Cleavage of Signal Regulatory Protein α (SIRPα) Enhances Inflammatory Signaling*

Signal regulatory protein α (SIRPα) is a membrane glycoprotein immunoreceptor abundant in cells of monocyte lineage. SIRPα ligation by a broadly expressed transmembrane protein, CD47, results in phosphorylation of the cytoplasmic immunoreceptor tyrosine-based inhibitory motifs, resulting in the inhibition of NF-κB signaling in macrophages. Here we observed that proteolysis of SIRPα during inflammation is regulated by a disintegrin and metalloproteinase domain-containing protein 10 (ADAM10), resulting in the generation of a membrane-associated cleavage fragment in both THP-1 monocytes and human lung epithelia. We mapped a charge-dependent putative cleavage site near the membrane-proximal domain necessary for ADAM10-mediated cleavage. In addition, a secondary proteolytic cleavage within the membrane-associated SIRPα fragment by γ-secretase was identified. Ectopic expression of a SIRPα mutant plasmid encoding a proteolytically resistant form in HeLa cells inhibited activation of the NF-κB pathway and suppressed STAT1 phosphorylation in response to TNFα to a greater extent than expression of wild-type SIRPα. Conversely, overexpression of plasmids encoding the proteolytically cleaved SIRPα fragments in cells resulted in enhanced STAT-1 and NF-κB pathway activation. Thus, the data suggest that combinatorial actions of ADAM10 and γ-secretase on SIRPα cleavage promote inflammatory signaling.

Proteolysis and release of the SIRP␣ NH 2 -terminal domain has been demonstrated in primary cultured neurons, melanoma cells, and macrophage cell lines. Cleavage of murine SIRP␣ through an MMP inhibitor-sensitive pathway was first observed in cultured cells engineered to express both SIRP␣ and an active form of Ras (8). In these cells, blocking SIRP␣ proteolysis resulted in inhibited cell migration, cell spreading, and cytoskeletal reorganization. In addition, SIRP␣ was shown to regulate synaptic activity through activation of MMP inhibitor-sensitive proteolysis in primary murine neurons wherein the NH 2 -terminal domain of postsynaptic SIRP␣ promoted the maturation of the presynaptic terminal through a CD47-dependent mechanism (9). However, the identification, characterization, and regulation of SIRP␣ cleavage products in inflammatory signaling have not been investigated in myeloid cells. In addition to SIRP␣ expression in brain tissue and in myeloid cells, SIRP␣ is also detected at lower levels in heart, placenta, lung, testis, ovary, colon, liver, small intestine, prostate, spleen, kidney, skeletal muscle, and pancreas. The role of SIRP␣ in the pulmonary epithelium has not been explored and may also play a role in the resolution of inflammatory lung injury.
In these studies, we describe the role of proteolysis of SIRP␣ in response to proinflammatory stimuli. We determined that the inflammatory mediators lipopolysaccharide (LPS) and tumor necrosis factor-␣ (TNF␣) increase SIRP␣ proteolysis in a THP-1 human monocytic cell line and in human lung epithelial cells. We mapped a putative ADAM10 cleavage region within the extracellular juxtamembrane region and demonstrate a requirement for this enzyme for proteolysis of human SIRP␣. Finally, we discovered that after primary proteolysis the SIRP␣ membrane-associated fragment is rapidly cleaved by ␥-secretase where it is released into the cytosol. Cellular expression of plasmids encoding these SIRP␣-derived putative fragments was observed to modulate cellular inflammatory signaling, thus providing a physiological role for SIRP␣ sequential proteolysis in human cells.
Immunoblotting-Cell lysis and immunoblotting were performed as described previously (10). Briefly, cells were lysed in PBS supplemented with protease inhibitors (Thermo Scientific), and lysates were sonicated and then centrifuged at 500 ϫ g to remove debris. An aliquot of sample containing total cellular protein was frozen at Ϫ80°C. For subcellular fractionation, cells were also centrifuged at 16,000 ϫ g to pellet membranes. The supernatant containing the cytoplasmic protein was removed and stored at Ϫ80°C. The resultant crude membrane pellet was resuspended in radioimmune precipitation assay buffer supplemented with protease inhibitors, briefly sonicated, and stored at Ϫ80°C. Total cellular, soluble, and membrane proteins were loaded on SDS-polyacrylamide gels and processed for immunoblotting using the following antibodies SIRP␣ COOH-terminal antibody targeting amino acids 487-503 (566310, Millipore), SIRP␣ COOH-terminal antibody targeting residues surrounding Pro-413 (D613M, Cell Signaling Technology), V5 (Life Technologies), FLAG M2 antibody (Sigma), GAPDH (Sigma), STAT-1 (9172P, Cell Signaling Technology), pSTAT1 (7649P, Cell Signaling Technology), ADAM10 (AB19026, EMD Millipore), and ␤-actin mouse monoclonal antibody (Sigma). Immunoblots were exposed to SuperSignal West Femto chemiluminescent substrate (Thermo Scientific). Extracellular proteins were isolated from culture medium that was concentrated using an Amicon Ultra-4 centrifugal filter unit with Ultracel-30 membrane (EMD Millipore) before processing for immunoblotting.
