Cholesterol Crystals Activate the Lectin Complement Pathway via Ficolin-2 and Mannose-Binding Lectin: Implications for the Progression of Atherosclerosis

Vascular diseases arising from atherosclerosis are among the leading causes of morbidity and death in Western countries. Cholesterol crystals (CC) appear early in the development of atherosclerosis and play an essential role in the formation of atherosclerotic plaques by inducing inflammation and functioning as an endogenous danger signal (1). Cholesterol is almost water insoluble, but it is solubilized in vivo in lipid bilayers or micelle systems, where it serves as a stabilizing component of cell membranes and a precursor of bile salts and steroid hormones. Unbalanced cholesterol metabolism results in undesired in vivo precipitation of CC. Cholesterol crystallizes in a bilayer structure with an end-for-end arrangement of approximately parallel molecules (2). CC are known to activate the complement system (3–5) and induce complement-dependent inflammasome activation and cytokine release in phagocytes in vitro (1, 6, 7).

Activation of the complement system is an important part of the innate immune response and is initiated when pattern recognition molecules (PRMs) recognize pathogen-associated molecular patterns on foreign cells or damaged host cells. The effector functions of the complement system include elimination of invading pathogens, regulation of adaptive immunity, and maintenance of tissue homeostasis (8). The complement system is activated through three different pathways: the classical pathway (CP), the lectin pathway (LP), and the alternative pathway (AP).

The PRM mannose-binding lectin (MBL) is a soluble collectin that is able to activate the LP (9, 10). MBL recognizes a wide variety of microorganisms and altered self via a carbohydrate recognition domain (CRD), and it distinguishes between normally present self-structures and foreign structures, including microbes and damaged self, based on the orientations of hydroxyl groups on carbohydrates (11).

The LP can also be activated by ficolins (12–14). Three types of ficolins have been characterized in humans: ficolin-1 (M-ficolin), ficolin-2 (l-ficolin), and ficolin-3 (H-ficolin). The ficolins are multimeric macromolecules, structurally and functionally related to collectins. Ficolin-1, ficolin-2, and ficolin-3 are found in serum (15–17), with ficolin-1 being the least abundant (15). Ficolins primarily recognize acetylated compounds via the fibrinogen-like domain, such as N-acetyl-d-glucosamine (GlcNAc) and acetylated BSA (AcBSA) (18, 19). The LP recognition molecules have also been shown to recognize and participate in the removal of altered and dying host cells (20–23). PRMs of the LP are found in complexes with the MBL/ficolin–associated serine proteases (MASPs). Three catalytically active serine proteases (MASP-1, MASP-2, and MASP-3) and two alternative splice products (MAP-1 [MAp44] and sMAP [MAp19)] are known (24).

The CP is activated by direct binding of the PRM C1q to various structures and ligands on pathogens or apoptotic cells, or indirectly via other molecules such as Igs or C-reactive protein. The AP is activated by spontaneous hydrolysis of the internal C3 thioester in the fluid phase or directly on foreign surfaces that are not protected against complement activation (25). Additionally, the AP functions as an amplification loop, enhancing CP- and/or LP-initiated complement activation substantially.

Complement activation and deposition of complement activation products occur both in experimental and human atherosclerosis (26–31). Monohydrate CC, similar to those found in atherosclerotic plaques, activate the CP and the AP (3–5, 7). CC activate the CP through binding of C1q, but whether C1q binds directly to the CC or indirectly via Igs, C-reactive protein, or other molecules is still not clear. C1q is present in atherosclerotic lesions and has been shown to play a protective role in early atherosclerotic development in a C1q-deficient mouse model of atherosclerosis (32). However, in this model, C5b-9 deposition in aortic lesions was not abolished, suggesting involvement of the LP and/or the AP.

Whether the LP is involved in the local inflammation initiated by CC in the pathophysiology of atherosclerosis is not known, but MBL is present in atherosclerotic lesions, and MBL gene variants leading to functional defects of MBL have been associated with myocardial infarction (33) and severity of atherosclerosis (34, 35). Cholesterol compounds have been shown to differ in their complement activation ability, depending on the position as well as the number of hydroxyl groups (4). Because MBL and the ficolins bind hydroxyl groups, we hypothesized that MBL and ficolins bind monohydrate CC similar to those found in atherosclerotic plaques, and they function as an upstream innate inflammatory signal in the pathophysiology of atherosclerosis. Thus, the objectives of this study were to investigate the binding of the pattern recognition molecules from the lectin complement pathway (MBL, ficolin-1, -2, and -3) and the associated serine proteases (MASPs) to CC in vitro and in vivo in human carotid plaques, and to determine the role of the LP in the complement activation initiated by CC. Furthermore, the purpose was to clarify whether the C1q-mediated CP activation on CC was initiated through direct binding of C1q or whether C1q bound indirectly via IgG, IgM, or both.

