Direct interaction between CD91 and C1q Karen Duus 1 , Erik W. Hansen 2 , Pascale Tacnet 3 , Philippe Frachet 3 , Gerard J. Arlaud 3 , Nicole M. Thielens 3 and Gunnar Houen 1 1 Department of Clinical Biochemistry and Immunology, Statens Serum Institut, Copenhagen, Denmark 2 Department of Pharmacology and Pharmacotherapy, Faculty of Pharmaceutical Sciences, University of Copenhagen, Denmark 3 Laboratoire d’Enzymologie Mole ´ culaire, Institut de Biologie Structurale Jean-Pierre Ebel, Grenoble, France Introduction The complement system is an important branch of innate immune defence and there are three major path- ways of complement activation that are currently recognized: (a) the classical (antibody-dependent) path- way; (b) the MBL (lectin) pathway; and (c) the alter- native pathway [1,2]. The classical pathway is activated upon antibody binding to target antigens and the recognition of immune complexes. C1q, the recogni- tion unit of C1, has an important role in immune com- plex clearance and complement activation. Binding of C1q to immune complexes is known to activate the C1q-associated proteases, C1r and C1s. Through cleavage of C4 and C2 by C1s, the C3- and C5-conver- tases are generated and the pore-like membrane attack complex (C5b-9) can be formed. In addition to its role in immune complex recognition, C1q has been shown to bind both necrotic and apoptotic cells and to play an important role in the scavenging of such cells [3–6]. Timely removal of apoptotic and necrotic cells is imperative to avoid initiation of autoimmune reactions [7] and C1q deficiency results in systemic lupus erythe- matosus [8]. C1q receptors have been suggested to play Keywords C1q; calreticulin; CD91; collectin; scavenger receptor Correspondence G. Houen, Department of Clinical Biochemistry and Immunology, Statens Serum Institut, Artillerivej 5, DK-2300 Copenhagen, Denmark Fax: +45 32683149 Tel: +45 32683276 E-mail: gh@ssi.dk (Received 18 November 2009, revised 27 May 2010, accepted 2 July 2010) doi:10.1111/j.1742-4658.2010.07762.x C1q-mediated removal of immune complexes and apoptotic cells plays an important role in tissue homeostasis and the prevention of autoimmune conditions. It has been suggested that C1q mediates phagocytosis of apop- totic cells through a receptor complex assembled from CD91 (a-2- macro- globulin receptor, or low-density lipoprotein receptor-related protein) and calreticulin, with CD91 being the transmembrane part and calreticulin act- ing as the C1q-binding molecule. In the present study, we observe that C1q binds cells from a CD91 expressing monocytic cell line as well as mono- cytes from human blood. C1q binding to monocytes was shown to be cor- related with CD91 expression and could be inhibited by the CD91 chaperone, receptor-associated protein. We also report data showing a direct interaction between CD91 and C1q. The interaction was investigated using various protein interaction assays. A direct interaction between puri- fied C1q and CD91 was observed both by ELISA and a surface plasmon resonance assay, with either C1q or CD91 immobilized. The interaction showed characteristics of specificity because it was time-dependent, satura- ble and could be inhibited by known ligands of both CD91 and C1q. The results obtained show for the first time that CD91 recognizes C1q directly. On the basis of these findings, we propose that CD91 is a receptor for C1q and that this multifunctional scavenger receptor uses a subset of its ligand- binding sites for clearance of C1q and C1q bound material. Abbreviations F-C1q, FITC-labelled C1q; FITC, fluorescein isothiocyanate; HSA, human serum albumin; LDL, low-density lipoprotein; LPS, lipopolysaccharide; MM6, Mono Mac 6; PBMC, peripheral blood mononuclear cell; PE, phycoerythrin; pNPP, para-nitrophenyl phosphate; RAP, receptor-associated protein; SAP, serum amyloid P; SPR, surface plasmon resonance. 3526 FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS an important role in the function of C1q and the prevention of autoimmune diseases [2,7,9] and several candidates have been described, including CD93 (C1qRp), CD35 (CR1), C1qRO 2 (CD59), a 2 b 1 inte- grin, gC1qR (p33), megalin and calreticulin (cC1qR) [9–16]. The most studied of the proposed receptors is calret- iculin. Calreticulin does not have a transmembrane domain and therefore is incapable of signalling for phagocytosis. Accordingly, calreticulin has been pro- posed to form a complex with CD91 on the surface of phagocytes [17–23]. On the other hand, CD91 of phag- ocytic cells has also been suggested to interact with calreticulin translocated from the endoplasmic reticu- lum to the cell surface of apoptotic cells [24,25]. CD91 is a member of the low-density lipoprotein (LDL) receptor gene family and comprises a receptor with more than 30 ligands [26–31]. Two chains of CD91 form a heterodimer: the transmembrane b-chain with a short cytoplasmic region and the extracellular a-chain with four ligand-binding clusters formed from 31 ligand-binding type repeats [26,32]. Of the haemato- poietic cells, only monocytes, their precursors and erythroblasts express CD91 [33]. In the present study, we investigated the direct inter- action of CD91 with C1q and show that CD91 itself recognizes C1q independently of calreticulin. Results C1q interaction with peripheral blood monocytes CD91 is a C1q receptor that is able to promote the ingestion of apoptotic material coated with C1q. We used human peripheral blood to confirm that the CD91 positive cells were C1q binding (Fig. 1A) and found that most of the C1q binding cells expressed CD91. CD91 has been reported to be present on mac- rophages and macrophage precursors [33] and this was confirmed by double staining of the cells with antibod- ies against CD91 and CD14. This