A Combined Proteome and Ultrastructural Localization Analysis of 14-3-3 Proteins in Transformed Human Amnion (AMA) Cells

The 14-3-3 proteins constitute a family of highly conserved and broadly expressed multifunctional polypeptides that are involved in a variety of important cellular processes that include cell cycle progression, growth, differentiation, and apoptosis. Although the exact cellular function(s) of 14-3-3 proteins is not fully elucidated, as a rule these proteins act by binding to protein ligands, thus regulating their activity; so far more than 300 cellular proteins have been reported to interact with 14-3-3 proteins. Binding to cognate interacting partners is isoform-specific, but redundancy also exists as several binding peptides can be recognized by all isoforms, and some functions can be carried out by any isoform indistinctly. Moreover by interacting with different ligands in a spatially and temporally regulated fashion the same isoform can play multiple possibly even opposing roles where the resultant cellular outcome will be determined by the integration of the various effects. Although there is a large body of literature on specific aspects of 14-3-3 biology, not much is known on the coordinated aspects of 14-3-3 isoform expression, post-translational modifications, and subcellular localization. To address the question of isoform-specific differences, we carried out a comparative analysis of the patterns of expression, phosphorylation, and subcellular localization of the 14-3-3 β, ε, σ, τ, and ζ protein isoforms in transformed human amnion (AMA) cells. To validate as well as broaden our observations we analyzed the occurrence of the various isoforms in a large number of established cell lines and mammary and urothelial tissue specimens. Given the systematic approach we undertook and our application of isoform-discriminating technologies to the analysis of various cellular systems, we expect the data presented in this study to serve as an enabling resource for researchers working with 14-3-3 proteins.

The 14-3-3 proteins constitute a family of highly conserved and broadly expressed multifunctional polypeptides that are involved in a variety of important cellular processes that include cell cycle progression, growth, differentiation, and apoptosis. Although the exact cellular function(s) of 14-3-3 proteins is not fully elucidated, as a rule these proteins act by binding to protein ligands, thus regulating their activity; so far more than 300 cellular proteins have been reported to interact with 14-3-3 proteins. Binding to cognate interacting partners is isoform-specific, but redundancy also exists as several binding peptides can be recognized by all isoforms, and some functions can be carried out by any isoform indistinctly. Moreover by interacting with different ligands in a spatially and temporally regulated fashion the same isoform can play multiple possibly even opposing roles where the resultant cellular outcome will be determined by the integration of the various effects. Although there is a large body of literature on specific aspects of 14-3-3 biology, not much is known on the coordinated aspects of 14-3-3 isoform expression, post-translational modifications, and subcellular localization. To address the question of isoform-specific differences, we carried out a comparative analysis of the patterns of expression, phosphorylation, and subcellular localization of the 14-3-3 ␤, , , , and protein isoforms in transformed human amnion (AMA) cells. To validate as well as broaden our observations we analyzed the occurrence of the various isoforms in a large number of established cell lines and mammary and urothelial tissue specimens. Given the systematic approach we undertook and our application of isoform-discriminating technologies to the analysis of various cellular systems, we expect the data presented in this study to serve as an enabling resource for researchers working with 14-3-3

proteins. Molecular & Cellular Proteomics 7:1225-1240, 2008.
The 14-3-3 proteins compose a multifunctional protein family comprising seven mammalian isoforms (␤, ␥, , , , , and ) that are involved in a multitude of biological processes and that play a regulatory role in processes such as apoptotic cell death, mitogenic signal transduction, and cell cycle control (for reviews, see Refs. [1][2][3]. Although the exact cellular function(s) of 14-3-3 proteins is not fully elucidated, these proteins act by binding to protein ligands, thus regulating their activity; so far more than 300 cellular proteins have been reported to interact with 14-3-3 proteins (2,4,5). Expression of the various isoforms can reportedly be tissue-, time-, development-, and stimuli-specific (Refs. 2 and 3 and references therein). Furthermore binding to cognate interacting partners can be isoform-specific with functional specialization occurring to a certain extent (6). Some redundancy also exists as it has been shown that several binding peptides can be recognized by all isoforms and that some functions can be carried out by more than one isoform (2,7,8). Thus, it is likely that some functions are family-specific, and any isoform can perform it, whereas others may be unique to a given isoform or isoforms. Moreover by interacting with different ligands in a spatially and temporally regulated fashion isoforms can play multiple or even opposing roles with the resulting cellular outcome being determined by the integration of the various effects.
There is a growing emergence of correlative data linking 14-3-3 proteins to various human diseases such as neurodegenerative conditions, cancer, and heart disease, although it has not yet been proven that 14-3-3 proteins play a causative role in the pathogenesis of any of these disorders (Ref. 9 and references therein). One member of the 14-3-3 protein family, the 14-3-3 isoform , also known as stratifin (10), has been the focus of intense research as it is expressed only by epithelial cells (10) and is frequently deregulated in human cancers, suggesting a putative involvement in the malignant transformation of epithelia. This protein inhibits G 2 /M cell cycle progression in a p53-regulated manner and is critical to uphold G 2 arrest upon DNA damage in colorectal cancer cells (11,12). Additionally 14-3-3 associates with Cdk2 and Cdk4, suggesting that it may also play a role in regulating G 1 /S progression (13). In primary human epidermal keratinocytes, down-regulation of 14-3-3 has been shown to result in evasion from senescence and immortalization (14). Overall these lines of evidence suggest that functional inactivation of 14-3-3 may be linked to carcinogenesis, a postulate that is supported by the discovery of 14-3-3 down-regulation in various human cancers, including breast, stomach, colon, lung, liver, pancreas, oral cavity, vulva, and bladder (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25). However, expression of 14-3-3 is also reportedly up-regulated in lung (26), head and neck (27), gastric (28), and pancreatic cancer (29,30). Additionally abnormally elevated levels of 14-3-3 have been reported in hyperproliferative skin diseases, such as psoriasis, condylomata acuminata, and actinic keratoses (31,32), showing that non-malignant human epidermal keratinocytes can express this protein at increased levels and yet display high proliferative indexes. These apparent discrepancies may stem from the different cellular roles carried out by 14-3-3 proteins in different cell types. To address the question of isoform-specific differences in ligand binding, distribution, and regulation we carried out a comprehensive proteomics study of the expression, subcellular localization, and phosphorylation of five 14-3-3 isoforms (␤, , , , and ) in the same cellular system, transformed human amnion (AMA) cells.
