE OTHER MEMBERS of the TGF- superfamily, activin has wide-ranging actions in regulating cellular growth, proliferation, and differentiation in both adult and embryonic tissues. It is therefore not surprising that a number of regulatory mechanisms have evolved to modulate activin signaling intracellularly and extracellularly. Among the extracellular regulators are inhibin, a negative feedback regulator of activin-mediated FSH biosynthesis, and follistatin (FST), a monomeric glycoprotein that irreversibly binds and neutralizes activin. Although originally identified in follicular fluid, FST is actually produced in many of the same tissues as activin and probably acts primarily as an autocrine/paracrine regulator of activin activity (reviewed in Refs.1 and 2).
The Fst gene is composed of six exons, each coding for distinct functional regions of the protein (3, 4), including a signal peptide, an N-domain that is critical for binding activin (4), three 10-cysteine (Cys) FST domains that also participate to varying degrees in activin binding (5), and a C-terminal acidic tail that is present in the full-length FS315 molecule. However, through alternative splicing of intron 5, a second Fst mRNA is produced that codes for a protein of 288 amino acids (FST288) that does not have the C-terminal tail region (3). The physiological significance of this deletion is based on the finding that the first FST domain contains a basic sequence that has been shown to bind cell surface, heparin-sulfated proteoglycans [heparin-binding sequence (HBS)] (6), a property that is inhibited by the acidic C-terminal tail in FST315 (7). Purification of FST from porcine follicular fluid identified FST303, a third FST protein that is apparently produced via proteolytic cleavage of FST315 within the C-terminal tail (7). Interestingly, FST303 has intermediate cell surface-associating properties compared with FST288 and FST315, suggesting that the acidic C-terminal tail length differentially modulates the ability of the HBS to associate with heparin sulfate proteoglycans (7). Taken together, these studies demonstrate that the Fst gene produces three protein forms that differ in cell surface binding and in at least some of their biological activities.
FST-like-3 (FSTL3), also known as FST-related gene (8) and FST-related protein (9), binds activin with nearly the same affinity and irreversibility as FST (10). The gene structure of FSTL3 is similar to that of FST, containing one less FST domain, but sharing overall 45% primary sequence identity (9). Thus, FSTL3 and FST share a number of structural and functional properties suggesting that there may be substantial overlap in their biological actions in vivo.
A number of differences between FSTL3 and FST have also been described, including the lack of an heparin-binding sequence, which prevents FSTL3 from binding cell surface proteoglycans such as FST (11). In addition, although FSTL3 has a signal sequence and is secreted from cells in which it is highly expressed or overexpressed (8, 10), FSTL3 protein has also been localized within the nucleus of numerous cell lines, primary granulosa cells (10), and tissue sections (12), indicating that at least some FSTL3 remains intracellular. FST, in contrast, has been observed in the cytoplasm of some cells, including granulosa cells, but not in the nucleus (10). These observations suggest that biosynthesis and intracellular trafficking of FSTL3 and FST are differentially regulated.
Several mechanisms have been described where proteins with secretory signal peptides, which direct translated proteins to the endoplasmic reticulum (ER), can also be transported to the nucleus or other intracellular organelles. Alternative splicing can alter a signal peptide or nuclear localization signal in some mRNAs, but not in others, allowing a protein to be transported to different compartments (13). In addition, translation of some proteins can initiate at different methionines (Met) or at non-AUG codons, creating proteins with variable or even no signal peptides (14, 15). Interestingly, both FST and FSTL3 have a second Met in their signal sequences, which, if used, would produce a protein with a shortened and possibly nonfunctional signal peptide. This raises the possibility that the use of alternate Mets to initiate FST or FSTL3 translation could allow a portion of translated protein to remain intracellular.
The current study was undertaken to compare the kinetics of FST isoform and FSTL3 biosynthesis, trafficking, and secretion in transfected cells as well as HeLa cells, where FSTL3 is naturally secreted, using pulse-chase metabolic labeling. Our results confirm the localization of substantial newly synthesized FSTL3 in the nucleus for up to 8 h. In contrast, FST315 is secreted most rapidly, followed by FST303 and FST288, with a portion of FST288 being retained in the cell for up to 4 h, but not in the nuclear fraction. Our results also show that endogenous nuclear FSTL3 is transported to the nucleus and is N-glycosylated, although not to the same extent as secreted FSTL3. There are two Mets in the signal sequence of FSTL3, and protein translated from the second Met, which has no signal peptide, is transported to the nucleus but is not glycosylated. This distinction in glycosylation suggests that under natural conditions, nuclear FSTL3 is not derived from translation initiation at the second Met, but probably enters the ER to get glycosylated, then transported back to the cytoplasm and eventually to the nucleus. Taken together, these results indicate that FSTL3 and FST are both secreted and retained within the cell, albeit in different compartments and by different mechanisms, consistent with the concept that they may have differential intracellular actions in addition to their known extracellular functions.
Materials and Methods
Medium and cell culture
Chinese hamster ovary (CHO) and HeLa cells (American Type Culture Collection, Manassas, VA) were cultured in medium containing 45% Ham’s F-12, 45% DMEM, 10% heat-inactivated fetal bovine serum (Invitrogen Life Technologies, Inc., Grand Island, NY), 1% L-glutamine, and antibiotics (100 IU/ml penicillin and 100 μg/ml streptomycin sulfate) and maintained in a humidified atmosphere of 5% CO2. HEK 293 cells were maintained in RPMI 1640 with 10% fetal bovine serum and supplements as described above.
Transfection and selection of stably secreting clones
Human FSTL3 (provided by Millennium Pharmaceuticals, Cambridge MA), human FST288 or human FST315 (gifts from Dr. Shimasaki, University of California, San Diego, CA) cDNAs were used as templates for PCR amplification of the coding sequence without a stop sequence, which were then subcloned into the pcDNA3.1-Myc-HIS (Invitrogen Life Technologies, Inc., Carlsbad, CA) vector. The FST303 expression construct was made using the FST315 template, but positioning the downstream primer to end at Gln303. CHO cells were transfected with expression constructs using Lipofectamine (Invitrogen Life Technologies, Inc.) according to the manufacturer’s instructions. Twenty-four hours after transfection, 600 μg/ml G418 (Mediatech, Herndon, VA) was added to the culture medium, and colonies were screened for secretion of FST or FSTL3 as appropriate. FST-transfected colonies were screened with our two-site FST immunoassay (16), whereas FSTL3 colonies screened by RIA as previously described (10). We originally examined both high and low secreting colonies, but no differences in intracellular trafficking were observed, so only the highest secreting clones were maintained for additional study.
Metabolic labeling studies
Cells were plated in 12-well (for the time-course experiment) or six-well (for the subcellular fractionation experiment) dishes (300,000 or 900,000 cells/well, respectively) in 1 or 3 ml medium (respectively). After 24 h, cells were washed twice and preincubated for 90 min with Cys-free DMEM (Invitrogen Life Technologies, Inc.), after which Cys-free DMEM containing 100 μCi/ml [35S]Cys (NEN Life Science Products, Boston, MA) was added for an additional 30 min. After the labeling step, the cells were washed once with medium and incubated for the indicated chase times (see figures) until harvesting.
Cell extraction and fractionation
At the indicated time points, medium was collected ("secreted") and replaced with fresh medium containing 10 μg/ml heparin sulfate (Sigma-Aldrich Corp., St. Louis, MO) for 5 min to release cell surface-bound FST ("cell surface"). All conditioned medium samples were centrifuged to remove nonadherent cells and debris before being stored for later analysis. After heparin treatment, cells were extracted in 1 ml TNE buffer [20 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA, 0.1% Nonidet P-40, and a protease inhibitor cocktail (Roche, Mannheim, Germany)]. After incubating for 15 min on ice, the cells were subjected to two rounds of freeze/thaw to facilitate membrane lysis and centrifuged. This supernatant was collected as cellular. No cell surface-associated FST315 or FSTL3 was observed (data not shown).
To determine subcellular localization of labeled FSTL3 and FST isoforms, nuclear and cytoplasmic extracts were prepared by differential centrifugation as described previously (17). After removing conditioned medium, which is shown in figures as secreted (S) protein, cells were dislodged from the dish and pelleted by centrifugation at 1000 x g for 15 min. The cells were then resuspended in 1 ml hypotonic buffer [10 mM Tris-HCl (pH 8), 3 mM MgCl2, 1 mM EDTA, plus protease inhibitors], allowed to sit on ice for 10 min to induce swelling, and vortexed to disrupt the cell membrane. Lysates were centrifuged at 1000 x g for 10 min, and the supernatant was collected as cellular (C). The nuclear pellet was resuspended in 1 ml TNE buffer and subjected to three freeze/thaw cycles. After a 5-min centrifugation at 1000 x g, the supernatant (soluble nuclear extract) was collected as nuclear (N).
To verify that the rapid fractionation method used for the time-course studies faithfully separated nuclear fractions from ER and other cytoplasmic proteins, nuclei free from ER membranes were isolated essentially as previously described (18). Two 10-cm plates of HeLa cells were grown to confluence, one was extracted directly, and the other was metabolically labeled for 30 min and chased for 2 h as described above. Both plates were then treated identically during the fractionation process. HeLa cells were collected by scraping, gently centrifuged, washed with hypotonic buffer [20 mM Tris (pH 7.5), 3 mM CaCl2, and 2 mM MgCl2], and then resuspended in 5 packed cell volumes of hypotonic buffer and incubated on ice for 10 min. Swollen cells were homogenized in a Dounce homogenizer (Kontes Co., Vineland, NJ) and centrifuged at low speed (1000 x g). The postnuclear supernatant was further processed as described below. The crude nuclear pellet was further purified by resuspension in 4 ml sucrose buffer I [0.32 M sucrose, 10 mM Tris (pH 8.0), 3 mM CaCl2, 2 mM magnesium acetate, 0.1 mM EDTA, 1 mM dithiothreitol, and 0.5% Nonidet P-40], incubation on ice for 15 min, and then layering the nuclear suspension over 4.4 ml sucrose buffer II [2 M sucrose, 10 mM Tris (pH 8.0), 5 mM magnesium acetate, 0.1 mM EDTA, and 1 mM dithiothreitol], and centrifugation at 30,000 x g for 45 min. Purified nuclei were resuspended in hypotonic buffer, after which high salt buffer [1.2 M NaCl and 20 mM Tris (pH 7.5)] was added for 30 min at 4 C. The resulting suspension was centrifuged for 20 min at 4 C, and the supernatant, containing soluble nuclear proteins, was collected, diluted with an equal volume of hypotonic buffer, and analyzed by SDS-PAGE at the nuclear fraction.
The postnuclear supernatant obtained above was centrifuged at 16,000 x g for 20 min at 4 C. The supernatant was collected as the cytosolic fraction, and the pellet was collected as the crude ER/membrane fraction by resuspending this pellet in modified RIPA buffer [20 mM Tris (pH 7.5), 150 mM sodium chloride, 50 mM sodium fluoride, 0.1 mM EDTA, 0.5% Nonidet P-40, 0.5% deoxycholate, and 0.2% sodium dodecyl sulfate] before SDS-PAGE analysis.
Cytosolic, ER/membrane, and purified nuclear fractions were analyzed by immunoprecipitation with the FSTL3 antibody as described above (35S-labeled plate) or by Western blotting for calnexin (Abcam, Cambridge, MA), an ER-resident transmembrane protein, to localize the ER membranes (nonlabeled plate).
Detection of FSTL3 and FST
FSTL3 and FST were immunoprecipitated from each fraction using 100 ng of our previously described polyclonal anti-FSTL3 antibody (10) for FSTL3 or 500 ng anti-Myc polyclonal antibody (Upstate Biotechnology, Inc., Charlottesville, VA) for FST288 and FST303. FST315 was undetectable using the Myc antibody, so 100 ng monoclonal anti-FST 7FS30 covalently cross-linked to paramagnetic particles that are typically used in the FS immunoassay (16) was used. To reduce nonspecific protein precipitation, samples were preincubated with either 10 μl protein A (Pierce Chemical Co., Rockford, IL) or 10 μl paramagnetic particles without antibody (FST315 only) for 1 h, followed by centrifugation. Antibodies were incubated with cleared supernatants overnight, followed by addition of 10 μl protein A (except FST315 samples) at room temperature for 2 h. After centrifugation, samples were washed six times with TNE buffer, then boiled in 2x SDS-PAGE buffer containing -mercaptoethanol (Invitrogen Life Technologies, Inc.), and reduced proteins were separated on 8–16% Novex Tris-glycine gels (Invitrogen Life Technologies, Inc.). Proteins immunoprecipitated from untransfected CHO cells are designated nonspecific in the figures and were not included in the analysis.
