The DEK Nuclear Autoantigen Is a Secreted Chemotactic Factor

Abstract

The nuclear DNA-binding protein DEK is an autoantigen that has been implicated in the regulation of transcription, chromatin architecture, and mRNA processing. We demonstrate here that DEK is actively secreted by macrophages and is also found in synovial fluid samples from patients with juvenile arthritis. Secretion of DEK is modulated by casein kinase 2, stimulated by interleukin-8, and inhibited by dexamethasone and cyclosporine A, consistent with a role as a proinflammatory molecule. DEK is secreted in both a free form and in exosomes, vesicular structures in which transcription-modulating factors such as DEK have not previously been found. Furthermore, DEK functions as a chemotactic factor, attracting neutrophils, CD8⁺ T lymphocytes, and natural killer cells. Therefore, the DEK autoantigen, previously described as a strictly nuclear protein, is secreted and can act as an extracellular chemoattractant, suggesting a direct role for DEK in inflammation.

DEK is a mammalian oncoprotein and putative autoantigen whose primary biological function has not been previously elucidated, despite literature linking it to the regulation of transcription, chromatin architecture, and mRNA processing (1, 2, 11, 15, 16, 24, 35, 51, 65). The DEK protein is widely conserved among species and is transcribed at high levels, especially in hematopoietically derived cells (14, 62-64). DEK was first characterized as part of the protein product of the DEK-CAN fusion oncogene, resulting from a t(6;9) translocation found in a subset of patients with acute myelogenous leukemia (64). DEK is overexpressed in several different malignancies, including melanoma, hepatocellular carcinoma, glioblastoma, retinoblastoma, bladder cancer, T-cell large granular lymphocyte leukemia, and acute myelogenous leukemia independent of the t(6;9) translocation (7, 10, 19, 20, 29, 30, 32, 39, 64). Although DEK has previously been described as a strictly nuclear protein, DEK autoreactivity has been identified as a major component of the autoantibody profile in patients with juvenile idiopathic arthritis (JIA) and is also seen in patients with other autoimmune diseases (8, 21, 38, 49, 54, 55). In the studies presented in this paper, we show that the nuclear protein DEK can be actively secreted by inflammatory cells, is found in synovial fluid samples from JIA patients, and can function as a chemoattractant for inflammatory cells, suggesting a potential role for DEK in immunity and/or autoimmunity.

MATERIALS AND METHODS

Cell preparation.

Monocyte-derived macrophages (MDM) were prepared as previously described (31). Briefly, heparinized venous blood samples were collected from healthy volunteers following a protocol approved by the institutional review board, and peripheral blood mononuclear cells were separated by Ficoll-Hypaque (Amersham Pharmacia Biotech AB, Uppsala, Sweden) density gradient centrifugation. MDM were purified by adherence to plastic for 2 h at 37°C at a concentration of 5 × 10⁵/ml. Adherent cells were consistently >90% monocytes (31). Adherence-purified human monocytes were cultured in X-Vivo medium supplemented with 40% human AB serum (Bio-Whittaker, Walkersville, MD), 100 units of penicillin per ml, and 50 units of streptomycin per ml (41).

Synovial fluid samples.

Synovial fluid samples were obtained from the Pediatric Rheumatology Tissue Repository of the Cincinnati Children's Hospital Medical Center through a protocol approved by the institutional review board. The samples were originally collected from 45 children undergoing therapeutic intra-articular injections for inflammatory arthritis and from 10 children undergoing diagnostic arthroscopy for presumed noninflammatory joint pain. Synovial fluid samples were diluted (1:1) in phosphate-buffered saline (PBS) and centrifuged at 200 × g for 30 min to separate cells from the fluid prior to freezing at −70°C.

Antibodies.

Rabbit polyclonal antiserum against an N-terminally truncated DEK protein (amino acids 68 to 375) was kindly provided by Gerard Grosveld (St. Jude Children's Research Hospital, Memphis, TN). Affinity-purified, goat polyclonal anti-DEK antibody, raised against a peptide from the carboxy terminus of DEK, and monoclonal antibody to vimentin (V-9) were purchased from Santa Cruz (Santa Cruz, CA). Monoclonal antibody and goat polyclonal antibody to CD81 were also purchased from Santa Cruz (Santa Cruz, CA). A monoclonal antibody was raised against the full-length DEK protein, which was overexpressed in Spodoptera frugiperda SF-9 insect cells infected with recombinant baculovirus (Clontech, Palo Alto, CA). DEK protein from baculovirus-infected SF-9 cells was isolated using His-Trap columns for purification of histidine-tagged proteins (Amersham Pharmacia Biotech) with further purification by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and electroelution. Cell fusion, hybridoma cell line development, and ascitic fluid production were all performed by the hybridoma core facility at the University of Michigan and the hybridoma development service at the Saint Louis University Health Sciences Center. CD56-PE (CD56 conjugated to phycoerythrin [PE]), CD19-FITC (CD19 conjugated to fluorescein isothiocyanate [FITC]), CD4-FITC, CD8-PE, and mouse immunoglobulin G1 isotype control antibodies for flow cytometry were purchased from BD Pharmingen (San Diego, CA).

Immunohistochemistry.

Human monocytes and MDM were cultured using a glass chamber slide system (Nalge Nunc International, Naperville, IL). Cells were washed with PBS, fixed for 10 min at 4°C with PBS containing 4% paraformaldehyde, and then washed again with PBS prior to blocking for 1 h with 0.2% bovine serum albumin in PBS. Slides were incubated with polyclonal goat or rabbit anti-DEK antibody diluted 1:100 in PBS with 0.1% saponin (to permeabilize cells) for 1 h. Slides were washed thoroughly with 0.1% saponin in PBS, and were blocked a second time by incubation with normal goat or rabbit serum for 1 h and then washed and incubated with 10 μg/ml Alexa Fluor 488 conjugated to goat anti-rabbit or rabbit anti-goat antibody (Molecular Probes, Eugene, OR). The slides were washed with distilled water, dried, and mounted with SlowFade Antifade kit (Molecular Probes). Fluorescence was viewed with a Leitz Orthoplan microscope or Bio-Rad MRC-600 laser-scanning confocal microscope. Photographs were taken with a Sony DKC5000 3CCD RGB camera.