revealed no popula- tion of single positive cells (Fig. 1B), thus indicating that CD14 positive monocytes express CD91. Some 45–70% of the CD14 positive cells bound C1q with some person-to-person variability (Fig. 1C). C1q bind- ing to other cell types was also observed (Fig. 1A) and Fig. 1. C1q interaction with peripheral blood monocytes. (A) Blood cells were stained with PE anti-CD91 (x-axis) and F-C1q (y-axis). The red blood cells were lyzed and remaining cells were analyzed by flow cytometry. Only monocytic cells gated by forward scatter and side scatter is shown. (B) Peripheral blood monocytes are highly CD91 positive. Blood cells were stained with flourophore- conjugated antibodies recognizing CD14 and CD91 (left) or an iso- type control (right). Red blood cells were lyzed and remaining cells were analyzed by flow cytometry. (C) C1q binding levels for three individuals. Blood cells were stained with F-C1q and PE anti-CD14. The red blood cells were lyzed and remaining cells were analyzed by flow cytometry. C1q binding levels of white blood cells are pre- sented on the x-axis with CD14 positivity on the y-axis. (D) PBMCs from human blood were stained with F-C1q with the addition of the CD91 blocking protein RAP (100-fold molar excess), soluble CD91 (10-fold molar excess) or the control protein HSA (100-fold molar excess). After staining, C1q-positive cells were quantified by flow cytometry. F-C1q stained cells without inhibition were set as 100%. K. Duus et al. Direct interaction between CD91 and C1q FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS 3527 therefore inhibition experiments were conducted with isolated peripheral blood mononuclear cells (PBMCs). These experiments showed that partial inhibition of the C1q–monocyte interaction could be obtained with soluble CD91 and the CD91 blocker, receptor-associ- ated protein (RAP) (Fig. 1D). Similar results were obtained by Lillis et al. [34], where only some phago- cytotic inhibition was observed with monocytes defi- cient for CD91. These results indicate that scavenging of C1q complexes relies on a highly redundant recep- tor system and that several receptors for C1q are likely to exist. Interaction of C1q with CD91 expressing cells The monocytic cell line Mono Mac 6 (MM6) was used to examine C1q interaction with monocytic cells. The cell line has previously been described to express CD91 [35]. However, in the present study, only a low level of CD91 was identified on the surface of MM6 cells (approximately 7% of cells were CD91 positive; data not shown). To increase surface expression of CD91, the cells were stimulated with lipopolysaccharide (LPS). As measured by flow cytometry, approximately 40% of the cells were found to be CD91 positive after LPS stimulation (Fig. 2A), and the mean fluorescent inten- sity of the cells stained with fluorescently-labelled CD91 antibody was observed to increase. Some 85% of the CD91 positive cells were highly CD14 positive (data not shown). When the stimulated MM6 cells were stained with fluorescein isothiocyanate (FITC)-labelled C1q (F-C1q), 1.2–9% of the cell population showed high fluorescence intensity, indicating C1q interaction. The percentage of positive cells could be inhibited by RAP and unlabelled C1q (Fig. 2B and results not shown). We used confocal scanning laser microscopy to visualize localization at the cell surface. CD91 antibody bound MM6 at the cell surface and some (but not complete) co-localization was observed between the added F-C1q and CD91 expressed by the cells (Fig. 2C), indicating that other receptor systems are likely to exist. Staining of the nuclei revealed no signs of apoptosis. Cell size Cell size PE anti-isotype control PE anti-CD91 0 FSC-H 45.52% 6.13% 1023 10 0 10 1 10 2 10 3 FL2-H 10 4 0 FSC-H 1023 10 0 10 1 10 2 10 3 FL2-H 10 4 2 1 Relative binding 0 300300 RAP/C1q molar ratio (i) (ii) (iii) A B C Fig. 2. CD91 expression and C1q binding on Mono Mac 6 cells. (A) Flow cytometry of stimulated MM6 cells showing CD91 expression. After LPS stimulation, CD91 expression was measured on MM6 cells by the addition of PE-labelled CD91 antibody (left) or an isotype control (right) and fluores- cence was measured by flow cytometry. Approximately 40% of the LPS-stimulated cells were determined to be CD91 positive (45.52–6.13% = 39.39%). (B) C1q interac- tion with stimulated MM6 cells and RAP inhibition. MM6 cells were stained with F-C1q with or without the addition of RAP. RAP showed inhibition of the C1q interac- tion in 30- and 300-fold molar excess. The results are presented with the relative bind- ing with F-C1q without the addition of RAP set as 1. (C) CD91 and C1q localization in MM6 cells analyzed by confocal laser microscopy. Cells were stained with biotiny- lated antibodies against CD91 and streptavi- din coupled with Alexa Fluor 546 (red colour, i) and the addition of F-C1q (green, ii). Images i and ii are merged to produce an overlay plot (iii). Scale bar = 10 l m. Direct interaction between CD91 and C1q K. Duus et al. 3528 FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS Evidence for a direct interaction between CD91 and C1q Evidence for a direct interaction was obtained by surface plasmon resonance (SPR) analysis. Using C1q as the soluble analyte and CD91 as the immobilized ligand, several concentrations of C1q were used to reveal a time- and concentration-dependent interaction (Fig. 3A). Furthermore, in the reversed configuration with C1q immobilized and CD91 as the soluble analyte, an interaction was observed (Fig. 3B). Using an ELISA binding assay, CD91 showed a direct interaction with C1q coated on a polystyrene surface (Fig. 4A). At a physiological salt concentration (0.15 m NaCl), the interaction was time-dependent, saturable and fast, being detectable