14-3-3σ associates with cell surface aminopeptidase N in the regulation of matrix metalloproteinase-1

The complexity of multicellular organisms demands extensive communication between cells of different germ layers in order to direct tissue development and repair. Increasing evidence suggests that bidirectional signaling between stromal and epithelial cells plays a crucial role in maintaining the integrity of the extracellular matrix (ECM) and that a disruption in this communication can lead to fibrogenesis (Deitch et al., 1983; Machesney et al., 1998; Niessen et al., 2001) or predispose cells to malignancy (Bissell and Radisky, 2001; Radisky and Bissell, 2004). Matrix metalloproteinases (MMPs) represent a group of diverse proteolytic enzymes that are responsible for degradation of ECM and are involved in a variety of biological processes, such as tissue repair and remodeling, embryonic development, bone growth/resorption and tumor metastasis (Butler and Overall, 2009b; Nagase et al., 2006; Rodriguez et al., 2010). The synthesis of MMPs by stromal fibroblasts is partially regulated by neighboring epithelial cells primarily through diffusible cytokines and growth factors (Gabison et al., 2005; Garner, 1998; Harrison et al., 2006).

Stratifin (SFN, or 14-3-3σ) is a member of a large family of highly conserved 14-3-3 proteins, which are known to function as intracellular chaperones in signal transduction, apoptosis and cell cycle regulation (Hermeking, 2003; Pozuelo et al., 2004). Several groups have reported on the presence of 14-3-3 proteins in the extracellular space (Butler and Overall, 2009a; Katz and Taichman, 1999; Kobayashi et al., 2009; Leffers et al., 1993; Satoh et al., 1999). Our group has recently identified a novel function for the keratinocyte-releasable form of SFN as a potent MMP-1-stimulatory factor in dermal fibroblasts (Ghahary et al., 2004). It was found that SFN released from keratinocytes induces the expression of MMP1, MMP3, MMP8 and MMP24 in fibroblasts through the p38 mitogen-activated protein kinase (MAPK) signaling pathway (Ghaffari et al., 2006; Ghahary et al., 2004; Lam et al., 2005). Furthermore, high levels of 14-3-3 eta (η) and gamma (γ) isoforms were detected in the synovial fluid of patients suffering from rheumatoid arthritis. The levels of 14-3-3η strongly correlated with the serum levels of MMP-1 and MMP-3, and the purified protein possessed MMP-1-stimulatory activity (Kilani et al., 2007).

The data presented indicate that releasable 14-3-3 might function as a ligand in a receptor-mediated transmembrane signaling pathway that leads to the induction of MMP-1 expression. Isoforms of 14-3-3 proteins have been reported to interact with the cytoplasmic domain of various plasma membrane proteins such as bullous pemphigoid antigen-2 or BP180 (Li et al., 2007), insulin-like growth factor I receptor (Craparo et al., 1997), muscle specific tyrosine kinase receptor (Strochlic et al., 2004), epidermal growth factor receptor (Oksvold et al., 2004) and type I transforming growth factor β receptor (McGonigle et al., 2001). However, no interaction between 14-3-3 proteins and the ectodomain of a cell surface receptor has been identified to date. In the current study, we report on the association between SFN and aminopeptidase N (APN) or CD13, a type 2 integral membrane protein, and highlight a transmembrane signaling mechanism for SFN-mediated MMP-1 expression in fibroblasts.

Human primary dermal fibroblasts, which are highly responsive to SFN activity (Ghahary et al., 2004), were used in a cell-binding assay to test the potential receptor binding activity of rSFN. To avoid nonspecific detection of intracellular 14-3-3 isoforms, a biotin-labeled recombinant rSFN (biotin–SFN) was prepared. Cells were incubated with biotin–SFN at 4°C for 15 minutes and analyzed with fluorescent-conjugated streptavidin. The results revealed that biotin–SFN efficiently binds to the cell surface of fibroblasts (Fig. 1A, top panels). To ensure specific binding, the assay was repeated by competing biotin–SFN with an excessive amount of unlabeled rSFN. This nearly eliminated the binding activity of biotin–SFN (Fig. 1A, middle panels). The assay was also repeated with a biotin-labeled GST protein (biotin–GST) expressed by the same plasmid as rSFN (Ghahary et al., 2004). No cell binding activity was detected in the presence of biotin–GST (Fig. 1A, bottom panels). SFN binding to the cell surface was further evaluated by using a ¹²⁵I-labelled SFN (¹²⁵I-SFN) in a ligand–receptor binding assay. Cells treated with ¹²⁵I-SFN were incubated in the presence of an increasing concentration of unlabelled SFN for 2 h and then harvested by lysis buffer to measure the total binding in a gamma counter. As shown in Fig. 1B, more than 90% of the ¹²⁵I-SFN binding was replaced with competing and unlabelled SFN, demonstrating a specific SFN binding to the cell surface. A plasma membrane (PM) isolation technique was utilized to investigate the localization of the exogenous rSFN that had been added to the cells. Following a short incubation with biotin–SFN at 4°C, the PM and cytoplasmic fractions were separated by SDS-PAGE and analyzed by immunoblotting to identify biotin–SFN. As expected, biotin–SFN was detected predominantly in the membrane fraction (Fig. 1C). The GAPDH was not observed in the PM fraction, which indicated that the membrane preparation was devoid of cytosolic protein contamination. Exogenous biotin–GST was used in a separate group to control for the potential nonspecific binding of a biotin-labeled protein.