Fluorescent Immunostaining-HeLa cells were inoculated into glass-bottomed 35-mm plates and transiently transfected with the indicated plasmids for 24 h. Cells were treated as indicated, washed with cold PBS twice, and fixed with 4% paraformaldehyde for 10 min prior to incubation of the fixed cells with staining solution (0.1% Triton X-100 in PBS with 1% goat serum) for 30 min. The cells were then probed with a SIRP␣ COOH-terminal monoclonal antibody (Cell Signaling Technology) (1:300) or V5 antibody (Invitrogen) (1:300) in staining solution overnight. Plates were washed three times and incubated with fluorescence-conjugated goat anti-rabbit secondary antibody (1:500) for 1 h. Plates were then washed three times for 10 min. DAPI was then added (1:5000) for 5 min. Images were acquired by a combination laser-scanning microscope system (Nikon A1, Nikon, Mellville, NY), and the results were analyzed using Nikon NIS-Elements software.
Bacterial Infection-Pseudomonas aeruginosa (PA103) were cultured as described previously (11). Briefly, inocula were freshly prepared prior to experiments from frozen stocks of PA103. Overnight plate cultures were then inoculated in tryptic soy broth supplemented with 1% glycerol and 100 mM sodium glutamate and grown by rotary shaking at 37°C to log phase. Cells were infected with PA103 as described.
Statistics-Student's t test was used for statistical analysis for comparison of two groups. The comparison of statistical significance among three or more groups was determined by oneway analysis of variance followed by pairwise comparisons using the Tukey t test.

SIRP␣ Is Cleaved by a Matrix Metalloproteinase in Response
to Inflammatory Mediators-To determine the half-life of SIRP␣, we incubated the human monocytic cell line THP-1 with the protein biosynthesis inhibitor CHX and measured protein lifespan; a t1 ⁄ 2 of ϳ2 h was indicated (Fig. 1a). Although we observed a decrease in SIRP␣ protein in the presence of CHX, this turnover did not appear to be affected by either a lysosomal inhibitor, leupeptin, or a proteasomal inhibitor, MG132 (Fig.  1a), suggesting that the full-length protein is processed by an alternative pathway. However, in these studies, we consistently detected the appearance of a faster migrating band at the expected size of a previously identified proteolytic fragment observed during MMP-mediated proteolysis of SIRP␣ that contains the cytoplasmic domain, the transmembrane domain, and a small portion of the extracellular domain. We termed this fragment SIRPcϩm (Fig. 1b) (8). This fragment was visualized with antibodies to two different SIRP␣ COOH-terminal epitopes. Interestingly, SIRPcϩm rapidly disappeared unless stabilized by the addition of MG132 in the culture medium and was visualized after prolonged exposure of immunoblots ( Fig. 1,  b and c). Leupeptin had no effect on SIRPcϩm stability (Fig. 1c). We isolated THP-1 total membrane and cytoplasmic fractions to determine the localization of the MG132-preserved SIRPcϩm fragment. SIRPcϩm was enriched in the total mem-brane fraction, suggesting that it remained membrane-associated and that proteolysis likely occurred in the extracellular membrane-proximal domain (Fig. 1d). This is consistent with prior observations in murine SIRP␣ (8). Because proteolysis of murine SIRP␣ is also enhanced by PMA stimulation (8), we incubated THP-1 with PMA and measured the degradation of full-length human SIRP␣ and the production of SIRPcϩm. Degradation of full-length SIRP␣ and the appearance of SIRPcϩm was dose-dependently increased by PMA (Fig. 1e). SIRP␣ proteolysis in RAW264.7 macrophages resulted in the production of a similar MG132-sensitive fragment (Fig. 1f). To determine whether SIRP␣ proteolysis plays a biologic role in macrophage signaling, we treated cells with the proinflammatory mediator LPS. In a time course analysis of SIRP␣ proteolysis in THP-1 cells, LPS treatment led to a decrease in abundance of full-length SIRP␣ at ϳ1 h post-treatment, an effect that was inhibited by treatment of cells with a broad spectrum MMP inhibitor, Compound 2 (12) (Fig. 2a). Although it is reasonable to predict from the LPS-induced disappearance of fulllength protein that SIRPcϩm was being generated, this fragment was not readily detected again unless stabilized by MG132 treatment (Fig. 2a, lower blots). In comparative analysis of proteolysis of SIRP␣, LPS and TNF␣ enhanced the accumulation of SIRPcϩm in THP-1 cells to a similar extent as actions of PMA with appearance of a fragment identical to the PMA-generated product (Fig. 2b). With each stimulus, we were unable to observe the production of SIRPcϩm without inclusion of MG132 in the culture medium to stabilize the fragment.
SIRP␣ Proteolysis Is Charge-dependent-To identify the site of proteolytic cleavage in human SIRP␣, we generated a series FIGURE 1. Proteolysis of SIRP␣ in monocytes. a, THP-1 cells were seeded in serum-free medium for 2 h and treated with CHX and either vehicle, leupeptin, or MG132. Cells were harvested at various times post-CHX treatment and probed using a SIRP␣ COOH-terminal antibody. b, THP-1 cells seeded in serum-free medium for 2 h followed by treatment with CHX. Full-length SIRP␣ (top) and a cleavage product, SIRPcϩm (middle), are shown after extended exposure. c, THP-1 cells seeded in serum-free medium for 2 h and treated with CHX 30 min after exposure to either vehicle, leupeptin (Leu), or MG132 (MG), and immunoblots were probed with antibodies as indicated. d, THP-1 cells were incubated with MG132, and total cellular lysates, membrane (Mem) and cytosolic (Cyto) proteins were isolated from cells and probed for SIRP␣. e, THP-1 cells were incubated for 4 h with vehicle or PMA at the indicated concentrations with or without MG132 to stabilize the cleaved fragment. Cell lysates were harvested and processed for immunoblotting as indicated. f, RAW264.7 macrophages were incubated with PMA for 4 h with or without inclusion of MG132 and probed using a SIRP␣ COOH-terminal antibody. Each panel is representative of two to three experiments.