Ultrapure cholesterol (C8667), 1-propanol (279544), RPMI 1640 medium, erythrosine, human serum albumin (A9731), BSA (A2153), EGTA (E3889), GlcNAc (A8625), d-mannose (M6020), and d-(+)-galactose (G0750) were purchased from Sigma-Aldrich. EDTA (324503) was purchased from Calbiochem. Glycergel mounting medium (C0563), Dako EnVision Plus autostainer, polymer EnVision-HRP anti-mouse, diaminobenzidine (DAB+) chromogen solution, Dako wash buffer (S3006), and serum-free protein block (X0909) was purchased from Dako (Glostrup, Denmark). Purified human C4 (A108) and purified human C1q (A099) were purchased from Complement Technology (Tyler, TX). Serum-free medium Chinese hamster ovary (CHO) CD1 was from Lonza (Vallensbaek Strand, Denmark). Tissue-Tek OCT compound was from Sakura Finetek (Tokyo, Japan). Hematoxylin and saffron was from Chemi-Teknik (Oslo, Norway). Alexa Fluor 647 anti-mouse IgG was from Thermo Fisher Scientific (Waltham, MA) and Mowiol 4-88 was from Hoechst (Frankfurt, Germany).

The following reagents were used as complement inhibitors: C5 inhibitor; eculizumab (Soliris, Alexion Pharmaceuticals), C3 inhibitory peptide; compstatin Cp40 (36), C1q inhibitory mAb mouse anti-human C1q clone 85 (IgG1) (MW1828, Sanquin); MASP-1 inhibitor SGMI-1 (37); and MASP-2 inhibitor SGMI-2 (37). The following Abs were used: FITC-conjugated polyclonal rabbit anti-rat Ab (F1763, Sigma-Aldrich), FITC-conjugated polyclonal goat anti-rabbit Ab (F1262, Sigma-Aldrich), FITC-conjugated polyclonal goat anti-mouse Ab (F0479, Dako), mouse IgG1κ isotype control (BD Biosciences), mouse IgG2a isotype control (BD Biosciences), rabbit IgG isotype control (Invitrogen), rat IgG1κ isotype control (BD Biosciences), mouse anti-human MBL mAb 131-10 (IgG1κ) (BioPorto Diagnostics), mouse anti-human MBL 131-01 (IgG1κ) (BioPorto Diagnostics), MBL inhibitory mAb mouse anti-human MBL 3F8 (IgG1) (38), MBL binding mAb mouse anti-human MBL 1C10 (IgG2) (39), rat anti-human MASP-2 mAb 8B5 (HM2190, Hycult Biotech), mouse anti-human complement component C5b-9 mAb (IgG2a) (011-01, Antibody Shop), polyclonal rabbit anti-human C4c Ab (Q0369, Dako), polyclonal rabbit anti-human C1q (A0136, Dako), rabbit anti-human IgM (0425, Dako), and rabbit anti-human IgG (0423, Dako). In-house–produced mAbs included: mouse anti-human ficolin-1 FCN106 (IgG1κ) (15), mouse anti-human ficolin-2 FCN219 (IgG2a) (16), ficolin-2 inhibitory Ab mouse anti-human ficolin-2 FCN212 (IgG1κ) (unpublished), mouse anti-human ficolin-3 FCN334 (IgG1κ) (17), and mouse anti-human MASP-1/-3/MAP-1 8B3 (IgG1κ) (40, 41).

Recombinant MBL, ficolin-1, -2, -3, and MASP-1, -2, and -3 were produced in our laboratory as previously described (42–44). In short, the proteins were expressed in CHO-DG44 cells cultivated in RPMI 1640 medium supplemented with 10% FCS, 100 U/ml penicillin, 0.1 mg/ml streptomycin, 2 mM l-glutamine, and 200 nM methotrexate or by using serum-free medium CHO CD1 supplemented with 200 nM methotrexate.

CC were prepared essentially as described by Samstad et al. (7). Ultrapure cholesterol (100 mg) was dissolved in 50 ml 1-propanol. Distilled water (75 ml) was added to the solution and it was left undisturbed for 15 min for the crystals to stabilize. The solution was centrifuged and the remaining 1-propanol was removed by evaporation. All steps were performed at room temperature (RT). The CC were resuspended in PBS/0.05% HSA and stored at 4°C in the dark.