After SDS-PAGE, gels were dried and quantitated by phosphorimager. For each point in the time course, analyzed samples include secreted (conditioned medium), cell extracts (cellular), and cell surface (released by heparin sulfate). Representative gels are shown in Fig. 1, and quantitative analyses are presented as a percentage of the total FST or FSTL3 of all three fractions at each time point, representing the mean of three replicate experiments, and are shown in Figs. 2 and 4. The relative proportion of labeled protein in each compartment was calculated for each time point as a percentage of the total counts at that time point.
Deglycosylation
The glycosylation status of some immunoprecipitated samples was assessed using recombinant N-glycanase (Glyko, Inc., Novato, CA) according to the manufacturer’s instructions, followed by SDS-PAGE as described above.
Creation and analysis of Met mutants
The cDNAs for FST288 and FSTL3 used to make stable cell lines (see above) were used as a template for site-directed mutagenesis (QuikChange, Stratagene, La Jolla, CA). Two FST288 mutants were created where either the first Met (M-29) or the second Met (M-8) in the signal sequence was changed to alanine (Ala). Similarly, the first Met (M-26) or the second Met (M1) in FSTL3 was changed to Ala. Because FSTL3 protein was produced from the FSTL3 M-26A mutant cDNA, a third mutant was prepared in which both Mets (M-26 and M1) were changed to Ala and used for in vitro translation to monitor translation in the absence of Mets in the signal sequence.
In vitro translation of FSTL3
Expression vectors containing FSTL3 wild-type (WT) or Met mutants were linearized downstream of the stop codon and used as templates in TnT reticulocyte lysate in vitro translation reactions in the presence of [35S]Cys according to the manufacturer’s protocol (Promega Corp., Madison WI). Translated products were resolved on 12% SDS-PAGE gels (Bio-Rad Laboratories, Hercules, CA) under reducing conditions and detected by autoradiography after 24-h exposure.
Electrophoresis and Western blot of signal peptide mutants
HEK 293 cells, which have a superior activin reporter response compared with CHO cells, were transiently transfected with 125 ng FST or FSTL3 mutant cDNAs and cultured for 48 h, after which they were extracted with passive lysis buffer (Promega Corp.). Equal volumes (10 μl) of whole cell extract (E) or 20 μl conditioned medium (M) were subjected to 12% SDS-PAGE using Tris-HCl Ready Gel from Bio-Rad Laboratories and were transferred onto polyvinylidene difluoride membrane (Bio-Rad Laboratories) following the manufacturer’s recommendation. After blocking in Tris-buffered saline supplemented with 10% dry milk and 0.2% Tween 20, blots were probed with anti-Myc antibodies (1 μg/ml; clone 4A6, Upstate Biotechnology, Inc., Lake Placid, NY), followed by goat antimouse horseradish peroxidase-conjugated antibody (1:7500 dilution; Santa Cruz Biotechnology, Santa Cruz, CA), and immunoreactivity was visualized using Western Lightning chemiluminescence reagent (PerkinElmer, Boston, MA).
Activin/FST in vitro bioassay
Transient transfection of 293 cells was performed in 24-well trays using Effectene (Qiagen, Valencia, CA) and a total of 225 ng DNA, which included 80 ng of the Smad3-responsive reporter CAGA-Luc (19), 15 ng pRL-TK (Promega Corp.), 5 ng activin-A cDNA, and 0.25–25 ng FST or FSTL3 mutant cDNA balanced with empty vector to a total of 125 ng. Culture medium was refreshed after 16 h, and the following day, cells were extracted with Passive Lysis Buffer (Promega Corp.) and assessed for firefly and Renilla luciferase activities using the Dual-Luciferase Reporter Assay System (Promega Corp.).
Immunocytochemistry
CHO cells were passaged onto sterilized glass coverslips (Fisher Scientific, Pittsburgh, PA) and transfected with cDNA constructs using Effectene reagent (Qiagen) according to the manufacturer’s protocol. After 24 h, transfected cells were fixed with 4% paraformaldehyde in PBS for 20 min and permeabilized with 0.1% sodium dodecyl sulfate in PBS for 4 min. Cells were then incubated with primary antibody diluted in 1% human -globulin (Sigma-Aldrich Corp.) in PBS. FSTL3 was detected with a purified rabbit polyclonal antibody (provided by Millennium Pharmaceuticals) at 5 μg/ml, and FST was detected using our monoclonal anti-FST antibody 7FS30 at 3.3 μg/ml. After a 1-h incubation, the cells were incubated with fluorescein-conjugated goat antirabbit IgG or Texas Red-conjugated goat antimouse (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) secondary antibodies (5 μg/ml). After 45 min, the cells were washed three with PBS, and coverslips were mounted onto glass slides with Vectashield (Vector Laboratories, Inc., Burlingame, CA). Immunofluorescent cells were visualized using an Axioskop microscope (Zeiss, New York, NY) equipped with a digital camera. Untransfected HeLa cells were treated in an identical manner to detect nuclear FSTL3.
Statistics
Time-course experiments were analyzed by two-way ANOVA for differences in both time and cell type; both had significant effects (P < 0.001). Differences between cell lines at each time point were identified by Tukey’s test. Similarly, results from cell fractionation analyses were analyzed by one-way ANOVA, and differences between cell lines for each fraction were determined using Tukey’s test. Differences of P < 0.05 were considered significant.
Results
Time course of FSTL3 and FST biosynthesis and secretion
Many cells that secrete FST contain both FST288 and FST315 mRNA transcripts and secrete multiple forms of FST protein. Therefore, to examine the biosynthesis and secretion of individual FST forms, cell lines were established by transfecting CHO cells (which do not secrete FST endogenously) with a single type of FST cDNA. During optimization experiments, we examined clones secreting different amounts of FST and found no differences in the fraction of secreted FST or FSTL3 relative to intracellular protein at different production levels (data not shown), so clones secreting maximal amounts of FST and FSTL3 were chosen for additional study. For comparison, we also examined HeLa cells, which naturally secrete FSTL3 as well as retain some FSTL3 in the nucleus (10).
Stable cell lines secreting individual FST isoforms (FST315, FST303, or FST315) or FSTL3 as well as HeLa cells that naturally secrete FSTL3 were metabolically labeled with [35S]Met for 30 min and analyzed by immunoprecipitation of secreted and cellular proteins at chase times up to 8 h. Representative autoradiograms for FST288, FST303, and FST315 are shown in Fig. 1, A–C, and for FSTL3 in HeLa and CHO cells in Fig. 1, D and E, respectively. FST315 was detected as two proteins of 40,000 and 46,000 in both conditioned medium (secreted) and cell extracts (cellular). As expected, cell surface-associated FST315 was undetectable in the medium after heparin release. FST 303 was secreted as two proteins of 38,000 and 42,000. Two specific bands were also detected in the lysates, but their electrophoretic migration was slightly faster than in the medium samples, suggesting differential glycosylation of intracellular and extracellular FST303. Although both secreted FST303 bands were detected in the membrane-associated (heparin-releasable) fraction, the amounts were substantially lower than membrane-associated FST288 and did not appear to change over time. Secreted FST288 consisted of three bands ranging in apparent molecular mass from 34,000–44,000, which were identical with those precipitated from cell lysates. Substantial cell surface-associated FST288 was released by heparin sulfate treatment. FST has been shown previously to be variably glycosylated at its two potential N-linked glycosylation sites (7), suggesting that the multiple specific FST bands we precipitated represent differentially glycosylated variants for each FST isoform.
FSTL3 was secreted as a single protein species of approximately 33,000 by HeLa and transfected CHO cells. FSTL3 was also detected in lysates from both cell lines as a single protein, but migrated more rapidly at 28,000. As expected, FSTL3 was never detected in association with cell surface proteoglycans (data not shown).
Results from three independent experiments for each cell line were combined by expressing immunoprecipitable FSTL3 or FST as a percentage of the total FSTL3 or FST in the sample. The time course for proteins secreted into the medium (Fig. 2A) revealed that FST315 was secreted fastest from cells, with more than half of the labeled protein in the medium by 30 min and more than 80% by 1 h. FST303 was secreted significantly more slowly (P < 0.01), reaching only 40% at 30 min and taking 2 h to release 80% of the labeled protein. Secretion rates into the medium for FST288 and FSTL3 (HeLa or CHO) were not significantly different from each other, but were significantly slower than those for FST315 and FST303 (P < 0.01). Only 20% of labeled FST288 or FSTL3 was detected in the medium at 30 min, but, like FST303, reached 80% by 2 h.
Consistent with the presence of a functional HBS on FST288 and FST303 (20), substantial amounts of secreted FST288 and FST303 were detected associated with cell membranes in the heparin-releasable fraction (Fig. 2B). By 30 min and for the duration of the experiment, FST288 associated with cell surface proteoglycans was significantly greater than FST303 (P < 0.01), and both were significantly greater (P < 0.01) than FST315, where the HBS was neutralized by the acidic C-terminal tail.
Because the cell surface FST most likely came from the secreted pool, we reanalyzed the secretory dynamics with the cell-associated proteins added to the medium proteins (Fig. 2, A and B) to create total secreted proteins (Fig. 2C). In this analysis, total secreted protein is not significantly different between the FST isoforms at 30 min. However, at 1 h, total secreted FST315 was significantly greater (90% vs. 65–70%; P < 0.01) than the other isoforms, indicating that this isoform is secreted more rapidly than the other isoforms over 60 min (Fig. 2C). In contrast, FSTL3 secretory rates were significantly slower than all three FST isoforms (P < 0.01) at both 30 (25%) and 60 (50–60%) min, indicating that intracellular trafficking of FSTL3 is quite distinct from that of all three FST forms. There were no significant differences in FSTL3 secretion dynamics between naturally secreting HeLa cells and transfected CHO cells at any time point except 1 h (P = 0.03), indicating that the slower secretory pattern of FSTL3 was not an artifact of transfection.
HeLa cells and FSTL3-transfected CHO cells had similar FSTL3 retention levels (>75%) and were both significantly (P < 0.01) greater than the 40–60% of the FST isoforms at 30 min (Fig. 2D). By 60 min, FST315 retention was significantly lower than that of all other cell lines (P < 0.01), with intracellular proteins amounting to only 10% of the total, whereas the others ranged from 30–40% of the total. Only at this 60 min point did FSTL3-transfected CHO cells contain more intracellular FSTL3 than endogenous FSTL3 in HeLa cells (P < 0.01). After 4 h, differences between the FST isoforms and FSTL3 were no longer significant and accounted for less than 10% of the total labeled FST.
Glycosylation of secreted and cellular FSTL3
To determine the cause of the slower migration of FSTL3 immunoprecipitated from medium compared with cell lysates, samples collected from CHO cells stably expressing FSTL3 as well as from HeLa cells, which naturally secrete FSTL3, were subjected to deglycosylation before SDS-PAGE analysis. The slower migrating FSTL3 species secreted from transfected CHO cells, which had an apparent molecular mass of 33,000 was reduced to approximately 26,000 after deglycosylation. The FSTL3 from cell lysates, which migrated at an apparent molecular mass of 28,000, was also reduced to 26,000 (Fig. 3A). These results, which were identical for HeLa cells (Fig. 3B), indicate that secreted FSTL3 is more heavily glycosylated than intracellular FSTL3, although intracellular FSTL3 is partially glycosylated, because the core deglycosylated proteins were of equal size.
Subcellular localization of retained FST and FSTL3
To determine the intracellular compartment in which FST and FSTL3 were retained, subcellular fractionation was performed before immunoprecipitation at chase times of 30 min and 2 h. Quantitative analysis of three replicate experiments is shown in Fig. 4, A and B, for chase times of 30 min and 2 h, respectively. At 30 min, more than 60% of FST315 was secreted, whereas less than 20% of FST288 was found in the medium. Approximately 20% of FST288 was located in the cytoplasmic fraction. At 2 h, less than 10% of FST315, but nearly 20% of FST288, was cytoplasmic, but neither was found in the nucleus. In contrast, more than 50% of FSTL3 in the HeLa cells was located in the nuclear fraction at 30 min and 25% at 2 h, significantly (P < 0.01) more than for FST288 or FST315. However, the transfected CHO cell FSTL3 appeared to be localized more to the cytoplasm than the nuclear fraction, similar to FST288. These results indicate that FSTL3 in naturally secreting HeLa cells was retained largely within the nuclear compartment, whereas FSTL3 in transfected CHO cells was found in both the nucleus and cytoplasm. Because our previous immunocytochemical study identified FSTL3 in the nucleus of transfected CHO cells (10), this difference may reflect the fact that the FSTL3 concentration within the nucleus may be limited so that excess production through transfection resulted mostly in secretion and cytoplasmic localization. This concept is supported by overexpressing FSTL3 in HeLa cells, which results in cytoplasmic localization of the extra FSTL3 (Fig. 4C) that obfuscates the nuclear localization of endogenous FSTL3 in these cells (Fig. 4D). Nevertheless, the transfected CHO cells had significantly more FSTL3 protein in the nuclear fraction than was present in the FST288- and FST315-secreting cells, indicating that newly synthesized FSTL3 was transported to the nucleus even in this CHO cell line.