Primary breast epithelial cell cultures were derived from breast tumors and cultured as described previously (33). Human dermal microvascular endothelial cells (CADMEC), primary epithelial amnion cells, human placental trophoblast cells, melanocytes, and human embryonic fibroblasts were isolated and analyzed as described previously (34,35).
Breast Tissue Biopsies-Breast tissue specimens were collected at the Copenhagen University Hospital, Copenhagen, Denmark as part of a large translational breast cancer research project involving high risk breast cancer patients. Fresh tissue samples were obtained from non-malignant areas of the surgical specimen distal to the tumor. Twenty to 30, 6-m cryostat sections of frozen tissue were dissolved in 0.1 ml of CLB1 lysis solution (Zeptosens AG, Witterswil, Switzerland) and kept at Ϫ20°C until used. The first and last sections of each sample were used for immunofluorescence analysis using a keratin 19 antibody as this epithelial marker is ubiquitously expressed by mammary epithelial cells. The availability of these pictures greatly facilitated the interpretation of the two-dimensional (2D) PAGE studies as it gave a rough estimate of the ratio of glands/cyst/tumor to stromal tissue. All patients had had no previous surgery to the breast and did not receive preoperative treatment. The project was approved by the Scientific and Ethical Committee of the Copenhagen and Frederiksberg Municipalities (KF 01-069/03).
Bladder Tissue Biopsies-Bladder specimens were collected over a period of 6 years at Skejby Hospital, Aarhus, Denmark. Random biopsies diagnosed as benign were dissected, cleaned of muscle and fat tissue, split into small pieces with the aid of a scalpel, and subsequently labeled with [ 35 S]methionine. Following labeling for 14 -16 h, the medium was carefully aspirated, and the pieces were dissolved in 0.3-0.4 ml of lysis solution (8 M urea, 100 mM DTT, 2% Nonidet P-40, 2% carrier ampholytes, pH 7-9) (35) with the aid of a hand glass homogenizer and subsequently stored at Ϫ20°C until used. The Scientific and Ethical Committee of Aarhus County approved the project.
Tumor Interstitial Fluid Collection-Approximately 0.25 g of clean fresh breast tissue biopsies were cut into small pieces, washed carefully in 5 ml of PBS, and incubated in a 10-ml conical plastic tube containing 0.8 ml of PBS for 1 h at 37°C in a humidified CO 2 incubator. Subsequently the samples were centrifuged at 1000 rpm for 2 min, and the supernatant was aspirated with the aid of an elongated Pasteur pipette. Samples were further centrifuged at 5000 rpm for 20 min in a refrigerated centrifuge (4°C). The final supernatant, with a protein concentration that ranged from 1 to 4 mg/ml, was freeze dried and resuspended in 0.5 ml of lysis solution for 2D gel analysis.
Indirect Immunofluorescence-Cells were grown on 18-mm glass coverslips and fixed with Cytoskelfix TM according to the manufacturer's instructions (R&D Enterprises, Lakewood, CO). Fixed cells were treated with Image-iT FX TM signal enhancer (Molecular Probes) to block nonspecific staining and subsequently incubated with the relevant primary antibodies overnight at 4°C. The sections were washed three times with cold PBS between incubations. Normal rabbit or mouse serum instead of primary antibody was used as a negative control. Double stainings were performed using appropriate Alexa Fluorா 488-and Alexa Fluor 594-labeled secondary antibodies (Molecular Probes). Image analysis was done using a laser scanning microscope (Zeiss 510LSM). The specificity of the 14-3-3 antibody for immunocytochemistry was verified by staining MRC5 fibroblast 1 The abbreviations used are: DMEM, Dulbecco's modified Eagle's medium; PCNA, proliferating cell nuclear antigen; 2D, two-dimensional; Bis-Tris, 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol; ER, endoplasmic reticulum; MT, microtubule; CSF, cerebrospinal fluid; GDI, GDP dissociation inhibitor; NIF, non-malignant mammary interstitial fluid; HMECs, human mammary epithelial cells. cells, which do not express the 14-3-3 protein (see Fig. 6C). Immunostaining of tissue sections was performed as described previously (14).
Proteomics Analysis and Quantitation of the Levels of 14-3-3-2D PAGE was performed as described previously (38). Gels were stained with silver nitrate or in some cases with Coomassie Brilliant Blue and subjected to autoradiography. Proteins were identified using a combination of procedures that included matrix-assisted laser desorption ionization time-of-flight mass spectrometry (Reflex IV, Bruker Daltonics), 2D PAGE Western immunoblotting, and comparison with the master 2D gel images of human keratinocytes and transitional cell carcinoma proteins (human 2D PAGE databases, Danish Centre for Translational Breast Cancer Research). For quantitation, 2D gel autoradiographs and silver-stained gels were scanned using a Molecular Imager device (Bio-Rad) and were analyzed using the PDQuest analysis software package (PDQuest version 7.3.0; Bio-Rad). Protein levels were normalized to actin, to the sum intensity of all valid spots on a gel, or to the sum intensity of all 14-3-3 isoforms on a gel, and the average means with corresponding standard deviations were determined for each case.