Western blots.

MDM were maintained for 12 h in serum-free conditioned media with added dexamethasone (0.5 μM or 1 μM) (Sigma, St. Louis, MO), cyclosporine A (CsA) (1 μg/ml) (Sigma, St. Louis, MO), carbonyl cyanide chlorophenylhydrazone (CCCP) (10 μM) (Sigma, St. Louis, MO), interleukin-8 (IL-8) (10 ng/ml) (R&D Systems, Minneapolis, MN), or 4,5,6,7-tetrabromobenzotriazole (TBB) (50 μM) (47) for 3 h. Following three washes with PBS, serum-free medium was added for 3 to 12 h before harvest. The viability of the cells incubated in serum-free medium with or without added drugs was >95% for up to 48 h as measured by a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT)-based colorimetric assay (Boehringer Mannheim, Indianapolis, IN). Supernatants were collected, centrifuged for 20 min at 200 × g, and then concentrated by a centrifugal filter device (Millipore) in the presence of Complete protease inhibitor cocktail tablets (Boehringer Mannheim, Indianapolis, IN). Equal amounts of protein (20 μg) were loaded under reducing conditions, and proteins were separated by SDS-PAGE. The proteins were subsequently transferred to nitrocellulose and probed with mouse monoclonal anti-DEK or rabbit or goat polyclonal anti-DEK antibodies. The bound primary antibody was then detected with horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit secondary antibody using the Super Signal West Pico system (Pierce Chemical Co, Rockford, IL).

Silver staining.

Total protein from the supernatants of 2.5 × 10⁷ day 12 MDM was concentrated and separated by SDS-PAGE. Bands were compared to 1 μg of purified histone protein (type III-SS calf thymus; Sigma) and total cell extract from 5 × 10⁵ cells. The proteins were analyzed by silver staining by the method of Wray et al. (70).

Isolation of exosomes.

Exosomes were isolated by differential centrifugation as described previously (9). Day 14 MDM were washed and incubated in serum-free RPMI 1640 medium for 48 h, and supernatant (12 ml) from 2.5 × 10⁷ cells was collected and centrifuged for 10 min at 200 × g. The supernatant was removed, recentrifuged for 10 min at 500 × g, and then sequentially centrifuged at 2,000 × g for 30 min, 10,000 × g for 30 min using an SS-34 rotor (Sorvall), and 70,000 × g for 60 min using a TY-65 rotor (Beckman Instruments, Inc., Fullerton, CA). All pellet fractions were solubilized under reducing conditions using standard SDS-PAGE loading buffer and then incubated for 5 min at 95°C. The samples from each pellet and the final supernatant were analyzed by SDS-PAGE and Western blotting (see above).

Proteinase K protection assay.

The 70,000 × g exosome fraction was divided into three 35-μl aliquots and then was either left untreated or treated with 200 ng proteinase K (Roche, Indianapolis, IN) in the presence or absence of 0.5% Triton X-100. The samples were incubated for 3 h at 37°C, and the reaction was terminated with phenylmethylsulfonyl fluoride (5 mM final concentration). Immediately after the addition of phenylmethylsulfonyl fluoride, SDS loading buffer was added, and the sample was heated at 95°C for 5 min and analyzed by SDS-PAGE and Western blotting (see above) (3, 13).

In vitro migration assay.

White blood cells were isolated from venous blood samples from healthy donors as described above and were depleted of monocytes by adherence to plastic for 2 h as described above in the “Cell preparation” section. Monocyte-depleted white blood cells (1 × 10⁶ cells/ml) were fluorescently labeled with 20 μM 2′,7′-bis-(2-carboxyethyl)-5-carboxyfluorescein acetoxymethyl (BCECF AM; Molecular Probes, Eugene, OR) according to the manufacturer's directions. The cells were washed and resuspended at 1 × 10⁶ cells/ml in serum-free RPMI 1640 medium, and 1 × 10⁵ labeled cells in 100 μl were added to the upper chamber of a 24-well Transwell chemotaxis insert with a pore size of 3 μm or 5 μm (Corning, Corning, NY). The lower chambers contained serum-free RPMI 1640 medium alone, recombinant DEK produced in the baculovirus system, or a deletion mutant (THP) of the β form of human GLI-2 (50) produced in baculovirus as control protein. After 30 min to 1 h, the number of fluorescently labeled migrating cells in the lower chamber was determined at 485- or 535-nm wavelength using a Tecan GENios plate reader (Phenix, Austria). The results were expressed as the average increase in the number of cells migrating toward lower chambers containing DEK divided by the number of cells migrating toward the control wells (medium alone).

Neutrophil purification.

Forty-milliliter venous blood samples were collected from healthy volunteers into 60-ml sterile syringes containing a mixture of 7 ml of 0.25 M citrate (0.17 M sodium citrate and 0.083 M citric acid) and 6% dextran in PBS buffer. The blood samples were incubated for 30 min at room temperature, and the upper phase was collected and further separated by Ficoll-Hypaque as described above. The neutrophil fraction was collected and washed with Hanks balanced salt solution buffer and then pelleted again. The remaining red blood cells were further lysed in 9 ml cold distilled water for 30 seconds, and the reaction was stopped by 1 ml of cold 10× PBS buffer. Cells were washed again with Hanks balanced salt solution and resuspended in RPMI 1640 medium as described above for the migration assay.

Flow cytometry.