A pool of normal human serum (NHS) was obtained by drawing venous blood from six healthy donors (three male and three female donors) into dry glass vials with no additive. The blood samples were left at RT for 2 h, before the serum was collected by centrifugation at 3000 × g for 15 min, pooled, and stored at −80°C, awaiting further analysis. A pool of umbilical cord serum (UCS) was obtained from umbilical cord blood collected from three individuals. The blood samples were left at RT for 2 h, before the serum was collected by centrifugation at 3000 × g for 15 min, pooled, and stored at −80°C, awaiting further analysis. Serum samples from a previously described MBL defect patient (42) and a C1q-deficient patient (45) were used in the flow cytometry experiments.

Relevant controls were included routinely in all experiments, including negative controls samples (without protein/serum), isotype controls (mouse IgG1κ isotype control, mouse IgG2a isotype control, rabbit IgG isotype control, rat IgG1κ isotype control), and background control samples (containing no Ab or only primary or secondary Ab). Samples were analyzed by flow cytometry using a Gallios flow cytometer (Beckman Coulter), and data were analyzed using Kaluza software (Beckman Coulter). CC were gated as a uniform population on the forward site scatter, containing 80–90% of all counts (see Supplemental Fig. 1).

CC (1.0 × 10⁶ particles/ml) were incubated with 5% NHS or MBL defect serum or 0.25–4 μg/ml recombinant protein diluted in barbital buffer (5 mM barbital sodium, 145 mM NaCl, 2 mM CaCl₂, 1 mM MgCl₂ [pH 7.4]) containing 0.5% heat-inactivated FCS (barbital/FCS) for 30 min at 37°C with shaking. The CC were washed in barbital/FCS after each step. Binding of MBL was detected with 0.5 μg/ml (recombinant protein) or 5 μg/ml (serum) mouse anti-human MBL mAb 131-10 (30 min at 4°C), followed by 2.5 μg/ml FITC-conjugated polyclonal goat anti-mouse Ab (20 min at 4°C). In some experiments CC were incubated with serum preincubated 5 min at RT with 10 μg/ml MBL inhibitory mAb mouse anti-human MBL 3F8, or 10 μg/ml MBL binding mAb mouse anti-human MBL 1C10, 10 mM EDTA, 10 mM EGTA/Mg, 10 mM GlcNAc, 10 mM d-mannose, or 10 mM d-(+)-galactose.

CC (1.0 × 10⁶ particles/ml) were incubated with 5% NHS or 0.25–4 μg/ml recombinant protein diluted in barbital/FCS for 30 min at 37°C with shaking. The CC were washed in barbital/FCS after each step. Binding of ficolin-1, -2, or -3 was detected using 5 μg/ml mouse anti-human ficolin-1 FCN106, 0.5 μg/ml (recombinant protein) or 5 μg/ml (serum) mouse anti-human ficolin-2 mAb FCN219, or 5 μg/ml mouse anti-human ficolin-3 mAb FCN334 (30 min at 4°C), followed by 2.5 μg/ml FITC-conjugated polyclonal goat anti-mouse Ab (20 min at 4°C). In some experiments, CC were incubated with serum preincubated 5 min at RT with 10 μg/ml ficolin-2 inhibitory mAb FCN212, 10 μg/ml AcBSA prepared as described in Hein et al. (46), 10 μg/ml BSA, 10 mM EDTA, 10 mM EGTA/Mg, 20 mM GlcNAc, 20 mM d-mannose, or 20 mM d-(+)-galactose.

CC (1.0 × 10⁶ particles/ml) were incubated with 5% NHS diluted in barbital/FCS for 30 min at 37°C with shaking. The CC were washed in barbital/FCS after each step. Binding of MASP-1/-3/MAP-1 was detected using 5 μg/ml mouse anti-human MASP-1/-3/MAP-1 8B3, and MASP-2 binding was detected using 5 μg/ml rat anti-human MASP-2 mAb 8B5 (30 min at 4°C), followed by 2.5 μg/ml FITC-conjugated polyclonal goat anti-mouse Ab or 2 μg/ml FITC-conjugated polyclonal rabbit anti-rat Ab (20 min at 4°C).