To identify a candidate receptor for SFN, we employed a cell surface biotin labeling technique followed by serial SFN– and avidin–sepharose affinity purification. The ectodomains of PM proteins were labeled with a membrane-impermeable and non-cleavable biotin (sulfo-NHS–biotin), which binds covalently with the exposed primary amine groups. The cell membrane was disrupted by a mild lysis buffer and passed through an SFN– or blank-sepharose (control) batch column to isolate any potential SFN binding partners. To further purify potential candidates, the proteins retained by SFN–sepharose were eluted at acidic pH (glycine pH=2), neutralized and then passed through an immobilized avidin affinity column. Final purified proteins were eluted by the SDS sample buffer, separated by SDS-PAGE and visualized by sliver staining. As shown in Fig. 2A, silver staining of affinity-purified complex revealed several potential SFN-binding proteins. Streptavidin immunoblotting of the same sample identified two main biotinylated proteins at 170 and 150 kDa. As the 170 kDa band was also seen in the control group (blank sepharose), the 150 kDa protein was the only biotinylated membrane protein with binding affinity for SFN. To identify this protein, the SDS-PAGE band was subjected to trypsin digestion and liquid chromatography-mass spectrometry (LC–MS–MS) protein identification. The peptide sequences were analyzed with the MASCOT search engine. The peptides were aligned with the amino acids 485 to 498 and 843 to 855 of the human aminopeptidase N (APN) or CD13 (Fig. 2B). The probability-based Mowse score of 87 indicated a statistically significant alignment.

To confirm the results from the MS protein identification, we repeated the SFN–sepharose affinity binding using fibroblast cell lysates and separated the eluted complex by SDS-PAGE and subjected it to immunoblotting using an antibody against APN. Consistent with our original finding, the APN protein demonstrated a high binding affinity to immobilized SFN, as shown in Fig. 2C. To demonstrate further that rSFN is a specific binding partner for APN, we used membrane-impermeable bis-sulfosuccinimidyl suberate (BS³) to crosslink cell surface binding partners. Cells were incubated with rSFN before lysis and then crosslinked with BS³. Immunoblotting by the antibody against APN revealed a clear shift in the APN position only in the presence of rSFN and the BS³ crosslinker (Fig. 2D). We also performed immunoprecipitation experiments with antibodies against each protein. Fibroblasts were incubated with rSFN, washed extensively and lysed before performing the immunoprecipitation. Two different antibodies against APN co-immunoprecipitated SFN, as detected by western blot analysis using an antibody against SFN (Fig. 2E, top panels). A reverse immunoprecipitation experiment further revealed that the antibody to SFN was able to co-precipitate APN protein, as shown in the lower panel of Fig. 2E. To study the co-distribution of APN and SFN, fixed cells were incubated with biotin–SFN at 4°C for 30 min before immunofluorescence analysis with a monoclonal antibody against APN and fluorescent-conjugated streptavidin. The representative image shown in Fig. 2, panel F, revealed a high level of co-distribution between biotin–SFN and APN at the leading edge of migrating fibroblasts. Cells that lacked an apparent leading edge (non-migrating) also displayed co-distribution between SFN and its receptor, although to a lesser extent (data not shown).

To test the presence of the candidate SFN cell surface receptor in vivo, we examined the expression of APN and SFN in the normal and wounded rabbit ear by immunofluorescence. Dermal wound samples were collected on 16 days after injury, with a non-injured area in the back of the same ear used as a control. As expected, minimal expression of APN and SFN were observed in the uninjured skin (Fig. 3, panels a and c). However, examination of the wound revealed a large number of APN-positive cells throughout the dermis on day 16 post injury (Fig. 3, panel b), with the majority of the positive cells exhibiting a spindle-shaped fibroblast-like morphology. This coincided with an increase in the expression of SFN in the epidermis (Fig. 3, panel d), confirming the coexpression of the ligand in the wound. Immunofluorescence analysis of vimentin, a mesenchymal marker, was used to reveal that the majority of the cells in the dermis of the wounded skin exhibit a fibroblast phenotype (Fig. 3, insert in panel b). To study the co-distribution of SFN and its receptor in the skin at different stages of tissue repair, a longitudinal rat wound healing model was utilized, where tissue lysates from days 8, 11, 14, 17 and 30 post wounding were analyzed by western blot. Although both SFN and APN were detected in the wound following re-epithelialization (day 11), there was a strong overlapping peak on day 30 post-wounding (Fig. 3, panel b). This provided evidence for the coexpression of this ligand and receptor during the remodeling phase of wound healing.