Cells migrating to the lower chambers (∼10⁶) collected from a 5-μm-pore-size Transwell chemotaxis insert were washed three times with fluorescence-activated cell sorting (FACS) buffer (Dulbecco's PBS, 3% fetal bovine serum, and 0.09% NaN₃) and then incubated on ice with antibodies in 100 μl of FACS buffer for 60 min. Cells were washed again three times in FACS buffer, fixed in 1 ml of PBS with 1% paraformaldehyde overnight at 4°C, washed three times with FACS buffer, and analyzed by EPICS XL Flow Cytometer System II software (Coulter, Miami, FL).

In vivo cell migration assay.

Eighty C57BL/6 male mice, 4 to 6 weeks old, were injected intraperitoneally with 50 μg of lipopolysaccharide (LPS), 50 μg of the THP isoform of human GLI-2 (a control protein) (50) produced in baculovirus under the same conditions as DEK), or 50 μg of DEK in 250 μl of PBS, or just PBS. Half of the mice were sacrificed after 4 h, and the other half were sacrificed 8 h after injection. Mouse peritoneal cavities were lavaged three times with 3 ml of cold PBS. White blood cell counts, percentages of polymorphonuclear leukocytes, and numbers of polymorphonuclear leukocytes per milliliter were determined from differential and hemocytometer counts performed on peritoneal lavage fluid samples (23). All experiments were performed in compliance with University of Michigan guidelines and were approved by the University Committee on the Use and Care of Animals.

RESULTS

DEK is present in the cytoplasm of MDM.

As DEK was previously considered to be an exclusively nuclear protein (1, 2, 14-16, 62-65), we initially wished to examine its subnuclear localization by immunofluorescence and confocal microscopy. For this purpose, we chose to employ a well-characterized model of monocyte-derived macrophages in which human monocytes differentiate into macrophages with incubation in 40% human serum (41, 43). This culture system induces the development of multiple differentiation pathways in the MDM, resulting in a phenotype similar to that of macrophages exposed to chronic inflammatory stimuli, with the cells acquiring functional properties such as increased phagocytic and secretory activity. After 7 to 10 days of serum-induced differentiation, MDM secrete large amounts of tissue-destructive proteinases and develop the ability to degrade elastin (41, 43).

While DEK has previously been described as an exclusively nuclear protein (14), during MDM differentiation, we observed a time-dependent increase in DEK's cytoplasmic distribution at days 3, 5, and 12 (Fig. 1a, c, and e). Additionally, with serum-induced differentiation, we noted that by day 12 DEK had disappeared from the nuclei of many of the MDM (69.5% ± 9.5% as calculated from 400 cells in a total of four different fields from four different donors) (Fig. 1e and f). The ability of DEK to translocate to the cytoplasm was further confirmed by using an enhanced green fluorescent protein hybrid construct (DEK-EGFP) in transiently transfected MDM. Live-cell microscopy images of selected cells showed distribution of DEK-EGFP in MDM in nuclear, perinuclear, cytoplasmic, and perimembranal locations (data not shown). These observations are consistent with our previous work demonstrating the disappearance of DEK from nuclear extracts of phorbol myristate acetate-differentiated U937 monocytic cells (11).

FIG. 1.

Intracellular DEK is not an exclusively nuclear protein. Human peripheral blood monocytes were isolated and cultured using a glass chamber slide system and were maintained in 40% human serum. Cells were placed in serum-free medium overnight before staining the next day. Day 3 MDM (panel a), day 5 MDM (panel c), or day 12 MDM (panel e) were incubated with DEK polyclonal antiserum and stained with FITC-conjugated goat anti-rabbit antibody for immunodetection by confocal microscopy. Nuclear staining with 4′,6′-diamidino-2-phenylindole (DAPI) is shown in panels b, d, and f (magnification, ×100). The white arrowheads in panel e show MDM with minimal or no nuclear staining for DEK. The corresponding nuclei are again indicated by the white arrowheads in panel f.

DEK is secreted by MDM via a process dependent on protein kinase CK2.

Because serum-differentiated inflammatory MDM actively secrete a variety of proteins (41, 43), we investigated whether DEK's movement into the cytoplasm was followed by its secretion. Cell-free supernatants from day 3, 7, and 12 MDM were analyzed for the presence of extracellular DEK. Starting at day 3 and continuing through day 12, as monocytes differentiated into MDM, we were able to identify increasing amounts of the 50-kDa form of DEK and trace amounts of the 35-kDa form in cell-free supernatants by Western blot analysis using a DEK-specific monoclonal antibody (Fig. 2A). As anticipated, a corresponding reduction in the amount of nuclear DEK was detected in the same time frame (Fig. 2A). (Intracellular DEK can be detected in multiple forms, especially in primary cells, probably due to heavy posttranslational phosphorylation and acetylation [6, 15, 16, 26, 27, 66], and intracellular DEK runs at sizes ranging from 35 to 55 kDa [15, 16]. The 35-kDa form is a known breakdown product lacking the N terminus [6, 8, 11, 14, 35, 49]). Overall, the presence of extracellular DEK in the supernatants of differentiated MDM was confirmed using three separate monoclonal antibodies and three separate polyclonal antibodies to DEK (data not shown).

FIG. 2.

DEK is secreted in a CK2-dependent manner as monocytes differentiate into MDM. Monocytes incubated in 40% human serum for 3, 7, and 12 days were washed and maintained in serum-free medium for 12 h prior to collection of cell supernatants. A total of 30 μg of protein was subjected to Western blot analysis using anti-DEK monoclonal antibody. The arrows in panels A and B indicate the 50- and 35-kDa forms of DEK. (A) Western blot of supernatants and nuclear extracts. Total nuclear protein loaded is shown by Ponceau S staining. (B) Day 8 MDM treated with a 50 μM concentration of the highly specific CK2 blocker TBB (+) or with the dimethyl sulfoxide control (−) for 3 h were washed and maintained in serum-free medium for the indicated time. Supernatants were then collected and analyzed by Western blotting using mouse monoclonal anti-DEK antibody, polyclonal CD81 antibody, or antivimentin antibody. (C) Increasing amounts of concentrated supernatant from day 12 MDM were collected and analyzed for the presence of histones by silver staining. For controls, purified histones and cell extract were run on the same gel.

