Human monoclonal Fab and human plasma antibodies to carbamyl-epitopes cross-react with malondialdehyde-adducts

Abstract

Oxidized low-density lipoprotein (OxLDL) plays a crucial role in the development of atherosclerosis. Carbamylated LDL has been suggested to promote atherogenesis in patients with chronic kidney disease. Here we observed that plasma IgG and IgM antibodies to carbamylated epitopes were associated with IgG and IgM antibodies to oxidation-specific epitopes (ρ = 0·65–0·86, P < 0·001) in healthy adults, suggesting a cross-reaction between antibodies recognizing carbamyl-epitopes and malondialdehyde (MDA)/malondialdehyde acetaldehyde (MAA) -adducts. We used a phage display technique to clone a human Fab antibody that bound to carbamylated LDL and other carbamylated proteins. Anti-carbamyl-Fab (Fab106) cross-reacted with oxidation-specific epitopes, especially with MDA-LDL and MAA-LDL. We showed that Fab106 bound to apoptotic Jurkat cells known to contain these oxidation-specific epitopes, and the binding was competed with soluble carbamylated and MDA-/MAA-modified LDL and BSA. In addition, Fab106 was able to block the uptake of carbamyl-LDL and MDA-LDL by macrophages and stained mouse atherosclerotic lesions. The observed cross-reaction between carbamylated and MDA-/MAA-modified LDL and its contribution to enhanced atherogenesis in uraemic patients require further investigation.

Introduction

Modified low-density lipoprotein (LDL) is a significant source of various antigens in atherosclerosis as well as in a number of other diseases characterized by increased oxidative stress. Oxidized LDL (OxLDL) is highly immunogenic and elicits antibody formation against oxidation-specific epitopes originating from various end-products of lipid peroxidation, including reactive aldehydes, such as malondialdehyde (MDA), 4-hydroxynonenal and 1-palmitoyl-2-(5-oxovaleroyl)-sn-glycero-3-phosphocholine.¹

Both IgM and IgG antibodies to OxLDL can be detected in the plasma of healthy human subjects as well as in animal models of atherosclerosis, and their role as potential markers and functional components in human disease has been of interest. It has been demonstrated in humans that low levels of IgM antibodies to epitopes of OxLDL associate with increased carotid artery atherosclerosis.² High levels of IgM antibodies against phosphocholine (PC), OxLDL and MDA-LDL predict a decreased rate of progression of atherosclerosis in hypertensive patients³ and low IgM antibody levels against PC predict the risk of death in haemodialysis patients.⁴ Recently, IgG and IgM antibodies to oxidation-specific epitopes were shown to predict cardiovascular disease and stroke in a 15-year follow-up study; IgG to copper oxidized LDL (CuOxLDL) predicted a higher cardiovascular disease event rate, whereas IgM to MDA-LDL predicted a lower event rate.⁵ The recent general view has been that B-1 cell activation and production of natural IgM are associated with atheroprotection.⁶ Animal studies have implied that the IgM antibodies to OxLDL are natural antibodies of the innate immune system, and they cross-react with a wide variety of common epitopes shared by microbes⁷ and apoptotic cells.⁸ One of the key characteristics of natural antibodies to OxLDL implicated in atherogenesis is their ability to inhibit the uptake of OxLDL by macrophage scavenger receptors.⁹ So far, the roles of the adaptive humoral immune response and IgG antibodies to OxLDL in atherogenesis have been less studied and the data remain inconclusive, even though some early studies suggested a pro-atherogenic role for IgG antibodies to OxLDL.¹⁰

Chronic kidney disease and uraemia are known to be associated with enhanced atherosclerosis, and one important mechanism suggested is post-translational modification of LDL by carbamylation (homocitrullination).^11–13 This reaction is based on covalent binding of isocyanic acid, a decomposition product of urea, to proteins, lipids, amino acids and peptides.^14–17 A minimal degree of carbamylated plasma proteins and haemoglobin is found in healthy subjects, whereas patients with renal insufficiency have increased levels of carbamylation of various plasma proteins due to elevated plasma urea levels.¹⁸ Higher levels of carbamylated proteins have recently been associated with higher mortality in patients with end-stage renal disease.^19,20 Carbamylation in vivo may also occur by myeloperoxidase-catalysed oxidation of thiocyanate. Elevated thiocyanate levels are especially found in smokers. Myeloperoxidase-catalysed carbamylation has been linked to inflammation and atherosclerosis.²¹