after only a few minutes of incubation. By contrast, increasing the NaCl concentration to 0.65 m strongly inhibited the interaction (Fig. 4A). In the opposite configuration, with CD91 immobilized, sol- uble C1q also bound and could be detected with C1q specific antibodies (Fig. 4B) although some nonspecific interaction was observed between the polyclonal anti- body directed against C1q and CD91. The C1q–CD91 interaction was also observed when C1q was bound to immobilized IgM, indicating that CD91 binding is possi- ble with ligand-bound C1q (Fig. 4C). Kinetic analysis of the SPR binding data, showed that the best fit for the interaction between immobi- lized CD91 and soluble C1q was provided by a two- state reaction model (Table 1), suggesting that several interaction phases take place. This was supported by the ELISA, which revealed that binding of soluble CD91 to C1q coated on the polystyrene surface was highly sensitive to 0.65 m NaCl in the early phase of the interaction (Fig. 4A). By contrast, if CD91 and C1q were allowed to interact under physiological salt concentrations before the addition of 0.5 m NaCl, then the interaction gradually became insensitive to high salt (Fig. 4D). This was suggestive of a two-state inter- action with an initial ionic binding step allowing subse- quent stable interaction. Interaction of CD91 with the globular domain of C1q and its collagen fragment was also analyzed by SPR analysis, using the latter proteins as soluble ligands. In the case of the globular domain, the two- state binding model also provided the best fit. More- over, the affinity of the interaction calculated by a Langmuir 1:1 model was approximately 20-fold lower compared to intact C1q (K D = 3.0 · 10 )7 m). The C1q collagen fragment also bound immobilized CD91, and the reaction followed a classical 1 : 1 Langmuir model, with a K D in the nanomolar range (Table 1). 0 100 200 0 40 80 120 Response difference (RU) Time (s) 50 nM 40 nM 30 nM 20 nM 10 nM 5 nM 2.5 nM Response difference (RU) 100 –100 0 0 100 200 Time (s) 50 nM 40 nM 30 nM 20 nM A B Fig. 3. Interaction of CD91 with C1q, demonstrated by SPR analy- sis. (A) CD91 was immobilized by coupling of amino groups to a carboxylated surface. C1q was injected for 120 s at concentrations in the range 2.5–50 n M. (B) CD91–C1q interaction. C1q was immo- bilized by amine coupling to a carboxylated surface and soluble CD91 was injected for 120 s at the indicated concentration in the range 20–50 n M. Table 1. Comparison of the kinetic constants of the CD91 interaction between C1q, C1q globular heads and C1q collagen tails. Ligand Model with best fit K a (K a1 , K a2 ) K d (K d1 , K d2 ) K D , Langmuir 1 : 1 model Chi squared C1q Two-state reaction with conformational change 4.04 · 10 5 M )1 s )1 6.61 · 10 )10 S )1 5.09 · 10 )3 s )1 0.464 s )1 – 6.14 C1q globular region Two-state reaction with conformational change 1.94 · 10 4 M )1 s )1 3.28 · 10 )9 S )1 5.97 · 10 )3 s )1 1.87 · 10 )7 s )1 – 5.93 C1q collagen region Langmuir 1 : 1 5.13 · 10 5 M )1 s )1 2.04 · 10 )3 s )1 3.95 · 10 )9 M 5.89 K. Duus et al. Direct interaction between CD91 and C1q FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS 3529 Several known interaction partners of CD91 and C1q inhibit their interaction To characterize the CD91–C1q interaction in more detail, the ability of known ligands of either protein to interfere with binding was investigated. RAP is a CD91 chaperone that has previously been shown to block the interaction of CD91 with all its known ligands [36,37]. In agreement with this characteristic, RAP also strongly inhibited binding of CD91 to immobilized C1q (Fig. 5A). a-2-macroglobulin and Pseudomonas exotoxin A are known ligands of CD91 and they also exerted some inhibition (15% and 40% inhibition, respectively) (data not shown). Calreticulin, a protein known to act as a ligand for the collagen region of C1q and the proposed C1q-binding compo- nent of the CD91 ⁄ calreticulin receptor complex, signifi- cantly inhibited binding of CD91 to immobilized C1q in the presence of 5 mm Ca 2+ ions, with almost 45% inhibition at a 300 : 1 calreticulin ⁄ CD91 molar ratio (Fig. 5B). Inhibition was also observed in the absence of Ca 2+ , although to a lesser extent. Serum amyloid P (SAP), another known ligand of C1q [38,39], abolished binding of CD91 to immobi- lized C1q in the presence of 5 mm Ca 2+ but had only a slight inhibitory effect in the absence of Ca 2+ (Fig. 5C). A significant inhibition of the interaction by the sulphated polysaccharide fucoidan was also observed, with approximately 65% inhibition at a 600 : 1 fucoidan ⁄ CD91 molar excess (Fig. 5D). The physiological partner proteases of C1q, C1r and C1s were also tested as inhibitors, with inhibition observed only for C1r (Fig. 5E). Inhibition was also observed with the globular region of C1q, confirming the results observed by the SPR data (results not shown). Taken together, these inhibition experiments indicate that C1q occupy known ligand-binding sites on CD91 and that CD91 presumably interacts with C1q at 0 1 2 3 4 1500 1000 5000 Time (min) Physiological NaCl (0.15 M ) 0.5 M additional NaCl (0.65 M ) CD91 binding (A 405 ) A 0 0.2 0.4 0.6 0.8 1 IgG (agg)IgMIgG b-CD91 C1q + b-CD91 CD91 binding (A 405 ) C 0 1 2 3 4 150 100 50 0 Time without NaCl (min) CD91 binding (A 405 ) D 0 1 2 3 CD91-C1q CD91-BSA CD91 Anti-C1q Anti-isotype control Antibody binding (A 405 ) B Fig. 4. Interaction between CD91 and C1q, demonstrated by ELISA. (A) Time dependence of the association between CD91 and C1q. Biotinylated CD91 (b-CD91) was added to microtitre plates coated with C1q. b-CD91 was allowed to incubate for the indicated time (from 10 min to 20 h) with a NaCl concentration of either 0.15 or 0.65 M. Bound b-CD91 was quantified by incubation