Reverse signaling through membrane-bound interleukin-15.

The results from this study implicate membrane-anchored interleukin (IL)-15 constitutively expressed on the cell surface of PC-3 human prostate carcinoma cells and interferon-gamma-activated human monocytes in reverse signaling upon stimulation with soluble IL-15 receptor-alpha or anti-IL-15 antibodies, mediating the outside-to-inside signal transduction that involves the activation of members of the MAPK family (ERK and p38) and focal adhesion kinase. The presence of membrane-bound IL-15 was not dependent on the expression of the trimeric IL-15 receptor complex by these cells and resisted treatment with acidic buffer or trypsin. Reverse signaling through membrane-bound IL-15 considerably increased the production of several pro-inflammatory cytokines by monocytes, such as IL-6, IL-8, and tumor necrosis factor-alpha, thereby indicating the relevance of this process to the complex immunomodulatory function of these cells. Furthermore, stimulation of transmembrane IL-15 also enhanced the transcription of IL-6 and IL-8 in the PC-3 cell line and promoted migration of PC-3 cells as well as LNCaP human prostate carcinoma cells stably expressing IL-15 on the cell surface. Thus, IL-15 can exist as a biologically active transmembrane molecule that possesses dual ligand-receptor qualities with a potential to induce bidirectional signaling. This fact highlights a new level of complexity in the biology of IL-15 and offers novel important insights into our understanding of the cellular responses modulated by this pleiotropic cytokine.

Interleukin (IL) 1 -15 is a pleiotropic cytokine that belongs to the four-helix bundle cytokine family and was first identified because of its ability to substitute IL-2 in supporting the growth of the murine IL-2-dependent CTLL cell line (1). IL-15 shares with IL-2 the IL-2 receptor (IL-2R)-␤ and IL-2R␥/␥c chains (2), but has a unique high affinity ␣ chain (IL-15R␣) responsible for the differential functional effects of IL-15 versus IL-2 on cells of the same type (3). IL-15 can replace IL-2 in most of its activities in the immune system, including induction of T cell proliferation and chemotaxis, stimulation of natural killer cell growth and interferon-␥ (IFN-␥) production, generation of cytotoxic effector cells, and co-stimulation of B cell growth and immunoglobulin synthesis (4 -6). IL-15 and IL-15R␣ knockout mice display a marked reduction in numbers of natural killer cells, memory phenotype CD8 ϩ T lymphocytes, and a distinct population of intestinal intraepithelial lymphocytes, suggesting an important role for IL-15 in the development and/or survival of these cells (7,8). In addition, IL-15 has a potent anti-apoptotic function, inhibiting apoptosis of activated T and B cells, keratinocytes, and melanoma cells in vitro and protecting mice from Fas-induced hepatic failure and multisystem apoptosis in vivo (9 -11).
IL-15 mRNA is constitutively expressed by a large variety of cell types and tissues, including monocytes/macrophages, fibroblasts, keratinocytes, kidney epithelial cells, nerve cells, placenta, heart, and skeletal muscle (4,7,(12)(13)(14)(15). Interestingly, most primary cells and cell lines that express IL-15 mRNA do not release detectable amounts of this cytokine into the culture medium. This discrepancy is explained by the fact that IL-15 has a complex, multifaceted control of expression with regulation at the levels of transcription, translation, and intracellular trafficking. The existence of two IL-15 isoforms that differ in the length of the signal peptide has been reported (16). These isoforms exhibit a differential intracellular trafficking, secretion, and endosomal localization, indicating an important role for the signal peptide in multiple mechanisms controlling IL-15 production (17,18). IL-15 associated with the short signal peptide (IL-15SSP) is not secreted but rather stored intracellularly in the nucleus and cytoplasm, whereas the alternative isoform characterized by the longer signal peptide (IL-15LSP) is located in the Golgi, early endosomes, and the endoplasmic reticulum and has been suggested to follow a pathway that may result in cytokine secretion (17,18).
In monocytes/macrophages, IL-15 is expressed in a biologically active membrane-bound form, and its mRNA expression can be up-regulated by exogenous stimuli such as IFN-␥ and lipopolysaccharide (12). Furthermore, tumor necrosis factor-␣ (TNF-␣)-stimulated dermal fibroblasts are able to sustain proliferation of activated T cells through expression of membranebound IL-15 (13). The presence of biologically active membrane-anchored IL-15 on the cell surface of normal human monocytes, several monocytic cell lines, and TNF-␣-stimulated dermal fibroblasts suggests that, under physiological conditions, IL-15 may be present mainly in a membrane-bound rather than secreted form. This study highlights the ability of membrane-anchored IL-15 to mediate outside-to-inside (reverse) signaling that activates focal adhesion kinase (FAK) and MAPKs in the PC-3 human prostate carcinoma cell line and IFN-␥-activated human monocytes. Furthermore, reverse signaling through membrane IL-15 promoted migration of prostate cancer cells and induced production of pro-inflammatory cytokines in both PC-3 cells and monocytes. Thus, our findings attribute yet another function to this pluripotent cytokine by demonstrating the capacity of IL-15 to function both as a ligand and a receptor molecule.