Small interfering RNAs (siRNAs) were employed to knockdown APN expression in fibroblasts in order to examine the role of the APN–SFN interaction in SFN-mediated MMP-1 expression. Two different siRNA sequences were used against the mRNA encoding APN (GenBank accession number: NM_001150). Fig. 4A shows the inhibitory effect of siRNA on APN expression, as detected by immunofluorescence analysis. The quantification of the APN fluorescence intensity, which was normalized against the number of nuclei in each field (DAPI stain), revealed 50% and 80% inhibition at 10 and 50 nanomolar siRNA concentrations, respectively (Fig. 4A, right panel). The cells were then transfected with either non-silencing or APN siRNAs for 72 h and then treated with rSFN to examine the MMP-1 expression. The siRNA knockdown of APN sufficiently blocked the SFN-mediated MMP-1 expression, in contrast to the controls (Fig. 4B). The non-silencing siRNA concentration was matched with the highest dose of APN siRNA at 50 M. To exclude the possibility that APN siRNA would directly suppress MMP-1 expression, fibroblasts were treated with recombinant human interleukin-1 (rIL-1), a potent MMP-1-stimulatory cytokine (Mackay et al., 1992). As expected, the knockdown of APN by siRNA did not block the IL-1-mediated MMP-1 expression (Fig. 4C). In addition, we screened various strains of fibroblasts for expression of APN. IMR-90, a human fetal lung fibroblast, does not express APN protein at levels detectable by immunoblotting. Interestingly, IMR-90 was also the only tested strain that was not responsive to SFN stimulation (Fig. 4D), even though these cells have been shown to express MMP-1 when stimulated by different collagen substrates (Abraham et al., 2007).

Previously, we have shown that stimulation with SFN leads to the transient activation of p38 MAPK (Lam et al., 2005). In the current study, we tested the intermediary role of APN in the activation of p38 by knocking down its expression before rSFN stimulation. The phosphorylation of p38 was analyzed by immunoblotting using a phospho-p38-specific antibody. Both non-transfected and control siRNA-transfected cells demonstrated phosphorylation of p38 following 90 min of stimulation by rSFN, whereas the APN knockdown significantly blocked the p38 activation by rSFN (Fig. 5). The relative APN and p-p38 levels from two separate experiments were quantified by densitometry analysis (Fig. 5, upper panel). This finding suggests that an interaction between SFN and APN is required to activate the p38 MAPK pathway and to induce MMP-1 expression in the dermal fibroblasts.

All 14-3-3 proteins, including SFN, comprise a dimerization region at the N-terminus as well as a target-binding region primarily concentrated at the C-terminus (Tzivion et al., 2001). To examine whether a specific region within SFN is responsible for its association with APN, a number of GST-fusion SFN fragments with deletions in the N- or C-termini were expressed in bacteria and purified with glutathione–sepharose (Fig. 6A, top panel). The binding capacity of APN with each GST–SFN fragment was then examined by a GST pulldown assay and analyzed by SDS-PAGE and western blot. The relative binding affinity of APN to each SFN fragment was quantified by densitometry analysis and normalized against the GST group (Fig. 6, panel A, bar graph). Deletions in the N-terminus (Nd50 and Nd100) did not affect the APN binding, in contrast to its binding with the full-sequence SFN. However, SFN fragments with 148 (Cd148) amino acid deletions from the C-terminus displayed a significant reduction in binding affinity to APN (Fig. 6A). Accordingly, the SFN MMP-1-stimulatory activity was only diminished in the Cd148 deletion in contrast to the full length or Nd100 deletion (Fig. 6B). These findings suggest that the SFN C-terminus is the potential binding site for APN and the active site for MMP-1-stimulatory activity.

14-3-3 proteins regulate many cellular processes by binding to phosphorylated sites in target proteins (Pozuelo et al., 2004). We therefore tested the phosphorylation dependence of SFN binding to APN in the presence and absence of a phosphatase. Suspended intact cells were incubated with an alkaline phosphatase at 200 U/ml for 30 min at 30°C before lysis. Cell lysates of phosphatase-treated and control groups were then passed through a sepharose–SFN affinity column, and the APN level of the eluate was analyzed by western blotting. Surprisingly, the binding affinity of APN to SFN was significantly reduced following the dephosphorylation of the cell surface proteins (Fig. 6C, top left panel). As phosphorylation of APN has not been reported previously, we immunoprecipitated APN protein from primary fibroblasts before its separation by SDS-PAGE and western blot analysis. As shown in Fig. 6C (top-right panel), an antibody against phoshphoSer/Thr/Tyr clearly detected a band corresponding to the size of APN, demonstrating the presence of a phosphorylated amino acid residue in this protein. Immunoblotting also detected APN in the same blot at a molecular mass that matched the previously identified phosphorylated protein. In a separate experiment, live cells were incubated with phosphatase for 30 minutes at 37°C and then washed extensively before being treated with rSFN. A western blot analysis of MMP-1 expression revealed that the SFN-mediated upregulation of MMP-1 was attenuated in a dose-dependent manner as a result of pre-incubation with the phosphatase enzyme (Fig. 6C, lower panel). To rule out the indirect effect of dephosphorylation, the IL-1-mediated MMP-1 expression was shown to be unchanged in the presence of the phosphatase enzyme.

APN is a zinc-dependent metalloproteinase and a type II integral membrane ectoenzyme (Riemann et al., 1999). Bestatin is a potent inhibitor of APN enzymatic activity that blocks the active site by directly binding with the zinc ion (Addlagatta et al., 2006). To test the role of APN enzymatic activity in SFN-mediated MMP-1 expression, fibroblasts were pre-incubated with various doses of bestatin before treatment with rSFN. An APN enzymatic assay revealed a dose-dependent inhibition by bestatin in fibroblasts (Fig. 6D, upper panel). However, the bestatin had no effect on the SFN-mediated MMP-1 expression (Fig. 6D, lower panel). This finding suggests that APN enzymatic activity does not play a role in regulation of MMP-1 expression and that the enzyme active site is most likely not the binding site for SFN as the presence of bestatin did not block interaction with the ligand.