of FLAG-tagged SIRP␣ mutants where we substituted amino acids in the full-length construct NH 2 -terminally adjacent to the transmembrane domain with the sequence DYDDDDK ( Fig. 3a). This sequence of amino acids, termed FLAG tag, is recognized by commercially available antibodies. We then measured the association of the FLAG domain using a specific antibody that will either associate with the secreted NH 2 -terminal domain or the membrane-bound COOH-terminal domain depending on the site of proteolysis. If the cleavage site is upstream of the FLAG domain resulting in the FLAG epitope that is attached to the SIRPcϩm domain, we should observe a reactive band at the size of the SIRPcϩm fragment when probing with either the FLAG antibody or the SIRP␣ COOH-terminal antibody generated against residues 487-503 (Fig. 3b, left). Alternatively, if the site of cleavage is downstream of the inserted FLAG domain, the FLAG epitope will be associated with the secreted NH 2 -terminal domain without reactivity with the FLAG antibody when probing cell lysates. However, in this latter scenario, a FLAG-reactive band should be detected in the medium at the predicted size of the SIRP␣ cleaved fragment (Fig. 3b, right). Because HeLa cells are readily transfectable, we first evaluated whether overexpressed wild-type (WT) SIRP␣ was subject to proteolysis by analyzing production of the SIRPcϩm fragment. As observed in THP-1, PMA enhanced the production of the SIRPcϩm. MG132 also increased the abundance of SIRPcϩm (Fig. 3c). HeLa cells were then transfected with the various FLAG-containing SIRP␣ plasmid constructs, and cells were then stimulated with PMA to activate proteolysis and MG132 to preserve the SIRPcϩm fragment, respectively. Cell lysates were then processed for immunoblotting using a COOH-terminal SIRP␣ antibody or an anti-FLAG antibody. We observed that expression of only the FLAG site 1 construct in cells resulted in a FLAG-reactive band in the SIRPcϩm fragment (Fig. 3d). Hence, these results suggest that SIRP␣ cleavage occurs between FLAG sites 1 and 2 ( Fig. 3, a and d). Due to the lack of an acceptable antibody against the hSIRP␣ NH 2 -terminal domain, we chose to characterize the proteolysis and release of the SIRP␣ NH 2 -terminal domain using our FLAG insertional mutants. After determining that FLAG site 3 was not associated with the SIRPcϩm domain and was cleaved normally, we transfected HeLa cells with a plasmid encoding FLAG site 3 and examined proteolysis. PMA treatment of cells increased the appearance of the SIRPcϩm fragment in cell lysates (Fig. 3e). We then confirmed the release of the SIRP␣ NH 2 terminus into the medium by probing for the FLAG domain in the supernatants. PMA was sufficient to trigger the release of the SIRP␣ NH 2 -terminal domain into the culture medium, an effect abrogated by MMP inhibition (Fig. 3e). Although MG132 significantly increased the stability of SIRPcϩm, it had little effect on the abundance of the extracellular fragment, suggesting that proteolysis itself is not altered by MG132. Thus, these data suggest that SIRP␣ is indeed clipped, resulting in the generation of both an extracellular released NH 2 -terminal fragment and a membrane-associated cleavage product. Having determined that the approximate site of proteolysis was between FLAG sites 1 and 2, we introduced a FLAG tag between these two regions (termed SIRP␣ FLAG site 1.5). Compared with expression of the FLAG site 1 and 2 constructs, cellular expression of the SIRP␣ FLAG site 1.5 variant dramatically inhibited both the appearance of SIRPcϩm and the extracellular release of the SIRP␣ NH 2 -terminal domain (Fig. 3f). Because alterations near the transmembrane domain could potentially disrupt protein folding or trafficking, we performed immunofluorescence imaging to characterize the membrane localization of FLAG site 1.5 after cellular expression. In HeLa cells, both the WT SIRP␣ and the FLAG site 1.5 mutant appear to associate with the plasma membrane to a similar extent (Fig. 3g).