To process the large number of samples necessitated by the time-course experiment, the subcellular fractionation protocol was weighted toward efficiency rather than complete separation of cellular compartments. To confirm that the [35S]FSTL3 was indeed located within the nucleus and not within ER membranes adjacent to the nucleus, a more stringent fractionation procedure was used to examine both 35S-labeled cells for FSTL3 localization as well as unlabeled cells for calnexin, a protein known to be inserted and resident within ER membranes. Greater than 95% of the calnexin was localized to the fraction containing ER and other membranes, whereas no radiolabeled FSTL3 was observed in this fraction (Fig. 5). Conversely, all the radiolabeled FSTL3 was found in the soluble nuclear fraction, whereas less than 5% of the calnexin was found in this material. These results demonstrate that newly synthesized FSTL3 is localized within the nucleus, confirming our time-course results.
Use of alternate translation start sites
N-Terminal sequencing of FSTL3 demonstrated that a majority of secreted (after cleavage of the signal peptide) FSTL3 begins with Met (10), which represents a nine-amino acid N-terminal extension compared with the first FST residue, which is Gly1 (Fig. 6A). In addition, FST has a second Met in its signal peptide 8 residues N terminal to Gly1. Thus, both FSTL3 and FST have two Mets in their signal sequence, the second of which almost completely deletes the signal peptide in FST and represents the beginning of mature FSTL3 (no signal peptide). We therefore hypothesized that one mechanism by which FSTL3 could be simultaneously targeted to the secretory pathway as well as to the nucleus would be if both Met residues were used to initiate translation. Moreover, because FST288 was partially retained in the cytoplasm (see above), a similar mechanism may explain its secretory dynamics.
To test this hypothesis, we made mutants of FST288 and FSTL3 with either the first or second Met changed to Ala. As shown in Fig. 6B, FST288 was detectable from both the WT and second Met mutant (M-8A) constructs in cell extracts (E) or as secreted FST288 in the medium (M). In contrast, no FST288 bands were detected in cell extracts or in the medium when the first Met was mutated (M-29A). Like FST, FSTL3 was detected in both cell extracts and conditioned medium in the WT and second Met mutant constructs (M1A). However, when the first Met was mutated, FSTL3 was still observed in cell extracts, although it was not secreted into the medium. These results suggest that FSTL3, but not FST288, could be translated from the second Met as well as the first.
To confirm this observation, we made an additional FSTL3 control construct with both Mets mutated (M-26,1A), and tested all four FSTL3 constructs for in vitro translational activity. Strong radiolabeled bands were observed for FSTL3 WT and second Met mutant constructs (M1A), consistent with the Western blot analysis (Fig. 7). Two weaker bands were observed for the first Met mutant (M-26A), one larger and one smaller than WT FSTL3. The double Met mutant construct produced a weak band that aligned with the upper FSTL3 band from the M-26A construct, indicating that this is not a product of a Met-initiated translation because there is no in-frame Met upstream of M-26. Thus, the lower band in the M-26A mutant is probably derived from translation initiation at the second Met, producing a protein shortened by the deletion of the 26-amino acid signal peptide.
We next tested whether the mutant FST288 or FSTL3 proteins retained the activin-inhibiting activity characteristic of WT proteins. As shown in Fig. 8, WT and second Met (FST M-8A) mutant FST288 were equally effective in inhibiting activin-induced reporter activity in 293 cells. However, the first Met mutant (FST M-29A) was inactive, consistent with the lack of detectable protein produced from this construct (see Fig. 7). In contrast, although both WT and second Met mutant (M1A) FSTL3 were equally effective in antagonizing activin-mediated reporter activity, the first Met mutant (M-26A) of FSTL3 did produce protein that was biologically active in inhibiting activin signaling. The reduced activity of this mutant may be the result of the lower expression levels, as shown in Fig. 6B. Thus, in contrast to FST, biologically active FSTL3 can be translated from the M-26A first Met mutant.
Immunocytochemical localization of Met mutant proteins
To determine whether FSTL3 translated from the second Met (FSTL3 M-26A) was targeted to the nucleus, we examined CHO cells transfected with each FSTL3 or FST288 mutant construct by immunocytochemistry. As previously reported (10), FSTL3 is weakly detectable in the nucleus of untransfected CHO cells (Fig. 9A). When transfected with WT FSTL3, very strong fluorescence was observed in the cytoplasm (Fig. 9B), which overwhelmed the weaker nuclear staining observed in controls. A similar pattern of immunofluorescence was observed when the second Met (M1A) was mutated (Fig. 9C). In contrast, when transfected with the first Met mutant construct (M-26A), these cells exhibited only nuclear staining (Fig. 9D) that was much enhanced compared with that of untransfected cells, indicating that FSTL3 produced by translation initiation at the second Met is transported to the nucleus. A similar analysis of FST288 Met mutants showed no background in untransfected cells (Fig. 9E), cytoplasmic FST in WT and second Met mutant (FST M-9A) constructs (Fig. 9, F and G), and no signal from the first Met mutant (FST M-9A) construct (Fig. 9H), consistent with lack of translation initiation for FST from the second Met in FST. For comparison, nuclear localization of naturally produced FSTL3 in HeLa cells is shown in Fig. 9, I (x40) and J (x100), where unstained nucleoli were observed surrounded by abundant nuclear FSTL3. Much weaker cytoplasmic FSTL3 staining was also observed at x40 (Fig. 9I). Taken together with the Western blot analysis and in vitro translation experiments, these results indicate that the use of alternative translation start sites might explain the simultaneous secretory and nuclear trafficking of FSTL3.
Glycosylation status of WT and Met mutant FSTL3
FSTL3 immunoprecipitated from cellular extracts of HeLa cells and transfected CHO cells was clearly glycosylated, although not to the same degree as secreted FSTL3 (see Fig. 3, A and B). However, translation initiation at the second Met in FSTL3 would completely eliminate the signal sequence, preventing FSTL3 from entering the ER and becoming glycosylated. Moreover, the M-26A band in Fig. 6B electrophoresed to a lower apparent molecular mass than WT, suggesting that this protein was not glycosylated. We therefore investigated whether cellular extracts from CHO cells transfected with WT, M-26A, or M1A mutant FSTL3 cDNAs produced glycosylated or nonglycosylated proteins. As expected, FSTL3 in cell extracts from WT or M1A cDNAs was glycosylated (Fig. 10). In contrast, FSTL3 from cells transfected with the M-26A FSTL3 cDNA (first Met mutant) did not have a change in apparent molecular mass after deglycosylation, indicating that this M-26A protein was not glycosylated and thus did not enter the ER. Because endogenous nuclear FSTL3 in HeLa cells is glycosylated (Fig. 3), this lack of glycosylation of M-26A FSTL3 suggests that under normal conditions, nuclear FSTL3 is not derived from protein translated from the second Met. Therefore, although FSTL3 translation can initiate from the second methionine, this mechanism does not appear to explain how endogenous (HeLa) or transfected (CHO) FSTL3 is transported to the nucleus.
Discussion
Using biosynthetic labeling, we determined that newly synthesized FSTL3 is both secreted and transported to the nucleus, and at 1 h, the amounts of FSTL3 in each pool are about equal. In addition, this nuclear FSTL3 is maintained for up to 8 h, suggesting that the nuclear FSTL3 may be turned over more slowly than secreted FSTL3. In comparison, the FST315 isoform, which is most closely related to FSTL3 because both are unable to bind membrane-associated proteoglycans, is secreted much more rapidly than FSTL3, with less than 10% retained within cells at 1 h. Thus, comparison of FST and FSTL3 exposes distinct differences in biosynthesis/secretory dynamics and regulation despite similar structures and bioactivities (21).
Both FST303 and FST288 isoforms, which are able to bind membrane-associated proteoglycans (7), are secreted more slowly than FST315, but faster than FSTL3. Moreover, both isoforms are partially retained in the cytoplasm for up to 8 h. Because intracellular membrane systems are virtually contiguous with the plasma membrane, this cytoplasmic FST288 and FST303 may be associated with intracellular membrane-associated proteoglycans, suggesting that intracellular FST288 and/or FST303 may have functions distinct from secreted FST. Interestingly, quantitative analysis of the biosynthesis time course demonstrates that significant amounts of both FST288 and FSTL3 remain inside cells for relatively long periods, although in different compartments (cytoplasmic vs. nuclear), suggesting that both FST288 and FSTL3 may have intracellular or intranuclear (respectively) actions in addition to their well-described extracellular roles in regulating activin and related TGF family members.
Both FSTL3 and FST have two Mets within their signal peptide. Thus, one potential mechanism for a protein to be simultaneously secreted and transported to the nucleus involves the use of either Met to initiate translation. In this model, transcripts starting at the first Met would be directed to the ER and ultimately secreted. In contrast, transcripts arising from the second Met would not have a signal peptide, and thus be translated in the cytoplasm and then directed to the nucleus via an as yet unidentified mechanism. We found that FSTL3 translation can initiate at either Met, but the first Met with a strong Kozak sequence produced more protein. When the first Met was replaced by Ala, FSTL3 protein was still produced, albeit less than from the first Met, and this FSTL3 was detectable in the nucleus, but not secreted. Interestingly, nuclear FSTL3 was capable of antagonizing activin-mediated reporter activity, suggesting that nuclear FSTL3 may have important biological roles regulating activin signaling, although the mechanism for this activity remains to be elucidated.
Immunocytochemical analysis demonstrated that endogenous FSTL3 (HeLa cells), transfected WT FSTL3, and transfected first Met mutant (M-26A) FSTL3 are all transported to the nucleus. Moreover, the deglycosylation studies for endogenous and WT FSTL3 are consistent with the secreted and nuclear FSTL3 being derived from different pools of protein, because they are differentially glycosylated. However, FSTL3 from the M-26A mutant, which was not secreted, is not glycosylated. These observations suggest that although they all can contribute to nuclear FSTL3, endogenous and transfected WT FSTL3 use a different mechanism than M-26A FSTL3. It is possible that the extracted intracellular FSTL3 represents protein extracted from ER before being fully glycosylated, whereas secreted FSTL3 has traversed the ER and Golgi and therefore has mature carbohydrate moieties. This would imply that under normal cellular conditions, all FSTL3 enters the ER, with a portion being shuttled out to the cytoplasm, perhaps by a chaperone, and then directed to the nucleus.
An alternative mechanism involved translation and glycosylation of FSTL3 in the cytoplasm (22), although this carbohydrate must be distinct from that found on secreted FSTL3 because they migrate differently on SDS-PAGE. This cytoplasmic FSTL3 might then be directed to the nucleus via an as yet unidentified chaperone, because FSTL3 contains no identifiable nuclear localization signals. Although proteins can be glycosylated on hydroxylated prolines in the cytoplasm in yeast (22), it is not likely that these carbohydrates would be removed using the enzymes that work on N-linked carbohydrates used in these studies. Thus, taken together with our observations that FSTL3 translated from the second Met (N-26A mutant) is not glycosylated, and extracts of WT FSTL3 clearly are glycosylated, our results favor a model in which FSTL3 is primarily translated from the first Met, enters the ER to get partially glycosylated, and then is removed from the ER before glycosylation is complete via a mechanism that might involve chaperone-mediated retrograde transport to the cytoplasm and then to the nucleus.