Although macrophages are known to be resistant to many apoptotic stimuli, we wished to be certain that the presence of DEK in the supernatants of differentiated MDM was not a consequence of cell damage or demise. Several methods were used to confirm that DEK is actively secreted, rather than simply leaked, into the extracellular compartment. Previously, we have shown unaltered cell viability and detected no cellular necrosis or apoptosis of the differentiated MDM, even in the absence of sera for up to 3 days, as measured by extracellular lactate dehydrogenase and other indicators (41, 43). We now examined the cell viability of day 12 MDM by the MTT assay after 12, 24, and 48 h in serum-free medium and found no detectable decline in viability over this period. Specifically, by the MTT assay, the mean absorbance values (A₅₇₀ − A₆₅₀) ± standard deviations were 1.19 ± 0.37 for the control cells (day 12 MDM in serum-containing medium) and 1.12 ± 0.25, 1.00 ± 0.26, and 1.31 ± 0.29 for day 12 MDM after 12, 24, and 48 hours in serum-free medium, respectively. A total of 1 × 10⁵ cells were used for each condition. The means are from three individual experiments, each performed in duplicate (n = 6).

To further substantiate the hypothesis of an active mechanism of secretion, we used several additional approaches. First, while DEK is found in abundance in the serum-free supernatant of day 8 MDM after just 3 hours in culture (Fig. 2B), secretion of DEK can be blocked by 4,5,6,7-tetrabromobenzotriazole (TBB), a specific inhibitor of the proapoptotic kinase CK2 (34, 46, 48) that has been previously shown to be the primary kinase involved in DEK phosphorylation (26). After a 3-hour treatment of day 8 MDM with TBB, we noted DEK to be absent in supernatants collected during the first 3 hours of incubation in serum-free medium (Fig. 2B). At 6 hours after removal of TBB, its inhibitory effect had begun a gradual decline but was still observed even after 12 h. This indicates that DEK is actively secreted by viable MDM and that DEK secretion requires CK2-mediated phosphorylation. The supernatant was also analyzed for the presence of the exosomal marker CD81 (Fig. 2B). As will be discussed below, a portion of secreted DEK is found in these secreted structures. TBB modestly decreased secretion of CD81, indicating that CK2 might have some general effect on the secretion of exosomes, in addition to a more pronounced and specific effect on DEK secretion. We have previously demonstrated that MDM can secrete the intermediate filament protein vimentin through a process involving the Golgi apparatus (37). Treatment of MDM with TBB had no effect on the secretion of vimentin (Fig. 2B), again demonstrating the specific effect of TBB on DEK secretion. In addition, no nuclear histone proteins were found in the supernatant of MDM using the sensitive silver stain method (Fig. 2C), further supporting the observation that DEK is actively secreted into the supernatant of activated MDM and is not merely leaked out with other nuclear proteins. In addition, treatment of MDM with the proapoptotic energy blocker CCCP completely blocked DEK secretion (see Fig. 4C and discussion below). Interestingly, we could not detect the presence of the nuclear protein HMGB-1 (high-mobility group box 1 protein) in MDM supernatants. HMGB-1 is a nuclear protein that is released from necrotic cells or actively secreted by monocytes in response to lipopolysaccharide, IL-1, or tumor necrosis factor alpha (TNF-α) (5, 67). Under serum-free cell culture conditions, Western blot analysis of the supernatants showed no evidence of HMGB-1 (data not shown). These data indicate that DEK secretion proceeds through a specific pathway and exclude a significant contribution by necrosis or apoptosis to the accumulation of DEK in the supernatants of MDM.

FIG. 4.

Immunosuppressive agents block the secretion of DEK. (A) Synovial fluid samples obtained from a patient with a noninflammatory Baker's cyst (lane 1) or patients with JIA (lanes 2 and 3) were analyzed for the presence of DEK by Western blotting after the fluid was separated from cells by centrifugation as described in Materials and Methods. DEK was detected by Western blotting using the mouse monoclonal anti-DEK antibody. (B) Day 11 serum-activated MDM were incubated in serum-free medium for 12 h alone or in the presence of 0.5 μM or 1 μM dexamethasone (Dex). (C) Day 12 serum-differentiated MDM were incubated in serum-free medium alone for 12 h or in the presence of 10 μM CCCP or 1 μg/ml CsA. In panels B and C, secreted DEK was again detected in the supernatant by Western blotting using DEK-specific monoclonal antibody.

DEK is secreted both in a free form and within exosomes.

Like several other secreted mammalian proteins, including IL-1α and IL-1β, basic fibroblast growth factor, and HMGB-1 (5, 36), DEK lacks an identifiable signal sequence. Furthermore, DEK does not appear to utilize the classical Golgi apparatus pathway for secretion, as we have observed that treatment of MDM with Golgi apparatus-trafficking inhibitors, such as brefeldin A and monensin, does not alter the amount of DEK found in culture supernatants (data not shown). Additionally, although secretion through the Golgi apparatus traditionally requires proteins to be glycosylated and DEK does have several potential glycosylation sites, we found that DEK secretion was not inhibited by treatment of MDM with tunicamycin (which inhibits N-glycosylation) (data not shown). Treatment of DEK protein with N-glycosidase F (which enzymatically removes the N-linked oligosaccharides) or withN-acetylneuraminidase II or O-glycosidase DS [which enzymatically remove all Ser/Thr-linked Gal(β1,3)GalNAc(α1)] also failed to prove actual glycosylation of DEK at potential sites as well (data not shown).