To further ascertain the specific molecular attack sites for SIRP␣ cleavage, we generated a series of point mutations within the FLAG domain of SIRP␣ based on predictions of which motifs may be important. Substitution of Ala at residue 359 did not alter cutting, but substitution of Lys at residue 359 enhanced proteolysis. Substitution of Ser-358 and Thr-360 with negatively charged residues resulted in reduced SIRP␣ proteolysis (data not shown). Because of these results and because the FLAG tags are enriched with negatively charged residues, we hypothesized that the inhibition of proteolysis by the FLAG site 1.5 domain was charge-mediated. Therefore, we generated additional FLAG mutant constructs retaining only the negatively charged Asp residues (Fig. 4a). FLAG site 1.5, FLAG-D1, and FLAG-D2 constructs effectively blocked SIRP␣ proteolysis compared with the WT FLAG domain (Fig. 4b). In contrast, replacing the same residues with a series of basic positively charged residues (Arg) (FLAG-R) drastically enhanced proteol- ysis, resulting in enhanced accumulation of the SIRPcϩm product (Fig. 4b). Because the site of cleavage we identified was flanked by two negatively charged Glu residues, we hypothesized that a centrally placed negative residue should significantly limit SIRP␣ cleavage. Mutagenesis of Asn-359 to a negatively charged residue (Asp or Glu) greatly reduced SIRP␣ proteolysis under baseline and PMA-induced conditions (Fig.  4, c and d). In contrast, replacement of Asn-359 with a positively charged residue (Lys or Arg) enhanced proteolysis reflected primarily by loss of the parent SIRP␣ (Fig. 4, c and d). Mutation of this residue to an uncharged amino acid had no major effect on proteolysis (Fig. 4, c and d).
To further examine charge-dependent regulation of SIRP␣, we analyzed proteolysis of mSIRP␣ to determine whether similar residues were important for cleavage. We transfected HeLa cells with plasmid encoding mSIRP␣ and measured PMA-me-diated proteolysis. Like hSIRP␣, mSIRP␣ was cleaved in response to PMA treatment, resulting in the production of a COOH-terminal fragment (Fig. 4e). In these studies, this fragment was subsequently determined to be cleaved by ␥-secretase (see Fig. 6), and thus both MG132 and DAPT were used interchangeably as ␥-secretase inhibitors to preserve the cleaved fragment. Murine SIRP␣ contains a Gln residue (Gln-361) in a similar location to the juxtamembrane Asn residue (Asn-359) in hSIRP␣. Because mutation of the amide-containing amino acid Asn in hSIRP␣ drastically altered proteolysis, we made a series of point mutants near the Gln residue in mSIRP␣, speculating that this residue may also be important for proteolysis (Fig. 4a). Consistent with our observations in hSIRP␣, the introduction of negatively charged residues in place of Gln significantly inhibited mSIRP␣ proteolysis, whereas the substitution of a neutral amino acid, Ala, had no effect. The introduc- tion of three Asp residues was even more efficient at inhibiting proteolysis, suggesting a charge-dependent cutting mechanism. These results suggest that charge-mediated proteolysis is conserved across species despite major differences in the juxtamembrane domain sequence (Fig. 4, a and e).
ADAM10 Is Responsible for SIRP␣ Proteolysis-Although cleavage of SIRP␣ is inhibited by broad spectrum MMP inhibitors, the specific protease(s) responsible has not been identified. Ohnishi et al. (8) suggested that proteolysis of murine SIRP␣ might be MMP9-dependent. We therefore interrogated the role of MMP9 in THP-1 using a specific MMP2/9 inhibitor. Although the broad spectrum MMP inhibitor GM6001 (GM) blocked SIRP␣ degradation, the MMP2/9 inhibitor had no effect (Fig. 5a). PMSF, a serine protease inhibitor, also had no effect on proteolysis. We then tested whether Compound 2, an MMP8 inhibitor, could block proteolysis because it had been shown previously to inhibit the disappearance of membrane SIRP␣ in apoptotic neutrophils (13). Although Compound 2 was effective in inhibiting SIRP␣ cleavage, this antagonist was subsequently shown to be promiscuous, inhibiting many MMP and ADAM proteases (Fig. 5b) (12). Given the membraneproximal site and rapid onset of proteolysis, we hypothesized that the membrane-bound ADAM protease family member was likely responsible for SIRP␣ shedding (14,15). The specific ADAM10 inhibitor GI254023X (GI) blocked LPS-stimulated shedding by ϳ60% at 0.1 M and completely blocked cleavage of SIRP␣ at 1 M (Fig. 5b). To further examine the effects of the GI inhibitor on proteolysis, we overexpressed hSIRP and mSIRP in HeLa cells pretreated with GI and measured proteolysis in response to PMA. GI, like the broad spectrum MMP inhibitor GM, significantly reduced hSIRP and mSIRP proteolysis (Fig. 5c). Interestingly, GI was equally effective at inhibiting proteolysis in the hSIRP␣ mutants N359Q and N359A and in the mSIRP␣ mutants Q361N and Q361A, suggesting that these residues were not essential for ADAM10 specificity (data not shown). To directly examine whether endogenous ADAM10 was the protease responsible for SIRP␣ degradation, we transfected THP-1 cells with DsiRNA targeting ADAM10. ADAM10 knockdown appeared sufficient to prevent LPS-induced cutting of SIRP␣ as evidenced by increased full-length SIRP␣, suggesting protection from proteolysis (Fig. 5d). To more thoroughly characterize the role of ADAM10 in SIRP␣ degradation, we transfected HeLa cells with DsiRNA against ADAM10; this was followed by transfection of the SIRP␣ FLAG site 3 mutant (Fig.  3e), which allowed us to monitor the extracellular release of the SIRP␣ NH 2 -terminal domain. Knockdown of ADAM10 decreased both the release of the NH 2 -terminal SIRP␣ fragment and the appearance of SIRPcϩm (Fig. 5e).