The 14-3-3 proteins interact with over 200 binding partners and therefore regulate numerous cellular processes, mainly by inducing conformational changes in the target protein, masking of specific binding or active sites or the colocalization of two proteins (Yaffe, 2002). Although 14-3-3 proteins are mainly intracellular, their presence in the secretome of various cell lines (Butler and Overall, 2009a) as well as primary keratinocytes has been reported by others (Katz and Taichman, 1999; Leffers et al., 1993), followed by the unexpected discovery of their MMP-1-stimulatory activity by our group (Ghahary et al., 2004). It was recently suggested that, owing to a lack of conventional signaling peptide, keratinocyte-releasable SFN follows a non-classical secretory pathway involving the release of exosomes (Chavez-Munoz et al., 2008). In fact, several well-known extracellular proteins, including IL-1β and FGF-2, are also released by non-classical (ER/Golgi independent) pathways (Keller et al., 2008; Nickel, 2007). Activation of the p38 MAPK signal transduction pathway followed by a rapid increase in the expression of MMP-1 in response to SFN (Lam et al., 2005), as well as its association with the cell surface of fibroblasts, all suggest the existence of a receptor-mediated transmembrane signaling pathway. In this study, biotin labeling of extracellular domains, serial affinity purification and LC–MS–MS protein sequencing have led us to identify APN or CD13 as a potential SFN cell surface receptor or co-receptor.

APN is an ectoenzyme metalloproteinase belonging to the M1 family and a type II integral membrane protein with a short N-terminal cytoplasmic domain. It is highly glycosylated, which contributes to at least 20% of its molecular mass. Furthermore, it is expressed in a wide variety of cells, such as myeloid cells, fibroblasts, renal and intestinal epithelia, and endothelial cells (Favaloro, 1991; Nomura et al., 2004; Riemann et al., 1999). APN has been referred to as a ‘moonlighting’ protein owing to its wide range of functions, such as the enzymatic regulation of peptides, viral receptor, tumor-homing peptide receptor, tumor cell invasion, proliferation and apoptosis, motility, antigen presentation, cholesterol uptake and signal transduction (Mina-Osorio, 2008). Our findings introduce a novel function for APN as a cell surface receptor for the SFN ligand, with a link to the signal transduction events leading to MMP-1 expression in fibroblasts. It will be interesting to see whether the interaction between APN and SFN also plays a role in the previously identified functions of APN in fibroblasts and other human cells.

The transmission of extracellular signals to MMP-1 gene expression in various cell types is mainly mediated by the family of serine/threonine-specific MAPKs, including ERK1/2 and p38 MAPK subgroups (Nagase and Woessner, 1999). SFN stimulation of fibroblasts leads to a transient activation of p38 MAPK as well as an elevation in Jun and Fos – components of the heterodimeric AP-1 transcription factor – resulting in transcription of the gene encoding MMP-1 (Lam et al., 2005). Our findings indicate that APN knockdown is sufficient to block the p38 activation in the presence of an rSFN stimulus, suggesting a link between the APN and MAPK signaling cascade. In fact, signal transduction has been proposed as a mechanism of action for a number of APN functions that operate independently of its enzymatic activity (Mina-Osorio, 2008). Santos and colleagues found that the ligation of APN with monoclonal antibodies (mAbs) increases the intracellular free-calcium ions in monocytes and induces phosphorylation of ERK1/2, JNK and p38 MAPK. The same study also demonstrates that inhibitors of tyrosine kinases and phosphoinositide 3-kinase block the increase in calcium influx as well as the phosphorylation of MAPKs in the presence of the mAb against APN (Santos et al., 2000). In addition, antibodies to APN that have little effect on enzymatic activity exhibit a high capacity to induce cell adhesion in monocytes; this effect can be inhibited by SB203580, a p38 MAPK inhibitor (Mina-Osorio et al., 2006). APN has a very short cytoplasmic domain without any known signaling motifs (Luan and Xu, 2007). Therefore, it is possible that the APN interacts with a yet unidentified integral or peripheral signaling molecule in the SFN-mediated activation of p38 MAPK. In fact, Mina-Osorio and colleagues recently demonstrated the co-immunoprecipitation of Grb2 and Sos with APN from lysates of monocytic cells, suggesting a possible link between APN, Ras and the ERK1/2 MAPK signaling pathway (Mina-Osorio et al., 2006).

The co-distribution of SFN with APN at the leading edge of fibroblasts was unpredicted. Interestingly, an in vitro wound scratch assay revealed higher expression of APN within the migratory fibroblasts, where it co-distributed with vinculin, a migration marker (Lai et al., 2010). In addition, the inhibition of cell migration by reducing or blocking APN levels in tumor cells, lymphocytes and endothelial cells has also been reported (Terauchi et al., 2007; Proost et al., 2007). Therefore, the higher expression of APN receptor at the leading edge of migratory cells could explain the enhanced co-distribution results observed with biotin–SFN. Furthermore, the movement of cells through ECM is highly dependent on the activity of proteases such as MMPs (Gianneli et al., 1997; Kessenbrock et al., 2010; Pilcher et al., 1997; Pirila et al., 2003). This raises the possibility that SFN interaction with APN and the subsequent increase in MMP-1 activity could play a role in the cell migration of fibroblasts.