Interestingly, immunofluorescence staining showed well-defined vesicular structures in the cytoplasm of MDM in which costaining of DEK and the known exosomal marker CD81 was seen (Fig. 3A, panel d), suggesting that DEK movement into the extracellular space may occur via secretory vesicles representing exosomal precursors. As a critical component of their biological function, macrophages and other cells of the hematopoietic lineage use secretory lysosomes or prelysosomal multivesicular compartments to deliver proteins to the plasma membrane (41, 52). Internal and membrane proteins are released from multivesicular compartments into the extracellular environment in small membrane vesicles called exosomes, the product of the fusion of multivesicular late endosomes with the plasma membrane (12, 25). No chromosomal material was detected in these structures, differentiating them from the nuclei (Fig. 3A, panel a).

FIG. 3.

DEK is secreted in exosomes and in an exosome-free form from differentiated MDM. (A) Colocalization of DEK and the exosomal marker CD81 by confocal microscopy. Day 13 MDM cultured in glass chamber slides as described in Materials and Methods were incubated with mouse anti-CD81 antibody and stained with an Alexa Fluor-conjugated rabbit anti-mouse antibody to detect CD81 (panel b) or incubated with goat polyclonal DEK antiserum and stained with a FITC-conjugated rabbit anti-goat antibody to detect DEK (panel c). Colocalization of DEK and CD81 (panel d, shown in yellow and indicated by white arrowheads) was detected in vesicular structures throughout the cells. Cell nuclei were stained with 4′,6′-diamidino-2-phenylindole (DAPI) (blue) (panel a). Magnification for all four panels, ×200. (B) Western blot analysis of exosomes in supernatants from day 15 MDM. Human peripheral blood monocytes were isolated and maintained in 40% serum for 15 days. Cells were placed in serum-free medium overnight before supernatant collection. Supernatants were centrifuged at 500 × g, 10,000 × g, and 70,000 × g as described in Materials and Methods. The pellets from specific fractions were solubilized in sample buffer for Western blot analysis. Exosomes are expected to be present in the 70,000 × g fraction (9). (C) Western blot analysis of the exosome-containing and exosome-free fractions from day 15 MDM supernatants. Concentrated supernatant fractions, with exosomes (70,000 × g) and without exosomes, were probed with mouse monoclonal anti-DEK antibody or goat polyclonal anti-CD81 antibody. (D) The 70,000 × g, exosome-containing fraction was used in a proteinase K protection assay. The sample in lane 1 was not treated, that in lane 2 was treated with proteinase K (prot K) alone, and that in lane 3 was treated with proteinase K and Triton X-100. Samples were analyzed for the presence of DEK and CD81 under reducing conditions by Western blot analysis, using specific goat polyclonal anti-DEK antibody or anti-CD81 antiserum. DEK is detected at 50 kDa and 35 kDa, and CD81 is detected at 20 kDa.

To further examine the possibility that DEK is secreted at least in part via exosomes, we isolated the exosomal fraction from the supernatants of day 13 to 15 MDM by serial centrifugation as described previously (9, 42) and looked for the presence of DEK. As shown in Fig. 3B, DEK is found in the pellet of the 70,000 × g centrifugation fraction, known to be highly enriched in exosomes (9, 42). DEK is detected in this fraction in both its 50- and 35-kDa forms. The slight difference between the apparent molecular mass of the intracellular DEK (Fig. 3B, cell extract) and exosomal DEK (Fig. 3B, 70,000 g) is likely due to differences in the phosphorylation or acetylation state of the protein (6, 15, 16, 26, 27, 66), which runs at sizes ranging from 35 to 55 kDa (6, 8, 11, 14-16, 35, 49). When the 70,000 × g fraction was layered onto a linear sucrose density gradient, DEK was found in fractions corresponding to densities of 1.142 g/ml and 1.159 g/ml (data not shown), previously demonstrated to be the exosome-containing fractions (42). In addition, we analyzed the exosome-containing fraction (70,000 × g) and the exosome-free fraction for the presence of DEK and CD81. Interestingly, while CD81 was found only in the exosome fraction (Fig. 3C, bottom panel), DEK is detected in both the exosome and exosome-free fractions (Fig. 3C, top panel). Thus, DEK is secreted both by an exosome-free mechanism and in exosomes. As can be seen in Fig. 3B and C, the proportion of free and exosome-associated DEK secreted by primary MDM shows some donor variability.

We next performed a proteinase K protection assay to determine whether DEK secreted in association with exosomes was located on the surface of the exosome or within it. As shown in Fig. 3D, analysis of the 70,000 × g fraction revealed that DEK was protected from digestion by proteinase K (Fig. 3D, top panel), whereas the surface membrane protein CD81 was completely digested (Fig. 3D, bottom panel). When the exosomes were permeabilized with a detergent, however, DEK was completely enzymatically digested by proteinase K, indicating that DEK is located inside the exosome and is not on the surface of the exosomal membrane. Although the 35-kDa form of DEK appears to be the predominant form in the enriched exosomal fraction (Fig. 3B and D), depending on the exosomal preparation, we have observed different relative levels of the 50- and 35-kDa DEK species (compare Fig. 3B and D to C). Thus, several different lines of evidence demonstrate that DEK can be released from MDM via exosomes as well as in a free form.

DEK is secreted into the joint space of children with active inflammatory arthritis.

The predominance of macrophages and macrophage-like synoviocytes in inflamed joints led us to consider whether DEK secretion might be seen in vivo. Several clinical studies have shown that approximately 40 to 60% of all JIA patients have circulating antibodies to DEK. DEK reactivity is present in a much higher proportion of children with pauciarticular onset JIA and is nearly omnipresent in children with JIA-associated iridocyclitis (38, 49, 54). To determine whether DEK is secreted into the joint space of children with active inflammatory arthritis, we used a monoclonal anti-DEK antibody to probe Western blots of synovial fluid samples from a childhood arthritis tissue repository. The DEK-specific monoclonal antibody identified a strong DEK band in synovial fluid samples (Fig. 4A, lanes 2 and 3), but only low levels of DEK were detected in the thick, clear noninflammatory joint fluid aspirated from a Baker's cyst in a patient whose polyarticular juvenile arthritis was in remission (Fig. 4A, lane 1). When we analyzed a greater number of patients, high levels of DEK were detected in 36 of 45 (80%) juvenile arthritis patients, compared to 3 of 10 (30%) synovial fluid samples from children undergoing arthroscopic surgery for presumed noninflammatory causes of joint effusion (P < 0.003 by Fisher's exact test). Taken together, our findings indicate that DEK is preferentially secreted into inflammatory joint effusions.