We have previously shown in mice that carbamyl-LDL immunization induces a specific IgG immune response, which is cross-reactive with MDA-LDL.²² We have also shown that levels of IgG antibodies to carbamylated proteins are elevated in conditions known to induce enhanced carbamylation, such as uraemia and smoking.²² The aim of the current study was to investigate humoral antibody cross-reaction between carbamylated LDL and OxLDL in humans. Both carbamyl- and MDA-epitopes are found in humans and associated with increased atherosclerosis,^{19,20,23–26} which prompted us to investigate the association of human plasma antibodies binding to carbamyl-epitopes and oxidation-specific MDA-epitopes. An additional aim was to clone human monoclonal anti-carbamyl-Fab antibody by phage display technique and investigate the binding properties and cross-reactivity to carbamyl- and oxidation-specific epitopes. Cross-reactive antibodies may provide important new knowledge concerning the enhanced atherogenesis in uraemic patients.

Materials and methods

Human samples

Human blood samples (n = 42 for antibody measurements, n = 4 for phage display library construction) were collected from healthy volunteers in the Clinical Research Center of Oulu University Hospital and informed written consent was obtained from each participant. The studies were approved by the Ethics Committee of Oulu University Hospital, Oulu, Finland (21/2006 and 159/2001) and followed the Declaration of Helsinki.

LDL isolation and modifications

Low-density lipoprotein fraction (density 1·019–1·063 g/ml) was isolated by sequential density gradient ultracentrifugation from pooled plasma of healthy donors.²⁷ The LDL was carbamylated in vitro with potassium cyanate as previously described.²² First, 20 μm butylated hydroxytoluene (BHT) and 0·27 mm EDTA were added into freshly isolated LDL to minimize oxidation. Then, 2 mg LDL was diluted to 1·5 times the original volume with 0·3 m Na₂B₄O₇, pH 8·0 buffer and 20 mg potassium cyanate was added per mg of LDL. The LDL was carbamylated for 6 hr at 37°. In addition, carbamyl-LDL preparation was further tested for the absence of thiobarbituric acid reactive substances, and tested with monoclonal antibodies for the absence of oxidized phospholipids. Carbamylated albumin (using BSA) was prepared similarly with 24 hr incubation. The extent of lysine modification was determined with the 2,4,6-trinitrobenzene sulphonic acid method²⁸ and the amount of homocitrulline (carbamyl-lysine) was verified by amino acid analysis.²²

Malondialdehyde-modified LDL (MDA-LDL) and malondialdehyde acetaldehyde-modified LDL (MAA-LDL) were prepared as described previously.²⁹ Freshly prepared 0·5 m MDA was used to modify LDL with EDTA and BHT for 3 hr at + 37°: 0·5 m MDA was prepared from 1,1,3,3-tetramethoxypropane malonaldehyde-bis(dimethyl acetal) in 0·6% HCl and incubated at + 37° for 10 min. The pH was adjusted to 6·0–7·0 with NaOH and sterile water was added to a final volume of 4 ml. Then, 150 μl of 0·5 m MDA solution was used for MDA conjugation of 1 mg LDL protein. For MAA-modification, 0·5 m MDA pH 4·8 was prepared; 310 μl PBS, 140 μl 20% acetaldehyde, 5 mg LDL and 300 μl 0·5 m MDA were mixed in order. The pH was re-adjusted to 4·8 and the mixture was incubated at + 37° for 2 hr. MDA-BSA and MAA-BSA were prepared similarly. The extent of lysine modification was verified with the 2,4,6-trinitrobenzene sulphonic acid method²⁸ and the absence of homocitrulline (carbamyl-lysine) in native, MDA- and MAA-modified proteins was verified by amino acid analysis as described previously.²²

For copper oxidation, LDL without BHT was first extensively dialysed to remove EDTA and then oxidized by incubating LDL 1 mg/ml in PBS with 4 mm CuSO₄ at 37° for 24 hr. The reaction was stopped by addition of EDTA to a 200 μm final concentration. Modified LDL and BSA preparations were dialysed against PBS with 0·27 mm EDTA and sterile filtered.