Our finding regarding the phosphorylation-dependent nature of SFN and APN binding is unexpected, mainly because APN is a type II plasma membrane protein with no previously reported phosphorylation sites. Nevertheless, a short incubation with a phosphatase is sufficient to block not only the binding between SFN and APN but also the expression of SFN-mediated MMP-1. An indirect effect on the signal transduction pathways was ruled out, as IL-1-mediated MMP-1 expression was not hindered by dephosphorylation of intact cells. Reports show that ECM proteins (osteopontin, sialoprotein, dentin matrix protein-1, phosphophoryn) as well as many secretory proteins (casein, enkephalin precursor, progastrin, chromogranin) become phosphorylated during their movement through the secretory pathway (Boskey et al., 2008; Turner et al., 1993). In addition, the extracellular phosphorylation of the ectodomains of integral proteins is not unprecedented. It has been shown that the CD36 ectodomain, a membrane receptor for thrombospondin and collagen, is phosphorylated by protein kinase C (PKC) (Asch et al., 1993). Ehrlich and Redegeld have also argued for the presence of adenosine triphosphate (ATP) and ectoprotein kinases in the extracellular space (Ehrlich et al., 1990; Redegeld et al., 1997). Although the detection of a phosphoSer/Thr/Tyr band points towards the existence of phosphorylated APN, this finding is still in its preliminary stages. Further work is under way to identify the exact phosphorylation site(s) in APN by mass spectroscopy analysis and to investigate its role in SFN activity by site-directed mutagenesis.

Seven mammalian isoforms of 14-3-3 proteins have been described previously, all of which form dimers that bind to high-affinity phosphorylation-dependent motifs R[S/x]x(P)SxP and Rxxx(P)SxP (Yaffe et al., 1997) as well as to atypical phosphorylated and unmodified proteins (Fuglsang et al., 1999; Furlanetto et al., 1997; Waterman et al., 1998). All 14-3-3 proteins comprise a dimerization region located at the N-terminus and a target-protein-binding region located primarily in the C-terminus (Tzivion et al., 2001; Wang et al., 1999). The fragmentation study presented here reveals attenuation in the APN binding affinity following deletions in the SFN C-terminus, which indicates a potential binding site for APN. The effect of dimerization of 14-3-3 proteins on binding to a target protein seems to be dependent on the strength of their binding affinity. It has been shown that partners with high-affinity binding sites are able to bind to 14-3-3 even in the absence of dimerization (Tzivion et al., 1998; Yaffe et al., 1997). Interestingly, the deletions in the SFN dimerization interface (N-terminus) did not appear to affect its APN binding affinity in our study. This could indicate the presence of a high-affinity binding site within the extracellular domain of APN, allowing it to interact with the SFN monomer that lacks the N-terminal dimerization region.

The degradation of ECM by collagenases (e.g. MMP-1) is an essential component of pathophysiological events such as wound repair and tumor invasion. In wound repair, a significant portion of MMP production is performed by stromal fibroblasts, partly under the paracrine influence of local epithelial and immune cells (Toriseva and Kahari, 2009). Although the exact nature of APN and SFN interaction in wound healing remains unknown, we show an apparent co-distribution of SFN and APN in both the rabbit and rat tissue during the healing process. The rise in their expression appears to take place following re-epithelialization and during the early stages of the remodeling phase, where MMPs play a significant role in reorganization of the newly deposited granulation tissue.

APN or CD13, in combination with CD34 and collagen, is used as a cell surface marker in the identification of fibrocytes. Fibrocytes are peripheral blood-borne fibroblast-like cells that rapidly enter the injured site and synthesize ECM (Abe et al., 2001; Bucala et al., 1994). However, Sorrell and colleagues have demonstrated that the frequency and distribution of these cells do not account for the APN pattern in adult skin, indicating that fibroblasts could serve as another potential source of APN (Sorrell et al., 2003). In line with our in vivo findings, Mori and colleagues demonstrated an increase in the APN levels in the wounded versus uninjured skin of balb/c mice (Mori et al., 2005). This effect might be attributed to the reported APN enzymatic activity in ECM degradation (Saiki et al., 1993) or the role of APN in cellular motility and migration (Chang et al., 2005; Fukasawa et al., 2006). Based on our findings, it is also reasonable to speculate that the increased expression of APN is a response that augments cellular sensitivity to MMP-stimulating factors such as keratinocyte-derived SFN.