Immunomodulation of DEK secretion.

In view of the above observations, we wondered whether modulators of inflammation would play a role in inducing DEK secretion in our MDM model. Indeed, dexamethasone inhibited DEK secretion from MDM in a dose-dependent fashion (Fig. 4B) and caused DEK to accumulate inside the cell (data not shown). Additionally, secretion of DEK by day 12 serum-differentiated MDM could be blocked by CsA, another powerful immunosuppressive agent (Fig. 4C). While the exact mechanism by which dexamethasone and CsA block DEK secretion remains to be defined, it is important to note that clinically employed immunomodulating agents can block the secretion of DEK.

To further investigate the mechanisms that regulate DEK secretion, we hypothesized that proinflammatory cytokines could have a role in stimulating DEK secretion in our MDM culture system and attempted to reproduce the cytokine/chemokine cascade that induces DEK's release from cultured macrophages. In order to assess the prosecretion effect of cytokines, we used day 10 MDM growing in 10% human serum, rather than in 40% human serum, as the latter already maximally secrete DEK. MDM were treated with mediators that are implicated in activation of monocytes, macrophages, and synovial macrophage-like cells, including TNF-α (5 to 100 ng/ml), gamma interferon (IFN-γ) (10 to 100 ng/ml), MCP-1 (monocyte chemoattractant protein 1) (10 to 100 ng/ml), and bioactive molecules such as lipopolysaccharide (500 to 1,000 ng/ml) alone and in combination with IFN-γ (17, 22). None of the inflammatory mediators listed above reproducibly stimulated DEK secretion by MDM (data not shown). In contrast, IL-8 (10 ng/ml) was effective in stimulating vigorous secretion of DEK from day 12 MDM grown in 10% human serum (Fig. 5A). As one of the major chemokines produced by synovial stromal cells, IL-8 plays an important role in attracting neutrophils and peripheral blood monocytes into the inflamed synovium (22, 40, 59, 60), again suggesting a potential link between DEK secretion and inflammatory joint disease.

FIG. 5.

DEK is a chemoattractant for monocyte-depleted peripheral white blood cells. (A) IL-8 induces the secretion of DEK from MDM. Day 12 MDM grown in 10% serum were maintained for 12 h in serum-free conditioned medium alone or with 10 ng/ml of recombinant IL-8. Supernatants were then harvested, and secreted DEK was detected in the supernatant by Western blotting using DEK-specific monoclonal antibody. (B) DEK is a chemoattractant. Monocyte-depleted peripheral white blood cells were prepared as described in Materials and Methods. A total of 1 × 10⁵ cells in 100 μl of RPMI 1640 medium were placed in the upper chambers of a 24-well chemotaxis microchamber plate with a pore size of 5 μm. The lower chambers contained RPMI 1640 medium with various concentrations of recombinant DEK (made in baculovirus), recombinant GLI-2 control protein (also made in baculovirus), or no protein control (C′). After 1 h, the number of fluorescently labeled migrating cells in the lower chamber was determined at 485- or 535-nm wavelengths by a Tecan GENios plate reader (Phenix, Austria). The results are expressed as the average increase ± standard deviation (error bar) of the number of cells migrating toward DEK or GLI-2 compared to control wells containing medium only. Each assay was run in duplicate. Data shown using 34 nM and 68 nM DEK are the averages from three different donors, while data using 136 nM DEK are the averages from six different donors. The increase in the migration of monocyte-depleted peripheral white blood cells toward 136 nM DEK versus GLI-2 or no-protein control is statistically significant (P = 0.006). (C) DEK attracts CD8⁺ T cells and CD56⁺ natural killer cells. Monocyte-depleted peripheral white blood cells were placed in the upper chamber of a 24-well chemotaxis microchamber plate with a pore size of 5 μm. Lower chambers contain either 79 nM DEK in RPMI 1640 medium or medium alone. Cells were collected 1 h after migration and double stained with immunofluorescent antibodies to CD8⁺ (PE) and CD4⁺ (FITC) or CD56 (PE) and CD19 (FITC). Fluorescence profiles were recorded by FACS analysis. The results are expressed as the average increase ± standard deviation (error bar) of the percentage of cells migrating toward DEK versus control wells containing medium only. Migration of CD8⁺ cells and CD56⁺ cells toward DEK is statistically significant (*, P = 0.034; **, P = 0.0028). The results shown represent the averages of three experiments using cells from three different donors.

DEK is a chemotactic factor.

The ability of IL-8 to stimulate DEK secretion prompted us to hypothesize further that secreted DEK may itself act as an immunomodulatory molecule and led us to investigate whether DEK itself may have proinflammatory or chemotactic activity. Accordingly, full-length recombinant DEK protein produced in baculovirus was tested for its ability to attract Ficoll-Hypaque-purified peripheral white blood cells. Indeed, migration of these white blood cells toward DEK was observed to increase in a dose-dependent fashion (Fig. 5B), reaching statistical significance (P < 0.01) when migration toward the highest concentration of DEK (136 nM) is compared with migration toward a control protein (a recombinant form of human GLI-2 (50) or toward medium alone. Similar results were seen with FLAG-tagged DEK protein purified from mammalian cells (data not shown). Flow cytometry analysis identified CD8⁺ T cells and CD56⁺ natural killer cells migrating toward DEK (Fig. 5C).