The conventional model of protein trafficking holds that proteins contain structural elements that regulate their transport to cellular compartments consistent with their functions. Thus, secretory and plasma membrane-bound proteins contain signal peptides that mediate transport of the nascent protein into the ER, where posttranslational modifications and multimerization can occur under regulated conditions. These proteins are then sorted through the Golgi into appropriate pathways and transported to the cell surface. In contrast, proteins destined for intracellular roles do not typically have signal peptides, are synthesized in the cytoplasm, and are then targeted to appropriate cellular compartments by chaperones or localization elements within their primary structure. A growing number of exceptions to this conventional model have been elucidated, allowing for secreted proteins to also be transported to various intracellular compartments, including the nucleus, where they may have actions distinct from those mediated by their cell surface receptors. For example, PTHrP inhibits proliferation in vascular smooth muscle cells when it interacts with its cell surface receptor, whereas nuclear PTHrP stimulates the proliferation of these cells (23). Mechanisms explaining differential biosynthesis and trafficking of PTHrP include 1) endocytosis of the hormone-receptor complex, where PTHrP can then be transported to the nucleus via a classical nuclear localization signal, 2) use of alternate translation start codons to produce transcripts that have shortened signal peptides and are unable to direct translocation to the ER, and 3) retrograde transport from the ER mediated by chaperones (14, 24). Other secreted ligands using one or more of these mechanisms include fibroblast growth factor, epidermal growth factor, and several ILs (reviewed in Refs.25 and 26).
Most of these examples involve hormones binding to membrane receptors, either on the cell surface or in another compartment such as endosomes. However, one unique feature of FSTL3 is that it is a secreted binding protein rather than a secreted hormone or growth factor and thus does not have its own cell surface receptor that could mediate transport back to the cytoplasm. Moreover, FSTL3 does not have a heparin-binding sequence and does not bind cell surface proteoglycans (11), so that a proposed mechanism for FST288 endocytosis mediated by proteoglycan binding (20, 27) is not likely to be operational for FSTL3. In addition, no known nuclear localization signal has been identified on FSTL3, and cellular targeting prediction programs (e.g. PSORT) fail to predict the likelihood of nuclear localization for FSTL3. Interestingly, nuclear localization was predicted for FSTL3 without its signal peptide. Our results, therefore, indicate that although FSTL3 can be translated from either Met, and FSTL3 from the second Met is translocated to the nucleus, this is probably not the major source of nuclear FSTL3 under natural conditions. Rather, nuclear FSTL3 in HeLa cells is glycosylated, but not to the same degree as secreted FSTL3, suggesting that nuclear FSTL3 is removed from the ER to the cytoplasm before it can be fully glycosylated and then transported to the nucleus.
We observed a difference between FSTL3 produced endogenously by HeLa cells compared with CHO cell lines stably expressing FSTL3 at high levels. Although the intracellular FSTL3 in HeLa cells was most prominently found in soluble nuclear extracts, the CHO FSTL3 was found in higher amounts in cytoplasmic extracts, which was confirmed by immunocytochemistry. This may be due to differences in actual protein production levels, because transfected CHO cells secreted 5- to 10-fold more FSTL3 than HeLa. Moreover, the capacity of the nuclear transport mechanism may be limited and cell type specific, so that additional FSTL3 production in the transfected CHO lines might be directed elsewhere. In contrast, when we were selecting CHO colonies we did not see any difference between low and high secreting clones in terms of relative percentages of intracellular vs. secreted FSTL3, suggesting that transport mechanisms were not obviously disordered in overexpression CHO clones. Nevertheless, it is clear from the HeLa cells that FSTL3 is transported to the nucleus, where it remains for substantial periods. This slow turnover may explain why we observed nuclear FSTL3 in many cells despite their low FSTL3 expression levels and lack of detectable secreted FSTL3 (10).
In contrast to FSTL3, FST was not translated from the second Met. It is not clear why this is so because neither FST or FSTL3 has a consensus Kozak sequence at the second Met. However, because FST cannot be translated from the second Met, nascent FST protein would probably enter the ER and secretory pathway, creating the rapid secretory profile we observed for FST315. This implies that the more slowly secreted FST288 and FST303 isoforms observed in the cytoplasm are due to biochemical associations acting on these isoforms. Both FST288 and FST303 can bind cell surface proteoglycans via the active HBS in FST domain 1, suggesting that interaction with membrane-associated proteoglycans on intracellular membranes might be acting to retard the release of at least some FST288 and FST303. It is thus clear that different mechanisms exist to regulate biosynthesis, transport, and secretion of FSTL3 and FST that portend different intracellular bioactivities, which remain to be established.
Our results show that newly synthesized FSTL3 is both secreted and actively retained within the nucleus. In contrast, the structurally and functionally related FST315 isoform is rapidly secreted whereas some FST288 and FST303 are retained intracellularly, but not in the nucleus. Although one potential mechanism for simultaneous secretion and nuclear transport of FSTL3 could involve translation from a second Met that deletes the signal peptide, our results indicate that this does not occur normally, because endogenous nuclear FSTL3 is glycosylated, whereas FSTL3 expressed from the second Met is not. In contrast, our results suggest that natural nuclear FSTL3 enters the ER, where it is glycosylated, then transported out of the ER through the cytoplasm and finally to the nucleus. Interestingly, nuclear FSTL3 retains activin antagonist activity, suggesting that this nuclear pool of FSTL3 may be a new, short-loop mechanism for regulating the activities of activin and related TGF ligands. Elucidation of this ER nuclear transport mechanism might also uncover a novel biosynthetic pathway important for the nuclear activity of secreted proteins. Moreover, the differential secretory dynamics of FSTL3 and FST isoforms may result in compartmentalization within cells as well as gradient formation within tissues and organs, thereby constituting an important and actively regulated aspect of their biology.
Acknowledgments
The expert technical assistance of Amy Schoen and Alicia Zaske is greatly appreciated. We also appreciate the technical advice and critical review of the manuscript by Dr. Henry Kronenberg.
Footnotes
This work was supported in part by National Institutes of Health Grants R01-DK-55838 and R01-HD-39777 (to A.L.S.) and a grant from the Japan Research Foundation for Clinical Pharmacology (to S.S.).
Current address for S.S.: Department of Obstetrics and Gynecology, Tokushima Red Cross Hospital, 28-1 Shingai Chuden-cho, Komatsujima City, Tokushima 773-8502, Japan.
First Published Online September 8, 2005
Abbreviations: CHO, Chinese hamster ovary; ER, endoplasmic reticulum; FST, follistatin; FSTL3, FST-like-3; HBS, heparin-binding sequence; WT, wild type.
Accepted for publication August 31, 2005.
References
Welt C, Sidis Y, Keutmann H, Schneyer A 2002 Activins, inhibins, and follistatins: from endocrinology to signaling. A paradigm for the new millennium. Exp Biol Med 227:724–752
Phillips DJ 2000 Regulation of activin’s access to the cell: why is mother nature such a control freak Bioessays 22:689–696
Shimasaki S, Koga M, Esch F, Cooksey K, Mercado M, Koba A, Ueno N, Ying SY, Ling N, Guillemin R 1988 Primary structure of the human follistatin precursor and its genomic organization. Proc Natl Acad Sci USA 85:4218–4222
Sidis Y, Schneyer AL, Sluss PM, Johnson LN, Keutmann HT 2001 Follistatin: essential role for the N-terminal domain in activin binding and neutralization. J Biol Chem 276:17718–17726
Keutmann HT, Schneyer AL, Sidis Y 2004 The role of follistatin domains in follistatin biological action. Mol Endocrinol 18:228–240
Inouye S, Ling N, Shimasaki S 1992 Localization of the heparin binding site of follistatin. Mol Cell Endocrinol 90:1–6
Sugino K, Kurosawa N, Nakamura T, Takio K, Shimasaki S, Ling N, Titani K, Sugino H 1993 Molecular heterogeneity of follistatin, an activin-binding protein. J Biol Chem 68:15579–15587
Hayette S, Gadoux M, Martel S, Bertrand S, Tigaud I, Magaud JP, Rimokh R 1998 FLRG (follistatin-related gene), a new target of chromosomal rearrangement in malignant blood disorders. Oncogene 16:2949–2954
Schneyer A, Tortoriello D, Sidis Y, Keutmann H, Matsuzaki T, Holmes W 2001 Follistatin-related protein (FSRP): a new member of the follistatin gene family. Mol Cell Endocrinol 180:33–38
Tortoriello DV, Sidis Y, Holtzman DA, Holmes WE, Schneyer AL 2001 Human follistatin-related protein: a structural homologue of follistatin with nuclear localization. Endocrinology 142:3426–3434
Sidis Y, Tortoriello DV, Holmes WE, Pan Y, Keutmann HT, Schneyer AL 2002 Follistatin-related protein and follistatin differentially neutralize endogenous vs. exogenous activin. Endocrinology 143:1613–1624
Tsuchida K, Nakatani M, Matsuzaki T, Yamakawa N, Liu Z, Bao Y, Arai KY, Murakami T, Takehara Y, Kurisaki A, Sugino H 2004 Novel factors in regulation of activin signaling. Mol Cell Endocrinol 225:1–8
Kurys G, Tagaya Y, Bamford R, Hanover JA, Waldmann TA 2000 The long signal peptide isoform and its alternative processing direct the intracellular trafficking of interleukin-15. J Biol Chem 275:30653–30659
Fiaschi-Taesch NM, Stewart AF 2003 Minireview: parathyroid hormone-related protein as an intracrine factor—trafficking mechanisms and functional consequences. Endocrinology 144:407–411
Yudt MR, Cidlowski JA 2001 Molecular identification and characterization of a and b forms of the glucocorticoid receptor. Mol Endocrinol 15:1093–1103
McConnell DS, Wang Q, Sluss PM, Bolf N, Khoury RH, Schneyer AL, Midgley Jr AR, Reame NE, Crowley Jr WF, Padmanabhan V 1998 A two-site chemiluminescent assay for activin-free follistatin reveals that most follistatin circulating in men and normal cycling women is in an activin-bound state. J Clin Endocrinol Metab 83:851–858
McKeon J, Khaledi MG 2001 Quantitative nuclear and cytoplasmic localization of antisense oligonucleotides by capillary electrophoresis with laser-induced fluorescence detection. Electrophoresis 22:3765–3770
Greenberg ME, Bender TP 1997 Identification of newly transcribed RNA. In: Ausubel FM, ed. Current protocols in molecular biology. New York: Wiley; 4.10.1–4.10.11
Dennler S, Itoh S, Vivien D, Ten Dijke P, Huet S, Gauthier JM 1998 Direct binding of Smad3 and Smad4 to critical TGF-inducible elements in the promoter of human plasminogen activator inhibitor-type 1 gene. EMBO J 17:3091–3100
Nakamura T, Sugino K, Titani K, Sugino H 1991 Follistatin, an activin-binding protein, associates with heparan sulfate chains of proteoglycans on follicular granulosa cells. J Biol Chem 266:19432–19437
Schneyer A, Sidis Y, Xia Y, Saito S, Re EE, Lin HY, Keutmann H 2004 Differential actions of follistatin and follistatin-like 3. Mol Cell Endocrinol 225:25–28
West CM, Van Der WH, Sassi S, Gaucher EA 2004 Cytoplasmic glycosylation of protein-hydroxyproline and its relationship to other glycosylation pathways. Biochim Biophys Acta 1673:29–44
Massfelder T, Dann P, Wu TL, Vasavada R, Helwig JJ, Stewart AF 1997 Opposing mitogenic and anti-mitogenic actions of parathyroid hormone-related protein in vascular smooth muscle cells: a critical role for nuclear targeting. Proc Natl Acad Sci USA 94:13630–13635
Jans DA, Thomas RJ, Gillespie MT 2003 Parathyroid hormone-related protein (PTHrP): a nucleocytoplasmic shuttling protein with distinct paracrine and intracrine roles. Vitam Horm 66:345–384
Johnson HM, Subramaniam PS, Olsnes S, Jans DA 2004 Trafficking and signaling pathways of nuclear localizing protein ligands and their receptors. Bioessays 26:993–1004
Jans DA, Hassan G 1998 Nuclear targeting by growth factors, cytokines, and their receptors: a role in signaling Bioessays 20:400–411
Sugino H, Sugino K, Hashimoto O, Shoji H, Nakamura T 1997 Follistatin and its role as an activin-binding protein. J Med Invest 44:1–14(Seiichiro Saito, Yisrael Sidis, Abir Muk)
Abstract
Follistatin (FST) and FST-like-3 (FSTL3) are structurally related proteins that bind and neutralize activin and closely related members of the TGF superfamily. Three FST isoforms (FST288, FST303, and FST315) are produced from the Fst gene that are primarily secreted proteins. FSTL3 is secreted, but is also observed within the nucleus of most cells. We used pulse-chase 35S labeling to examine the biosynthetic and intracellular transport patterns that lead to differential secretion and intracellular retention of these proteins. Among the FST isoforms, FST315 was secreted fastest and FST288 was secreted more slowly, with some remaining intracellular. In contrast, FSTL3 was secreted the slowest, with newly synthesized proteins being both secreted and trafficked to the nucleus. This nuclear FSTL3 was N-glycosylated, although not to the same degree as secreted FSTL3. Both FST and FSTL3 have two Mets in their signal sequence. Mutation of the first Met in FST288 eliminated protein translation, whereas FSTL3 could be translated from either Met. However, although FSTL3 translated from the second Met, which had no signal sequence, was confined to the nucleus, it was not glycosylated. Interestingly, this FSTL3 retained activin-antagonizing activity. Thus, although bioactive, nuclear FSTL3 can be translated from the second Met when the first Met is mutated, the glycosylated nuclear FSTL3 produced endogenously indicates that a different mechanism must be used under natural conditions that apparently includes N-glycosylation. Moreover, the differential biosynthetic and intracellular transport patterns for FST288 and FSTL3 suggest that these two activin-binding proteins may have distinct intracellular roles.