Cell Culture, Stimulation, Acidic Treatment, and Transfection Conditions-PC-3 (American Type Culture Collection CRL-1435) and LN-CaP (American Type Culture Collection CRL-1740) human prostate carcinoma cell lines were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. Human monocytes were obtained from healthy donors by an elutriation procedure. Monocytes were incubated with 10 ng/ml IFN-␥ for 24 h prior to stimulation to increase the membrane IL-15 expression. Before stimulation, cells were washed twice with Dulbecco's phosphate-buffered saline (PBS) and serum-starved for 4 h. For each assay, 5 ϫ 10 6 cells/ml were stimulated with sIL-15R␣ (1 ng/ml) or anti-IL-15 Abs (100 ng/ml) for 5 or 15 min at 37°C. Activation was interrupted by adding 8 -10 volumes of ice-cold PBS with 10 mM EDTA and 100 mM sodium vanadate. Cells were pelleted and frozen at Ϫ80°C before electrophoresis. Acidic treatment was performed as described (20). Cells were incubated in ice-cold 25 mM glycine and 150 mM NaCl (pH 3.0) for 10 min.
For immunoprecipitation studies, lysates containing 500 g of proteins were immunoprecipitated overnight at 4°C by incubation in 0.5% Nonidet P-40 buffer with 2 g/ml Abs. Immunocomplexes were captured on protein A/G-agarose. After washing, pellets were resuspended in SDS-PAGE sample buffer, boiled for 5 min, and analyzed by 10% SDS-PAGE. The resolved proteins were transferred onto nitrocellulose (Bio-Rad). Blots were blocked for 1 h in PBS with 0.05% Tween 20 and 3% bovine serum albumin (Sigma). After incubations with primary and secondary Abs and washing with PBS/Tween, visualization of specific proteins was carried out by an enhanced chemiluminescence method using ECL Western blotting detection reagents (Amersham Biosciences) according to the manufacturer's instructions.
Samples were amplified in a DNA Thermocycler (Eppendorf, Hamburg) for 30 cycles. Each cycle consisted of denaturation at 94°C for 15 s, annealing at 60°C for 30 s, and elongation at 72°C for 30 s, preceded by initial denaturation at 94°C for 5 min and followed by a final extension step at 72°C for 5 min. To evaluate mRNA expression semiquantitatively, in addition to the PCR product from 30 cycles, 15 l of the PCR products from the 26, 28, and 32 cycles were run simultaneously. Aliquots of PCR products were electrophoresed on 1.5% agarose gel and visualized by ethidium bromide staining. ␤-Actin message was used to normalize the cDNA amount to be used. A mock PCR (without cDNA) was included to exclude contamination in all experiments.
Wound Healing Assay-Exponentially growing cells (2 ϫ 10 6 ) were plated onto cell culture plates coated with rat tail collagen (10 g/ml; Roche Applied Science, Heidelberg) in complete growth medium. After 8 h, the monolayers of cells were wounded by manual scratching with a pipette tip, washed with PBS, placed into complete growth medium, and photographed with a Nikon Diaphot 300 phase-contrast microscope. Matched pair-marked wound regions were photographed again after 18 (PC-3 cells) or 8 (LNCaP cells) h.
Flow Cytometric Analysis-Cells were stained with monoclonal Abs or IL-15-IgG 2b fusion protein as described previously (21) and analyzed on a FACSCalibur (BD Biosciences) using CELL Quest software. Negative controls consisted of isotype-matched, directly conjugated, nonspecific Abs (Pharmingen).
Data Analysis-All experiments were performed in at least three independent assays, which yielded highly comparable results. Protein sequences were analyzed using ProteinPredict software for sequence analysis and prediction of protein function and structure (www.emblheidelberg.de/predictprotein/predictprotein.html). Blots were quantitated using ImageQuant TL software (Amersham Biosciences). Data are summarized as means Ϯ S.D. Statistical analysis of the results was performed by Student's t test for unpaired samples. p Ͻ 0.05 was considered statistically significant.

PC-3 Cells and IFN-␥-activated Monocytes Express Membrane-bound IL-15-
The biologically active form of membranebound IL-15 is constitutively expressed on the cell surface of human monocytes and several monocytic cell lines (12,22), whereas IFN-␥ stimulation further up-regulates its surface expression (12). Therefore, monocytes were stimulated with IFN-␥ for 24 h in all experiments described hereafter. Furthermore, we found that the PC-3 prostate carcinoma cell line expresses only the membrane-bound form of IL-15, whereas the IL-15R␣, IL-2R␤, and IL-2R␥/␥c chains are absent in these cells. To assess the expression of IL-15 in PC-3 cells and monocytes in more detail, we first analyzed its mRNA content by RT-PCR using a pair of primers that recognize two distinct IL-15 isoforms corresponding to the cytokine with the short (21 amino acids) and long (48 amino acids) signal peptides, respec-tively (16). As shown at Fig. 1A, PC-3 cells express only the IL-15LSP isoform, which is 513 bp in length, whereas monocytes express both the IL-15LSP and IL-15SSP (494 bp plus additional 119 bp) isoforms. The LNCaP prostate carcinoma cell line, which does not express IL-15, was used as a control. Next, we analyzed the presence of IL-15 on the cell surface of PC-3 cells and IFN-␥-activated monocytes by flow cytometry using anti-IL-15 Abs. These experiments confirmed that both PC-3 cells and monocytes express IL-15 on the cell membrane (Fig. 1B). However, the amount of membrane-bound IL-15 on monocytes was higher than in PC-3 cells. The expression of IL-15 in PC-3 cells was further corroborated by Western blot experiments using immunoprecipitation with anti-IL-15 Abs (Fig. 1C). Isotype-matched Abs did not precipitate IL-15 and served as a control. The presence of IL-15 on the cell surface of PC-3 cells was also confirmed by confocal microscopy (Fig. 1D). IL-15LSP associated with the cell membrane and was also found in the nuclei, which is in agreement with recent findings (11). Thus, only the IL-15LSP isoform is present in PC-3 cells, whereas both IL-15LSP and IL-15SSP are found in IFN-␥activated monocytes.