Protein Identification by Mass Spectrometry-Protein spots were excised from dry gels, and the gel pieces were rehydrated in water. Gel pieces were detached from the cellophane film and cut into approximately 1-mm 2 pieces followed by "in-gel" digestion of proteins as described previously (39). Samples were prepared for analysis by applying 2 l of digested and extracted peptides on the surface of a 400/384 AnchorChip target (Bruker Daltonics GmbH) followed by co-crystallization with ␣-cyano matrix (40).
Mass spectra were acquired on using a Reflex IV MALDI-TOF mass spectrometer equipped with a Scout 384 ion source recording peptide masses between 700 and 3500 Da. All spectra were obtained in positive reflector mode with delayed extraction using an accelerating voltage of 28 kV. The resulting mass spectra were internally calibrated using autodigested tryptic mass values visible in all spectra (805.417, 1046.596, 1153.574, 1433.721, 2185.075, and 2273.160). Calibrated spectra were processed by the Xmass 5.1.1 and BioTools 2.1 software packages (Bruker Daltonics GmbH). Irrelevant masses (matrix, metal adducts, and autodigested tryptic masses as well as masses of tryptic peptides from keratins) were excluded from the analysis by manual examination of all spectra by pairwise comparison. The spectrum of interest was superimposed with the spectrum obtained from the negative control (set of peptides from the empty gel piece treated in parallel) to exclude the most common contaminations. No restriction on the protein molecular mass was applied. The taxonomy parameter was restricted to Mammalia, for trypsin cleavage specificity one missed cleavage was allowed, and peptide tolerance was limited to 50 ppm. A number of fixed (acrylamide-modified cysteine, i.e. propionamide/carbamidomethylation) and variable modifications (methionine oxidation and protein N terminus acetylation) were included in the search parameters. For protein identification, peptide masses were transferred to the BioTools 2.0 interface (Bruker Daltonics) to search in the National Center for Biotechnology Information non-redundant (NCBInr) (version, September 22, 2007; 670,092 mammalian entries) database as a first search using the MASCOT search engine (version 2.2, released February 28, 2007; Matrix Science Ltd.). Often the peptides identified matched equally well to multiple database entries using the NCBInr database, and that is why the second/final search was performed using the same parameters but using the UniProtKB/Swiss-Prot 54.2 (17,252 human entries) database, and if the number and the sequence of the recognized peptides were identical to the first search, the Swiss-Prot accession number was assigned for the identified protein. If peptides matched to multiple members of a protein family to eliminate the redundancy (for selecting which member to report) an additional manual analysis of the individual spectrum was performed. Isoform-specific peptides as well as their intensities were visually inspected and verified in each individual spectrum, and the presence of additional isoform-specific peptides was checked (see supplemental figures for all identifications) in an attempt to unambiguously identify the correct protein family member. Additionally the identity of all identified isoforms was confirmed by Western blot analysis using isoform-specific antibodies. The protein identifications were considered to be confident when the protein score of the hit exceeded the threshold significance score of 70 (p Ͻ 0.05) and not less than six peptides were recognized.

RESULTS AND DISCUSSION
The extraordinarily high sequence conservation between 14-3-3 protein isoforms, characterized by the presence of highly conserved sequence blocks composed of long stretches of invariant amino acids (41), poses a significant technological challenge to researchers working with these proteins. Many reports addressing functional aspects of 14-3-3 biology do not differentiate between the various isoforms mostly because of methodological limitations. One of the great advantages of 2D gel-based proteomics comes from its ability to distinguish protein isoforms based on their apparent molecular weights and isoelectric points. The wealth of information that can be mined from a 2D PAGE pattern showing discrete protein spots for isoforms belonging to a given protein family makes this approach invaluable for isoform analysis. Fig. 1 shows a representative 2D PAGE-based protein expression profile of transformed human amnion (AMA) cells. A cluster of polypeptide spots localizing to the region we previously identified as containing the 14-3-3 protein family (10) was analyzed by mass spectrometry. Protein identities were established for all spots in this region ( Fig. 1, right-hand panel). We identified a complement of five 14-3-3 isoforms (␤, , , , and ), of which only two (␤ and ) did not separate into discrete spots, as being expressed by AMA cells. Isoforms ␥ and were not detected indicating that AMA cells do not express these isoforms.
Expression of 14-3-3 Isoforms during the Cell Cycle of AMA Cells-A number of genetic and biochemical studies in several organisms have shown that 14-3-3 proteins play a key role in cell cycle control (2,49). It is conceivable that transient regulatory changes in expression of 14-3-3 occur only at a specific point(s) during the cell cycle. To address this possibility we investigated the expression levels of the various 14-3-3 isoforms throughout the cell cycle of transformed human epithelial amnion (AMA) cells. Synchronized cells were harvested at different time points, labeled for 30 min with [ 35 S]methionine, and subjected to 2D PAGE analysis. As can be seen in Fig. 2A, the levels of expression of the various 14-3-3 isoforms were rather constant throughout the cell cycle as estimated by their ratio to nearby migrating housekeeping gene products (Table I) (Table I), validating our analysis. Interphase AMA cells also expressed the 14-3-3 isoforms at unchanged levels and exhibited nearly the same levels as unsynchronized AMA cells.
Overall our results showed that all 14-3-3 isoforms detected by 2D PAGE were expressed at analogous levels throughout the cell cycle making it unlikely that cell cycle-mediated differential expression of 14-3-3 proteins constitutes a major regulatory mechanism of the function of these proteins at least in AMA cells. However, given the dynamic expression patterns shown by some 14-3-3 isoforms during development and in response to specific stimuli (11,12,52,53) we cannot exclude the possibility that under certain growth conditions 14-3-3 expression may vary during the cell cycle. Homeostatic cells most likely express 14-3-3 at well defined levels, which may be characteristic for that particular cell type, and only when challenged by given external stimuli will cells react by changing protein expression levels of a given 14-3-3 isoform.