The ADAM10 Cleaved Fragment Is Processed by ␥-Secretase-Thus far, the data support that ADAM10 proteolysis of SIRP␣ results in the appearance of a transient, membrane-associated cleaved fragment (SIRPcϩm) that was stabilized by MG132. However, when we further examined the stability of SIRPcϩm in the presence of MG132, we realized that this effect was only observed at higher concentrations, indicative of a proteasomeindependent effect of the inhibitor given its nonspecificity. These results were confirmed by the more specific proteasome inhibitor bortezomib that did not stabilize SIRPcϩm (Fig. 6a). MG132 has also been shown to inhibit calpains, cathepsins, and ␥-secretase activity in the micromolar range (16). We tested the ability of the calpain inhibitor PD150606 to prevent degradation of the SIRPcϩm and observed that it had no effect on preserving the stability of the SIRPcϩm fragment (Fig. 6b). ADAM10 proteolysis of substrates often leads to secondary cleavage by ␥-secretase (17). Because MG132 is known to inhibit ␥-secretase activity, we tested whether the ␥-secretase inhibitor DAPT could prevent the disappearance of SIRPcϩm. In contrast to calpain inhibitors, DAPT prevented the disappearance of SIRPcϩm (Fig. 6c).
To further understand cleavage of SIRP␣, we sought to characterize the fragment produced after ␥-secretase-mediated SIRPcϩm proteolysis. However, cytosolic fragments that are produced after ␥-secretase cleavage are very unstable under native conditions (18 -20). Because we also did not detect a ␥-secretase-cleaved SIRP␣ product, we sought to enhance the stabilization of the cytoplasmic SIRP␣ fragment with the proteasome inhibitor lactacystin. HeLa cells were transfected with WT SIRP␣ followed by treatment with lactacystin. The proteasome inhibitors bortezomib and lactacystin were used interchangeably throughout the experiments as these agents exerted similar effects. Proteasomal inhibition resulted in the accumulation of a small, low intensity band of the approximate pre-dicted size of SIRPcyto-C ( Fig. 6d; * indicates the SIRPcyto-C fragment). Furthermore, we did not observe the appearance of this fragment after treatment with DAPT, suggesting that ␥-secretase proteolysis is necessary for the release of the SIRPcyto-C. To better visualize the proteolysis of SIRPcϩm by ␥-secretase, we cloned a construct resembling the ADAM10cleaved SIRP␣ fragment (pSIRPcϩm). pSIRPcϩm associated with the membrane fraction in HeLa cells just as the WT SIRP␣ cleavage product did (Fig. 6e). pSIRPcϩm was subject to spontaneous proteolysis by ␥-secretase, leading to the release of the cytoplasmic domain (Fig. 6f). To determine whether SIRPcyto-C is produced by endogenously expressed SIRP␣, we treated THP-1 cells with the proteasomal inhibitors lactacystin (data not shown) and bortezomib. Both inhibitors led to the production of a DAPT-sensitive cleavage fragment of the predicted size (Fig. 6g). We also confirmed the generation of SIRPcyto-C in RAW267.3 murine macrophages (Fig. 6h). These data indicate that SIRP␣ proteolysis results in the secondary cleavage by ␥-secretase and the generation of a cytosolic COOH-terminal fragment.
SIRP␣ Is Expressed and Proteolytically Regulated in the Lung Epithelium-Although SIRP␣ signaling has been well described in myeloid cells and the brain, SIRP␣ activity has been detected in other cell types and tissues. Because the lung epithelium is uniquely exposed to airborne stimuli and pathogens that trigger inflammatory signaling, we examined SIRP␣ expression in these cells. First, we determined that SIRP␣ protein is expressed abundantly in primary normal human bronchial epithelial cells, the transformed bronchial epithelial cell line BEAS-2B, and the alveolar type II-like adenocarcinoma cell line A549 (Fig. 7a). Furthermore, we observed a small fragment the size of SIRPcϩm in one normal human bronchial epithelial sample isolated from three subjects, underscoring the biological relevance of SIRP␣ proteolysis (Fig. 7b). To confirm that SIRP␣ is proteolytically regulated in the human lung epithelium, we treated BEAS-2B cells with PMA to stimulate cleavage and with DAPT to preserve the SIRPcϩm fragment. PMA ϩ DAPT treatment led to the accumulation of a 25-kDa band that resembled SIRPcϩm. Furthermore, generation of this band was significantly inhibited with the broad spectrum MMP inhibitor GM6001 (Fig. 7c). To determine whether ␥-secretase cleavage produced the SIRPcyto-C in human bronchial epithelial cells, we pretreated BEAS-2B with the proteasome inhibitor bortezomib to stabilize the fragment. PMA treatment in the presence of bortezomib led to the accumulation of a 17-kDa band that did not appear in the presence of the ␥-secretase inhibitor DAPT (Fig. 7d). A key LPS-containing bacterial pathogen that interacts with lung epithelium to incite inflammation is P. aeruginosa. We therefore exposed BEAS-2B cells to a P. aeruginosa strain, PA103. Infection with PA103 significantly increased SIRP␣ proteolysis in the lung epithelium (Fig. 7e). Of note, we detected the appearance of SIRPcyto-C in PA103treated BEAS-2B cells despite the lack of proteasomal inhibition. Furthermore, in THP-1 cells infected with PA103, we confirmed that bacterial infection enhanced SIRP␣ proteolysis (Fig. 7f). To assess whether P. aeruginosa enhanced cleavage through ADAM10, we pretreated BEAS-2B cells with the ADAM10 inhibitor GI, the broad spectrum MMP inhibitor GM, or the MMP2/9 inhibitor prior to infection with PA103. Although both GM and GI inhibited P. aeruginosa-induced proteolysis, MMP2/9 inhibition had limited effect (Fig. 7g). Furthermore, knockdown of ADAM10 in BEAS-2B cells also significantly inhibited SIRP␣ proteolysis compared with control DsiRNA (Fig. 7h). These findings confirm that SIRP␣ is both expressed and subject to ADAM10-and ␥-secretase-mediated proteolysis in the lung and that this process is increased in response to inflammatory pathogens.