PC-3 Cells Do Not Express Components of the IL-15R
Complex-To exclude the presence of the IL-15R␣, IL-2R␤, and IL-2R␥/␥c chains in PC-3 cells, we assessed their expression by RT-PCR and FACS analysis. The expression pattern of these receptor subunits in monocytes was evaluated in parallel. Indeed, IL-15R␣, IL-2R␤, and IL-2R␥/␥c mRNAs were absent in PC-3 cells ( Fig. 2A), and staining with specific Abs did not reveal the presence of these receptor chains on the cell membrane (Fig. 2B). On the other hand, activated monocytes expressed all subunits of the IL-15R complex. The presence of IL-15-binding sites on the cell membrane was convincingly demonstrated by the ability of monocytes to bind IL-15-IgG 2b fusion protein, whereas no IL-15-IgG 2b binding activity was detected in PC-3 cells (Fig. 2B).
A recent report by Musso et al. (12) showed the presence of membrane-bound IL-15 on monocytes, which did not elute with acetate buffer or trypsin treatment, suggesting that it is an integral membrane protein not associated with the IL-15R complex. Additionally, another study demonstrated the expression of biologically active IL-15 on the cell surface of TNF-␣stimulated dermal fibroblasts (13). Conversely, Dubois et al. (20) reported that IL-15 is noncovalently attached to the cell surface via its interaction with IL-15R␣ and is released after acidic treatment (pH 3.0) of monocytes. Thus, we tested whether IL-15 on PC-3 cells and monocytes could be removed from the cell surface by such acidic treatment. Fig. 2C shows that no changes in the median fluorescence intensity were detected in PC-3 cells after treatment with acidic buffer and staining with anti-IL-15 Abs compared with control cells incubated in PBS. Notwithstanding, we observed a decrease in the amount of cell surface-bound IL-15 on monocytes treated in a similar way. However, activated monocytes clearly retained membrane-bound IL-15 after acidic treatment as confirmed by staining with specific Abs, whereas no bound IL-15-IgG 2b fusion protein was detected. These results strongly support the idea that the presence of membrane-bound IL-15 is not dependent on the IL-15R complex. The experiments using trypsin further corroborated this concept. Trypsin treatment caused the IL-15R complex to lose its ability to bind IL-15-IgG 2b fusion protein, which was used as a receptor detector (12). Thus, it was reasonable to expect that if IL-15 detected on the cell membrane was indeed bound to its receptor, the whole ligand⅐receptor complex should be disrupted by trypsin, and either IL-15-IgG 2b fusion protein or anti-IL-15 antibody should not bind to trypsinized cells. Although anti-IL-15 Abs still detected IL-15 on the cell membrane, IL-15-IgG 2b did not bind any longer to the cell surface (data not shown). Thus, these findings suggest that IL-15 likely exists on the cell surface of monocytes in both the membrane-anchored and IL-15R␣-bound forms, whereas PC-3 cells express only membrane-bound IL-15. Although treatment with acidic buffer or trypsin did not affect the expression of transmembrane IL-15, it eluted noncovalently bound IL-15 from IL-15R␣, resulting in the decrease in median fluorescence intensity.