CC (1.0 × 10⁶ particles/ml) were incubated with recombinant MASP-1 (4 μg/ml), MASP-2 (1 μg/ml), or MASP-3 (1 μg/ml) or recombinant MASP-1, -2, or -3 in the presence of recombinant ficolin-2 (2 μg/ml) or ficolin-3 (2 μg/ml) diluted in barbital/FCS for 30 min at 37°C with shaking. The CC were washed in barbital/FCS after each step. Binding of MASP-1 or MASP-3 was detected using 5 μg/ml mouse anti-human MASP-1/-3/MAP-1 mAb 8B3 (30 min at 4°C), followed by 2.5 μg/ml FITC-conjugated polyclonal goat anti-mouse Ab (20 min at 4°C). Binding of MASP-2 was detected using 5 μg/ml rat anti-human MASP-2 mAb 8B5 (30 min at 4°C), followed by 2 μg/ml FITC-conjugated polyclonal rabbit anti-rat Ab (20 min at 4°C).

To assess ficolin-2–mediated complement activation, CC (1.0 × 10⁶ particles/ml) were incubated with recombinant ficolin-2 (2 μg/ml), recombinant MASP-2 (0.5 μg/ml), or a mixture of the proteins (30 min at 37°C with shaking), followed by incubation with purified human C4 (10 μg/ml) (30 min at 37°C with shaking). The CC were washed in barbital/FCS after each step. Deposition of C4b on the CC was detected using 0.5 μg/ml polyclonal rabbit anti-human C4c Ab (30 min at 4°C), followed by 2 μg/ml FITC-conjugated polyclonal goat anti-rabbit Ab (20 min at 4°C).

Complement activation from serum was assessed by incubating CC (1.0 × 10⁶ particles/ml) with 10% NHS, C1q-deficient serum with or without reconstitution with 10 μg/ml purified C1q, MBL defect serum, or UCS diluted in barbital/FCS (30 min at 37°C with shaking). The CC were washed in barbital/FCS after each step. Deposition of C5b-9 was detected using 2 μg/ml mouse anti-human complement component C5b-9 mAb, followed by 2.5 μg/ml FITC-conjugated polyclonal goat anti-mouse Ab (20 min at 4°C). In some experiments CC were incubated with serum preincubated 5 min at RT with 50 μg/ml C5 inhibitor (eculizumab), 20 μM C3 inhibitory peptide (compstatin Cp40), 10 μg/ml C1q inhibitory mAb mouse anti-human C1q clone 85, 5 μM MASP-1 inhibitor SGMI-1, 5 μM MASP-2 inhibitor SGMI-2, 10 μg/ml MBL inhibitory mAb mouse anti-human MBL 3F8, or 10 μg/ml MBL binding mAb mouse anti-human MBL 1C10.

CC (1.0 × 10⁶ particles/ml) were incubated with 5% NHS, C1q-deficient serum with or without reconstitution with 10 μg/ml purified C1q, MBL defect serum, or UCS diluted in barbital/FCS for 30 min at 37°C with shaking. The CC were washed in barbital/FCS after each step. Binding of C1q was detected with polyclonal rabbit anti-human C1q (30 min at 4°C), followed by 2 μg/ml FITC-conjugated polyclonal goat anti-rabbit Ab (20 min at 4°C). Binding of IgM or IgG was detected using 1 μg/ml rabbit anti-human IgM or 0.1 μg/ml rabbit anti-human IgG, followed by 2 μg/ml FITC-conjugated polyclonal goat anti-rabbit Ab (20 min at 4°C). In some experiments CC were incubated with serum preincubated 5 min at RT with 10 μg/ml C1q inhibitory mAb mouse anti-human C1q clone 85.

Microscopy was performed using a Zeiss Axio Observer through a ×63/1.40 oil DIC Plan-Apochromat objective. Imaging conditions were kept constant when acquiring images to be compared. Relevant controls were included routinely in all experiments, including negative controls samples (without protein), isotype controls (mouse IgG1κ isotype control, mouse IgG2a isotype control), and background control samples (containing no Ab or only primary or secondary Ab).

For binding of recombinant MBL and ficolin-2, CC (2.0 × 10⁶ particles/ml) were incubated with recombinant MBL (20 μg/ml) or ficolin-2 (8 μg/ml) diluted in barbital/FCS for 30 min at 37°C with shaking. The CC were washed in barbital/FCS after each step. Binding of MBL and ficolin-2 was detected with 5 μg/ml mouse anti-human MBL mAb 131-10 or 0.5 μg/ml mouse anti-human ficolin-2 mAb FCN219 (30 min at 4°C), followed by 2.5 μg/ml FITC-conjugated polyclonal goat anti-mouse Ab (20 min at 4°C). CC were placed on slides by cytospin (centrifugation for 5 min at 300 × g) and mounted with Glycergel mounting medium.