Introduction
LIKE OTHER MEMBERS of the TGF- superfamily, activin has wide-ranging actions in regulating cellular growth, proliferation, and differentiation in both adult and embryonic tissues. It is therefore not surprising that a number of regulatory mechanisms have evolved to modulate activin signaling intracellularly and extracellularly. Among the extracellular regulators are inhibin, a negative feedback regulator of activin-mediated FSH biosynthesis, and follistatin (FST), a monomeric glycoprotein that irreversibly binds and neutralizes activin. Although originally identified in follicular fluid, FST is actually produced in many of the same tissues as activin and probably acts primarily as an autocrine/paracrine regulator of activin activity (reviewed in Refs.1 and 2).
The Fst gene is composed of six exons, each coding for distinct functional regions of the protein (3, 4), including a signal peptide, an N-domain that is critical for binding activin (4), three 10-cysteine (Cys) FST domains that also participate to varying degrees in activin binding (5), and a C-terminal acidic tail that is present in the full-length FS315 molecule. However, through alternative splicing of intron 5, a second Fst mRNA is produced that codes for a protein of 288 amino acids (FST288) that does not have the C-terminal tail region (3). The physiological significance of this deletion is based on the finding that the first FST domain contains a basic sequence that has been shown to bind cell surface, heparin-sulfated proteoglycans [heparin-binding sequence (HBS)] (6), a property that is inhibited by the acidic C-terminal tail in FST315 (7). Purification of FST from porcine follicular fluid identified FST303, a third FST protein that is apparently produced via proteolytic cleavage of FST315 within the C-terminal tail (7). Interestingly, FST303 has intermediate cell surface-associating properties compared with FST288 and FST315, suggesting that the acidic C-terminal tail length differentially modulates the ability of the HBS to associate with heparin sulfate proteoglycans (7). Taken together, these studies demonstrate that the Fst gene produces three protein forms that differ in cell surface binding and in at least some of their biological activities.
FST-like-3 (FSTL3), also known as FST-related gene (8) and FST-related protein (9), binds activin with nearly the same affinity and irreversibility as FST (10). The gene structure of FSTL3 is similar to that of FST, containing one less FST domain, but sharing overall 45% primary sequence identity (9). Thus, FSTL3 and FST share a number of structural and functional properties suggesting that there may be substantial overlap in their biological actions in vivo.
A number of differences between FSTL3 and FST have also been described, including the lack of an heparin-binding sequence, which prevents FSTL3 from binding cell surface proteoglycans such as FST (11). In addition, although FSTL3 has a signal sequence and is secreted from cells in which it is highly expressed or overexpressed (8, 10), FSTL3 protein has also been localized within the nucleus of numerous cell lines, primary granulosa cells (10), and tissue sections (12), indicating that at least some FSTL3 remains intracellular. FST, in contrast, has been observed in the cytoplasm of some cells, including granulosa cells, but not in the nucleus (10). These observations suggest that biosynthesis and intracellular trafficking of FSTL3 and FST are differentially regulated.
Several mechanisms have been described where proteins with secretory signal peptides, which direct translated proteins to the endoplasmic reticulum (ER), can also be transported to the nucleus or other intracellular organelles. Alternative splicing can alter a signal peptide or nuclear localization signal in some mRNAs, but not in others, allowing a protein to be transported to different compartments (13). In addition, translation of some proteins can initiate at different methionines (Met) or at non-AUG codons, creating proteins with variable or even no signal peptides (14, 15). Interestingly, both FST and FSTL3 have a second Met in their signal sequences, which, if used, would produce a protein with a shortened and possibly nonfunctional signal peptide. This raises the possibility that the use of alternate Mets to initiate FST or FSTL3 translation could allow a portion of translated protein to remain intracellular.
The current study was undertaken to compare the kinetics of FST isoform and FSTL3 biosynthesis, trafficking, and secretion in transfected cells as well as HeLa cells, where FSTL3 is naturally secreted, using pulse-chase metabolic labeling. Our results confirm the localization of substantial newly synthesized FSTL3 in the nucleus for up to 8 h. In contrast, FST315 is secreted most rapidly, followed by FST303 and FST288, with a portion of FST288 being retained in the cell for up to 4 h, but not in the nuclear fraction. Our results also show that endogenous nuclear FSTL3 is transported to the nucleus and is N-glycosylated, although not to the same extent as secreted FSTL3. There are two Mets in the signal sequence of FSTL3, and protein translated from the second Met, which has no signal peptide, is transported to the nucleus but is not glycosylated. This distinction in glycosylation suggests that under natural conditions, nuclear FSTL3 is not derived from translation initiation at the second Met, but probably enters the ER to get glycosylated, then transported back to the cytoplasm and eventually to the nucleus. Taken together, these results indicate that FSTL3 and FST are both secreted and retained within the cell, albeit in different compartments and by different mechanisms, consistent with the concept that they may have differential intracellular actions in addition to their known extracellular functions.
Materials and Methods
Medium and cell culture
Chinese hamster ovary (CHO) and HeLa cells (American Type Culture Collection, Manassas, VA) were cultured in medium containing 45% Ham’s F-12, 45% DMEM, 10% heat-inactivated fetal bovine serum (Invitrogen Life Technologies, Inc., Grand Island, NY), 1% L-glutamine, and antibiotics (100 IU/ml penicillin and 100 μg/ml streptomycin sulfate) and maintained in a humidified atmosphere of 5% CO2. HEK 293 cells were maintained in RPMI 1640 with 10% fetal bovine serum and supplements as described above.
Transfection and selection of stably secreting clones
Human FSTL3 (provided by Millennium Pharmaceuticals, Cambridge MA), human FST288 or human FST315 (gifts from Dr. Shimasaki, University of California, San Diego, CA) cDNAs were used as templates for PCR amplification of the coding sequence without a stop sequence, which were then subcloned into the pcDNA3.1-Myc-HIS (Invitrogen Life Technologies, Inc., Carlsbad, CA) vector. The FST303 expression construct was made using the FST315 template, but positioning the downstream primer to end at Gln303. CHO cells were transfected with expression constructs using Lipofectamine (Invitrogen Life Technologies, Inc.) according to the manufacturer’s instructions. Twenty-four hours after transfection, 600 μg/ml G418 (Mediatech, Herndon, VA) was added to the culture medium, and colonies were screened for secretion of FST or FSTL3 as appropriate. FST-transfected colonies were screened with our two-site FST immunoassay (16), whereas FSTL3 colonies screened by RIA as previously described (10). We originally examined both high and low secreting colonies, but no differences in intracellular trafficking were observed, so only the highest secreting clones were maintained for additional study.
Metabolic labeling studies
Cells were plated in 12-well (for the time-course experiment) or six-well (for the subcellular fractionation experiment) dishes (300,000 or 900,000 cells/well, respectively) in 1 or 3 ml medium (respectively). After 24 h, cells were washed twice and preincubated for 90 min with Cys-free DMEM (Invitrogen Life Technologies, Inc.), after which Cys-free DMEM containing 100 μCi/ml [35S]Cys (NEN Life Science Products, Boston, MA) was added for an additional 30 min. After the labeling step, the cells were washed once with medium and incubated for the indicated chase times (see figures) until harvesting.
Cell extraction and fractionation
At the indicated time points, medium was collected ("secreted") and replaced with fresh medium containing 10 μg/ml heparin sulfate (Sigma-Aldrich Corp., St. Louis, MO) for 5 min to release cell surface-bound FST ("cell surface"). All conditioned medium samples were centrifuged to remove nonadherent cells and debris before being stored for later analysis. After heparin treatment, cells were extracted in 1 ml TNE buffer [20 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA, 0.1% Nonidet P-40, and a protease inhibitor cocktail (Roche, Mannheim, Germany)]. After incubating for 15 min on ice, the cells were subjected to two rounds of freeze/thaw to facilitate membrane lysis and centrifuged. This supernatant was collected as cellular. No cell surface-associated FST315 or FSTL3 was observed (data not shown).
To determine subcellular localization of labeled FSTL3 and FST isoforms, nuclear and cytoplasmic extracts were prepared by differential centrifugation as described previously (17). After removing conditioned medium, which is shown in figures as secreted (S) protein, cells were dislodged from the dish and pelleted by centrifugation at 1000 x g for 15 min. The cells were then resuspended in 1 ml hypotonic buffer [10 mM Tris-HCl (pH 8), 3 mM MgCl2, 1 mM EDTA, plus protease inhibitors], allowed to sit on ice for 10 min to induce swelling, and vortexed to disrupt the cell membrane. Lysates were centrifuged at 1000 x g for 10 min, and the supernatant was collected as cellular (C). The nuclear pellet was resuspended in 1 ml TNE buffer and subjected to three freeze/thaw cycles. After a 5-min centrifugation at 1000 x g, the supernatant (soluble nuclear extract) was collected as nuclear (N).
To verify that the rapid fractionation method used for the time-course studies faithfully separated nuclear fractions from ER and other cytoplasmic proteins, nuclei free from ER membranes were isolated essentially as previously described (18). Two 10-cm plates of HeLa cells were grown to confluence, one was extracted directly, and the other was metabolically labeled for 30 min and chased for 2 h as described above. Both plates were then treated identically during the fractionation process. HeLa cells were collected by scraping, gently centrifuged, washed with hypotonic buffer [20 mM Tris (pH 7.5), 3 mM CaCl2, and 2 mM MgCl2], and then resuspended in 5 packed cell volumes of hypotonic buffer and incubated on ice for 10 min. Swollen cells were homogenized in a Dounce homogenizer (Kontes Co., Vineland, NJ) and centrifuged at low speed (1000 x g). The postnuclear supernatant was further processed as described below. The crude nuclear pellet was further purified by resuspension in 4 ml sucrose buffer I [0.32 M sucrose, 10 mM Tris (pH 8.0), 3 mM CaCl2, 2 mM magnesium acetate, 0.1 mM EDTA, 1 mM dithiothreitol, and 0.5% Nonidet P-40], incubation on ice for 15 min, and then layering the nuclear suspension over 4.4 ml sucrose buffer II [2 M sucrose, 10 mM Tris (pH 8.0), 5 mM magnesium acetate, 0.1 mM EDTA, and 1 mM dithiothreitol], and centrifugation at 30,000 x g for 45 min. Purified nuclei were resuspended in hypotonic buffer, after which high salt buffer [1.2 M NaCl and 20 mM Tris (pH 7.5)] was added for 30 min at 4 C. The resulting suspension was centrifuged for 20 min at 4 C, and the supernatant, containing soluble nuclear proteins, was collected, diluted with an equal volume of hypotonic buffer, and analyzed by SDS-PAGE at the nuclear fraction.
The postnuclear supernatant obtained above was centrifuged at 16,000 x g for 20 min at 4 C. The supernatant was collected as the cytosolic fraction, and the pellet was collected as the crude ER/membrane fraction by resuspending this pellet in modified RIPA buffer [20 mM Tris (pH 7.5), 150 mM sodium chloride, 50 mM sodium fluoride, 0.1 mM EDTA, 0.5% Nonidet P-40, 0.5% deoxycholate, and 0.2% sodium dodecyl sulfate] before SDS-PAGE analysis.
Cytosolic, ER/membrane, and purified nuclear fractions were analyzed by immunoprecipitation with the FSTL3 antibody as described above (35S-labeled plate) or by Western blotting for calnexin (Abcam, Cambridge, MA), an ER-resident transmembrane protein, to localize the ER membranes (nonlabeled plate).