Construction of phage display library

Total RNA was isolated from peripheral blood lymphocytes with the RNeasy Mini Kit (Qiagen, Hilden, Germany) and used to synthesize cDNA with Moloney murine leukaemia virus reverse transcriptase and oligo(dT)₁₈ primers included in the First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Waltham, MA). The cDNAs were used for generation of antibody libraries. The phage display library was constructed in three rounds of PCR with human Fab primers.³⁰ Heavy chain variable regions, κ-light chain variable regions and λ-light chain variable regions were first amplified separately using cDNA as a template. The constant regions of heavy chain and κ- and λ-light chains were amplified from the cloned human Fab template plasmids pComb3XTT and pComb3Xλ obtained from Dr C.F. Barbas III at the Scripps Research Institute, La Jolla, CA. In the second-round PCR the heavy and light chain overlap products were generated separately from the pooled first-round PCR products. The final full-length Fab-coding fragments were assembled in the third PCR from the second-round products. The Fab-coding fragments were digested with SfiI and cloned into the pComb3X phagemid vector. The precipitated and resuspended ligation mixtures were transformed into XL1-Blue Escherichia coli (Agilent Technologies, Santa Clara, CA) by electroporation (Gene Pulser electroporator and cuvette with 0·2-cm gap; Bio-Rad, Hercules, CA). After transformation, the bacteria were amplified and infected with the VCSM13 helper phage (Agilent Technologies). Phage particles were obtained from the overnight culture medium by 4% (weight/volume) polyethylene glycol-8000/3% (weight/volume) sodium chloride precipitation and centrifugation at 15 000 g for 15 min. Five rounds of panning against carbamyl-LDL were performed, and individual clones binding to carbamyl-LDL were selected with chemiluminescence immunoassay. The phagemid DNAs were isolated with the QIAprep Spin Miniprep kit (Qiagen, Hilden, Germany). The nucleotide sequences were confirmed and aligned to the germline genes with the IMGT/V-QUEST sequence alignment tool (http://www.imgt.org).^31,32

Chemiluminescence immunoassay method

Chemiluminescence immunoassays to detect plasma antibodies binding to specific antigens, or to test human Fab antibody binding to antigens were performed as previously described.^22,29,33 Antigens were immobilized on microtitre plates (Nunc Microfluor2, Thermo Fisher Scientific, Waltham, MA) overnight at + 4° in PBS with 0·27 mm EDTA. The plates were blocked with 0·5% gelatine in PBS with 0·27 mm EDTA for 30 min. The primary antibodies or human plasma samples (1 : 500) were diluted in 0·5% gelatine in PBS with 0·27 mm EDTA and incubated for 1 hr at room temperature or overnight at + 4°. The amount of bound molecules was measured with alkaline phosphatase-conjugated secondary antibodies, or alkaline phosphatase-conjugated Neutravidin (for biotinylated antibodies), LumiPhos 530 (Lumigen, Southfield, MI) as a substrate and a Wallac Victor³ (Perkin Elmer, Waltham, MA). The results are expressed as relative light units measured in 100 ms (RLU/100 ms). The specificity of human plasma antibodies and Fab106 antibody to carbamyl-LDL was tested using liquid-phase competition immunoassay. Plasma or antibody dilutions were incubated overnight at + 4° in the presence and absence of competitors (0–200 μg/ml). The immunocomplexes were pelleted by centrifugation for 30 min at 16 000 g, + 4° and antibodies remaining in the liquid phase were analysed using the chemiluminescence immunoassay described above.

Analysis of modified BSA on SDS–PAGE and Western blot

Native BSA, carbamyl-BSA, MDA-BSA and MAA-BSA (2 μg each) were reduced with β-mercaptoethanol and analysed on 10% SDS–PAGE and by Western blotting. Precision Plus Protein™ All Blue standard (Bio-Rad) was used as the molecular weight marker. The samples were transferred from the gel onto nitrocellulose membrane and blocked with 3% gelatine in PBS. The membrane was incubated with Fab106 antibody (2 μg/ml) for 1 hr at room temperature and the binding was detected with Fab-specific goat anti-human IgG (Sigma, St Louis, MO) and AlexaFluor 680-labelled donkey anti-goat (Invitrogen, Carlsbad, CA) antibodies (0·5 μg/ml 1 hr and 0·2 μg/ml 45 min at room temperature, respectively). The membrane was scanned with an Odyssey infrared scanner (LI-COR Biosciences, Lincoln, NE).