with alkaline phosphatase-conjugated streptavidin and pNPP. (B) CD91 was immobilized on a microtitre plate and C1q or the control protein bovine serum albumin was allowed to interact. Bound C1q was detected with rabbit antibodies recognizing C1q and alkaline phos- phatase-conjugated secondary antibodies. Data are presented as the mean ± SD of two individual wells for one representative experiment. (C) A microtitre plate was coated with the indicated protein; either IgG, IgM or IgG aggregated [IgG(agg)] by incubation at 60 °C for 20 min, b-CD91 was either added directly (grey bars) or after a layer of C1q (white bars) and bound CD91 was detected by incubation with alkaline phosphatase-conjugated streptavidin and pNPP. (D) The time-dependent influence of 0.5 M NaCl on CD91 binding to C1q. b-CD91 was added to a microtitre plate coated with C1q. After the indicated amounts of time, 0.5 M NaCl was added and the incubation continued for a total incubation time of 2 h, before the bound b-CD91 was quantified by incubation with alkaline phosphatase-conjugated streptavidin and pNPP. Direct interaction between CD91 and C1q K. Duus et al. 3530 FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS several sites, with one of them being at or near the C1r attachment site. Calreticulin was only able to inhi- bit the interaction to 45%, despite the use of a very large excess. Considering that calreticulin is known to interact strongly with C1q, this suggests that only a part of the binding sites on C1q are shared by calreti- culin and CD91. No interaction between CD91 and calreticulin could be observed, regardless of which of them was immobilized (results not shown). Discussion It is well established that C1q initiates the clearance of apoptotic material, although the identification of direct C1q-receptors has remained elusive. CD91 has previ- ously been described to function as a receptor complex together with calreticulin. In this receptor complex, calreticulin is suggested to act as the C1q-recognition molecule and CD91 as the phagocytic, transmembrane molecule. To determine whether C1q binding could occur on human blood cells, we first confirmed the high CD91 level on human blood monocytes. We then analyzed C1q binding to human blood cells, confirming a mono- cyte interaction. The interaction between C1q and PBMCs could be partially inhibited by RAP and solu- ble CD91. The interaction between C1q and CD91 was also investigated using the monocytic cell line MM6 known to present CD91 at the cell surface. Fluorescent-labelled C1q was found to bind these cells in a manner overlapping the areas of CD91 expression. The partial co-localization suggests that other C1q receptors exist. This was confirmed by the inhibition experiments where the interaction between C1q and MM6 cells could be partially inhibited by the addition of RAP, suggesting that only part of the C1q interac- tion occurs specifically through CD91 or other mem- bers of the LDL receptor superfamily. CD91 and calreticulin have previously been described to function as a C1q receptor complex with calreticulin as the recognition unit of the complex. Despite these reports, in the present study, we provide evidence that CD91 directly recognizes C1q indepen- 0 1 2 3 4 100100 RAP/CD91 molar ratio RAP Ovalbumin CD91 binding (A 405 ) 0 1 2 3 4 0 30 300 Calreticulin/CD91 molar ratio Without calcium 5 m M calcium CD91 binding (A 405 )CD91 binding (A 405 ) 0 1 2 3 4 600600 SAP/CD91 molar ratio Without calcium 5 m M calcium 0 1 2 3 600600 Fucoidan/CD91 molar ratio CD91 binding (A 405 ) 0 0.5 1 1.5 500 C1s C1r BSA CD91 binding (A 405 ) CD91/serine protease molar ratio A B C D E Fig. 5. Inhibition of the CD91–C1q interaction. Biotinylated CD91 was added to microtitre plates coated with C1q, together with the indicated molar excess of inhibitor. (A) RAP and the control protein ovalbumin; (B) calreticulin with and without the addition of 5 m M calcium; (C) SAP with and without 5 mM calcium; (D) fucoidan; and (E) C1r and C1s. After 2 h of incubation, the amount of bound bioti- nylated CD91 was quantified by incubation with alkaline phospha- tase-conjugated streptavidin and pNPP. Data are the mean ± SD of two individual wells for one representative experiment. K. Duus et al. Direct interaction between CD91 and C1q FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS 3531 dently of calreticulin. The CD91–C1q interaction showed all of the signs for being specific because the interaction was time- and concentration-dependent. Furthermore, the interaction was inhibited by several known ligands of either protein. Interaction was observed both with commercially available C1q ⁄ CD91 and purified C1q ⁄ CD91. In several assays, the interac- tion showed signs of two-state binding. First, the inter- action appeared to be salt sensitive at the early stages but became salt insensitive at later stages. Second, when the interaction was investigated using SPR, a two-state binding was indicated and, third, an inter- action was observed both for the globular region and with the collagen tail of C1q. A likely hypothesis is that CD91 binds C1q first through its collagen tail in a salt-sensitive manner and then mediates a salt-insensi- tive interaction with the globular head or vice versa. CD91 contains four ligand-binding clusters of which RAP is known to block at least three [36,40]. RAP is specific for the receptors of the ‘LDL receptor super- family’ and blocks all tested ligands of CD91 [36]. RAP completely inhibited binding between CD91 and C1q, indicating that the interaction occurs through one of more of these ligand-binding clusters. Some inhibition was also observed with the CD91 ligands a-2-macroglobulin and exotoxin A. These ligands occupy the CD91 ligand-binding cluster II and