Membrane-anchored IL-15 Mediates Reverse Signaling That Involves Protein Phosphorylation and Activation of MAPKs and FAK in PC-3 Cells and Monocytes-
The phenomenon of reverse signaling or the ability of a membrane-bound ligand to induce the activation of intracellular mediators has surfaced recently as an important mechanism to regulate qualitatively distinct cellular responses to specific stimuli (23). Many ligand-receptor pairs have been shown to be capable of bidirectional signal transduction (24 -29). The presence of membrane-bound IL-15 on the cell surface of PC-3 cells and monocytes strongly invited us to investigate its biological relevance for host cells. To test whether membrane IL-15 may mediate reverse signaling events, PC-3 and monocytes were treated with recombinant sIL-15R␣ for different time intervals, and the pattern of phos-phorylation of cellular proteins was assessed. Notably, the concentration of endotoxin in all sIL-15R␣ preparations was extremely low, and lipopolysaccharide in such concentration was not able to induce any signaling in PC-3 cells and monocytes (data not shown). Nevertheless, to verify that the observed effects of membrane IL-15 stimulation were genuine and not due to nonspecific activation through contamination with endotoxin associated with sIL-15R␣ preparations, we also stimulated cells with anti-IL-15 Abs, whereas treatment with isotype-matched Abs was used as a control. Fig. 3A illustrates that stimulation with sIL-15R␣ or anti-IL-15 Abs clearly induced tyrosine phosphorylation of several proteins with molecular masses ranging from 30 to 120 kDa in PC-3 cells and monocytes within the first minutes of action. Notably, the phosphorylation pattern in the cells stimulated with sIL-15R␣ was slightly different from that in the cells treated with anti-IL-15 Abs. Preincubation of sIL-15R␣ with a 200-fold excess of IL-15 to saturate IL-15-binding sites or anti-IL-15R␣ Abs to neutralize sIL-15R␣ prior to stimulation of PC-3 cells or monocytes with sIL-15R␣, respectively, abolished the ability of sIL-15R␣ to induce protein phosphorylation in both cell types (Fig.  3A). Furthermore, the phosphorylation of these cellular proteins was still observed in the cells treated with acidic buffer Next, we sought to establish the identity of the phosphorylated molecules. To this end, the membranes were reprobed with Abs directed against several cellular proteins that reportedly play important roles in mediating downstream signaling from a wide range of membrane-expressed receptor molecules. These experiments revealed that stimulation with sIL-15R␣ induced the phosphorylation of ERK and p38 (Fig. 3C). Activation of ERK was evaluated by employment of phospho-specific Abs directed against the C termini of ERK1 (p44) and ERK2 (p42) (both phosphorylated at Tyr 204 ) (30). Interestingly, the predominant tyrosine phosphorylation of ERK2 (p42) was observed in human monocytes and, to a lesser extent, in PC-3 cells (Fig. 3C). Moreover, FAK phosphorylation was detected. However, the phosphorylation status of JNK (Fig. 3C) and the FAK-related kinase Pyk2 (data not shown) was not altered. Similar results were obtained after stimulation with anti-IL-15 Abs, whereas treatment with acidic buffer or trypsin did not abolish these effects (data not shown). Noteworthy, stimulation with a lower concentration of sIL-15R␣ induced a stronger change in the phosphorylation of MAPKs and FAK compared with anti-IL-15 Abs (1 ng/ml sIL-15R␣ versus 100 ng/ml anti-IL-15 Abs) (data not shown).
Given the ability of membrane-bound IL-15 to mediate the phosphorylation of these intracellular signaling molecules upon stimulation with sIL-15R␣ or anti-IL-15 Abs, a series of immunoprecipitation experiments were performed to test whether membrane IL-15 may physically associate with FAK, ERK, and p38 to induce their activation. However, no direct interactions between IL-15 and FAK or MAPKs were detected (data not shown). Notwithstanding, we found that stimulation with sIL-15R␣ or anti-IL-15 Abs resulted in the phosphorylation of IL-15 at serine (but not tyrosine) residues ( Fig. 3C and data not shown). Taken together, these experiments demonstrate that membrane IL-15 is phosphorylated upon stimulation with sIL-15R␣ or anti-IL-15 Abs and mediates tyrosine phosphorylation of MAPKs and FAK in both PC-3 cells and activated monocytes.
Membrane IL-15 Mediates Signaling in the LNCaP Prostate Cancer Cell Line-To validate the ability of membrane-anchored IL-15 to mediate the activation of signaling molecules, the LNCaP cell line, which does not express endogenous IL-15 as well as components of the IL-15R complex (Figs. 1 and 2), was transiently or stably transfected with a vector coding for IL-15LSP or IL-15SSP. Fig. 4A shows that transfected cells expressed mRNA for the respective IL-15 construct. Interestingly, transient transfection with IL-15LSP clearly rendered the cells responsive to sIL-15R␣ treatment, resulting in the characteristic pattern of phosphorylation, whereas the cells transfected with IL-15SSP were unresponsive (Fig. 4B). However, when cells were stably transfected with IL-15SSP or IL-15LSP, tyrosine-phosphorylated proteins were observed in cells expressing both constructs, although the phosphorylation

FIG. 3. Membrane-bound IL-15 mediates reverse signaling in PC-3 cells and monocytes. A, PC-3 cells and human monocytes (hMo)
were activated with sIL-15R␣ (1 ng/ml) or anti-IL-15 Abs (100 ng/ml) for 5 and 15 min. Then, the cell lysates were subjected to Western blotting, and tyrosine-phosphorylated proteins were detected on the membrane using anti-phosphotyrosine Abs. Incubation of sIL-15R␣ with a 200-fold excess of IL-15 or anti-IL-15R␣ Abs prior to stimulation of PC-3 cells or monocytes with sIL-15R␣, respectively, was used to saturate IL-15-binding sites and to block sIL-15R␣-induced signaling. B, acidic treatment does not affect protein phosphorylation mediated by membrane-bound IL-15. Cells were treated with acidic buffer or PBS (as a control) for 10 min prior to treatment with sIL-15R␣. Tyrosine-phosphorylated proteins were subjected to 10% SDS-PAGE, transferred onto nitrocellulose membrane, and detected using anti-phosphotyrosine Abs. C, cells were stimulated with sIL-15R␣, and the phosphorylation of ERK, JNK, and p38 kinases was detected using phospho-specific Abs. The positions of phosphorylated kinases are indicated on the right. The results are expressed as means Ϯ S.D. *, p Ͻ 0.05 compared with the control. The equal amounts of ERK and p38 are shown as controls for loading. FAK or IL-15 was immunoprecipitated (IP) from cell lysates using specific Abs, and phosphorylation of FAK or IL-15 was detected by probing the membrane with anti-phosphotyrosine or anti-phosphoserine Abs, respectively. Western blots (WB) were reprobed with anti-FAK Abs after stripping to prove equal loading of precipitated FAK protein.