Phosphorylation Status of 14-3-3 Isoforms during the Cell Cycle-14-3-3 proteins are known to control progression through the cell cycle by binding in a phosphorylation-dependent manner to key cell cycle regulators modulating their activity (2,13). As phosphorylation of the various isoforms themselves may modulate their function, we investigated levels of 14-3-3 phosphorylation throughout the cell cycle of AMA cells labeled for a short time with [ 32 P]orthophosphate (Fig. 2B). We observed constitutive levels of phosphorylation of the ␤/ isoforms in all phases of the cell cycle and no phosphorylation of the and isoforms, and interestingly, we found that the isoform was phosphorylated at low but detectable levels in G 1 and mitosis and at higher levels in S phase (Fig. 2B).
Aitken et al. (54) have shown previously that the ␤ and isoforms are post-translationally modified in vivo. The phosphorylated forms of these isoforms showed a nearly 2-fold increase in the ability to inhibit protein kinase C consistent with the suggestion that phosphorylation of 14-3-3 plays a role in the regulation of ligand binding. The 14-3-3 isoform was shown previously to be phosphorylated in vivo by the breakpoint cluster region protein BCR kinase (55), a multifunctional regulator of the Rho GTP-binding protein family involved in cellular signaling (56,57). We established that phosphorylation of the ␤ and isoforms also occurred in AMA cells (Fig. 2B), and we found that, although the levels of phosphorylated 14-3-3 ␤ and were constant, phosphorylation of 14-3-3 varied with cell cycle progression, increasing markedly during S phase (Fig. 2B). This indicates that modulation of the phosphorylation levels of 14-3-3 is a cell cycleassociated regulatory mechanism for this protein in AMA cells. Although several reports have identified post-translational modifications of 14-3-3 (58, 59) we could find no evidence for phosphorylation of the 14-3-3 isoform in AMA cells (Fig. 2B). Given that AMA cells express 14-3-3 at relatively low levels, it was conceivable that phosphorylation of this isoform occurred but only in a small fraction of the total protein available, and consequently was below the detection limit of our assay. To address this question we repeated the assay using non-cultured primary unfractionated keratinocytes. These cells express 14-3-3 at levels similar to those of actin (Fig. 2C). But even with these very high levels of expression we were unable to detect phosphorylation of 14-3-3. As before, we could readily identify phosphorylation of the ␤/ isoforms.

Cellular Localization of 14-3-3 Isoforms
One possible regulatory mechanism that could account for isoform-specific cellular function(s) is differential subcellular localization of the different 14-3-3 isoforms as there is some evidence indicating that certain isoforms have a specific subcellular localization (60,61), presumably reflecting differences in biological function(s).
Subcellular Distribution of 14-3-3 Isoforms-To study the subcellular localization of the 14-3-3 ␤, , , , and isoforms in AMA cells we performed immunostaining of fixed cells using isoform-specific antibodies. For this purpose we first established the specificity of different antibodies to be used for this study by 2D Western analysis of whole cell extracts of AMA cells. At the dilutions used, with the exception of SA-481, all antibodies showed single isoform specificity (Fig. 3). However, in the case of antibodies 22-II-D8B and F46820, some cross-reactivity to other isoforms could be observed at higher antibody titers (data not shown; see also Ref. 62). No signal was detected using antibodies recognizing isoforms ␥ and (data not shown) indicating that these isoforms are not expressed by AMA cells; this was consistent with the 2D PAGE data (Fig. 1).
Immunostaining of AMA cells showed that the cellular localization of 14-3-3 was cytoplasmic, mostly of a diffuse nature but with some filament-like organization and no detectable nuclear presence (Fig. 4A, ). The isoform showed diffuse cytoplasmic staining with strong perinuclear localization (Fig. 4A, , yellow arrowhead) and significant nuclear staining. The ␤ isoform showed diffuse cytoplasmic localiza-tion, weak nuclear staining, and a very strong Golgi presence (Fig. 4A, ␤, yellow arrowhead), indicating that this isoform may play a role in Golgi function. Indeed 14-3-3 isoform ␤ has been shown to specifically compete with Coatomer protein complex I vesicular proteins for binding to proteins with dibasic retention motifs, presumably mediating their release from the ER to allow forward transport through the cis-Golgi complex (63). The isoform showed diffuse cytoplasmic localization and punctate staining (Fig. 4A, , white arrowhead), perinuclear localization (Fig. 4A, , yellow arrowhead), and a significant nuclear presence. The isoform showed a diffuse cytoplasmic presence, structured perinuclear localization (Fig.  4A, , white arrowhead), cytoplasmic membrane localization (Fig. 4A, , yellow arrowhead), and no detectable nuclear presence.
14-3-3 Is Associated with the Cytoskeleton-To address the nature of the filamentous organization we had observed for 14-3-3, we performed double immunofluorescence analysis using a 14-3-3 antibody in combination with one of various antibodies recognizing proteins known to form cellular filamentous structures such as class I intermediate filaments (cytokeratins 5, 7, 8, 13, 14, 15, 16, 17, 18, 19, and 20), class  . We also found that 14-3-3 was associated with tubulin, another known 14-3-3-interacting protein. Interestingly the staining pattern of 14-3-3 although filament-like was focal rather than continuous suggesting deposition of this protein along microtubules (MTs) at given interaction sites (data not shown; see also Fig. 5, panels a and c). These results show that, although 14-3-3 is predominantly located in the cytoplasm in a diffuse manner, it also has a filamentous-like organization that correlates with CK8/18 and tubulin, two well known 14-3-3 interactors (4, 64 -66). Localization of 14-3-3 during Mitosis-14-3-3 proteins are known to bind to several key cell cycle regulators, modulating their activity (2,13). Binding occurs during interphase but not during mitosis, suggesting that cellular distribution, and functionality, might vary during the cell cycle. Thus, we investigated the subcellular localization of 14-3-3 isoforms throughout the cell cycle of AMA cells.