IV, thereby suggesting that C1q might interact with one or both of these sites on CD91 [41,42]. Several interaction partners of C1q were also tested for their ability to inhibit the interaction between CD91 and C1q. Fucoidan yielded inhibition of CD91 binding to C1q. Fucoidan was reported to bind C1q at a site that abolishes attachment of the serine proteases [43]. It is therefore possible that CD91 interacts with C1q at a site near or at the serine protease binding site. This theory was supported by inhibition as a result of the serine protease C1r, where a 50-fold excess inhibited the interaction between CD91 and C1q by approximately 40%. C1r binds C1q at the col- lagen tail at a site where the collagen tail is trimeric without obtaining higher-order oligomeric forms [44,45]. Complete inhibition was also observed with SAP. SAP is known to bind C1q at sites on both the collagen stalk and the globular region [38,39] and it is therefore possible that CD91 interacts at one or both of these sites. The possibility that both the globular region and the collagen region could be involved in this interaction was strengthened by SPR data, where either region showed an interaction with CD91. Taken together, these inhibition experiments imply that the CD91–C1q interaction occurs at one or more of the ligand-binding clusters of CD91. On the C1q molecule, the interaction site may be near or at the serine prote- ase attachment site, although several sites on the C1q molecule are assumed to be involved. Calreticulin is known to bind C1q and is a reported co-receptor of CD91. Only 45% inhibition was obtained in the case where calreticulin was used as an inhibitor of the C1q–CD91 interaction. The calreticulin inhibition indicates that only a part of the binding sites on C1q is shared between CD91 and calreticulin. No binding was detected between CD91 and calreticulin in the ELISA, regardless of which of them was immobi- lized or in solution. These results are somewhat intriguing because calreticulin has been reported to function as a co-receptor for CD91 in the binding of C1q and collectins [20,22,23]. However, Donnelly et al. [46] also were unable to detect distinct binding between calreticulin and CD91 and Walters and Berwin [47] reported that the uptake of calreticulin is different from another CD91 ligand, thereby suggesting that calreticulin is not a traditional ligand of CD91. In conclusion, the data obtained in the present study support the hypothesis that CD91 is a receptor for C1q and presumably detects and internalizes ligand bound C1q in the same way that it internalizes other ligands. We therefore conclude that CD91 itself is a receptor for C1q. On the basis of these findings, we propose a binding model for CD91 scavenging of C1q that is independent of calreticulin (Fig. 6). Materials and methods Proteins and chemicals Ovalbumin, bovine serum albumin, p-nitrophenyl-phos- phate (pNPP) substrate tablets, FITC, N-hydroxy-succinim- idobiotin, a-2-macroglobulin, exotoxin A, fucoidan, LPS (Pseudomonas aeruginosa) and C1q were obtained from Sigma (St Louis, MO, USA). Alkaline phosphatase-conju- gated streptavidin and IgM were obtained from Dako (Glostrup, Denmark). MaxiSorp microtitre plates were obtained from Nunc (Roskilde, Denmark). CD91 (contain- ing RAP) and biotin-labelled CD91 antibody were obtained from BioMac (Chamalie ` res, France). RAP was obtained from Innovative Research (Southfield, MI, USA). SAP was a gift from N. H. H. Heega ˚ rd (Statens Serum Institut, Copenhagen, Denmark). Adiponectin (trimeric form) was obtained from Abcam (Cambridge, UK). Human IgG and human serum albumin (HSA) were obtained from SSI (Copenhagen, Denmark). CM-5 sensorchips, surfactant P20, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide, N-hydro- xysuccinimide and ethanolamine were obtained from GE healthcare, BIAcore (Uppsala, Sweden). Excell medium was obtained from Safc Biosciences (Hampshire, UK). Direct interaction between CD91 and C1q K. Duus et al. 3532 FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS Penicillin, streptomycin and glutamine were obtained from Gibco-BRL (NY, USA). Flowbuffer, phycoerytrin (PE) mouse anti-human CD91, PE mouse anti-human CD14, FITC mouse anti-human CD91, FITC mouse anti-human CD14 and corresponding isotype control antibodies were obtained from BD Biosciences (Franklin Lakes, NJ, USA). C1r and C1s were obtained from R&D Systems (Abingdon, UK). Rabbit anti human C1q polyclonal antibody was obtained from Dako. High-yield Lyse and streptavidin labelled with Alexa Fluor 546 were obtained from Invitro- gen (Carlsbad, CA, USA). Vectashield mounting medium was from Vector Laboratories (Burlingame, CA, USA). Purification of human placenta calreticulin Human placenta calreticulin was purified using a previously described procedure [48,49]. The purified protein showed a single band with an apparent molecular weight of 60 kDa by SDS ⁄ PAGE and a single peak by MALDI-TOF ⁄ TOF mass spectrometry. Purification of C1q and C1q-derived fragments C1q was purified from human plasma as described previ- ously [50]. The collagen-like fragments of C1q were obtained by pepsin digestion and purified as described pre- viously [51]. The fragments corresponding to the C1q glob- ular domains were generated by treatment of C1q with collagenase and purified by high-pressure gel filtration chro- matography as described previously [51]. Purification of CD91 Commercially obtained CD91 (BioMac) was used when not stated otherwise. CD91 containing no RAP was a gift from S. K. Moestrup (Institute of Medical Biochemistry, Univer- sity of A ˚ rhus, A ˚ rhus, Denmark), purified as described pre- viously [52], and was used to confirm the C1q interaction. Protein biotinylation The proteins subjected to biotinylation were