was significantly stronger in the IL-15LSP-transfected cells (Fig. 4C). In accordance with these findings, reprobing the membrane with specific Abs allowed us to identify ERK as one of the phosphorylated substrates (Fig. 4, B and C). We did not observe JNK or p38 phosphorylation in these cells upon sIL-15R␣ stimulation (data not shown). Stimulation with anti-IL-15 Abs mimicked the action of sIL-15R␣, whereas isotypematched Abs were ineffective (data not shown). Notably, the protein tyrosine phosphorylation was most prominent within the first 5-15 min of sIL-15R␣ action, although the phosphorylation of ERK1/2 was sustained in both stably transfected cell lines for 30 min. Treatment with acidic buffer or trypsin did not abrogate these effects (data not shown). Confocal analysis demonstrated that, upon stable overexpression in LNCaP cells, both IL-15LSP and IL-15SSP were detected on the cell membrane and in the cytoplasm and nuclei (Fig. 4D), although the amount of IL-15LSP protein associated with the cell membrane was higher. The presence of IL-15SSP on the cell membrane provides a plausible explanation for the ability of this isoform to mediate signaling in LNCaP cells stably overexpressing IL-15SSP protein.
Reverse Signaling through Membrane-anchored IL-15 Triggers Cytokine Production-The MAPK cascade represents a key signaling pathway critical for linking membrane receptors to cytoplasmic and nuclear effectors and regulates a wide range of cellular functions. Given that treatment with sIL-15R␣ clearly induced dramatic changes in the level of phosphorylated proteins in PC-3 cells and monocytes, including ERK and p38, we next tested the production of pro-inflammatory cytokines by these cells. Stimulation with sIL-15R␣ or anti-IL-15 Abs for 24 h considerably enhanced the expression of TNF-␣, IL-6, and IL-8 by monocytes, as detected by RT-PCR (Fig. 5A) and ELISA (Fig. 5B), whereas treatment with acidic buffer did not abolish these responses (Fig. 5B). Isotype-matched control Abs were without effect.
In contrast to monocytes, PC-3 cells abundantly release a number of cytokines into the culture medium in the absence of stimulation, and we were not able to detect changes in the cytokine secretion pattern of these cells by ELISA in response to stimulation of membrane IL-15 (data not shown). However, a semiquantitative RT-PCR analysis of IL-6, IL-8, and TNF-␣ mRNA expression showed that stimulation of PC-3 cells for 8 h with sIL-15R␣ or anti-IL-15 Abs enhanced the transcription level of IL-6 and IL-8 (Fig. 5C) while not affecting TNF-␣ expression (data not shown). Thus, reverse signaling through membrane-bound IL-15 leads to signal transduction and gene expression, significantly up-regulating the production of IL-6, IL-8, and TNF-␣ by human monocytes and increasing the transcription level of IL-6 and IL-8 in PC-3 cells.
Reverse Signaling through Membrane-bound IL-15 Promotes Migration of Prostate Cancer Cells-A number of studies have demonstrated strong correlations between elevated FAK expression and the increased invasive potential of human tumors by providing support for a role of this kinase in cell migration (31,32). It has been shown that FAK is required for both integrin-and growth factor-stimulated cell motility (32,33). FAK is highly tyrosine-phosphorylated at a number of different sites either in growing integrin-stimulated cells or in migrating cells (31). Given the observed phosphorylation of FAK upon stimulation of membrane-bound IL-15, we performed a wound healing scratch assay to determine whether reverse signaling through membrane IL-15 would affect cell migration. To this end, equal numbers of growing prostate cancer cells were placed into 6-well plates. After 18 h of culture, cells were cleared within a defined area by scratching with a pipette tip, washed with PBS, placed into complete growth medium, photographed in phase-contrast (Fig. 6, 0 h), and allowed to migrate into the cleared area in the presence of sIL-15R␣ or anti-IL-15 Abs. Cells cultured in the medium or treated with isotype-matched Abs were used as controls. In fact, PC-3 cells treated with sIL-15R␣ or anti-IL-15 Abs had separated from the monolayer at the wound edges, and significant numbers of these cells started to migrate into the cleared area after 18 h compared with untreated cells or cells treated with isotype-matched Abs (Fig. 6, left panels). These results were further confirmed by experiments using LNCaP cells transfected with IL-15LSP, whereas parental LNCaP cells showed reduced mi-

FIG. 5. Reverse signaling through membrane IL-15 induces production of cytokines in human monocytes and PC-3 cells.