Detection of FSTL3 and FST
FSTL3 and FST were immunoprecipitated from each fraction using 100 ng of our previously described polyclonal anti-FSTL3 antibody (10) for FSTL3 or 500 ng anti-Myc polyclonal antibody (Upstate Biotechnology, Inc., Charlottesville, VA) for FST288 and FST303. FST315 was undetectable using the Myc antibody, so 100 ng monoclonal anti-FST 7FS30 covalently cross-linked to paramagnetic particles that are typically used in the FS immunoassay (16) was used. To reduce nonspecific protein precipitation, samples were preincubated with either 10 μl protein A (Pierce Chemical Co., Rockford, IL) or 10 μl paramagnetic particles without antibody (FST315 only) for 1 h, followed by centrifugation. Antibodies were incubated with cleared supernatants overnight, followed by addition of 10 μl protein A (except FST315 samples) at room temperature for 2 h. After centrifugation, samples were washed six times with TNE buffer, then boiled in 2x SDS-PAGE buffer containing -mercaptoethanol (Invitrogen Life Technologies, Inc.), and reduced proteins were separated on 8–16% Novex Tris-glycine gels (Invitrogen Life Technologies, Inc.). Proteins immunoprecipitated from untransfected CHO cells are designated nonspecific in the figures and were not included in the analysis.
After SDS-PAGE, gels were dried and quantitated by phosphorimager. For each point in the time course, analyzed samples include secreted (conditioned medium), cell extracts (cellular), and cell surface (released by heparin sulfate). Representative gels are shown in Fig. 1, and quantitative analyses are presented as a percentage of the total FST or FSTL3 of all three fractions at each time point, representing the mean of three replicate experiments, and are shown in Figs. 2 and 4. The relative proportion of labeled protein in each compartment was calculated for each time point as a percentage of the total counts at that time point.
Deglycosylation
The glycosylation status of some immunoprecipitated samples was assessed using recombinant N-glycanase (Glyko, Inc., Novato, CA) according to the manufacturer’s instructions, followed by SDS-PAGE as described above.
Creation and analysis of Met mutants
The cDNAs for FST288 and FSTL3 used to make stable cell lines (see above) were used as a template for site-directed mutagenesis (QuikChange, Stratagene, La Jolla, CA). Two FST288 mutants were created where either the first Met (M-29) or the second Met (M-8) in the signal sequence was changed to alanine (Ala). Similarly, the first Met (M-26) or the second Met (M1) in FSTL3 was changed to Ala. Because FSTL3 protein was produced from the FSTL3 M-26A mutant cDNA, a third mutant was prepared in which both Mets (M-26 and M1) were changed to Ala and used for in vitro translation to monitor translation in the absence of Mets in the signal sequence.
In vitro translation of FSTL3
Expression vectors containing FSTL3 wild-type (WT) or Met mutants were linearized downstream of the stop codon and used as templates in TnT reticulocyte lysate in vitro translation reactions in the presence of [35S]Cys according to the manufacturer’s protocol (Promega Corp., Madison WI). Translated products were resolved on 12% SDS-PAGE gels (Bio-Rad Laboratories, Hercules, CA) under reducing conditions and detected by autoradiography after 24-h exposure.
Electrophoresis and Western blot of signal peptide mutants
HEK 293 cells, which have a superior activin reporter response compared with CHO cells, were transiently transfected with 125 ng FST or FSTL3 mutant cDNAs and cultured for 48 h, after which they were extracted with passive lysis buffer (Promega Corp.). Equal volumes (10 μl) of whole cell extract (E) or 20 μl conditioned medium (M) were subjected to 12% SDS-PAGE using Tris-HCl Ready Gel from Bio-Rad Laboratories and were transferred onto polyvinylidene difluoride membrane (Bio-Rad Laboratories) following the manufacturer’s recommendation. After blocking in Tris-buffered saline supplemented with 10% dry milk and 0.2% Tween 20, blots were probed with anti-Myc antibodies (1 μg/ml; clone 4A6, Upstate Biotechnology, Inc., Lake Placid, NY), followed by goat antimouse horseradish peroxidase-conjugated antibody (1:7500 dilution; Santa Cruz Biotechnology, Santa Cruz, CA), and immunoreactivity was visualized using Western Lightning chemiluminescence reagent (PerkinElmer, Boston, MA).
Activin/FST in vitro bioassay
Transient transfection of 293 cells was performed in 24-well trays using Effectene (Qiagen, Valencia, CA) and a total of 225 ng DNA, which included 80 ng of the Smad3-responsive reporter CAGA-Luc (19), 15 ng pRL-TK (Promega Corp.), 5 ng activin-A cDNA, and 0.25–25 ng FST or FSTL3 mutant cDNA balanced with empty vector to a total of 125 ng. Culture medium was refreshed after 16 h, and the following day, cells were extracted with Passive Lysis Buffer (Promega Corp.) and assessed for firefly and Renilla luciferase activities using the Dual-Luciferase Reporter Assay System (Promega Corp.).
Immunocytochemistry
CHO cells were passaged onto sterilized glass coverslips (Fisher Scientific, Pittsburgh, PA) and transfected with cDNA constructs using Effectene reagent (Qiagen) according to the manufacturer’s protocol. After 24 h, transfected cells were fixed with 4% paraformaldehyde in PBS for 20 min and permeabilized with 0.1% sodium dodecyl sulfate in PBS for 4 min. Cells were then incubated with primary antibody diluted in 1% human -globulin (Sigma-Aldrich Corp.) in PBS. FSTL3 was detected with a purified rabbit polyclonal antibody (provided by Millennium Pharmaceuticals) at 5 μg/ml, and FST was detected using our monoclonal anti-FST antibody 7FS30 at 3.3 μg/ml. After a 1-h incubation, the cells were incubated with fluorescein-conjugated goat antirabbit IgG or Texas Red-conjugated goat antimouse (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) secondary antibodies (5 μg/ml). After 45 min, the cells were washed three with PBS, and coverslips were mounted onto glass slides with Vectashield (Vector Laboratories, Inc., Burlingame, CA). Immunofluorescent cells were visualized using an Axioskop microscope (Zeiss, New York, NY) equipped with a digital camera. Untransfected HeLa cells were treated in an identical manner to detect nuclear FSTL3.
Statistics
Time-course experiments were analyzed by two-way ANOVA for differences in both time and cell type; both had significant effects (P < 0.001). Differences between cell lines at each time point were identified by Tukey’s test. Similarly, results from cell fractionation analyses were analyzed by one-way ANOVA, and differences between cell lines for each fraction were determined using Tukey’s test. Differences of P < 0.05 were considered significant.
Results
Time course of FSTL3 and FST biosynthesis and secretion
Many cells that secrete FST contain both FST288 and FST315 mRNA transcripts and secrete multiple forms of FST protein. Therefore, to examine the biosynthesis and secretion of individual FST forms, cell lines were established by transfecting CHO cells (which do not secrete FST endogenously) with a single type of FST cDNA. During optimization experiments, we examined clones secreting different amounts of FST and found no differences in the fraction of secreted FST or FSTL3 relative to intracellular protein at different production levels (data not shown), so clones secreting maximal amounts of FST and FSTL3 were chosen for additional study. For comparison, we also examined HeLa cells, which naturally secrete FSTL3 as well as retain some FSTL3 in the nucleus (10).
Stable cell lines secreting individual FST isoforms (FST315, FST303, or FST315) or FSTL3 as well as HeLa cells that naturally secrete FSTL3 were metabolically labeled with [35S]Met for 30 min and analyzed by immunoprecipitation of secreted and cellular proteins at chase times up to 8 h. Representative autoradiograms for FST288, FST303, and FST315 are shown in Fig. 1, A–C, and for FSTL3 in HeLa and CHO cells in Fig. 1, D and E, respectively. FST315 was detected as two proteins of 40,000 and 46,000 in both conditioned medium (secreted) and cell extracts (cellular). As expected, cell surface-associated FST315 was undetectable in the medium after heparin release. FST 303 was secreted as two proteins of 38,000 and 42,000. Two specific bands were also detected in the lysates, but their electrophoretic migration was slightly faster than in the medium samples, suggesting differential glycosylation of intracellular and extracellular FST303. Although both secreted FST303 bands were detected in the membrane-associated (heparin-releasable) fraction, the amounts were substantially lower than membrane-associated FST288 and did not appear to change over time. Secreted FST288 consisted of three bands ranging in apparent molecular mass from 34,000–44,000, which were identical with those precipitated from cell lysates. Substantial cell surface-associated FST288 was released by heparin sulfate treatment. FST has been shown previously to be variably glycosylated at its two potential N-linked glycosylation sites (7), suggesting that the multiple specific FST bands we precipitated represent differentially glycosylated variants for each FST isoform.
FSTL3 was secreted as a single protein species of approximately 33,000 by HeLa and transfected CHO cells. FSTL3 was also detected in lysates from both cell lines as a single protein, but migrated more rapidly at 28,000. As expected, FSTL3 was never detected in association with cell surface proteoglycans (data not shown).
Results from three independent experiments for each cell line were combined by expressing immunoprecipitable FSTL3 or FST as a percentage of the total FSTL3 or FST in the sample. The time course for proteins secreted into the medium (Fig. 2A) revealed that FST315 was secreted fastest from cells, with more than half of the labeled protein in the medium by 30 min and more than 80% by 1 h. FST303 was secreted significantly more slowly (P < 0.01), reaching only 40% at 30 min and taking 2 h to release 80% of the labeled protein. Secretion rates into the medium for FST288 and FSTL3 (HeLa or CHO) were not significantly different from each other, but were significantly slower than those for FST315 and FST303 (P < 0.01). Only 20% of labeled FST288 or FSTL3 was detected in the medium at 30 min, but, like FST303, reached 80% by 2 h.
Consistent with the presence of a functional HBS on FST288 and FST303 (20), substantial amounts of secreted FST288 and FST303 were detected associated with cell membranes in the heparin-releasable fraction (Fig. 2B). By 30 min and for the duration of the experiment, FST288 associated with cell surface proteoglycans was significantly greater than FST303 (P < 0.01), and both were significantly greater (P < 0.01) than FST315, where the HBS was neutralized by the acidic C-terminal tail.
Because the cell surface FST most likely came from the secreted pool, we reanalyzed the secretory dynamics with the cell-associated proteins added to the medium proteins (Fig. 2, A and B) to create total secreted proteins (Fig. 2C). In this analysis, total secreted protein is not significantly different between the FST isoforms at 30 min. However, at 1 h, total secreted FST315 was significantly greater (90% vs. 65–70%; P < 0.01) than the other isoforms, indicating that this isoform is secreted more rapidly than the other isoforms over 60 min (Fig. 2C). In contrast, FSTL3 secretory rates were significantly slower than all three FST isoforms (P < 0.01) at both 30 (25%) and 60 (50–60%) min, indicating that intracellular trafficking of FSTL3 is quite distinct from that of all three FST forms. There were no significant differences in FSTL3 secretion dynamics between naturally secreting HeLa cells and transfected CHO cells at any time point except 1 h (P = 0.03), indicating that the slower secretory pattern of FSTL3 was not an artifact of transfection.
HeLa cells and FSTL3-transfected CHO cells had similar FSTL3 retention levels (>75%) and were both significantly (P < 0.01) greater than the 40–60% of the FST isoforms at 30 min (Fig. 2D). By 60 min, FST315 retention was significantly lower than that of all other cell lines (P < 0.01), with intracellular proteins amounting to only 10% of the total, whereas the others ranged from 30–40% of the total. Only at this 60 min point did FSTL3-transfected CHO cells contain more intracellular FSTL3 than endogenous FSTL3 in HeLa cells (P < 0.01). After 4 h, differences between the FST isoforms and FSTL3 were no longer significant and accounted for less than 10% of the total labeled FST.
Glycosylation of secreted and cellular FSTL3
To determine the cause of the slower migration of FSTL3 immunoprecipitated from medium compared with cell lysates, samples collected from CHO cells stably expressing FSTL3 as well as from HeLa cells, which naturally secrete FSTL3, were subjected to deglycosylation before SDS-PAGE analysis. The slower migrating FSTL3 species secreted from transfected CHO cells, which had an apparent molecular mass of 33,000 was reduced to approximately 26,000 after deglycosylation. The FSTL3 from cell lysates, which migrated at an apparent molecular mass of 28,000, was also reduced to 26,000 (Fig. 3A). These results, which were identical for HeLa cells (Fig. 3B), indicate that secreted FSTL3 is more heavily glycosylated than intracellular FSTL3, although intracellular FSTL3 is partially glycosylated, because the core deglycosylated proteins were of equal size.