Proteolysis of SIRP␣ Enhances Inflammatory Signaling-To determine whether proteolysis modifies SIRP␣ signaling, we measured inflammatory signaling in HeLa cells transfected with SIRP␣ mutant constructs. The IB kinase complex (IKK) is an upstream NF-B regulator. Phosphorylated IKK␣ and IKK␤ result in dissociation of the inhibitory IB␣ protein from NF-B, which is then transported into the nucleus where it transcriptionally activates gene targets (21). HeLa cells were minimally responsive to LPS; however, treatment of HeLa cells with TNF␣ led to a transient induction of pIKK␣/␤ phosphor- ylation (Fig. 8a). Because SIRP␣ has been shown previously to modify the IKK signaling pathway (5,22,23), we examined whether SIRP␣ could modify TNF␣-induced IKK␣/␤ activity. We determined that TNF␣ treatment in HeLa cells led to a modest increase in SIRP␣ proteolysis that was significantly inhibited by the ADAM10 inhibitor GI (Fig. 8, b and c). Cellular expression of a plasmid encoding a proteolysis-resistant SIRP mutant (FLAG-1.5) resulted in significant protection of SIRP␣ from TNF-mediated cleavage (Fig. 8c) The FLAG 1.5 site SIRP␣ modestly inhibited IKK phosphorylation compared to WT SIRP␣. In contrast, expression of a plasmid encoding the SIRPcyto-C fragment enhanced IKK phosphorylation (Fig. 8d). In addition, another proteolysis-resistant mutant, FLAG-D2, also inhibited pIKK, whereas the highly cleaved receptor FLAG-R showed enhanced IKK phosphorylation (Fig. 8e). Because SIRP␣ has been implicated in JAK/STAT signaling, we next examined the role of SIRP␣ proteolysis in STAT1 signaling (7,(23)(24)(25). In HeLa cells, TNF␣ treatment enhanced STAT1 phosphorylation over time (Fig. 8f). Phosphorylation of STAT1 was markedly reduced in HeLa cells expressing proteoly-sis-resistant SIRP␣ (FLAG-1.5) (Fig. 8f). This effect appeared to a greater extent than in cells expressing WT SIRP␣, whereas expression of a plasmid encoding the SIRPcϩm proteolysis fragment enhanced STAT1 phosphorylation (Fig. 8g). Due to notable inhibition of STAT1 activation observed in HeLa cells, we examined whether SIRP␣ proteolysis in THP-1 cells was linked to STAT1 signaling. Pretreatment of THP-1 with the ADAM10 inhibitor GI254023X decreased the time-dependent phosphorylation of STAT1 in response to LPS, implicating a role for ADAM10-mediated proteolysis in the STAT1 activation pathway (Fig. 8h). To confirm these results, we examined the induction of pSTAT1 in BEAS-2B in response to PA103. PA103 resulted in enhanced pSTAT1 4 -6 h after infection that was significantly attenuated by pretreatment with GI (Fig. 8i). These experiments suggest that SIRP␣ proteolysis contributes to inflammatory signaling.

Discussion
The new contributions of this study include (i) that SIRP␣ proteolysis and generation of degradation products enhance inflammatory signaling, (ii) the identification of ADAM10 as the metalloproteinase responsible for human SIRP␣ proteolysis, (iii) that cleaved SIRP␣ is targeted for secondary proteolysis by ␥-secretase releasing a COOH-terminal fragment, and (iv) that SIRP␣ processing in this manner is widespread including in primary human lung epithelia, suggesting that SIRP␣ proteolysis may be biologically important. SIRP␣ proteolysis may modify numerous diseases where SIRP␣ signaling has been implicated including cancer (26), renal ischemia reperfusion injury (27), stroke (28), Crohn disease, (29), and allergic airway inflammation (30).
Based on the central inhibitory role of SIRP␣ in the inflammatory response in macrophages, we investigated SIRP␣ protein turnover in myeloid cells. Although we did not observe proteasomal or lysosomal degradation of the full-length protein, we did detect the appearance of a small fragment that resembled a previously described, MMP inhibitor-sensitive degradation product (Fig. 1b) (8). To determine the pathobiological role of inflammatory signaling linked to degradation of SIRP␣, we treated THP-1 cells with known proinflammatory stimuli, TNF␣ and LPS. Kong et al. (5) suggested that SIRP␣ disappearance in RAW264.7 macrophages in response to LPS was mediated by decreased transcription and lysosomal degradation of the immunoreceptor. In contrast, we observed that the LPS-mediated disappearance of SIRP␣ could be completely prevented by broad spectrum MMP inhibitors.