Carotid plaques used for immunohistochemistry (IHC) and fluorescence microscopy were obtained from patients with carotid stenosis (47). Signed informed consent for participation in the study was obtained from all individuals. Plaque for paraffin sections was immediately after excision put in formalin for fixation before being embedded in paraffin. Plaque for frozen sections was taken and immediately put in Tissue-Tek OCT compound and frozen at −80°C.

Hematoxylin and erythrosine staining was prepared by a histological routine staining protocol. Briefly, formalin-fixed, paraffin-embedded sections of carotid plaques (4 μm) were dried at 60°C and stained in an automatic slide stainer (Sakura Tissue-Tek Prisma). After deparaffinization and rehydration to water the slides were stained with hematoxylin and erythrosine and rinsed in water for removal of excess dye. The sections were then dehydrated by adding ascending grades of ethanol and stained in saffron, rinsed in several baths of absolute ethanol, and cleared in Tissue-Clear before coverslipping in Sakura Tissue-Tek Glas automatic cover slipper.

Frozen sections (5 μm) on SuperFrost Plus glass were dried for 10 min and washed in Dako 3006 wash buffer before incubation in serum-free protein block for 30 min at RT. Mouse anti-human MBL mAb 131-01 (10 μg/ml), mouse anti-human ficolin-2 mAb FCN219 (8 μg/ml), and corresponding isotype controls were added and incubated overnight at 4°C. After washing, the sections were incubated with Alexa Fluor 647 anti-mouse IgG (3.3 μg/ml) for 60 min at RT before mounting in Mowiol 4-88. The paraffin-embedded and frozen sections were examined in an EVOS FL auto microscope (Thermo Fisher Scientific).

Dried paraffin-embedded sections of carotid plaques (4 μm) were deparaffinized and rehydrated by increasing amounts of water in a Sakura Tissue-Tek Prisma. The sections were then stained in a Dako EnVision Plus autostainer for mouse Abs according to the manufacturer’s protocol. Briefly, endogenous peroxidase activity was blocked by peroxidase block and incubated with mouse anti-human MBL mAb 131-01 (10 μg/ml), mouse anti-human ficolin-2 mAb FCN219, or an isotype control overnight at 4°C followed by polymer EnVision/HRP anti-mouse for 30 min. A DAB+ chromogen solution was used to develop the color. The sections were then stained with hematoxylin, dehydrated, and prepared as described for the hematoxylin and saffron staining.

GraphPad Prism version 6 (Graphpad Software) was used for statistical analysis. Data are expressed as mean ± SEM. Statistical analysis was performed on three independent experiments using two-tailed paired t test. A p value < 0.05 was considered statistically significant.

The study was approved by the regional Health Ethics Committee in the Capital Region of Denmark (reference no. H2-2011-133). Carotid plaques used for IHC and fluorescence microscopy were obtained from patients with carotid stenosis (47). The protocols were approved by the Regional Committee for Medical and Research Ethics, South-East, Norway (reference no. S-0923a 2009/6065).

In the first set of experiments we investigated the binding of the PRMs from the LP to CC in vitro. No binding of ficolin-1 or ficolin-3 from NHS to CC was observed (Fig. 1A, 1C). Ficolin-2 bound strongly to CC when incubated with NHS (Fig. 1B) or recombinant ficolin-2 (Figs. 1E, 2). Ficolin-2 binding was inhibited when serum was preincubated with a ficolin-2 inhibitory Ab (p < 0.01) or specific ligands AcBSA (p < 0.01) or GlcNAc (p < 0.05) (Fig. 1B). Removal of calcium by addition of EDTA or EGTA/Mg had no inhibitory effect on ficolin-2 binding, but EDTA significantly (p < 0.05) enhanced ficolin-2 binding. Strong binding of MBL to the CC was observed when incubated with NHS (Fig. 1D) or recombinant MBL (Figs. 1F, 2). MBL binding was significantly inhibited when serum was preincubated with an MBL inhibitory Ab (p < 0.01), specific ligands GlcNAc (p < 0.01) or d-mannose (p < 0.05), or in the absence of calcium (EDTA or EGTA/Mg) (p < 0.01) (Fig. 1D). No binding of MBL was observed from MBL defect serum. These data show that both ficolin-2 and MBL bind specifically to CC incubated with serum or recombinant proteins.