Subcellular localization of retained FST and FSTL3
To determine the intracellular compartment in which FST and FSTL3 were retained, subcellular fractionation was performed before immunoprecipitation at chase times of 30 min and 2 h. Quantitative analysis of three replicate experiments is shown in Fig. 4, A and B, for chase times of 30 min and 2 h, respectively. At 30 min, more than 60% of FST315 was secreted, whereas less than 20% of FST288 was found in the medium. Approximately 20% of FST288 was located in the cytoplasmic fraction. At 2 h, less than 10% of FST315, but nearly 20% of FST288, was cytoplasmic, but neither was found in the nucleus. In contrast, more than 50% of FSTL3 in the HeLa cells was located in the nuclear fraction at 30 min and 25% at 2 h, significantly (P < 0.01) more than for FST288 or FST315. However, the transfected CHO cell FSTL3 appeared to be localized more to the cytoplasm than the nuclear fraction, similar to FST288. These results indicate that FSTL3 in naturally secreting HeLa cells was retained largely within the nuclear compartment, whereas FSTL3 in transfected CHO cells was found in both the nucleus and cytoplasm. Because our previous immunocytochemical study identified FSTL3 in the nucleus of transfected CHO cells (10), this difference may reflect the fact that the FSTL3 concentration within the nucleus may be limited so that excess production through transfection resulted mostly in secretion and cytoplasmic localization. This concept is supported by overexpressing FSTL3 in HeLa cells, which results in cytoplasmic localization of the extra FSTL3 (Fig. 4C) that obfuscates the nuclear localization of endogenous FSTL3 in these cells (Fig. 4D). Nevertheless, the transfected CHO cells had significantly more FSTL3 protein in the nuclear fraction than was present in the FST288- and FST315-secreting cells, indicating that newly synthesized FSTL3 was transported to the nucleus even in this CHO cell line.
To process the large number of samples necessitated by the time-course experiment, the subcellular fractionation protocol was weighted toward efficiency rather than complete separation of cellular compartments. To confirm that the [35S]FSTL3 was indeed located within the nucleus and not within ER membranes adjacent to the nucleus, a more stringent fractionation procedure was used to examine both 35S-labeled cells for FSTL3 localization as well as unlabeled cells for calnexin, a protein known to be inserted and resident within ER membranes. Greater than 95% of the calnexin was localized to the fraction containing ER and other membranes, whereas no radiolabeled FSTL3 was observed in this fraction (Fig. 5). Conversely, all the radiolabeled FSTL3 was found in the soluble nuclear fraction, whereas less than 5% of the calnexin was found in this material. These results demonstrate that newly synthesized FSTL3 is localized within the nucleus, confirming our time-course results.
Use of alternate translation start sites
N-Terminal sequencing of FSTL3 demonstrated that a majority of secreted (after cleavage of the signal peptide) FSTL3 begins with Met (10), which represents a nine-amino acid N-terminal extension compared with the first FST residue, which is Gly1 (Fig. 6A). In addition, FST has a second Met in its signal peptide 8 residues N terminal to Gly1. Thus, both FSTL3 and FST have two Mets in their signal sequence, the second of which almost completely deletes the signal peptide in FST and represents the beginning of mature FSTL3 (no signal peptide). We therefore hypothesized that one mechanism by which FSTL3 could be simultaneously targeted to the secretory pathway as well as to the nucleus would be if both Met residues were used to initiate translation. Moreover, because FST288 was partially retained in the cytoplasm (see above), a similar mechanism may explain its secretory dynamics.
To test this hypothesis, we made mutants of FST288 and FSTL3 with either the first or second Met changed to Ala. As shown in Fig. 6B, FST288 was detectable from both the WT and second Met mutant (M-8A) constructs in cell extracts (E) or as secreted FST288 in the medium (M). In contrast, no FST288 bands were detected in cell extracts or in the medium when the first Met was mutated (M-29A). Like FST, FSTL3 was detected in both cell extracts and conditioned medium in the WT and second Met mutant constructs (M1A). However, when the first Met was mutated, FSTL3 was still observed in cell extracts, although it was not secreted into the medium. These results suggest that FSTL3, but not FST288, could be translated from the second Met as well as the first.
To confirm this observation, we made an additional FSTL3 control construct with both Mets mutated (M-26,1A), and tested all four FSTL3 constructs for in vitro translational activity. Strong radiolabeled bands were observed for FSTL3 WT and second Met mutant constructs (M1A), consistent with the Western blot analysis (Fig. 7). Two weaker bands were observed for the first Met mutant (M-26A), one larger and one smaller than WT FSTL3. The double Met mutant construct produced a weak band that aligned with the upper FSTL3 band from the M-26A construct, indicating that this is not a product of a Met-initiated translation because there is no in-frame Met upstream of M-26. Thus, the lower band in the M-26A mutant is probably derived from translation initiation at the second Met, producing a protein shortened by the deletion of the 26-amino acid signal peptide.
We next tested whether the mutant FST288 or FSTL3 proteins retained the activin-inhibiting activity characteristic of WT proteins. As shown in Fig. 8, WT and second Met (FST M-8A) mutant FST288 were equally effective in inhibiting activin-induced reporter activity in 293 cells. However, the first Met mutant (FST M-29A) was inactive, consistent with the lack of detectable protein produced from this construct (see Fig. 7). In contrast, although both WT and second Met mutant (M1A) FSTL3 were equally effective in antagonizing activin-mediated reporter activity, the first Met mutant (M-26A) of FSTL3 did produce protein that was biologically active in inhibiting activin signaling. The reduced activity of this mutant may be the result of the lower expression levels, as shown in Fig. 6B. Thus, in contrast to FST, biologically active FSTL3 can be translated from the M-26A first Met mutant.
Immunocytochemical localization of Met mutant proteins
To determine whether FSTL3 translated from the second Met (FSTL3 M-26A) was targeted to the nucleus, we examined CHO cells transfected with each FSTL3 or FST288 mutant construct by immunocytochemistry. As previously reported (10), FSTL3 is weakly detectable in the nucleus of untransfected CHO cells (Fig. 9A). When transfected with WT FSTL3, very strong fluorescence was observed in the cytoplasm (Fig. 9B), which overwhelmed the weaker nuclear staining observed in controls. A similar pattern of immunofluorescence was observed when the second Met (M1A) was mutated (Fig. 9C). In contrast, when transfected with the first Met mutant construct (M-26A), these cells exhibited only nuclear staining (Fig. 9D) that was much enhanced compared with that of untransfected cells, indicating that FSTL3 produced by translation initiation at the second Met is transported to the nucleus. A similar analysis of FST288 Met mutants showed no background in untransfected cells (Fig. 9E), cytoplasmic FST in WT and second Met mutant (FST M-9A) constructs (Fig. 9, F and G), and no signal from the first Met mutant (FST M-9A) construct (Fig. 9H), consistent with lack of translation initiation for FST from the second Met in FST. For comparison, nuclear localization of naturally produced FSTL3 in HeLa cells is shown in Fig. 9, I (x40) and J (x100), where unstained nucleoli were observed surrounded by abundant nuclear FSTL3. Much weaker cytoplasmic FSTL3 staining was also observed at x40 (Fig. 9I). Taken together with the Western blot analysis and in vitro translation experiments, these results indicate that the use of alternative translation start sites might explain the simultaneous secretory and nuclear trafficking of FSTL3.
Glycosylation status of WT and Met mutant FSTL3
FSTL3 immunoprecipitated from cellular extracts of HeLa cells and transfected CHO cells was clearly glycosylated, although not to the same degree as secreted FSTL3 (see Fig. 3, A and B). However, translation initiation at the second Met in FSTL3 would completely eliminate the signal sequence, preventing FSTL3 from entering the ER and becoming glycosylated. Moreover, the M-26A band in Fig. 6B electrophoresed to a lower apparent molecular mass than WT, suggesting that this protein was not glycosylated. We therefore investigated whether cellular extracts from CHO cells transfected with WT, M-26A, or M1A mutant FSTL3 cDNAs produced glycosylated or nonglycosylated proteins. As expected, FSTL3 in cell extracts from WT or M1A cDNAs was glycosylated (Fig. 10). In contrast, FSTL3 from cells transfected with the M-26A FSTL3 cDNA (first Met mutant) did not have a change in apparent molecular mass after deglycosylation, indicating that this M-26A protein was not glycosylated and thus did not enter the ER. Because endogenous nuclear FSTL3 in HeLa cells is glycosylated (Fig. 3), this lack of glycosylation of M-26A FSTL3 suggests that under normal conditions, nuclear FSTL3 is not derived from protein translated from the second Met. Therefore, although FSTL3 translation can initiate from the second methionine, this mechanism does not appear to explain how endogenous (HeLa) or transfected (CHO) FSTL3 is transported to the nucleus.
Discussion
Using biosynthetic labeling, we determined that newly synthesized FSTL3 is both secreted and transported to the nucleus, and at 1 h, the amounts of FSTL3 in each pool are about equal. In addition, this nuclear FSTL3 is maintained for up to 8 h, suggesting that the nuclear FSTL3 may be turned over more slowly than secreted FSTL3. In comparison, the FST315 isoform, which is most closely related to FSTL3 because both are unable to bind membrane-associated proteoglycans, is secreted much more rapidly than FSTL3, with less than 10% retained within cells at 1 h. Thus, comparison of FST and FSTL3 exposes distinct differences in biosynthesis/secretory dynamics and regulation despite similar structures and bioactivities (21).
Both FST303 and FST288 isoforms, which are able to bind membrane-associated proteoglycans (7), are secreted more slowly than FST315, but faster than FSTL3. Moreover, both isoforms are partially retained in the cytoplasm for up to 8 h. Because intracellular membrane systems are virtually contiguous with the plasma membrane, this cytoplasmic FST288 and FST303 may be associated with intracellular membrane-associated proteoglycans, suggesting that intracellular FST288 and/or FST303 may have functions distinct from secreted FST. Interestingly, quantitative analysis of the biosynthesis time course demonstrates that significant amounts of both FST288 and FSTL3 remain inside cells for relatively long periods, although in different compartments (cytoplasmic vs. nuclear), suggesting that both FST288 and FSTL3 may have intracellular or intranuclear (respectively) actions in addition to their well-described extracellular roles in regulating activin and related TGF family members.
Both FSTL3 and FST have two Mets within their signal peptide. Thus, one potential mechanism for a protein to be simultaneously secreted and transported to the nucleus involves the use of either Met to initiate translation. In this model, transcripts starting at the first Met would be directed to the ER and ultimately secreted. In contrast, transcripts arising from the second Met would not have a signal peptide, and thus be translated in the cytoplasm and then directed to the nucleus via an as yet unidentified mechanism. We found that FSTL3 translation can initiate at either Met, but the first Met with a strong Kozak sequence produced more protein. When the first Met was replaced by Ala, FSTL3 protein was still produced, albeit less than from the first Met, and this FSTL3 was detectable in the nucleus, but not secreted. Interestingly, nuclear FSTL3 was capable of antagonizing activin-mediated reporter activity, suggesting that nuclear FSTL3 may have important biological roles regulating activin signaling, although the mechanism for this activity remains to be elucidated.
Immunocytochemical analysis demonstrated that endogenous FSTL3 (HeLa cells), transfected WT FSTL3, and transfected first Met mutant (M-26A) FSTL3 are all transported to the nucleus. Moreover, the deglycosylation studies for endogenous and WT FSTL3 are consistent with the secreted and nuclear FSTL3 being derived from different pools of protein, because they are differentially glycosylated. However, FSTL3 from the M-26A mutant, which was not secreted, is not glycosylated. These observations suggest that although they all can contribute to nuclear FSTL3, endogenous and transfected WT FSTL3 use a different mechanism than M-26A FSTL3. It is possible that the extracted intracellular FSTL3 represents protein extracted from ER before being fully glycosylated, whereas secreted FSTL3 has traversed the ER and Golgi and therefore has mature carbohydrate moieties. This would imply that under normal cellular conditions, all FSTL3 enters the ER, with a portion being shuttled out to the cytoplasm, perhaps by a chaperone, and then directed to the nucleus.
An alternative mechanism involved translation and glycosylation of FSTL3 in the cytoplasm (22), although this carbohydrate must be distinct from that found on secreted FSTL3 because they migrate differently on SDS-PAGE. This cytoplasmic FSTL3 might then be directed to the nucleus via an as yet unidentified chaperone, because FSTL3 contains no identifiable nuclear localization signals. Although proteins can be glycosylated on hydroxylated prolines in the cytoplasm in yeast (22), it is not likely that these carbohydrates would be removed using the enzymes that work on N-linked carbohydrates used in these studies. Thus, taken together with our observations that FSTL3 translated from the second Met (N-26A mutant) is not glycosylated, and extracts of WT FSTL3 clearly are glycosylated, our results favor a model in which FSTL3 is primarily translated from the first Met, enters the ER to get partially glycosylated, and then is removed from the ER before glycosylation is complete via a mechanism that might involve chaperone-mediated retrograde transport to the cytoplasm and then to the nucleus.