Although the proteolytic degradation of SIRP␣ had been described previously, the only evidence of SIRP␣ cleavage by specific proteases comes from in vitro degradation of purified murine SIRP␣ with recombinant MMP1 and MMP9 (8). However, there is no evidence that MMP9 cleavage actually occurs in intact cells. Our experiments suggest that MMP9 is not responsible for SIRP␣ degradation in either THP-1 (Fig. 4a), HeLa cells transfected with SIRP␣ (data not shown), or BEAS-2B because a specific MMP2/9 inhibitor did not prevent proteolysis. MMP8 was also characterized as a protease responsible for SIRP␣ proteolysis in primary neutrophils (13). However, this may not be the case because the inhibitor used in those experiments was shown to be promiscuous, inhibiting many MMPs and ADAM proteases (12,31). Using specific inhibitors and RNAi, we determined that ADAM10 was involved in SIRP␣ proteolysis in epithelial and monocytic cells. Prior work in neuronal cells provides a rationale that ADAM10 may be involved in the proteolysis of SIRP␣. Toth et al. (9) found that neural activity-dependent calcium entry followed by Ca 2ϩ /calmodulin-dependent protein kinase activation was necessary for SIRP␣ proteolysis in hippocampal cultures. ADAM10 is canonically activated by calcium influx, plays a role in the degradation of many neuronal substrates including amyloid precursor protein (32), and binds calmodulin, a sensor frequently demonstrated to be activated by the influx of calcium (33). Whether calcium flux modulates SIRP␣ in myeloid-derived cells requires further investigation.
Although the approximate site of SIRP␣ proteolysis in the murine immunoreceptor had been traced to the juxtamembrane region in murine SIRP␣, this region has very little sequence homology to human SIRP␣. Therefore, we performed a series of experiments where we inserted FLAG domain residues in the proximal extracellular membrane that allowed us to determine the approximate region of proteolysis in the human. Surprisingly, human SIRP␣ cleavage by ADAM10 appeared to be inhibited by the presence of negatively charged residues near FIGURE 8. SIRP␣ proteolysis modulates inflammatory signaling. a, HeLa cells were incubated with TNF␣ (10 ng/ml) for the indicated times. Cells were harvested and probed for pIKK␣/␤ and total IKK␣. b, HeLa cells were transfected with WT SIRP␣ and treated with TNF␣ for 6 h to measure SIRP␣ proteolysis. c, WT SIRP␣-and FLAG-1.5-transfected HeLa cells were treated TNF␣ in the presence or absence of GI. d, HeLa cells were transfected with an empty vector (EV), WT SIRP␣, protease-resistant SIRP␣ (FLAG-1.5), or the proteolysis mutant SIRPcϩm or SIRPcyto-C (Cyto-C). At 48 h post-transfection, cells were incubated with TNF␣ for 15 min. e, HeLa cells were transfected with an empty vector (EV), WT SIRP␣, protease-resistant SIRP␣ mutants (FLAG-1.5 and FLAG-D2), and the enhanced proteolysis mutant (FLAG-R). Cells were treated as described above. f, HeLa cells were transfected with an empty vector (EV), protease-resistant SIRP␣ (FLAG-1.5), or the proteolysis mutant SIRPcϩm (CϩM) and treated with TNF␣ for 4 and 8 h. Cells were then harvested and immunoblotted for pSTAT1 and total STAT1. g, HeLa cells were transfected with empty vector, WT SIRP␣, protease-resistant SIRP␣ (FLAG-1.5), or proteolysis mutant SIRPcϩm. At 48 h posttransfection, cells were incubated with TNF␣ for 6 h, harvested, and treated as above. h, THP-1 cells were preincubated with GI followed by treatment with LPS for the indicated times. Cells were harvested and probed for pSTAT1 and total STAT1. i, BEAS-2B cells were preincubated with GI followed by incubation with PA103 (multiplicity of infection, 20) for the indicated times. Cells were harvested and probed for pSTAT1 and total STAT1. Each panel is representative of two to three experiments. Veh, vehicle. the site of proteolysis, whereas positively charged residues enhanced cutting. This is the first report to our knowledge that suggests a charge-dependent mechanism for MMP-mediated proteolysis. Both human and murine SIRP␣ have a conserved Gln or Asn near the predicted cleavage site, indicating that amide-containing amino acids in this region might convey ADAM10 specificity. We therefore examined the proteolysis of murine SIRP␣ overexpressed in HeLa cells. Like hSIRP, mutation of mSIRP Gln to Asp significantly inhibited proteolysis (Fig. 4). However, in both hSIRP and mSIRP, mutation of these residues to Ala neither altered proteolysis nor resulted in decreased inhibition by an ADAM10 inhibitor. This suggests that amide-containing amino acids although centrally located in the cleavage site are unlikely to convey ADAM10 specificity in this system. Additional studies are necessary to determine the structural role of these residues in ADAM-mediated proteolysis.
The molecular interplay between ADAM10 and ␥-secretase in other systems led us to hypothesize that SIRP␣ might also be a target of ␥-secretase for sequential processing. This protease belongs to the family of intramembrane-cleaving proteases that target type I membrane proteins. After initial proteolysis, ␥-secretase recognizes truncated proteins as a substrate and cleaves the transmembrane domain within the lipid bilayer (17). Several proteins have been described as substrates of ␥-secretase cleavage including amyloid precursor protein, Notch, E-cadherin, ErbB4, and CD44 (17). DAPT, a specific ␥-secretase inhibitor, prevented disappearance of SIRPcϩm, confirming SIRP␣ as a substrate (35). In addition, we detected a DAPT-sensitive band at the predicted size of a ␥-secretasecleaved SIRPcϩm. Secretase cleavage of Notch, amyloid precursor protein, and other proteins leads to nuclear transport of the cytosolic fragment and alteration of gene transcriptional networks (17)(18)(19). We hypothesized that a similar mechanism may occur in our system. Interestingly, Shen et al. (25), who sought to identify binding partners of SIRP␣, found that the COOH-terminal domain specifically interacted with several nuclear transcription factors, lending support to our theory.