dialysed against 0.1 m NaHCO 3 (pH 9.0) at 4 °C, followed by the addition of N-hydroxysuccinimidobiotin in dimethyl sulphoxide (10 mgÆmL )1 ) to a final concentration of 4 mgÆ mg )1 CD91. The solution was incubated for 2 h at room temperature with end-over-end agitation, and then dialysed against NaCl ⁄ P i (10 mm NaH 2 PO 4 ⁄ Na 2 HPO 4 , pH 7.3, 0.15 m NaCl) at 4 °C. Biotinylated CD91 was stored at )20 °C until use. FITC labelling Commercially available C1q and HSA were dialysed against 50 mm sodium carbonate (pH 9.5). FITC (100 lgÆ mg )1 pro- tein) was added and the reaction mixture was incubated with end-over-end agitation for 1 h in the dark. Excess fluo- rescein was removed by dialysis against NaCl ⁄ P i at 4 °C. ELISA Unless otherwise stated, incubations and washings were performed at room temperature on a shaking table using 100 lL per well for incubation and 200 lL per well for washing and blocking. TTN buffer (0.025 m Tris–HCl, 0.5% Tween 20, 0.15 m NaCl, pH 7.5) was used for block- ing, incubation and washing. Unless otherwise stated, the C1q used was from Sigma (St Louis, MO, USA). Proteins were coated at 1 lgÆmL )1 onto the surface of the microtitre plates using 0.05 m sodium carbonate (pH 9.6) as coating buffer. After coating overnight at 4 °C, plates were washed three times for 1 min, followed by blocking for 30 min. Subsequently, incubation with or without biotinylated CD91 diluted to 1 lgÆmL )1 was carried out for 2 h, followed by another three washes. Finally, the plates were incubated for 1 h with alkaline phosphatase-conjugated streptavidin diluted 1 : 1000. After another three washes, bound CD91 was quantified using pNPP (1 mgÆmL )1 )in1m diethanolamine, 0.5 mm MgCl 2 (pH 9.8). A 405 was read with background subtraction at Epidermal growth factor-type repeat Ligand-binding repeat Beta-propeller C1q Ligand bound C1q CD91 Apoptotic material Phagocyte Fig. 6. Proposed model for CD91 interaction with C1q. CD91 is present on phagocytes and consists of two noncovalently bound polypeptide chains; with the 85 kDa b-chain as the transmembrane and the a-chain of 515 kDa with four ligand-binding clusters (blue areas). Ligand interaction occurs through 31 similar ligand-binding repeats (blue squares) distributed unequally between the four ligand-binding clusters. The CD91 a-chain also consists of epider- mal growth factor-type repeats (grey circles) and b-propellers (black stars). Interaction occurs at several sites on C1q and possibly through one or more ligand-binding domains at CD91. CD91 is a phagocytic receptor and is likely to internalize C1q bound material. K. Duus et al. Direct interaction between CD91 and C1q FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS 3533 650 nm on a VERSAmax microplate reader (Molecular Devices, Sunnyvale, CA, USA). All experiments were per- formed at least twice and the results are presented as the mean ± SD of two wells for one representative experiment. Inhibition ELISA Inhibition of the CD91–C1q interaction was tested by mix- ing biotinylated CD91, diluted to a concentration of 1 lgÆmL )1 , with varying amounts of the indicated inhibitor for 1 h. The mixture was then added for 2 h to plates coated overnight with C1q. Subsequently, the plates were washed, incubated with alkaline phosphatase-conjugated streptavidin and developed as described above. SPR experiments CD91 and C1q were immobilized on a BIAcore CM5 sensor- chip in NaCl ⁄ Hepes-P20 (10 mm Hepes, 0.15 m NaCl, 3.4 mm EDTA, 0.005% surfactant P20) in accordance with the manufacturer’s instructions, resulting in an immobiliza- tion level of 13 000 and 17 000 relative units, respectively. Interaction between immobilized CD91 and C1q, the C1q globular domain and collagen fragment was investigated on a BIAcore 3000 instrument using NaCl ⁄ Tris (10 mm Tris–HCl, 0.15 m NaCl, pH 7.4) as running buffer. Inter- action between immobilized C1q and soluble CD91 was investigated on a BIAcore 1000 instrument with the same running buffer. The analyte was injected at a flow rate of 20 lLÆmin )1 for 120 s. Background was subtracted from a quenched (activated ⁄ deactivated) surface showing an absence of nonspecific binding. The surface was fully regen- erated using low concentrations of NaOH. Kinetics were determined using the software biaevaluation, version 3.1 (BIAcore). Two independent experiments were performed with all concentrations. Data are reported for one representative experiment. Culture of MM6 cells MM6 was used as a monocytic cell line. Cells were cultured at 37 °C under 5% CO 2 in Excell medium (to avoid serum) with the addition of 1% glutamine and 1% penstrep. These cells were stimulated twice with 10 ngÆmL )1 LPS with a 3-day interval and harvested on day 6 or 7. Flow cytometry analysis of MM6 cells A sample containing 10 5 cells was centrifuged to remove the culture medium and the cells were washed with NaCl ⁄ P i . The cell pellet was resuspended in 10 lLof NaCl ⁄ P i and stained by adding 2 lL of F-C1q or FITC- labelled HSA for 20 min at room temperature. Staining with CD91 antibody, CD14 antibody or isotype control antibodies was performed in accordance with the manufac- turer’s instructions. After staining, cells were washed with NaCl ⁄ P i and resuspended in 1 mL of the buffer. Flow cytometry was performed using a Calibur instrument (BD Biosciences), counting 10 000 cells and using cellquest (BD Biosciences) and winmdi (http://facs.scripps.edu) for the data analysis. Flow cytometry analysis of whole blood Two microlitres of F-C1q, F-HSA or antibodies labelled with PE or flourescein were added to 12.5 lL of heparin- ized blood. After incubation for 15 min, red blood cells were lysed by the addition of 1 mL of High-yield Lyse (Invitrogen). Cells were analyzed on a Calibur instrument (BD