Cellular localization of the isoform was unvaryingly diffuse throughout the cell cycle (illustrated in Fig. 4B, ). We did not observe any significant localization of 14-3-3 to the spindle suggesting that association with tubulin is not maintained during mitosis at least to detectable levels (Fig. 4D, panel a). Our studies, however, revealed a strong staining of the midbody within the area of constriction between the two daughter cells (Fig. 4C, panels a and b) and suggested a possible novel role of this protein in mitotic exit. Interestingly although 14-3-3 is present at considerable levels in the midbody (Fig. 4C,  panels b and c), it is barely visible in the cleavage furrow and incipient midbody (Fig. 4C, panel a). Interestingly we also FIG. 4

. Subcellular distribution patterns of 14-3-3 in AMA cells.
A, cellular distribution of 14-3-3 isoforms ␤, , , , and in interphase cells. Yellow arrowheads indicate perinuclear localization of 14-3-3, an area of strong 14-3-3␤ presence showing association with the Golgi, punctate staining of 14-3-3, and membrane localization of 14-3-3, respectively. White arrowheads indicate perinuclear localization of 14-3-3 and 14-3-3. Scale bar, 10 m. B, localization of 14-3-3 isoforms during the cell cycle. Representative AMA cells at different points in the cell cycle (, , and ␤, anaphase; , metaphase; , prometaphase) are shown for illustration. Yellow arrowheads indicate vesicular staining of 14-3-3␤, punctate staining of 14-3-3, and membrane localization of 14-3-3, respectively. The white arrowhead indicates perinuclear localization of 14-3-3. Scale bar, 5 m. C, increased localization of 14-3-3 to the midbody (yellow arrowheads in panels a-c) at cytokinesis. Localization to the midbody was a -specific event as no other 14-3-3 could be detected (illustrated for ␤ in panel d, yellow arrowhead). Scale bar, 2 m except in panel a (10 m). D, 14-3-3 does not localize to tubulin filaments during mitosis (panel a), but it is associated with centrosomes and localizes to the midbody (panel b; yellow arrowheads indicate centrosomes and the white arrowhead indicates the incipient midbody). Nuclear material was counterstained with TO-PRO-3. observed localization of the isoform to the centrosomes (Fig. 4D, panel b, yellow arrowheads). A similar observation was originally reported by Pietromonaco et al. (67) who found the and ␥ isoforms to be specifically associated with centrosomes and the spindle apparatus in mouse leukemic FDCP cells but could not detect the presence of the ␤ and isoforms in these structures. In addition, these authors also found differential localization of centrosomal 14-3-3 in mouse 3T3 cells with quiescent 3T3 cells lacking centrosomal 14-3-3 and serum-stimulated 3T3 cells showing centrosomal 14-3-3. The presence of 14-3-3 and ␥ isoforms in the human centrosome was confirmed by proteomics characterization of centrosomal preparations from the human lymphoblastic cell line KE-37 (68). Interestingly both studies failed to identify the presence of 14-3-3, an absence presumably ascribable to the fact that non-epithelial and consequently non-expressing 14-3-3 cell types were used in both cases. Proteins showing a similar localization pattern, i.e. present at the centrosome throughout mitosis by interacting with centrosome components rather than with microtubules and appearing at the cleavage furrow in late anaphase and in the midbody in cytokinesis, would represent putative centrosomal ligands of 14-3-3. One example of such a protein is Cep55 (centrosome protein 55 kDa). Cep55 has a mode III 14-3-3-binding motif, (pS/pT)X 1-2 -COOH where pS is phosphoserine and pT is phosphothreonine (56); locates to the centrosome; and upon mitotic entry undergoes multiple phosphorylation events that trigger its relocation to the midbody where it plays a key role in cytokinesis. Cells expressing phosphorylation-deficient mutant forms of Cep55 undergo cytokinesis failure (69).
Just like for the isoform, cellular localization of the isoform was unvaryingly diffuse throughout the cell cycle (exemplified in Fig. 4B, ) with no particular localization to discrete mitotic structures. In the case of 14-3-3␤, however, in addition to a diffuse cellular staining we could observe strong vesicular staining (Fig. 4B, ␤, yellow arrowheads) presumably resulting from the fragmentation of Golgi cisternae during mitosis (for a review, see Ref. 70). Localization of the isoform was diffuse with punctate staining throughout the cell cycle (exemplified in Fig.  4B, ) with no particular localization to discrete mitotic structures. The isoform showed strong membrane localization (Fig.  4B, , yellow arrowhead) throughout the cell cycle.