A, monocytes were stimulated with sIL-15R␣ (1 ng/ml) or anti-IL-15 Abs (100 ng/ml) for 24 h. Unstimulated cells or cells stimulated with isotype-matched Abs were used as controls (cont). Total RNA was extracted from cells and reverse-transcribed, and TNF-␣, IL-6, and IL-8 expression was detected by PCR using specific primers. The amount of cDNA analyzed was similar in different samples, as shown by PCR amplification of ␤-actin. B, untreated or acidic buffer-treated monocytes were stimulated with sIL-15R␣ (1 ng/ml) or anti-IL-15 Abs (100 ng/ml) for 24 h. Supernatants from 24-h cell cultures were analyzed for TNF-␣, IL-6, and IL-8 release by ELISA. The results represent means Ϯ S.D. of three independent experiments. *, p Ͻ 0.05. C, PC-3 cells were cultured for 8 h in medium (control) or in the presence of sIL-15R␣, anti-IL-15 Abs, or isotypematched Abs. Then, total RNA was extracted and reverse-transcribed to cDNA. For semiquantitative analysis, in addition to 30 cycles, 15-ml aliquots of the PCR products from 26, 28, and 32 cycles were also evaluated. The image shows the amplified bands after 28 cycles. ␤-Actin message was used to equalize the amount of cDNA used (lower panel). A mock PCR (no cDNA (ϪDNA)) was included as a negative control. The data represent three separate experiments with comparable results. gration properties (Fig. 6, middle and right panels). Notably, the migration of LNCaP cells expressing IL-15LSP into the scratched area was already detectable after 8 h. The time course of total wound closure was significantly shorter for PC-3 and IL-15LSP-transfected LNCaP cells after stimulation with sIL-15R␣ or anti-IL-15 Abs (ϳ26 or 16 h, respectively) compared with control cells, in which wound closure was observed considerably later presumably due to continued cell proliferation. These results show that reverse signaling through membrane-bound IL-15 promotes cell migration and may contribute to the enhanced invasive properties of prostate cancer cells. DISCUSSION In this study, we have provided experimental evidence that membrane-bound IL-15 is constitutively expressed on the cell surface of the PC-3 human prostate carcinoma cell line and IFN-␥-activated human monocytes and mediates reverse signaling events in both cell types upon stimulation with recombinant sIL-15R␣ or anti-IL-15 Abs. This signaling scenario involves the phosphorylation of members of the MAPK family (ERK and p38) and FAK. Furthermore, reverse signaling through transmembrane IL-15 up-regulated the transcription level of IL-6 and IL-8 and dramatically enhanced the migratory properties of prostate cancer cells. Moreover, stimulation of membrane-bound IL-15 resulted in positive feedback also in monocytes, leading to a significant increase in the production of several pro-inflammatory cytokines such as IL-6, IL-8, and TNF-␣. Given that mediators released by these innate immune response cells can profoundly influence adaptive immunity, and the ability of transmembrane IL-15 to activate neighboring cells, bidirectional signaling through membrane-bound IL-15 provides another striking example of a unique role of this pleiotropic cytokine in immune homeostasis.
Our results suggest that IL-15 on activated monocytes is directly anchored to the cell membrane rather than bound to the IL-15R␣ chain (12,20). This suggestion is in agreement with two recent studies that also reported the presence of biologically active membrane-bound IL-15 on monocytes (12,22). The bioactivity of IL-15 was convincingly demonstrated by the ability of mitomycin-treated or fixed monocytes to support proliferation of concanavalin A-stimulated human T cells or an IL-15-dependent CTLL cell line, respectively (12,22). The presence of biologically active membrane IL-15 was also shown upon the cell surface of TNF-␣-stimulated dermal fibroblasts (13). Moreover, stimulation with sIL-15R␣ or specific Abs did not induce any detectable changes in the phosphorylation pattern of intracellular molecular targets in experiments using macrophages from IL-15 Ϫ/Ϫ mice (data not shown). However, these results should be interpreted with caution and cannot be directly extended into the human system. Notwithstanding, the decrease in the amount of membrane-bound IL-15 on activated monocytes observed after acidic treatment indicates that, in addition to transmembrane IL-15, a certain number of IL-15 molecules are bound to IL-15R␣ in the absence of exogenous IL-15. This binding reportedly enables IL-15R␣ to recycle between endosomes and the cell membrane and to present IL-15 in trans to neighboring target cells that express either only the ␤␥ or IL-15R␣␤␥ complex, thus prolonging their survival (20). Given, however, that human IL-15 has at least two binding sites for recombinant sIL-15R␣ (34), it is tempting to speculate that the binding of the IL-15R␣⅐IL-15 complex on monocytes to IL-15R␣ on cells that express this high affinity chain might result in the formation of a receptor-ligand-receptor structure, where IL-15 is captured by two IL-15R␣ molecules expressed on the surface of different cell types. This complex structure may theoretically be able to induce signaling in both directions, triggering activation of IL-15R-mediated signaling events in both cell types. Future studies must address the plausibility of such bidirectional signaling and the contribution of the IL-2R␤ and IL-2R␥/␥c subunits to this process.