Biosciences), counting 5000 or 10 000 events. PBMC isolation and inhibition experiments Human peripheral blood was collected and mixed with NaCl ⁄ P i (1 : 1). The mixture was overlayed on Ficoll and centrifuged (800 g for 20 min at 4 °C). The interface of mononuclear cells was harvested and stained with 2 l Lof F-C1q and the described antibodies for 10 lL of PBMC, with or without the addition of inhibitor, for 20 min at room temperature. Cells were analyzed on a Calibur instru- ment (BD Biosciences), counting 10 000 events. Confocal scanning laser microscopy MM6 cells were grown on glass cover slips for 6 days and stimulated with LPS (10 ngÆmL )1 ) on days 1 and 3. The cells were fixed with 4% paraformaldehyde for 20 min at 37 °C and permeabilized with 0.1% Triton X-100 for 5 min after three washes with NaCl ⁄ P i . After four washes, the cells were stained for 30 min with biotinylated mouse anti-CD91 (dilu- tion 1 : 25) in Hanks balanced salt solution containing 5% inactivated fetal bovine serum and 40 mm Hepes. After four washes with NaCl ⁄ P i , the cells were stained with F-C1q (10 lgÆmL )1 ) and streptavidin-Alexa 546 (dilution 1 : 50) in the same staining buffer. The coverslips were mounted with 4¢,6¢-diamidino-2-phenylindole-containing mounting solu- tion and pictures were taken with a Leica, SP2 confocal microscope (Leica Microsystems, Wetzlar, Germany). Pic- tures were processed using image j [53]. Acknowledgements Dorthe Tange Olsen and Anne Mortensen are thanked for their excellent technical work. Søren Kragh Moest- rup and Christian Jacobsen are thanked for providing purified CD91 as well as helpful suggestions. Hans Jør- gen Jensen and Susana Aznar are thanked for help with the confocal microscopy. Direct interaction between CD91 and C1q K. Duus et al. 3534 FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS References 1 Goldfarb RD & Parrillo JE (2005) Complement. Crit Care Med 33, S482–S484. 2 Nauta AJ, Roos A & Daha MR (2004) A regulatory role for complement in innate immunity and autoimmu- nity. Int Arch Allergy Immunol 134, 310–323. 3 Bottcher A, Gaipl US, Furnrohr BG, Herrmann M, Girkontaite I, Kalden JR & Voll RE (2006) Involve- ment of phosphatidylserine, alphavbeta3, CD14, CD36, and complement C1q in the phagocytosis of primary necrotic lymphocytes by macrophages. Arthritis Rheum 54, 927–938. 4 Korb LC & Ahearn JM (1997) C1q binds directly and specifically to surface blebs of apoptotic human keratinocytes: complement deficiency and systemic lupus erythematosus revisited. J Immunol 158, 4525– 4528. 5 Nauta AJ, Trouw LA, Daha MR, Tijsma O, Nieuwland R, Schwaeble WJ, Gingras AR, Mantovani A, Hack EC & Roos A (2002) Direct binding of C1q to apopto- tic cells and cell blebs induces complement activation. Eur J Immunol 32, 1726–1736. 6 Navratil JS, Watkins SC, Wisnieski JJ & Ahearn JM (2001) The globular heads of C1q specifically recognize surface blebs of apoptotic vascular endothelial cells. J Immunol 166, 3231–3239. 7 Pittoni V & Valesini G (2002) The clearance of apopto- tic cells: implications for autoimmunity. Autoimmun Rev 1, 154–161. 8 Sontheimer RD, Racila E & Racila DM (2005) C1q: its functions within the innate and adaptive immune responses and its role in lupus autoimmunity. J Invest Dermatol 125, 14–23. 9 Tarr J & Eggleton P (2005) Immune function of C1q and its modulators CD91 and CD93. Crit Rev Immunol 25, 305–330. 10 Edelson BT, Stricker TP, Li Z, Dickeson SK, Shepherd VL, Santoro SA & Zutter MM (2006) Novel collec- tin ⁄ C1q receptor mediates mast cell activation and innate immunity. Blood 107, 143–150. 11 Eggleton P, Reid KB & Tenner AJ (1998) C1q – how many functions? How many receptors? Trends Cell Biol 8, 428–431. 12 Eggleton P, Tenner AJ & Reid KB (2000) C1q recep- tors. Clin Exp Immunol 120, 406–412. 13 Ghebrehiwet B, Silvestri L & McDevitt C (1984) Identification of the Raji cell membrane-derived C1q inhibitor as a receptor for human C1q. Purification and immunochemical characterization. J Exp Med 160, 1375–1389. 14 Ghiran I, Tyagi SR, Klickstein LB & Nicholson-Weller A (2002) Expression and function of C1q receptors and C1q binding proteins at the cell surface. Immunobiology 205, 407–420. 15 McGreal E & Gasque P (2002) Structure–function stud- ies of the receptors for complement C1q. Biochem Soc Trans 30, 1010–1014. 16 Sim RB, Moestrup SK, Stuart GR, Lynch NJ, Lu J, Schwaeble WJ & Malhotra R (1998) Interaction of C1q and the collectins with the potential receptors calreticu- lin (cC1qR ⁄ collectin receptor) and megalin. Immunobiol- ogy 199, 208–224. 17 Basu S, Binder RJ, Ramalingam T & Srivastava PK (2001) CD91 is a common receptor for heat shock pro- teins gp96, hsp90, hsp70, and calreticulin. Immunity 14, 303–313. 18 Binder RJ, Han DK & Srivastava PK (2000) CD91: a receptor for heat shock protein gp96. Nat Immunol 1, 151–155. 19 Binder RJ, Vatner R & Srivastava P (2004) The heat- shock protein receptors: some answers and more ques- tions. Tissue Antigens 64, 442–451. 20 Gardai SJ, Xiao YQ, Dickinson M, Nick JA, Voelker DR, Greene KE & Henson PM (2003) By binding SIRPalpha or calreticulin ⁄ CD91, lung collectins act as dual function surveillance molecules to suppress or enhance inflammation. Cell 115, 13–23. 21 Ogden CA, deCathelineau A, Hoffmann PR, Bratton D, Ghebrehiwet B, Fadok VA & Henson PM (2001) C1q and mannose binding lectin engagement of cell surface calreticulin and CD91 initiates macropinocytosis and uptake of apoptotic cells. J Exp Med 194, 781–795. 22 Orr AW, Pedraza CE, Pallero MA, Elzie CA, Goicoe- chea S, Strickland DK & Murphy-Ullrich JE (2003) Low density lipoprotein receptor-related protein is a calreticulin coreceptor that signals focal adhesion disas- sembly. J Cell Biol 161, 1179–1189. 