In addition to the well studied role of SIRP␣ in macrophage signaling, we examined whether SIRP␣ was expressed and proteolytically regulated in the airway epithelium. Yao et al. (27) showed that thrombospondin enhances reactive oxygen species production, leading to renal ischemia-reperfusion injury. SIRP␣ also protects against cardiac hypertrophy in neonatal rat cardiomyocytes and in murine cardiac tissue (36). Although initial experiments suggested that SIRP␣ is expressed in the lung, the specific cell types for expression of this immunoreceptor were unclear. Here we observed significant SIRP␣ expression in a bronchial epithelial cell line (BEAS-2B) and primary human bronchial epithelial cells as well as in an alveolar-like cell line (A549). SIRP␣ proteolysis was enhanced by bacterial infection in an ADAM10-dependent manner, leading to ␥-secretase-mediated secondary proteolysis. This confirms that the myeloid SIRP␣ and pulmonary SIRP␣ are similarly regulated and that cleavage of SIRP␣ may be a widely conserved mechanism in human cells for enhancing inflammatory signaling.
These results beg the question of whether phosphorylation of the SIRP␣ ITIM domain can alter SIRP␣ sensitivity to proteolysis. Although a number of stimuli lead to SIRP␣ phosphorylation, CD47 is the canonical ligand implicated in SIRP␣ signaling. Ohnishi et al. (8) suggested that ligation of SIRP␣ by CD47 had no effect on MMP-mediated proteolysis in CHO Ras mutant cells overexpressing SIRP␣. However, other SIRP␣ binding partners may also alter proteolysis. Thrombospondin and surfactant have both been shown to bind SIRP␣ (27). Surfactant D in particular has been shown to bind the region near the juxtamembrane region and inhibit phagocytosis and therefore is an especially attractive target (37)(38)(39).
Ligand binding to SIRP␣ has been shown to enhance phosphorylation of cytoplasmic ITIM domains through the activation of SRC kinases (4,40). ITIM phosphorylation recruits phosphatases such as SHP-1 and SHP-2, leading to the inhibition of NF-B signaling. SIRP␣ likely interferes with Toll-like receptor signaling specifically by binding SHP-2 and preventing its interaction with IKKs (5). In our experiments, proinflammatory stimuli such as LPS, TNF␣, and PA103 resulted in the ADAM10-mediated proteolysis of SIRP␣. We also observed that proteolysis-resistant SIRP␣ inhibited IKK␣/␤ more extensively than did WT SIRP␣ (Fig. 8). Interestingly, SHP-2 has been shown to be constitutively bound to SIRP␣, suggesting that proteolysis may liberate sequestered SHP-2 (5,7,41). In addition, studies by Shen et al. (25) identifying binding partners of SIRP␣ found that the COOH-terminal domain directly bound STAT1, suggesting that SIRP␣ proteolysis may lead to STAT1 release and nuclear trafficking. Our findings that the SIRP␣ COOH-terminal domain enhances inflammatory signaling is not without precedent. Neznanov et al. (43), by using an unbiased screen of a retroviral library of randomly fragmented cDNA from mouse fibroblasts, found that a cDNA encoding most of the cytoplasmic domain of SIRP␣ enhanced NF-B activation. We have determined that this fragment is produced in vivo in response to inflammatory signals through unique mechanisms (Fig. 9).
Interestingly, inflammatory signaling in neutrophils enhances the truncation of the SIRP␣ COOH terminus of neutrophils in a serine protease-and IL-17-dependent manner (42). Neutrophils with cleaved SIRP␣ display an enhanced proinflammatory phenotype in line with our findings, suggesting that SIRP␣ cleavage is mechanistically relevant and widespread. The authors hypothesized that after serine protease-mediated cleavage of a portion of the COOH-terminal domain the receptor would act as a decoy on the membrane, able to bind SIRP␣ ligands but unable to signal. Due to the loss of the entire extracellular domain, this mechanism is not at play in our system, but it is plausible, however, that liberation of SHP-2 or STAT1 could be relevant in this model as well through the shedding of COOH-terminally associated SHPs or STAT1. In addition, although serine protease cleavage of SIRP␣ is observed in neutrophils (42), MMP-mediated cleavage of this receptor in these cells also occurs (13).
In summary, we have demonstrated that SIRP␣ proteolysis enhances immune activation in response to inflammatory signaling (Fig. 9). Furthermore, we identified a protease that cleaves SIRP␣ in response to inflammation and the relevant residues necessary for proteolysis. Finally, ␥-secretase cleavage and COOH-terminal fragment release may offer an alternative signaling mechanism for this receptor. Collectively, these data suggest that the proteolytic regulation SIRP␣ plays an important role in the inflammatory pathway and that specific protease inhibitors may be able to inhibit inflammation.