Since DEK secretion can be stimulated by IL-8, a neutrophil chemoattractant, and because some of the migrating white blood cells also appeared morphologically to be neutrophils, we directly tested whether DEK is also capable of attracting neutrophils. We found that purified fresh human neutrophils specifically migrate toward DEK in a dose-dependent manner, with a bell-shaped dose-response curve similar to that of IL-8 (56) (Fig. 6A). In order to confirm that DEK is a chemoattractant, we used an in vivo murine model. Four groups of 20 4- to 6-week-old male C57BL/6 mice were injected intraperitoneally with 50 μg of DEK, 50 μg of the recombinant control protein GLI-2, also prepared in baculovirus, 50 μg of LPS, or PBS alone. In comparison to GLI-2 and PBS controls, differential white blood cell counts done on peritoneal lavage fluid samples showed that DEK significantly induces neutrophil migration into the peritoneum at 8 h following injection (P = 0.011 in comparison to the GLI-2 control; Fig. 6B). These results again strongly suggest that DEK may act as a proinflammatory chemotactic factor.

FIG. 6.

DEK is a chemoattractant for neutrophils. (A) Neutrophils purified as described in Materials and Methods were placed in the upper chambers of a 24-well chemotaxis microchamber plate with a pore size of 3 μm as described in the legend to Fig. 5. The lower chambers contained RPMI 1640 medium with various concentrations of recombinant DEK, as indicated. The bar labeled “control” contained only medium in the lower chamber. The bar labeled “DEK up” had DEK in the upper chamber and medium in the lower chamber. Data shown represent the averages ± standard deviations (error bars) for two experiments with cells from two different donors. The increase in neutrophil migration toward 79 nM DEK protein versus control is statistically significant (*, P = 0.0067). (B) Groups of mice received intraperitoneal injections of either PBS or 50 μg of either LPS, recombinant DEK, or a recombinant protein control prepared in parallel to DEK in baculovirus (an isoform of human GLI-2). Twenty mice were tested in each of the four groups. Half of the mice in each group were sacrificed after 4 h (white bars), and the other half were sacrificed at 8 h (black bars) after intraperitoneal injection and peritoneal lavage. The bars represent the average numbers ± standard deviations (error bars) of neutrophils migrating into the peritoneum of each group of mice. DEK significantly increased the infiltration of neutrophils compared to the control protein human GLI-2 (P = 0.011). PMNs, polymorphonuclear leukocytes.

DISCUSSION

DEK has previously been described in the literature as a 50- to 35-kDa phosphoprotein that binds DNA, represses transcription, is exclusively localized to the nucleus, and is expressed at high levels in a number of tissues (1, 2, 14-16, 62-65). In addition to DNA binding and transcriptional modulation, DEK has been implicated in two other separate but possibly related functions: (i) alterations in the topology of DNA and chromatin that can affect DNA replication and (ii) RNA processing (2, 26, 27, 33, 35, 66), although the latter is controversial.

We have previously shown that differentiation of the U937 monocytic cell line with phorbol myristate acetate results in the disappearance of DEK from nuclear extracts (11). To further investigate DEK's intracellular location in monocytes, we used a model in which primary monocytes develop into MDM that have a destructive phenotype similar to macrophages exposed to a chronic inflammatory stimulus (41, 43). While, as expected, immunofluorescence and confocal microscopy localized DEK to the nucleus in monocytes, we found that differentiation of monocytes into MDM was accompanied by movement of DEK from the nucleus to the cytoplasm as a function of time. This observation suggests that DEK could participate in cytoplasmic functions, including RNA transport and processing, under certain conditions.

Surprisingly, we further found that upon differentiation of monocytes into MDM, a significant amount of DEK not only leaves the nucleus but is actually secreted. To our knowledge, DEK is one of the few human nuclear DNA-binding proteins known to be actively secreted (see discussion of HMGB-1 below) and the first transcription-modulating factor to be secreted within exosomes. Multiple studies confirmed that DEK was not simply being leaked by apoptotic or necrotic cells but instead was being actively secreted. This conclusion is based on the MTT assay results, visual inspection, lack of histone protein in the supernatant, lactate dehydrogenase measurements in previous studies using this model (41, 43), the absence of HMGB-1 (which is released following necrosis of monocytes) in the supernatants, the rapid accumulation of DEK in the supernatant (as early as 3 h), and the observation that both TBB and CCCP, which are actually proapoptotic agents, specifically block DEK secretion. Interestingly, DEK does not have a signal sequence, and indeed, it does not appear to be secreted through the Golgi apparatus, as multiple attempts to block the secretion of DEK using Golgi complex blockers were unsuccessful. This is in marked contrast to our recent findings demonstrating, in the same macrophage model, that vimentin is secreted through the Golgi apparatus (37). Further studies indicated that DEK secretion is seen not only in our model of differentiated MDM but is also seen in the synovial fluid samples from patients with active juvenile arthritis. Importantly, in the latter samples, Western blot analysis did not reveal the presence of the cytoskeletal protein actin, largely excluding the contribution of cell death to DEK's secretion (data not shown). Thus, it appears that secretion of DEK is detected not only in a tissue culture model of primary macrophages but also in patients with inflammatory disease. This observation is of considerable interest in view of the strong association between antibodies to DEK and autoimmune diseases, especially JIA and JIA-associated uveitis (8, 21, 38, 54, 55).