However, PC-3 human prostate cancer cells were clearly devoid of components of the IL-15R complex and did not exhibit any reduction in the number of membrane-bound IL-15 molecules after acidic treatment or trypsinization. The highly invasive PC-3 human prostate cancer cell line has also been found to express the ␣ v ␤ 3 integrin; in contrast, the noninvasive LN-CaP prostate cancer cell line does not express this adhesion molecule (32) as well as membrane IL-15. Prostate carcinoma has been estimated to be the second leading cause of death due to cancer among men in the United States (35). A strong correlation between elevated FAK expression, which controls cell motility, and an increased invasive potential of human tumors has already been demonstrated (36). The observed phosphorylation of FAK in PC-3 cells may serve as a hallmark of the oncoming changes in the cellular behavior that promote cancer cell invasion. In fact, stimulation of PC-3 cells with sIL-15R␣ or anti-IL-15 Abs clearly enhanced the migratory properties of these cells, as assessed by their ability to migrate into the scratched area in a wound healing assay. Moreover, sIL-15R␣ or anti-IL-15 Abs were also able to promote migration of LN-CaP cells genetically modified to stably express membranebound IL-15LSP. In addition, members of the MAPK family play a role in modulating integrin-mediated cell migration (37,38). Thus, activation of the FAK and MAPK signaling pathways by membrane-anchored IL-15 causes prostate carcinoma cells to acquire a migratory phenotype in vitro. Furthermore, semiquantitative RT-PCR showed that stimulation of PC-3 cells with sIL-15R␣ or anti-IL-15 Abs induced an increase in the transcription level of IL-6 and IL-8, thereby indicating that activation of the intracellular signaling molecules in PC-3 cells enhances gene expression. It remains to be elucidated whether reverse signaling through transmembrane IL-15 may also affect the malignant properties of prostate carcinoma cells, such as survival, proliferation, and tumor invasion, leading to an accelerated development of cancer and enhanced metastatic potential in vivo.
Moreover, the existence of a natural soluble form of the high affinity IL-15R␣ chain in mouse serum and cell-conditioned medium (39) strongly suggests that, in addition to the interaction between the soluble ligand and the membrane-bound receptor or the membrane-coupled ligand-receptor pair, the binding of natural sIL-15R␣ to IL-15 that is expressed in the membrane-bound form might potentially represent a physiologically relevant mechanism to regulate distinct cellular responses. Importantly, stimulation with sIL-15R␣ induced a greater change in the phosphorylation of MAPKs and FAK at lower concentrations. A plausible explanation for this difference is provided by recent study (34) that identified four regions in human IL-15. The first one is located in the C-D loop and is recognized by a set of non-inhibitory Abs, whereas the second region is located in helix D and is recognized by two Abs that are inhibitory for IL-15 bioactivity. The two remaining regions are located in the B and C helixes, respectively, and react with recombinant sIL-15R␣. Thus, it seems likely that distinct IL-15 binding properties of sIL-15R␣ versus anti-IL-15 Abs might account for the ability of sIL-15R␣ to induce a greater change in the phosphorylation of MAPKs and FAK. Furthermore, the preferential tyrosine phosphorylation of ERK2 (p42) in response to stimulation with sIL-15R␣ or anti-IL-15 Abs is intriguing, and additional experiments are required to address its biological significance. Notably, ERK1 and ERK2 have also been shown to be selectively or differentially activated by distinct stimuli, including epidermal growth factor (40), TNF-␣ (41), insulin (42), and membrane immunoglobulin crosslinking (43).
However, the molecular mechanisms underlying the ability of transmembrane IL-15 to mediate the activation of signaling molecules are unclear and deserve systemic exploration in follow-up experiments. By analogy with TNF-␣, IL-15 has been predicted to have several consensus sequences, including potential phosphorylation sites for casein kinase II and protein kinase C (data not shown). Importantly, we observed the phosphorylation of transmembrane IL-15 at serine (but not tyrosine) residues, although no physical association between IL-15 and MAPKs or FAK was detected in the immunoprecipitation experiments upon stimulation with sIL-15R␣ or anti-IL-15 Abs. Noteworthy, the predicted cytoplasmic portion of IL-15LSP contains several serine residues (Fig. 7). Current studies in our laboratory are focused on the mutational analysis of IL-15LSP to identify which regions of this protein might be important for downstream signaling. Furthermore, it seems plausible that transmembrane IL-15 may be internalized upon binding of recombinant sIL-15R␣ or anti-IL-15 Abs to mediate the subsequent signaling events. Internalization from the plasma membrane has been shown to occur either via clathrincoated pits to sorting endosomes and the recycling compartment or via caveolae to smooth endoplasmic reticulum tubules (44). On the other hand, IL-15R␣ has been demonstrated to mediate recycling of IL-15R␣⅐IL-15 complexes between endosomes and the cell surface, leading to the persistence of IL-15 after withdrawal of the cytokine from the culture medium (20). Although the essential role of the IL-15R␣ cytoplasmic domain in the recycling process has been demonstrated (20), it is not clear yet whether membrane-bound IL-15 may reappear again on the plasma membrane after the interaction with its natural soluble receptor, contributing to the total amount of IL-15 upon the cell surface.