In cancer, it is the neighboring stromal fibroblasts in addition to malignant epithelial cells that contribute to the repertoire of MMPs responsible for the proteolytic breakdown of the peritumoral ECM (Egeblad and Werb, 2002; Overall and Lopez-Otin, 2002). Furthermore, stromal cells in tumors appear to be activated by paracrine factors released from malignant epithelial cells (Mueller and Fusenig, 2004). Cell surface APN has been found to play a crucial role in tumor cell metastasis and is overexpressed in a wide range of solid tumors (Luan and Xu, 2007). Interestingly, an immunohistochemical examination of human lung carcinoma tissue revealed that the majority of APN-positive cells were in fact stromal fibroblasts (Ichimura et al., 2006). It is therefore tempting to speculate that epithelial-released factors such as SFN can act through the APN receptor to induce MMP production by neighboring stromal fibroblasts. If this hypothesis is correct, then the reported increases in expression of both APN and MMPs in the tumor microenvironment are connected events and not simply a coincidence. In fact, new data from our laboratory show a significant increase in the expression of APN by fibroblasts when co-cultured with keratinocytes or treated with keratinocyte-derived conditioned medium (Lai et al., 2010).

In conclusion, the data presented in this paper identify APN as a novel cell surface receptor for the SFN ligand and enable us to propose a new function for this type II integral membrane protein in the SFN-mediated activation of p38 MAPK and MMP-1 expression in fibroblasts. As indicated in Fig. 7, our findings also raise a few key questions about the exact mechanism of APN activation by SFN. Does APN act alone or as a co-receptor? Where is the SFN binding site on the APN ectodomain? Where exactly are the phosphorylation sites on APN? Answers to these questions could pave the way for targeting APN in the treatment of disorders marked by an imbalance in MMP expression.

Following informed consent, foreskin samples were obtained from patients undergoing elective surgery under local anesthesia. Written informed consent was obtained from each participant, and the study was approved by the University of British Columbia Hospital Human Ethics Committee and conducted according to the Declaration of Helsinki Principles. Human primary dermal fibroblasts were harvested as described previously (Ghaffari et al., 2006). Fibroblasts, balb/3T3 and 293T cells were grown in DMEM supplemented with 10% fetal bovine serum (FBS) and antibiotic-antimycotic preparation (100 U/ml penicillin, 100 μg/ml streptomycin and 0.25 μg/ml amphotericin B). The human fetal lung fibroblast cell line IMR-90 (ATCC, Manassas, VA) was cultured in DMEM plus 10% FBS. The CCD-1064Sk cell lysate (Santa Cruz Biotechnology, Santa Cruz, CA) was used as a positive control for APN protein expression. Calf intestinal alkaline phosphatase (AP) (New England Biolabs, Ipswich, MA) was used for the protein dephosphorylation experiment according to the manufacturer's instructions. Bestatin hydrochloride (Sigma Chemicals, Oakville, ON, Canada) was used as a competitive and specific inhibitor of APN enzymatic activity. Cell lines isolated from human squamous cell carcinoma (SCC) were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and used as positive controls for MMP-1 expression.

Expression vectors for human stratifin have been described previously (Li et al., 2007). In brief, the human gene encoding full-length stratifin was subcloned into prokaryotic expression vectors pGEX-6P-1 (GE Healthcare Bio-Sciences, Quebec, Canada) and pET-28a(+) (Novagen, Madison, WI) for fusion protein tagged with glutathione S-transferase (GST) and histidine (His), respectively. The His-tagged SFN was purified using a Ni-Sepharose column according to the manufacturer's instructions (GE Healthcare Bio-Sciences, Quebec, Canada). Different fragments of SFN fused to GST were also prepared as described previously (Li et al., 2007). Fusion proteins containing peptide fragments of SFN were referred to as Nd50 (amino acids 51–246), Nd100 (amino acids 101–246), Cd98 (amino acids 1–150) and Cd148 (amino acids 1–100). GST-tagged SFN fragments bound to glutathione-sepharose (GE Healthcare Bio-Sciences, Quebec, Canada) were then used in affinity binding experiments.

To examine the cell binding activity of SFN protein, 5×10⁴ primary human fibroblasts were plated in a 12-well plate on a glass coverslip overnight. The cells were then serum starved for 3 h before the addition of protein. Recombinant stratifin (rSFN) in PBS was labeled with Sulfo-NHS-LC-Biotin (Pierce, Rockford, IL) in a 3:1 ratio (Biotin:rSFN), as instructed by the manufacturer. Cells were blocked with 1.0% bovine serum albumin (BSA) in DMEM for 30 minutes before the addition of biotin–SFN at 1.0 μg/ml for 15 minutes at 4°C and then washed extensively with PBS. Biotin-labeled glutathione S-transferase (biotin–GST) was used as a negative control. The competition assay was performed by pre-incubating cells with 200 μg/ml of rSFN for 30 minutes at 4°C before adding biotin–SFN. Biotin was detected by immunofluorescence microscopy using streptavidin–alexa 568 (Invitrogen, Eugene, OR). To perform a ligand–receptor binding assay, rSFN was labeled with ¹²⁵I (New England Nuclear, Boston, MA) and Iodo-Gen reagent (Pierce, Rockford, Illinois) according to the manufacturer's instructions and as described previously (Uludag et al., 1999). Following a wash with cold PBS, fibroblasts were incubated with ¹²⁵I-SFN (5×10⁵ c.p.m.) in the absence and presence of an increasing concentration of unlabeled rSFN for 2 h at 4°C. Unincorporated radiolabel was then washed out with cold PBS, the cells were harvested by lysis buffer and the total binding of radiolabel was measured in a liquid scintillation counter (Beckman coulter, Mississauga, ON, Canada).