To further investigate the role of the LP, we looked into the binding of the MASPs and MAP-1 to CC incubated with NHS (Fig. 3). Binding of MASP-1/-3/MAP-1 was detected using specific mouse mAb 8B3, recognizing the common H chain on all three molecules (40). A significant binding (p < 0.05) of native MASP-1/-3/MAP-1 to CC was observed (Fig. 3A), whereas no binding of MASP-2 from NHS was observed (Fig. 3B). To support this, we determined whether the MASPs were able to form complexes with ficolin-2 on the surface of the CC (Fig. 4). Recombinant MASPs bound to CC, and the binding of MASP-1 and MASP-3 was significantly increased (p < 0.05) when coincubated with ficolin-2, but not when coincubated with the homolog PRM ficolin-3 (Fig. 4A). Binding of MASP-2 was also increased when coincubated with ficolin-2, but the increase did not reach statistical significance (Fig. 4A).

We then determined whether ficolin-2/MASP-2 complexes were able to activate complement by inducing C4 cleavage. Ficolin-2/MASP-2 complexes significantly increased C4 activation and deposition on the CC compared with CC incubated with C4 alone (p < 0.05), ficolin-2 followed by C4 (p < 0.01), or MASP-2 followed by C4 (p < 0.05) (Fig. 4B). MASP-2 alone could be absorbed to the CC without ficolin-2 and enhance the C4 activation and deposition, but the increase did not reach statistical significance. These results indicate that the MASPs bind to CC and form functional complexes with ficolin-2.

To confirm the physiological significance of the binding of ficolin-2 and MBL to the CC in vitro, we examined the deposition of ficolin-2 and MBL in human carotid plaques by IHC on paraffin-embedded sections and immunofluorescence on frozen sections. Ficolin-2 and MBL were deposited in frozen sections of carotid plaques (Fig. 5). Furthermore, immunohistochemically staining of MBL in carotid plaques showed colocalization of MBL and CC, visualized as MBL deposition around CC clefts (Fig. 6). No staining was observed on any isotype controls. We were not able to detect the presence of ficolin-2 by IHC on paraffin sections, probably because of lack of epitope recognition of the ficolin-2–specific Ab. These results show that ficolin-2 and MBL are present in human carotid plaques and that MBL is localized around CC clefts.

To further address the involvement of the LP in the CC-induced complement activation, we measured the in vitro deposition of the complement activation product C5b-9 on CC after incubation with NHS, MBL defect serum, or C1q-deficient serum by flow cytometry (Fig. 7). In some experiments serum was preincubated with control Ab or different complement inhibitors, including C3 inhibitor (compstatin Cp40), C5 inhibitor (eculizumab), C1q inhibitory Ab, MASP-1 inhibitor, MASP-2 inhibitor, MBL control Ab, or MBL inhibitory Ab. C5b-9 deposition on CC was significantly inhibited with the C3 inhibitor compstatin Cp40 (p < 0.01) and completely removed by inhibition with the C5 inhibitor eculizumab (p < 0.01) (Fig. 7A).

C5b-9 deposition on CC from NHS was partly but significantly inhibited (p < 0.01) by the C1q inhibitory Ab (Fig. 7B, 7C). The remaining complement activation in NHS was inhibited when combining the C1q inhibitory Ab with an LP inhibitor (MBL inhibitory Ab or MASP inhibitor) (Fig. 7B, 7C). The MBL inhibitory Ab in combination with the C1q inhibitory Ab significantly reduced (p < 0.01) the C5b-9 deposition on CC compared with inhibition with the C1q inhibitory Ab alone or the C1q inhibitory Ab and the MBL binding Ab control (Fig. 7B). The MBL inhibitory Ab had no effect on the level of C5b-9 deposition on CC when incubated with NHS without the addition of the C1q inhibitory Ab. The MASP-1 inhibitor in combination with the C1q inhibitory Ab significantly reduced (p < 0.05) the level of C5b-9 deposition on CC when incubated with NHS compared with the C1q inhibitory Ab alone (Fig. 7C). The MASP-1 inhibitor alone reduced the level of complement activation, but the decrease did not reach statistical significance. The MASP-2 inhibitor alone did not have an effect on the complement activation on CC, but when incubated in combination with the C1q inhibitory Ab, it decreased the level of C5b-9 deposition, but the decrease did not reach statistical significance (Fig. 7C). In the MBL defect serum, complement activation on CC was significantly reduced (p < 0.01) by preincubating serum with the C1q inhibitory Ab (Fig. 7D). Complement activation on CC from C1q-deficient serum was inhibited by the LP inhibitors as in NHS. The MBL inhibitory Ab significantly decreased (p < 0.05) the C5b-9 deposition on the CC compared with buffer (Fig. 7E). Both the MASP-1 (p < 0.001) and the MASP-2 inhibitor decreased the complement deposition on CC from C1q-deficient serum (Fig. 7F), but the decrease did not reach statistical significance using the MASP-2 inhibitor. C1q reconstitution in the C1q-deficient serum significantly increased (p < 0.05) the level of complement activation on CC to the level in NHS (Fig. 7E). Inhibition of complement activation with the ficolin-2 inhibitory Ab alone or in combination with the C1q inhibitory Ab did not decrease the C5b-9 deposition in NHS (data not shown). Collectively, these data demonstrated that both the CP, initiated by C1q, and the LP, initiated by MBL, play important roles in the initiation of complement activation occurring on CC.