Despite widespread expression of IL-15 mRNA, detection of significant amounts of IL-15 in cell culture supernatants has proven to be extremely difficult (18). It has been suggested that the production of IL-15 is regulated at the level of protein translation and intracellular trafficking rather than transcription. Two distinct isoforms of IL-15 exhibit different patterns of intracellular distribution and trafficking (16 -18). Whereas IL-15LSP is found in the Golgi and early endosomes (17) and in the endoplasmic reticulum (18), IL-15SSP is present in the cytosol but does not co-localize with the Golgi, early endosomes, or the endoplasmic reticulum (17,18). However, both IL-15LSP and IL-15SSP may exhibit nuclear localization (11). Although the stable overexpression of IL-15SSP in LNCaP cells resulted in the appearance of IL-15 on the cell membrane, it seems likely that, under physiological conditions, only IL-15LSP is expressed in a transmembrane form. This notion gains further support from the observation that PC-3 cells expressed only IL-15LSP on the cell surface.
The ability of a membrane protein to induce bidirectional signal transduction was discovered several years ago. Most members of the TNF ligand (L) family exist as transmembrane proteins with relatively long intracellular domains, and many of them are involved in reverse signaling. This phenomenon has been shown for such members of the TNFL family as TNF-␣, CD30L, CD40L, OX40L, CD137L, and FasL (24 -29, 45). In addition, transmembrane ligands of the EPH family receptor Nuk have been implicated in the regulation of axon guidance, fasciculation, and compartment definition (46). Outside-to-inside signal transduction via membrane TNF-␣ induces the expression of E-selectin (CD62) on activated human CD4 ϩ T cells (47). Another member of the TNFL family, CD137L, was shown to induce monocyte activation through bidirectional signaling (28). A casein kinase I motif present in the cytoplasmic domain of members of the TNFL family has been implicated in reverse signaling (48). Members of the TNFL family, with the exception of lymphotoxin-␣, are type II transmembrane proteins (24 -29). They have a large and evolutionary highly conserved leader sequence that spans the cell membrane and a small N-terminal cytoplasmic portion. The presence of the extended signal peptide in IL-15LSP (48 amino acids) is rather unusual for secreted cytokines. Analysis of the amino acid sequences of membrane-bound IL-15LSP and TNF-␣ showed striking similarities in the protein structure organization (Fig. 7). Moreover, as mentioned above, IL-15 has been predicted to contain several consensus sequences, including potential phosphorylation sites for casein kinase II and protein kinase C, thereby sharing not only structural similarities with TNF-␣ (data not shown). Thus, IL-15LSP may naturally exist as a membrane-bound type II protein and, similar to members of the TNFL family, can mediate biologically relevant reverse signaling events. Despite the structural relationship with IL-2 and the fact that both IL-15 and IL-2 belong to a four-helix bundle cytokine family (1, 2), post-translational modifications and functional features make it tempting to speculate that IL-15 may be closer to TNF-␣ than to other cytokines. Members of the TNFL family are not secreted but are rather shed from the cell membrane by TNF-␣-converting enzyme, as previously demonstrated for TNF-␣ (49). Experiments are underway to explore the ability of TNF-␣-converting enzyme to cleave membrane-bound IL-15.
Abnormalities in IL-15 expression have been reported in diverse neoplastic and inflammatory diseases, including adult T cell leukemia and certain autoimmune disorders such as rheumatoid arthritis, inflammatory bowel disease, multiple sclerosis, chronic liver disease, and pulmonary sarcoidosis (4). It has been suggested that IL-15 may be at the apex of a cytokine cascade that includes downstream production of IL-1, IL-6, granulocyte/macrophage colony-stimulating factor, and other biologically active substances (50). Our results demonstrate that highly invasive human prostate cancer cells and IFN-␥-activated human monocytes express membrane-anchored IL-15, which is able to activate a reverse signaling cascade that involves the phosphorylation of FAK and MAPKs. Moreover, agonistic stimulation of membrane-bound IL-15 promotes migration of prostate cancer cells and increases tran- Both cytokines are predicted to exhibit a high degree of structural similarity, having rather short cytoplasmic and transmembrane domains and relatively long extracellular regions (predicted using Pro-teinPredict software). h, human. scription or release of pro-inflammatory cytokines in PC-3 cells or human monocytes, respectively. Thus, IL-15 acts as a bipolar molecule, inducing bidirectional signaling both as a ligand and as a novel receptor protein that transmits signals to activate the cells expressing it on their surface. Given the pivotal role of IL-15 in many physiological processes or their pathological deviations and the bioavailability of the high affinity sIL-15R␣ chain, reverse signaling through membrane-bound IL-15 provides novel insights into our understanding of the pleiotropic properties of IL-15 within and beyond the immune system.