We found no evidence for the presence of 14-3-3 ␤, , or in centrosomes, MTs, spindle, or midbody (exemplified for 14-3-3␤ in Fig. 4C, panel d), indicating that under our experimental conditions association with centrosomes and midbody are isoform-specific events, respectively. To further investigate the association of 14-3-3 to MTs, we studied the dynamics of 14-3-3 localization during microtubule reassembly in cultured cells following nocodazole-induced microtubule depolymerization. As expected, upon exposure to the drug we observed microtubule depolymerization (Fig. 5, panels a and b) with concomitant loss of filamentous organization of cytoplasmic 14-3-3 but an increased nuclear presence of this protein (Fig. 5, panels a and b, white arrowheads). Upon prolonged treatment with nocodazole and following complete MT depolymerization we detected, in addition to the usual diffuse cytoplasmic staining, a substantial presence of 14-3-3 at the centrosome and nucleus (Fig. 3, panel b, yellow and white arrowheads, respectively). Fifteen minutes after drug washout, we observed reassembly of a radiating array of cytoplasmic MTs emanating from the centrosome (Fig. 5, panel c, yellow arrowhead) and filamentous organization of 14-3-3 (Fig. 5, panel c, white arrowhead). These data suggest that association of 14-3-3 with tubulin is dependent on MT assembly and that equilibrium exists in the distribution of the various biochemical fractions of 14-3-3 because MT breakdown led to increased abundance of the nuclear form of the 14-3-3 protein. This observation is in agreement with a previously reported study that showed that 14-3-3 mutants that are unable to bind their interaction partners remain preferentially localized to the nucleus implying that interaction to cognate ligands is required for retention to the cytoplasm (71).
Localization of the isoform was at all times distinct from that of the other isoforms (data not shown and Fig. 3, panel b) with the possible exception of centrosomes where we observed some colocalization of the and ␤ isoforms (Fig. 5,  panels b and c, yellow arrowheads). Presumably the latter is just a reflection of the localization of the ␤ isoform to the Golgi/centrosome region rather than a functional interaction between the two isoforms.

FIG. 5. Microtubule polymerization dynamics and 14-3-3 localization.
AMA wells were exposed to nocodazole (10 M) for 6 h, washed thrice with cold culture medium, and allowed to recover in drug-free medium at 37°C. Cells were fixed at different time points (0 and 15 min; panels a or b and c, respectively) and processed for double immunofluorescence labeling using antibodies against 14-3-3 and 14-3-3␤ or tubulin. Scale bar, 5 m. White arrowheads (panels a and b) indicate an increased nuclear presence of 14-3-3, whereas yellow arrowheads show localization with MTs (panel a) and with centrosome/MT-organizing centers (panels b and c).

Extracellular Localization of 14-3-3 Isoforms
In addition to their intracellular localization 14-3-3 proteins are known to occur in extracellular environments, most notoriously in the cerebrospinal fluid (CSF). The presence of 14-3-3 proteins in CSF is generally associated with human prion diseases and other neurodegenerative disorders, such as Creutzfeldt-Jakob disease, mitochondrial encephalopathy with lactic acidosis and stroke-like episodes, multiple sclerosis, or encephalitis, making 14-3-3 CSF protein(s) a useful biological marker of brain damage in neurological disorders (72)(73)(74)(75). But it was only recently that evidence for a specific extracellular function of 14-3-3 proteins was provided. Ghahary et al. (76) demonstrated that extracellular 14-3-3 stimulates collagenase production by fibroblasts. Fig. 6A shows representative 2D gels of whole cellular extracts (left-hand panel) and corresponding externalized proteins (right-hand panel) present in the medium of cultured AMA cells. As can be seen, a fraction of all 14-3-3 isoforms could be found extracellularly. Given the simultaneous lack of relatively abundant cellular proteins such as cytokeratin 19, Rho GDI, or Annexin V in the medium, we can exclude the possibility that 14-3-3 proteins were simply released by cell death or lysis, indicating that they were externalized to the extracellular environment by a cellular mechanism.
We showed that AMA cells externalize all 14-3-3 isoforms examined to the growth medium, and our laboratory had previously identified 14-3-3 proteins in the extracellular environment of bladder and breast cancer tissues as well as of cultured cell lines (20,39). However, given the characteristic monolayer growth of cultured cell lines and the distinctive secretory properties of malignant cells, the presence of extracellular 14-3-3 in cancerous tissues and growth medium of cultured cell lines may be an epiphenomenon rather than reflect a possible functional role of extracellular 14-3-3. Fig.  6B shows representative 2D gels of whole cellular extracts and corresponding externalized proteins present in the interstitial fluid of a benign breast specimen (non-malignant mammary interstitial fluid (NIF)). As can be seen, all 14-3-3 isoforms were externalized to the NIF (Fig. 6B, AMA cell lysate and externalized proteins, respectively). Given the lack of cellular proteins such as cytokeratin 19, Rho GDI, or Annexin V in the NIF we can exclude the possibility that 14-3-3 were

FIG. 6. Analysis of extracellular 14-3-3 proteins in AMA cells (A) and in mammary tissue (B).
A, a crude cell lysate prepared from AMA cells was analyzed by IEF 2D PAGE (AMA cell lysate) and matched with the pattern of proteins present in the corresponding growth medium (externalized proteins). B, interstitial fluid extracted from mammary biopsies was analyzed by IEF 2D PAGE (externalized proteins) and matched with the pattern for the corresponding tissue lysate (mammary tissue lysate). In all cases the positions for the 14-3-3 isoforms and for some cellular proteins (cytokeratin 19, Annexin V, and Rho GDI) are indicated for reference. The identity of all indicated proteins was confirmed by MS. C, levels of expression of 14-3-3 in cultured primary HMECs. HMECs were cultured to near confluency at which point culture medium was replaced by fresh medium containing different additives (brefeldin A (BfA), prolactin (PRL), or nocodazole (Noco)). Cells and corresponding medium were collected after 4 or 24 h, and total cell lysates and corresponding medium were prepared for analysis of 14-3-3 expression by Western blot analysis. Levels of ␤-actin were used as a normalizing factor for total amount of protein loaded. D, immunohistochemistry analysis of 14-3-3 in a breast specimen. Normal breast epithelium from a reduction mammoplasty with epithelial cells staining positive and with significant staining of myoepithelial cells (ϫ200) is shown. Apical snouts containing 14-3-3 and being shed off to the lumen of the duct (arrows) were frequently detected suggesting this to be a mechanism for externalization of 14-3-3. simply released by cell death/lysis, indicating that they were externalized to the extracellular environment and confirming externalization of 14-3-3 as a physiological process.