23 Vandivier RW, Ogden CA, Fadok VA, Hoffmann PR, Brown KK, Botto M, Walport MJ, Fisher JH, Henson PM & Greene KE (2002) Role of surfactant proteins A, D, and C1q in the clearance of apoptotic cells in vivo and in vitro: calreticulin and CD91 as a common collec- tin receptor complex. J Immunol 169, 3978–3986. 24 Obeid M, Tesniere A, Ghiringhelli F, Fimia GM, Apetoh L, Perfettini JL, Castedo M, Mignot G, Panaretakis T, Casares N et al. (2007) Calreticulin exposure dictates the immunogenicity of cancer cell death. Nat Med 13, 54–61. 25 Gardai SJ, McPhillips KA, Frasch SC, Janssen WJ, Starefeldt A, Murphy-Ullrich JE, Bratton DL, Oldenborg PA, Michalak M & Henson PM (2005) Cell-surface calreticulin initiates clearance of viable or apoptotic cells through trans-activation of LRP on the phagocyte. Cell 123, 321–334. 26 Herz J & Strickland DK (2001) LRP: a multifunctional scavenger and signaling receptor. J Clin Invest 108, 779–784. 27 Krieger M & Herz J (1994) Structures and functions of multiligand lipoprotein receptors: macrophage K. Duus et al. Direct interaction between CD91 and C1q FEBS Journal 277 (2010) 3526–3537 ª 2010 The Authors Journal compilation ª 2010 FEBS 3535 [...]... compilation ª 2010 FEBS K Duus et al binding site for A beta on the C1q globular regions J Immunol 167, 6374–6381 52 Moestrup SK, Kaltoft K, Sottrup-Jensen L & Gliemann J (1990) The human alpha 2-macroglobulin receptor contains high affinity calcium binding sites important Direct interaction between CD91 and C1q for receptor conformation and ligand recognition J Biol Chem 265, 12623–12628 53 Collins TJ (2007).. .Direct interaction between CD91 and C1q 28 29 30 31 32 33 34 35 36 37 38 K Duus et al scavenger receptors and LDL receptor-related protein (LRP) Annu Rev Biochem 63, 601–637 Moestrup SK & Gliemann J (1991) Analysis of ligand recognition by the purified alpha 2-macroglobulin receptor (low density lipoprotein receptor-related... CD91 and CD91 for all J Antimicrob Chemother 53, 1–3 Strickland DK, Ashcom JD, Williams S, Burgess WH, Migliorini M & Argraves WS (1990) Sequence identity between the alpha 2-macroglobulin receptor and low density lipoprotein receptor-related protein suggests that this molecule is a multifunctional receptor J Biol Chem 265, 17401–17404 Willnow TE, Orth K & Herz J (1994) Molecular dissection of ligand... Trends Immunol 25, 368–373 45 Gal P & Ambrus G (2001) Structure and function of complement activating enzyme complexes: C1 and MBL-MASPs Curr Protein Pept Sci 2, 43–59 46 Donnelly S, Roake W, Brown S, Young P, Naik H, Wordsworth P, Isenberg DA, Reid KB & Eggleton P (2006) Impaired recognition of apoptotic neutrophils by the C1q ⁄ calreticulin and CD91 pathway in systemic lupus erythematosus Arthritis Rheum... JJ & Berwin B (2005) Differential CD91 dependence for calreticulin and Pseudomonas exotoxin-A endocytosis Traffic 6, 1173–1182 48 Hojrup P, Roepstorff P & Houen G (2001) Human placental calreticulin-characterization of domain structure and post-translational modifications Eur J Biochem 268, 2558–2565 49 Steino A, Jorgensen CS, Laursen I & Houen G (2004) Interaction of C1q with the receptor calreticulin... oligomers bind and activate the classical complement pathway via residues 14-26 and 76-92 of the A chain collagen-like region of C1q J Immunol 150, 169–176 3536 39 Zahedi K (1996) Characterization of the binding of serum amyloid P to type IV collagen J Biol Chem 271, 14897–14902 40 Obermoeller-McCormick LM, Li Y, Osaka H, FitzGerald DJ, Schwartz AL & Bu G (2001) Dissection of receptor folding and ligand-binding... 49 Steino A, Jorgensen CS, Laursen I & Houen G (2004) Interaction of C1q with the receptor calreticulin requires a conformational change in C1q Scand J Immunol 59, 485–495 50 Arlaud GJ, Reboul A, Sim RB & Colomb MG (1979) Interaction of C1-inhibitor with the C1r and C1s subcomponents in human C1 Biochim Biophys Acta 576, 151–162 51 Tacnet-Delorme P, Chevallier S & Arlaud GJ (2001) Beta-amyloid fibrils... in human tissues Cell Tissue Res 269, 375–382 Lillis AP, Greenlee MC, Mikhailenko I, Pizzo SV, Tenner AJ, Strickland DK & Bohlson SS (2008) Murine low-density lipoprotein receptor-related protein 1 (LRP) is required for phagocytosis of targets bearing LRP ligands but is not required for C1q- triggered enhancement of phagocytosis J Immunol 181, 364– 373 Schulz S, Birkenmeier G, Schagdarsurengin U, Wenzel... J (1994) Molecular dissection of ligand binding sites on the low density lipoprotein receptor-related protein J Biol Chem 269, 15827–15832 Strickland DK, Kounnas MZ & Argraves WS (1995) LDL receptor-related protein: a multiligand receptor for lipoprotein and proteinase catabolism FASEB J 9, 890–898 Moestrup SK, Gliemann J & Pallesen G (1992) Distribution of the alpha 2-macroglobulin receptor ⁄ low... Molecular analysis of ligand binding to the second cluster of complement-type repeats of the low density lipoprotein receptor-related protein Evidence for an allosteric component in receptor-associated protein-mediated inhibition of ligand binding J Biol Chem 272, 13608–13613 42 Neels JG, van den Berg BM, Lookene A, Olivecrona G, Pannekoek H & van Zonneveld AJ (1999) The second and fourth cluster of class . FEBS Evidence for a direct interaction between CD91 and C1q Evidence for a direct interaction was obtained by surface plasmon resonance (SPR) analysis. Using C1q as the soluble analyte and CD91 as the. (A 405 ) C 0 1 2 3 4 150 100 50 0 Time without NaCl (min) CD91 binding (A 405 ) D 0 1 2 3 CD91- C1q CD91- BSA CD91 Anti -C1q Anti-isotype control Antibody binding (A 405 ) B Fig. 4. Interaction between CD91 and C1q, demonstrated by ELISA analyzed C1q binding to human blood cells, confirming a mono- cyte interaction. The interaction between C1q and PBMCs could be partially inhibited by RAP and solu- ble CD91. The interaction between C1q