It has previously been shown that C1q binds to CC (7). To confirm this in our system and to further study the binding of C1q, we examined the binding of C1q, IgM, and IgG to CC from NHS, C1q-deficient serum, MBL defect serum, and UCS with low amounts of IgM. C1q bound strongly to the CC when incubated with serum (Fig. 8). The C1q binding from NHS, MBL defect serum, and UCS was significantly inhibited (p < 0.05) with a C1q inhibitory Ab (Fig. 8A). No binding of C1q was observed from C1q-deficient serum. The C1q binding was reconstituted to the level of NHS when adding purified C1q to the C1q-deficient serum. The C1q binding from UCS with low amounts of IgM was significantly lower (p < 0.01) than from NHS (Fig. 8A). Supporting this, IgM binding to CC was significantly lower (p < 0.01) from UCS compared with NHS. IgM binding from C1q-deficient serum was significantly increased (p < 0.05) compared with NHS. To investigate this, we inhibited C1q binding in NHS and reconstituted C1q in C1q-deficient serum. When inhibiting C1q binding in NHS, the detected level of IgM binding increased, whereas reconstitution of C1q in the C1q-deficient serum significantly reduced (p < 0.05) the detected level of IgM binding to the level of NHS. Inhibition of C1q in MBL defect serum had almost the same effect as in NHS, whereas the level in UCS only slightly increased (Fig. 8B). No binding of IgG to CC was observed (Fig. 8C). These data demonstrate that IgM but not IgG binds to CC when incubated with serum, and they indicate that C1q binds indirectly to the CC via IgM.

CC are found in early atherosclerotic lesions (1) and the presence of CC is regarded as a hallmark of atherosclerosis (48). CC activate the CP and the AP and induce complement-dependent inflammasome activation and cytokine release (1, 3–7). In the present study we report novel and important mechanisms by which CC induce complement activation, and, to our knowledge for the first time, show that the LP is involved in the CC-induced inflammatory response.

Complement activation is involved in the development of atherosclerotic lesions. Complement activation to the level of C5b-9 has long been known to occur in atherosclerotic plaques (28, 31), and several Igs, complement components, and regulators are present in atherosclerotic lesions, including IgG, IgM, C1q, and MBL (30, 31). In C3- (26, 49), C5- (50), C6- (29), and C1q-deficient (32) mice or rabbit models, reduced complement activity affected atherosclerotic development. C6 complement deficiency was shown to protect against development of diet-induced atherosclerosis in rabbits (29), whereas total aortic atherosclerosis, determined by lipid staining, was greater in aortas from C3-deficient mice compared with controls (26). Taken together, these results indicate that complement may have proatherogenic or antiatherogenic roles depending on the stage of atherosclerotic development. The role of the CP in atherosclerotic development has been demonstrated in a C1q-deficient mouse model of atherosclerosis (32). C1q was found to reduce atherosclerosis development and mediate removal of apoptotic cells. However, deposition of C5b-9 in mouse aortic root lesions was not abolished in the absence of C1q, suggesting involvement of the LP and/or the AP (32).

MBL is present and produced locally by myeloid cells in early experimental atherosclerotic lesions (35). MBL deposition has also been detected in late stage human atherosclerotic lesions, but no local expression of MBL was observed, suggesting a plasma origin MBL in late stage atherosclerosis (35). Our data demonstrate that MBL recognizes CC, similar to those found in atherosclerotic lesions, in a specific and calcium-dependent manner, leading to full complement activation and deposition of C5b-9 on the CC. This is surprising because MBL is regarded as a typical lectin, recognizing hydroxyl groups on carbohydrates. However, we suggest that MBL binds monohydrate CC via the CRD by recognizing hydroxyl groups exposed on the crystals. The binding of MBL to the CC demonstrated in this study was highly calcium-dependent and was inhibited by a specific inhibitor (an MBL inhibitory Ab) or known MBL ligands (GlcNAc or d-mannose), strongly suggesting that this is a specific interaction via the CRD. No binding of MBL was detected from MBL defect serum, which in fact contains low molecular MBL unable to bind and mediate complement activation via the CRD (51). To support these findings we examined the presence of MBL in human carotid plaques. MBL was deposited in carotid plaques and localized around CC clefts. This confirms the physiological significance of the in vitro findings and demonstrates a potential role of the MBL-mediated LP in the pathophysiology of atherosclerosis.