To address the mechanism of externalization of 14-3-3 we analyzed cultures of primary human mammary epithelial cells (HMECs). The presence of 14-3-3 in the medium could be detected after 4 h of culturing, reaching a maximum after 24 h of conditioning (Fig. 6C). Interestingly we observed an increase in the externalization of 14-3-3 in mammary epithelial cells exposed to nocodazole (10 M) (Fig. 6C, Noco), whereas we found no difference in the extracellular levels of 14-3-3 following prolactin-induced proliferative stimulation of breast cells (100 ng/ml) (Fig. 6C, PRL). Although 14-3-3 does not have any typical ER export signal, it is conceivable that secretion of this protein occurs via the canonical ER-Golgi secretory pathway. However, treatment of mammary cells with brefeldin A (5 g/ml), which is known to block ER-to-Golgi transport, did not inhibit 14-3-3 externalization in mammary cells but rather increased it (Fig. 6C, BfA) indicating that 14-3-3 is externalized through a non-classical mechanism of secretion. At least four distinct types of nonclassical export have been proposed: direct import into intracellular recycling endosomes, direct translocation across the plasma membrane, membrane flip-flop following protein anchorage to the membrane, or through the formation of exosomes, vesicles that form on the outer surface of the cell in a process known as membrane blebbing (for a review, see Ref. 77). Immunohistochemical analysis of benign breast tissue (Fig. 6D) showed the presence of 14-3-3 protein in vesicles disposed in the apical plasma membrane of mammary epithelial cells that protrude from the cell surface into the alveolar lumen, indicating that externalization of 14-3-3 may occur via formation of exosomes, a possibility supported by the proteomics analysis of exosomes isolated from human urine and of dendritic cell-derived exosomes (78,79) that identified 14-3-3 proteins as exosomal constituents.

Correlation of Cell Line-based Analysis with Tissue Biopsies
The overwhelming majority of studies on the biology of 14-3-3 proteins have been done using immortal cell lines. To address the question of how well the observations we report here using AMA cells correlate to the situation in vivo, we tried to reproduce some of our analyses in patient-derived samples. From different analyses performed on tumor tissues as well as other cell types we know that the complement of 14-3-3 proteins expressed in different samples can vary significantly, and to aid the interpretation of the data presented below we present in Fig. 7A a synthetic gel image summarizing data collected from different sources. Isoforms ␥ and , when expressed, focus together with and , respectively. In some cases, modified forms of (*) and (*) are also expressed at sufficiently high levels to allow direct visualization.
We determined the relative expression of 14-3-3 proteins in bladder and breast tissues by 2D gel analysis of 12 bladder and 16 mammary benign tissue samples obtained from patients that underwent tumor resection (a representative 2D gel of each tissue type is presented in Fig. 7B, bladder tissue and breast tissue, respectively). Samples were derived from nonmalignant epithelia harvested at sites distal to the carcinoma and judged histologically normal. We found that the relative expression of 14-3-3 isoforms was rather homogeneous within the same tissue type in the clinical specimens we analyzed with modest interpatient variation (Table II, breast and bladder tissue). But whereas the ␤/, ␥/, and / isoforms showed comparable expression levels between urinary bladder and mammary tissue samples, the isoform showed consistently lower expression in mammary epithelia (10-fold lower) as compared with bladder epithelium (Table II and Fig.  7B). By contrast, in AMA cells (compare Fig. 1 with Fig. 7B), one can see that expression of the whole complement of 14-3-3 isoforms is somewhat different in these cells from bladder and breast tissues.
It was conceivable that the differences in the expression of the various isoforms observed between AMA cells and bladder or breast tissue samples were cell-type specific. However, comparison of 14-3-3 expression profiles in a large set of cell lines suggests that this is not the case (Fig. 7B and Table III).   14-3-3

isoforms in tissue samples
Raw pixel intensity values were determined for each 14-3-3 isoform and normalized to the sum of the intensities of all 14-3-3 isoforms. Expression of each isoform is presented as a percentage of the total net expression of 14-3-3 proteins.

Cellular Analysis of 14-3-3 Proteins
For example, HCC1569, a breast carcinoma cell line, showed impaired expression of the / isoforms relative to the breast tissue samples analyzed. By contrast, expression of the remaining 14-3-3 isoforms was comparable between HCC1569 and the breast benign tissues we examined (illustrated in Fig.  7B) including 14-3-3, an isoform whose lack of expression has often been associated with malignancy (for a review, see Ref. 4). Conversely in the non-malignant bladder epithelium cell lines Hu609 and HCV29, expression of isoforms ␤/, ␥/, and / was analogous to what we observed in bladder tissues (Fig. 7A), but we could not detect expression of the isoform. The large number of different cell lines examined allowed us to get an idea of the variability in expression of the complement of 14-3-3 isoforms in different cell types. The isoform was not expressed in non-epithelial cell types (e.g. Raji, MRC5, or melanocytes) and was highly expressed in squamous cell lineages (HaCaT) and mammary myoepithelial cells (MCF-10A) confirming what we and others have shown, that expression of the isoform, besides being epithelialspecific, also has a cell lineage component with epithelial cells of squamous cell lineage greatly overexpressing this protein and mammary luminal epithelial cells showing 10-fold lower 14-3-3 expression as compared with mammary myoepithelial cells (10,20,80). Only the isoform varied so dramatically in expression. The other isoforms were expressed at more analogous levels in all cell lines with occasional variation, but all cell lines examined expressed them. Not only was expression of the complement of 14-3-3 isoforms variable in different cell lines, but also subcellular localization was not uniform for all isoforms in the various cell a Raw pixel intensity values were determined for each 14-3-3 isoform and normalized to the sum of the intensities of all 14-3-3 isoforms. Expression of each isoform is presented as a percentage of the total net expression of 14-3-3 proteins. b ND, non-detectable; below detection level of the assay.