For total cell lysates, following an extensive wash with PBS, cells were disrupted in lysis buffer [50 mM Tris-HCl, 150 mM NaCl, 1.0 mM EDTA, 1.0 mM EGTA, 0.025% NaN₃, 1% Triton X-100, 0.5% IGEPAL, protease inhibitor cocktail (Sigma Chemicals, Oakville, ON, Canada), at pH 7.5] and centrifuged at 16,000 g for 15 min. For preparation of the cell fractions, cells were released by EDTA and collected in tissue disruption buffer (150 mM NaCl, 20 mM Na₂HPO₄, 2 mM NaH₂PO₄, 20% glycerol, 2 mM sodium orthovanadate and protease inhibitor cocktail) and homogenized in a Dounce homogenizer glass (Fisher, Ottawa, ON, Canada). Homogenized preparations were further disrupted by sonication on ice and then centrifuged at 19,000 g to remove nuclei and cell debris. The supernatant was then centrifuged at 100,000 g for 90 minutes at 4°C. The soluble cytoplasmic fraction (supernatant) was collected, and the remaining pellet (total membrane fraction) was solubilized in lysis buffer.

Following release of fibroblasts with EDTA, cell surface proteins of intact cells were biotinylated using 2 mM water-soluble and membrane-impermeable EZ-Link sulfo-NHS-LC-biotin reagent (Pierce, Rockford, IL) in PBS for 30 minutes at 4°C. Free biotin was quenched and removed by three washes of cold PBS plus 100 mM glycine. To obtain the total cell lysate, cells were disrupted in lysis buffer (50 mM Tris-HCl; 10 mM EDTA; 5 mM EGTA; 0.025% NaN₃; 1% Triton-X100; 0.5% IGEPAL CA-630; protease inhibitor) for 30 minutes by gentle rotation at 4°C and then cleared by centrifugation at 18,000 g for 10 min at 4°C. Affinity-purified His–SFN was dialyzed against PBS and then conjugated with CNBR-activated Sepharose 4B (GE Healthcare Bio-Sciences, Quebec, Canada) at a ratio of 10 μmole ligand to 1.0 ml of sepharose, according to the manufacturer's instructions. A deactivated and blank Sepharose-4B was used as a control. Biotin-labeled cell lysate was then applied to a 30 μl bed of sepharose–SFN or sepharose alone for 2 h at 4°C with gentle rotation. Beads were then washed seven times with PBS plus 1% Tween-20, eluted with 100 mM glycine at pH 2.0. The glycine sample was neutralized with 1 M Tris solution, applied to 30 μl of Avidin–Agarose (Sigma-Aldrich, Oakville, ON, Canada) for 1 h at 4°C and washed three times with PBS plus 1% Tween-20. Biotinylated proteins were then eluted by SDS loading buffer [0.35 M Tris-HCl, 10.28% (w/v) SDS, 36% (v/v) glycerol, 5% 2-mercaptoethanol, 0.012% (w/v) bromophenol blue] with a 10 min incubation at room temperature followed by a 5 min boil. Eluted proteins were resolved by 10% precast keratin-free SDS-PAGE (Bio-Rad, Hercules, CA) and either immunoblotted with streptavidin-HRP or stained with ProteoSilver silver (Sigma-Aldrich, Oakville, ON, Canada). For protein sequencing, purified biotinylated proteins excised from silver-stained gels were subjected to trypsin digestion and identified by API Q STAR PULSARi Hybrid LC–MS–MS at the University of British Columbia MSL/LMB Proteomics Core Facility. The peptide sequences were analyzed using the Mascot search engine and Analyst software (Applied Biosystems, Foster City, CA) against the non-redundant NCBI database.

Cell lysates, cell fractions or purified proteins were transferred to PVDF membrane with Mini Trans-Blot Cell (Bio-Rad, Hercules, CA) after separation by SDS-PAGE. Immunoblotting was carried out with the following the antibodies: 1:5000 streptavidin–HRP (Pierce, Rockford, IL); 1:5000 anti-GAPDH (R&D Systems, Minneapolis, MN); 1:500 anti-APN [3D8] (Santa Cruz Biotechnology, Santa Cruz, CA); 1:1000 anti-stratifin (Medicorp, Montreal, QC, Canada); 1:500 anti-MMP-1 (R&D Systems, Minneapolis, MN); 1:1000 anti-phosphoserine/threonine/tyrosine (Abcam, Cambridge, MA); 1:20,000 anti-β-actin (Sigma-Aldrich, Oakville, ON, Canada); 1:3000 anti-mouse IgG-HRP and anti-rabbit IgG-HRP (Bio-Rad, Hercules, CA). Protein bands were visualized by an ECL detection system (Santa Cruz Biotechnology, Santa Cruz, CA).

Fibroblasts were released by EDTA and washed with PBS before incubating with rSFN for 30 minutes at 4°C. Soluble cell lysate was collected in lysis buffer, cleared by centrifugation and immunoprecipitated by 2 μg/ml monoclonal anti-APN [3D8], 2 μg/ml polyclonal anti-APN [H300] (Santa Cruz Biotechnology, Santa Cruz, CA) or anti-SFN antibody at 4 μg/ml for 1 h at 4°C with gentle rotation. Non-immunized rabbit and mouse IgG were used as controls. Immune complexes were collected on protein G–Sepharose beads (50 μl/ml) (GE Healthcare Bio-Sciences, Quebec, Canada) for 1 h at 4°C, washed five times with PBS.T (0.1% Tween-20), eluted by boiling in sample buffer and resolved by SDS-PAGE for western blot analysis.