The conventional model of protein trafficking holds that proteins contain structural elements that regulate their transport to cellular compartments consistent with their functions. Thus, secretory and plasma membrane-bound proteins contain signal peptides that mediate transport of the nascent protein into the ER, where posttranslational modifications and multimerization can occur under regulated conditions. These proteins are then sorted through the Golgi into appropriate pathways and transported to the cell surface. In contrast, proteins destined for intracellular roles do not typically have signal peptides, are synthesized in the cytoplasm, and are then targeted to appropriate cellular compartments by chaperones or localization elements within their primary structure. A growing number of exceptions to this conventional model have been elucidated, allowing for secreted proteins to also be transported to various intracellular compartments, including the nucleus, where they may have actions distinct from those mediated by their cell surface receptors. For example, PTHrP inhibits proliferation in vascular smooth muscle cells when it interacts with its cell surface receptor, whereas nuclear PTHrP stimulates the proliferation of these cells (23). Mechanisms explaining differential biosynthesis and trafficking of PTHrP include 1) endocytosis of the hormone-receptor complex, where PTHrP can then be transported to the nucleus via a classical nuclear localization signal, 2) use of alternate translation start codons to produce transcripts that have shortened signal peptides and are unable to direct translocation to the ER, and 3) retrograde transport from the ER mediated by chaperones (14, 24). Other secreted ligands using one or more of these mechanisms include fibroblast growth factor, epidermal growth factor, and several ILs (reviewed in Refs.25 and 26).
Most of these examples involve hormones binding to membrane receptors, either on the cell surface or in another compartment such as endosomes. However, one unique feature of FSTL3 is that it is a secreted binding protein rather than a secreted hormone or growth factor and thus does not have its own cell surface receptor that could mediate transport back to the cytoplasm. Moreover, FSTL3 does not have a heparin-binding sequence and does not bind cell surface proteoglycans (11), so that a proposed mechanism for FST288 endocytosis mediated by proteoglycan binding (20, 27) is not likely to be operational for FSTL3. In addition, no known nuclear localization signal has been identified on FSTL3, and cellular targeting prediction programs (e.g. PSORT) fail to predict the likelihood of nuclear localization for FSTL3. Interestingly, nuclear localization was predicted for FSTL3 without its signal peptide. Our results, therefore, indicate that although FSTL3 can be translated from either Met, and FSTL3 from the second Met is translocated to the nucleus, this is probably not the major source of nuclear FSTL3 under natural conditions. Rather, nuclear FSTL3 in HeLa cells is glycosylated, but not to the same degree as secreted FSTL3, suggesting that nuclear FSTL3 is removed from the ER to the cytoplasm before it can be fully glycosylated and then transported to the nucleus.
We observed a difference between FSTL3 produced endogenously by HeLa cells compared with CHO cell lines stably expressing FSTL3 at high levels. Although the intracellular FSTL3 in HeLa cells was most prominently found in soluble nuclear extracts, the CHO FSTL3 was found in higher amounts in cytoplasmic extracts, which was confirmed by immunocytochemistry. This may be due to differences in actual protein production levels, because transfected CHO cells secreted 5- to 10-fold more FSTL3 than HeLa. Moreover, the capacity of the nuclear transport mechanism may be limited and cell type specific, so that additional FSTL3 production in the transfected CHO lines might be directed elsewhere. In contrast, when we were selecting CHO colonies we did not see any difference between low and high secreting clones in terms of relative percentages of intracellular vs. secreted FSTL3, suggesting that transport mechanisms were not obviously disordered in overexpression CHO clones. Nevertheless, it is clear from the HeLa cells that FSTL3 is transported to the nucleus, where it remains for substantial periods. This slow turnover may explain why we observed nuclear FSTL3 in many cells despite their low FSTL3 expression levels and lack of detectable secreted FSTL3 (10).
In contrast to FSTL3, FST was not translated from the second Met. It is not clear why this is so because neither FST or FSTL3 has a consensus Kozak sequence at the second Met. However, because FST cannot be translated from the second Met, nascent FST protein would probably enter the ER and secretory pathway, creating the rapid secretory profile we observed for FST315. This implies that the more slowly secreted FST288 and FST303 isoforms observed in the cytoplasm are due to biochemical associations acting on these isoforms. Both FST288 and FST303 can bind cell surface proteoglycans via the active HBS in FST domain 1, suggesting that interaction with membrane-associated proteoglycans on intracellular membranes might be acting to retard the release of at least some FST288 and FST303. It is thus clear that different mechanisms exist to regulate biosynthesis, transport, and secretion of FSTL3 and FST that portend different intracellular bioactivities, which remain to be established.
Our results show that newly synthesized FSTL3 is both secreted and actively retained within the nucleus. In contrast, the structurally and functionally related FST315 isoform is rapidly secreted whereas some FST288 and FST303 are retained intracellularly, but not in the nucleus. Although one potential mechanism for simultaneous secretion and nuclear transport of FSTL3 could involve translation from a second Met that deletes the signal peptide, our results indicate that this does not occur normally, because endogenous nuclear FSTL3 is glycosylated, whereas FSTL3 expressed from the second Met is not. In contrast, our results suggest that natural nuclear FSTL3 enters the ER, where it is glycosylated, then transported out of the ER through the cytoplasm and finally to the nucleus. Interestingly, nuclear FSTL3 retains activin antagonist activity, suggesting that this nuclear pool of FSTL3 may be a new, short-loop mechanism for regulating the activities of activin and related TGF ligands. Elucidation of this ER nuclear transport mechanism might also uncover a novel biosynthetic pathway important for the nuclear activity of secreted proteins. Moreover, the differential secretory dynamics of FSTL3 and FST isoforms may result in compartmentalization within cells as well as gradient formation within tissues and organs, thereby constituting an important and actively regulated aspect of their biology.
Acknowledgments
The expert technical assistance of Amy Schoen and Alicia Zaske is greatly appreciated. We also appreciate the technical advice and critical review of the manuscript by Dr. Henry Kronenberg.
Footnotes
This work was supported in part by National Institutes of Health Grants R01-DK-55838 and R01-HD-39777 (to A.L.S.) and a grant from the Japan Research Foundation for Clinical Pharmacology (to S.S.).
Current address for S.S.: Department of Obstetrics and Gynecology, Tokushima Red Cross Hospital, 28-1 Shingai Chuden-cho, Komatsujima City, Tokushima 773-8502, Japan.
First Published Online September 8, 2005
Abbreviations: CHO, Chinese hamster ovary; ER, endoplasmic reticulum; FST, follistatin; FSTL3, FST-like-3; HBS, heparin-binding sequence; WT, wild type.
Accepted for publication August 31, 2005.
References
Welt C, Sidis Y, Keutmann H, Schneyer A 2002 Activins, inhibins, and follistatins: from endocrinology to signaling. A paradigm for the new millennium. Exp Biol Med 227:724–752
Phillips DJ 2000 Regulation of activin’s access to the cell: why is mother nature such a control freak Bioessays 22:689–696
Shimasaki S, Koga M, Esch F, Cooksey K, Mercado M, Koba A, Ueno N, Ying SY, Ling N, Guillemin R 1988 Primary structure of the human follistatin precursor and its genomic organization. Proc Natl Acad Sci USA 85:4218–4222
Sidis Y, Schneyer AL, Sluss PM, Johnson LN, Keutmann HT 2001 Follistatin: essential role for the N-terminal domain in activin binding and neutralization. J Biol Chem 276:17718–17726
Keutmann HT, Schneyer AL, Sidis Y 2004 The role of follistatin domains in follistatin biological action. Mol Endocrinol 18:228–240
Inouye S, Ling N, Shimasaki S 1992 Localization of the heparin binding site of follistatin. Mol Cell Endocrinol 90:1–6
Sugino K, Kurosawa N, Nakamura T, Takio K, Shimasaki S, Ling N, Titani K, Sugino H 1993 Molecular heterogeneity of follistatin, an activin-binding protein. J Biol Chem 68:15579–15587
Hayette S, Gadoux M, Martel S, Bertrand S, Tigaud I, Magaud JP, Rimokh R 1998 FLRG (follistatin-related gene), a new target of chromosomal rearrangement in malignant blood disorders. Oncogene 16:2949–2954
Schneyer A, Tortoriello D, Sidis Y, Keutmann H, Matsuzaki T, Holmes W 2001 Follistatin-related protein (FSRP): a new member of the follistatin gene family. Mol Cell Endocrinol 180:33–38
Tortoriello DV, Sidis Y, Holtzman DA, Holmes WE, Schneyer AL 2001 Human follistatin-related protein: a structural homologue of follistatin with nuclear localization. Endocrinology 142:3426–3434
Sidis Y, Tortoriello DV, Holmes WE, Pan Y, Keutmann HT, Schneyer AL 2002 Follistatin-related protein and follistatin differentially neutralize endogenous vs. exogenous activin. Endocrinology 143:1613–1624
Tsuchida K, Nakatani M, Matsuzaki T, Yamakawa N, Liu Z, Bao Y, Arai KY, Murakami T, Takehara Y, Kurisaki A, Sugino H 2004 Novel factors in regulation of activin signaling. Mol Cell Endocrinol 225:1–8
Kurys G, Tagaya Y, Bamford R, Hanover JA, Waldmann TA 2000 The long signal peptide isoform and its alternative processing direct the intracellular trafficking of interleukin-15. J Biol Chem 275:30653–30659
Fiaschi-Taesch NM, Stewart AF 2003 Minireview: parathyroid hormone-related protein as an intracrine factor—trafficking mechanisms and functional consequences. Endocrinology 144:407–411
Yudt MR, Cidlowski JA 2001 Molecular identification and characterization of a and b forms of the glucocorticoid receptor. Mol Endocrinol 15:1093–1103
McConnell DS, Wang Q, Sluss PM, Bolf N, Khoury RH, Schneyer AL, Midgley Jr AR, Reame NE, Crowley Jr WF, Padmanabhan V 1998 A two-site chemiluminescent assay for activin-free follistatin reveals that most follistatin circulating in men and normal cycling women is in an activin-bound state. J Clin Endocrinol Metab 83:851–858
McKeon J, Khaledi MG 2001 Quantitative nuclear and cytoplasmic localization of antisense oligonucleotides by capillary electrophoresis with laser-induced fluorescence detection. Electrophoresis 22:3765–3770
Greenberg ME, Bender TP 1997 Identification of newly transcribed RNA. In: Ausubel FM, ed. Current protocols in molecular biology. New York: Wiley; 4.10.1–4.10.11
Dennler S, Itoh S, Vivien D, Ten Dijke P, Huet S, Gauthier JM 1998 Direct binding of Smad3 and Smad4 to critical TGF-inducible elements in the promoter of human plasminogen activator inhibitor-type 1 gene. EMBO J 17:3091–3100
Nakamura T, Sugino K, Titani K, Sugino H 1991 Follistatin, an activin-binding protein, associates with heparan sulfate chains of proteoglycans on follicular granulosa cells. J Biol Chem 266:19432–19437
Schneyer A, Sidis Y, Xia Y, Saito S, Re EE, Lin HY, Keutmann H 2004 Differential actions of follistatin and follistatin-like 3. Mol Cell Endocrinol 225:25–28
West CM, Van Der WH, Sassi S, Gaucher EA 2004 Cytoplasmic glycosylation of protein-hydroxyproline and its relationship to other glycosylation pathways. Biochim Biophys Acta 1673:29–44
Massfelder T, Dann P, Wu TL, Vasavada R, Helwig JJ, Stewart AF 1997 Opposing mitogenic and anti-mitogenic actions of parathyroid hormone-related protein in vascular smooth muscle cells: a critical role for nuclear targeting. Proc Natl Acad Sci USA 94:13630–13635
Jans DA, Thomas RJ, Gillespie MT 2003 Parathyroid hormone-related protein (PTHrP): a nucleocytoplasmic shuttling protein with distinct paracrine and intracrine roles. Vitam Horm 66:345–384
Johnson HM, Subramaniam PS, Olsnes S, Jans DA 2004 Trafficking and signaling pathways of nuclear localizing protein ligands and their receptors. Bioessays 26:993–1004
Jans DA, Hassan G 1998 Nuclear targeting by growth factors, cytokines, and their receptors: a role in signaling Bioessays 20:400–411
Sugino H, Sugino K, Hashimoto O, Shoji H, Nakamura T 1997 Follistatin and its role as an activin-binding protein. J Med Invest 44:1–14(Seiichiro Saito, Yisrael Sidis, Abir Muk)