Several population-based studies, analyzing either MBL genotypes or resulting serum levels, suggest a role for MBL in cardiovascular disease (34, 52–58). However, the results are contradictory, suggesting either a proatherogenic or antiatherogenic role of MBL in cardiovascular disease.

Both MBL and C1q have been found to bind modified low-density lipoprotein and enhance clearance by monocytes and macrophages (59, 60). In early atherosclerotic lesions MBL may facilitate clearance of endogenous danger signals, such as late apoptotic or necrotic cells, cellular debris, low-density lipoprotein, and CC, which are associated with atherosclerosis development and plaque stability (20, 21, 61). In more advanced stages of atherosclerosis, MBL may have a proinflammatory role; hence, downregulation of MBL expression in late human lesions may illustrate control of complement activation and inflammation (35).

Our data also demonstrate that ficolin-2 binds CC independent of calcium and that ficolin-2/MASPs complexes on CC are able to activate C4, leading to C4b deposition on the CC in vitro. Furthermore, ficolin-2 was deposited in human carotid plaques. Whether ficolin-2 mainly functions as an opsonin for phagocytosis independent of complement activation or participates in systemic complement activation as C1q and MBL remains to be established. No binding of ficolin-1 or ficolin-3 from serum to the CC was observed in our experiments. Ficolin-1 is mainly found on the surface of monocytes and granulocytes and in secretory granules of neutrophils and monocytes, but it has been shown to circulate in serum in very low concentrations (14, 15, 62). Owing to the low level of ficolin-1 in serum, the potential contribution of locally produced ficolin-1 in CC-induced inflammation occurring in atherosclerosis remains to be investigated. The ficolins primarily recognize acetylated compounds, such as GlcNAc and AcBSA, via their fibrinogen-like domain (18, 19). Furthermore, ficolin-2 and ficolin-3 have been shown to bind a wide variety of exogenous and endogenous molecular patterns, including several strains of encapsulated Gram-positive bacteria (63, 64), LPS from Gram-negative bacteria (65), DNA (22), mitochondria (66), and apoptotic cells (23). Ficolin-2 recognizes and binds ligands through four different binding groves with distinct binding properties (S1, S2, S3, and S4), giving ficolin-2 unusual recognition versatility. In comparison, ficolin-1 and ficolin-3 only contain one binding grove (S1) (19, 67). The more diverse binding capacity of ficolin-2 could explain the calcium-independent binding to the CC. Calcium-independent interactions of ficolin-2 have also been shown in recognition of pathogens such as Aspergillus fumigatus and in interactions with endogenous ligands such as pentraxin-3 (68).

The results from the present study demonstrate an additive role of the MBL and the C1q-mediated complement activation on CC. In NHS the level of complement activation was inhibited by a C1q inhibitory Ab, but the complement activation was not abolished. The remaining C5b-9 deposition was removed when inhibiting both C1q- and MBL-mediated complement activation. Furthermore, we found deposition of ficolin-2 and MBL in human carotid plaques, and MBL was colocalized to CC clefts. These data suggest that both LP and the CP play an important role in the CC-mediated complement activation occurring in atherosclerotic lesions, and that complement inhibition could be an interesting target for treatment of cardiovascular disease. C1q has been shown to bind CC, but the binding mechanism has previously been unknown (7). Our data confirm the binding of C1q to CC and further demonstrate that IgM, but not IgG, binds to CC, and that the C1q binding is mediated by IgM.

Complement activation on CC was inhibited less effective in our setup with the C3 inhibitor compstatin Cp40 compared with the C5 inhibitor eculizumab. This could partly be explained by the existence of a C3 bypass mechanism resulting in direct cleavage of C5 without activation of C3 as has been proposed by Huber-Lang and colleagues (69–71).

In conclusion, our study provides evidence for an important role for the LP in the inflammatory response induced by CC, and in particular we emphasize the role of ficolin-2 and MBL in the CC-mediated inflammation occurring during atherosclerotic plaque development.

We thank Prof. John D. Lambris (University of Pennsylvania) for providing us with C3 inhibitory peptide compstatin Cp40. We also thank Jytte Bryde Clausen for technical assistance.

The authors have no financial conflicts of interest.