APN protein knockdown was tested by two different siRNA oligonucleotide sequences: si_ANPEP_1, CCGAAATGCCACACTGGTCAA, and si-ANPEP_2, CCGGGTGAACTACGACGAAGA purchased from Qiagen (Valencia, CA). A non-silencing siRNA, with the same GC content as that of APN siRNA, conjugated with Alexa Fluor 488 was used as a control and for monitoring transfection by fluorescence microscopy. HiPerfect transfection reagent was used according to the manufacturer's recommendations (Qiagen). Proliferating fibroblasts were harvested and 1×10⁵ cells plated in six-well dishes containing 5, 10 or 50 nM siRNA oligonucleotide. The medium was replaced 24 h post-transfection, and the cells were treated with rSFN (2.5 μg/ml) at 72 h post-transfection.

All animals were cared for in accordance with the guidelines of the Canadian Council on Animal Care and with approval from the University of British Columbia Animal Welfare and Policy Committee. Female New Zealand white rabbits weighing approximately 4–5 kg were used in the rabbit ear fibrotic wound model. After the rabbits were anaesthetized, 8-mm diameter full-thickness wounds down to the perichondrial membrane were created, as described previously (Lee et al., 2005; Morris et al., 1997). On day 16 post wounding, when reepithelialization was complete, the rabbits were euthanized and the wounds harvested and fixed in 2% PFA before embedding in paraffin. Normal and non-injured skin on the back of the same ear was used as the control. Sprague–Dawley rats were used as described previously in a longitudinal wound healing study to monitor levels of APN and SFN at different stages of dermal repair (Li et al., 2006). In brief, six full-thickness excisional wounds (6.0 mm diameter) were made on the dorsal surface of rat skin and allowed to heal for up to 30 days. One animal was euthanized on each post-wound day 8, 11, 14, 17 and 30, and a 4.0 mm punch biopsy was obtained from the injured area and frozen at −80°C. Tissue samples were weighed-out and homogenized in lysis buffer with a tissue homogenizer (Qiagen, Valencia, CA) and glass pestle, followed by sonication. A Bradford protein assay was performed, and an equal amount of tissue lysate (75 μg) was separated by SDS-PAGE, as described above. Tissue lysates from three random wounds were pulled together for this experiment.

Cultured fibroblasts were grown on glass coverslips, fixed with 1% paraformaldehyde (PFA) for 15 min on ice and blocked with 2% BSA in PBS. For staining the plasma membrane, wheat germ agglutinin–Alexa 488 (Invitrogen, Eugene, OR) was added (1:2000 dilution) before fixing the cells. APN was detected using mouse anti-APN 3D8 antibody (Santa Cruz) followed by Alexa-488-conjugated goat anti-mouse IgG (Invitrogen, Eugene, OR). Biotin-labeled SFN was detected with Alexa-568-conjugated streptavidin (Invitrogen, Eugene, OR). To observe the APN protein expression in the normal and injured rabbit skin tissue, paraffin-embedded sections fixed in 2% PFA were incubated with a 1:50 dilution of 3D8 antibody against APN followed by goat anti-mouse Alexa 488 secondary antibody (1:2500 dilution). Samples were mounted in ProLong Gold antifade reagent with DAPI (Invitrogen, Eugene, OR) and examined by a Zeiss Axioplan-2 fluorescence microscope and Northern Eclipse image analysis software. A second tissue section was stained with haematoxylin and eosin (H&E) for histological analysis.

The quantitative assays for APN on cultured human dermal fibroblasts were performed according to previously published methods (Raynaud et al., 1992). In brief, all assays were performed in triplicate using Ala-4 nitroanilide (Ala-pNA, Sigma-Aldrich, Oakville, ON, Canada) in phosphate buffer pH 7.4 as a chromogenic substrate. Ala-pNA (100 μl) was added at a concentration of 5 mM to 10⁵ cells. Reactions were stopped by adding 900 μl acetate buffer (1.0 M, pH 4.4) after 60 min of incubation at 37°C. After centrifugation (2 min at 10,000 g), the absorbance of the supernatant was detected spectrophotometrically at 405 nm.

Data were expressed as means ± s.d. and analyzed with one-way analysis of variance with a Tukey–Kramer multiple comparison test among different groups where indicated. The comparisons of more than two groups with low sample size were carried out by nonparametric Kruskal–Wallis and Dunn's multiple comparisons tests. P-values less than 0.05 were considered statistically significant in this study. All experiments were repeated three times, unless indicated otherwise.

We are grateful for the generous support of Anthony Behrmann and staff, who provided skin tissue samples. We thank Suzanne Perry at the UBC MSL Proteomics Core Facility for her assistance with MS–MS peptide sequencing. We extend special thanks to Elham Rahmani and Claudia Chavez for their support with the rat animal study. This study was supported by a Canadian Institute of Health Research grant (CIHR-MOP-13387). A.G. holds a CIHR Canada Graduate Scholarships Doctoral Award (CGD-85362).