Specific pathways were found to regulate the secretion of DEK from MDM. First, we found that CK2 activity is necessary for the secretion of DEK. It remains to be clarified whether the specific CK2 blocker TBB inhibits secretion by blocking CK2-mediated phosphorylation of DEK itself, as expected (26), or through its effect on an associated protein(s). We have also shown that IL-8 stimulates DEK secretion. IL-8 is a major proinflammatory cytokine that has been implicated in the recruitment of macrophages, neutrophils, and macrophage-like synoviocytes to sustain an apparently autonomous cytokine network in the inflamed synovium and joint space of patients with inflammatory arthritis (22, 59). Thus, the induction of DEK secretion by IL-8 is consistent with its chemoattractant function. Surprisingly, despite its history of being associated with nuclear functions only, DEK can attract inflammatory cells. DEK attracts cells known to express CXCR1 and CXCR2, including CD8⁺ T cells and natural killer cells, as well as neutrophils. Interestingly, when we analyzed the sequence of DEK, we discovered that it has an ELR (Glu-Leu-Arg) motif. The ELR motif is typically found on chemokines, such as GRO-α and IL-8, and mediates chemotaxis through the interaction of these chemokines with the CXCR2 and CXCR1 receptors (18, 53, 68). Further studies will assess whether DEK's chemoattractant function is mediated through specific interaction with CXCR2 and/or CXCR1 receptors. While the secreted nuclear protein HMGB1 is a chemoattractant for smooth muscle cells (28), our findings with DEK are the first demonstration that a nuclear DNA-binding protein can be secreted in response to inflammatory signals and then might perpetuate the inflammatory response by recruiting proinflammatory cells.

Macrophages use secretion as a critical component of their biological function, and like other cells of hematopoietic lineage, they use secretory lysosomes or prelysosomal multivesicular compartments to deliver proteins to the plasma membrane (41, 52). The multivesicular compartments also release small membrane vesicles called exosomes that result from fusion of multivesicular late endosomes with the plasma membrane and exocytose both internal and membrane proteins into the extracellular environment (12, 25, 57, 58). In this paper, we present data showing that the exosome-rich fraction of supernatants from activated MDM contains DEK and that DEK is found within the lumen of the exosomes. We were not able to detect HMGB-1 in these fractions (see discussion below). Although most of the known exosome-associated proteins are membrane proteins, several cytosolic proteins are associated with exosomes, including some proteins that demonstrate strong immunostimulatory properties (e.g., galectin-3 and heat shock protein hcs73) (57, 58). To our knowledge, no transcription factor other than DEK has been shown to be secreted within exosomes. While the full biological significance of their function is not yet known, exosomes appear to contribute both to antigen-dependent and -independent modulation of the immune response by effects that are at least partly mediated by antigen-presenting cells. Incubation of exosomes with dendritic cells has been reported to stimulate specific T cells with great efficiency (61, 69), suggesting that DEK's physical association with exosomes may facilitate its presentation to dendritic cells, thus providing a crucial link between this nuclear protein and its ability to elicit a potent autoantibody response. This hypothesis, while plausible, remains to be tested experimentally.

The similarity between the biology of DEK and HMGB proteins, irrespective of secretion, has been noted previously (66). Our discovery that DEK is a nuclear protein that translocates to the cytoplasm and is secreted has parallels to a body of recent literature relating to HMGB-1, a highly expressed, ubiquitous DNA-binding nuclear protein that is secreted by macrophages as a cytokine and late mediator of LPS lethality (4, 28, 67, 71). HMGB-1 secretion is induced by LPS or IFN-γ and can be partially blocked by inhibiting TNF-α (44), whereas DEK secretion is stimulated by IL-8 and does not appear to be affected by these other inflammatory mediators. Interestingly, it has recently been shown that intracellular DEK is a negative regulator of NF-κB that can repress transcription driven by the IL-8 promoter (45). In view of the findings presented here and our recent preliminary observation that DEK can be taken up by cells, it seems plausible that DEK and IL-8 are involved in a feedback loop.

Neither HMGB-1 nor DEK contains a signal sequence, and secretion occurs via nonclassical pathways. Whereas HMGB-1 is secreted via secretory lysosomes or released by necrotic macrophages (5), DEK secretion does not appear to involve the lysosomal compartment, since it is impervious to treatment with ammonium chloride (up to 50 mM), which alkalinizes intralysosomal pH (data not shown). Additionally, we have shown in this report that DEK is secreted by healthy, not necrotic, macrophages via exosomes and another yet-to-be-defined Golgi apparatus-independent secretory pathway. Thus, both substantial differences and similarities are seen between DEK and HMGB-1, which to our knowledge are the only nuclear DNA-binding proteins thus far reported to be secreted from cells.

Our studies imply that DEK is secreted by two or more Golgi apparatus-independent pathways in MDM and suggest that secreted DEK could contribute to immunity or autoimmunity in at least two distinct manners. First, DEK secreted via exosomes could be taken up by dendritic cells, processed, and then presented as a foreign antigen to B cells, leading to the production of autoantibodies (Fig. 7). Concomitantly, DEK can be secreted by another, yet unidentified, nonclassical pathway, and then act as a direct chemoattractant for inflammatory cells (Fig. 7).

FIG. 7.

Schematic model of DEK secretion by MDM. DEK is mainly a DNA-binding nuclear protein in undifferentiated monocytes/macrophages (step 1). With the acquisition of a more differentiated phenotype by MDM, DEK is incorporated in secretory vesicles in the multivesicular compartment and secreted via exosomes (step 2). Possibly through an intermediate step involving exosomal pickup by antigen-presenting cells (APC) (step 3), B cells produce autoantibodies against DEK (step 4). Alternatively, via an incompletely defined, Golgi apparatus-independent pathway (step 5), DEK is directly secreted into the extracellular milieu, where it acts as a chemoattractant for leukocytes (step 6). PMNs, polymorphonuclear leukocytes.

In summary, the DEK protein, previously described as a strictly nuclear protein but a known autoantigen in JIA patients, can be secreted by monocytic cells and is found in the joints of patients with autoimmune disease. Surprisingly, DEK is also a chemotactic factor that attracts neutrophils, CD8⁺ T cells, and natural killer cells. Studies presented here thus support the idea that, in addition to its role(s) in regulating gene expression, DEK can function as an extracellular proinflammatory factor.