lines. This is exemplified in Fig. 7C, showing that we could readily detect the presence of in the midbody of AMA cells undergoing cell division but were unable to do so in T47D breast cancer cells even though these two cell lines express 14-3-3 at similar levels (cf. Figs. 1 and 7B). We took advantage of the lack of expression of 14-3-3 in cell lines that in view of their cellular lineage should express this isoform (Hu609 and HCV29) and examined the cellular localization of the remaining 14-3-3 isoforms (, ␤, , and ) in these cells to determine whether some compensatory effects could be observed. As illustrated in Fig. 7D, the cellular staining patterns for 14-3-3 isoforms were identical to what we had observed previously in AMA cells (cf. Fig. 4A with 7D), indicating that if some functional compensation occurs, the phenotype is too subtle to be discovered in this manner.
These data suggest that the cluster of 14-3-3 isoforms present in a given cell line is characteristic for that cell line, and it does not necessarily reflect the pathological state of that cell line or even the expression patterns of 14-3-3 proteins from the tissue of origin. It is therefore doubtful that any mechanistic conclusions can be drawn alone from the presence or absence of a given 14-3-3 isoform in a cell line. On the other hand, changes in expression levels of a given isoform by different cell lines provide a helpful tool for studying the biological function of that isoform by isoform-specific in-depth studies.
One interesting possibility raised by our observations is that the ratio between the specific isoforms and the total 14-3-3 protein pool could be a regulatory determinant of 14-3-3 function because in many of the cell lines not expressing 14-3-3 or expressing other isoforms at relatively low levels there seems to be a compensatory mechanism. For example compare expression of 14-3-3␥/ in HCC1569 and MDA-MB-231, two cell lines with no expression and very low levels of expression of 14-3-3, respectively, with T47D, BT20, or MCF-7.

Conclusions
We performed an extensive 2D gel-based analysis of 14-3-3 isoform expression (Fig. 7B) and immunocytochemistrybased isoform subcellular localization in a panel of over 20 established cell lines and cultured primary cells. We observed that expression and localization of the complement of 14-3-3 isoforms was not uniform in the different cell lines (illustrated in Fig. 7C), underscoring the importance of cellular context and the care one has to take when extrapolating a cellular function by combining experimental data obtained in different cellular systems. To establish an experimental framework that could allow novel insights into possible differential biological roles for the different 14-3-3 isoforms we investigated the expression, subcellular localization, and post-translational modifications of 14-3-3 proteins in one single well defined biological background. For this we used a combination of proteomics and cell biology approaches to systematically analyze human amnion epithelial AMA cells corroborating the relevance of our findings in other cellular backgrounds such as urothelial and mammary tissue samples. We found that whereas 14-3-3 shows highly dynamic expression patterns in the various experimental systems examined and no detectable post-translational modifications the ␤, , , and isoforms were expressed at relatively uniform levels, and of these, ␤, , and were modified by phosphorylation, indicating that these isoforms are controlled by different regulatory mechanisms. In this respect, it is especially intriguing that the ratio of the pools of free cytoplasmic protein to bound protein seems to be regulated as the release of tubulin-associated cytoplasmic 14-3-3 following MT depolymerization resulted in a transient increase in the presence of nuclear and extracellular 14-3-3.
Compartmentalization of the different isoforms was also diverse. Thus, 14-3-3 co-localized with CK8/18 or tubulin in MTs, intermediate filaments, and centrosomes, whereas the ␤ isoform was predominantly associated with the Golgi apparatus. In addition we identified the transient presence of 14-3-3 in the midbody. A recent study has also shown that depletion of 14-3-3 results in reduced mitosis-specific expression of the internal ribosomal entry site-dependent form of the cyclin-dependent kinase Cdk11, leading to loss of Polo-like kinase-1 at the midbody and impaired cytokinesis (81). These data are suggestive of a regulatory role for 14-3-3 in cytokinesis at two different levels: controlling expression of Cdk11, a critical mitotic regulator, and transiently binding to cognate ligand(s) at the midbody during mitotic exit. We also showed that all isoforms examined occur extracellularly, pointing toward a function in the extracellular environment.
One important aspect of our study is the striking diversity we observed in the expression of the different 14-3-3 isoforms in particular when compared with tissue samples. One corollary of this observation is that great care has to be taken when interpreting data generated using established cell lines. For example, HCC1569, a widely used breast carcinoma cell line, expresses the / isoforms at low levels comparatively with breast epithelium, making it unsuitable for studying the normal functional role of these isoform. However, this cell line could provide a reasonable approximation to mammary epithelium for researchers studying the isoform. Another important aspect of the work presented here relates to the fact that most studies that have been reported on 14-3-3 proteins, because of technological limitations, do not distinguish between isoforms to identify which are relevant for the study in question. Worse still, in many cases 14-3-3 isoforms are assumed to be functionally equivalent. The technological approach we took combining 2D PAGE analysis and immunocytochemistry using isoform-specific antibodies has the advantage that one can assign any observed cellular effect to a given isoform or at the very worst to a pair of isoforms.
On a final note it should be stressed that our work was based on cellular systems that are in a given differentiation state and under defined growth conditions. Thus, we cannot exclude that differences in expression, post-translational modifications, and/or subcellular localization not seen in this study may take place for some or all isoforms under specific conditions. In fact, some reports have shown that this is the case. Expression of the ␤, ␥, and isoform are known to change during neuronal development (82), and expression of 14-3-3 is strongly induced by ionizing radiation and other DNA-damaging agents in colorectal cancer cells (12).