Biacore interaction analysis

Studies were performed with a Biacore T200 instrument at 25° using a standard CM5 sensor chip (GE Healthcare, Uppsala, Sweden). Dulbecco's PBS with 0·27 mm EDTA was used as a running buffer. The Biacore T200 instrument was cleaned according to the manufacturer's recommended methods before each measurement. Fab106 was immobilized on a Biacore CM5 chip. First, Fab106 was diluted into Biacore immobilization buffer (10 mm sodium acetate pH 5·0) to obtain a final concentration of 38 μg/ml. The new inserted CM5 chip was primed with immobilization buffer and equilibrated to 25°. The chip surface was activated by a 7-min injection of an earlier prepared mixture of 0·1 m N-hydroxysuccinimide and 0·4 m N-ethyl-N-(3-dimethylaminopropyl)carbodiimide in water with a flow rate of 10 μl/min. The Fab106 sample was then injected during 60 seconds, and residual activated groups on the chip surface were blocked by a 7-min injection of 1 m ethanolamine (pH 8·5). On average, 1500 resonance units (RU) were immobilized. The channel Fc2 was used for modification, channel Fc1 was used as a reference and left unmodified. Native BSA, carbamyl-BSA, MDA-BSA, MAA-BSA, native LDL, carbamyl-LDL, MDA-LDL and MAA-LDL were diluted into running buffer to obtain a final concentration within the range of 20 nm to 50 μm. Interaction analysis was performed by injecting each sample for 3 min with the flow rate of 30 μl/min. The chip was regenerated with a 30-second pulse of 1 m sodium chloride and a 30-second pulse of 0·05% digitonin. Data were analysed using Biacore T200 evaluation software v1.0.

Binding of Fab106 to apoptotic human T cells

Apoptosis was induced in human Jurkat T cells on a Petri dish by UV irradiation 51 mJ/cm² (UV Stratalinker; Stratagene, Santa Clara, CA) followed by overnight incubation in a humidified atmosphere with 5% CO₂ at + 37°. The cells were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum, 10 mm HEPES, 1 mm sodium pyruvate, 2 mm l-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. Then, 5 × 10⁵ cells were stained with anti-carbamyl-Fab106 antibody (5 μg/ml) and the binding was detected with FITC-labelled anti-human IgG (Fab-specific) (1 μg/ml, Sigma). Jurkat cells were also counterstained with 4′,6-diamidino-2-phenylindole (DAPI) nucleic acid stain according to the manufacturer's protocol (Invitrogen). Cells were imaged using a Zeiss Axio Imager.D2 microscope with a Colibri LED light source attached to an AxioCam MRm camera and AxioVision 4.8 software.

Fab106 binding to apoptotic cells (1 × 10⁶ cells) was verified by flow cytometry analysis. Direct staining with antibody (2·5 μg/ml) and competitive assay to investigate the binding specificity were performed. Fab106 in a final concentration of 3 μg/ml was incubated with or without competitors (250 μg/ml) carbamyl-LDL, MDA-LDL, native LDL, carbamyl-BSA, MDA-BSA, MAA-BSA and native BSA diluted in FACS buffer (0·1% BSA in PBS) overnight at + 4°. Samples were prepared in triplicates. Apoptotic Jurkat cells were washed with FACS buffer and centrifuged at 1800 g for 5 min. The competition samples were added and incubated for 40 min at + 4°. Cells were washed and incubated with FITC-labelled secondary anti-human IgG (Fab-specific) antibody (0·5 μg/ml, Sigma) for 40 min at + 4°. The washing was repeated and apoptotic cells were identified with propidium iodide (PI) staining (1 μg/ml). Binding of Fab106 to apoptotic T cells was analysed with a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA) and the data were analysed using FCS Express V3 software (De Novo Software, Los Angeles, CA).

Modified LDL uptake into J774A.1 mouse macrophages

Macrophage uptake assay was performed with ³H-labelled carbamyl-LDL and IRDye800-conjugated MDA-LDL as described earlier³⁴ with slight modifications. To prepare ³H-carbamyl-LDL, 40 mCi of tritium-labelled cholesteryl oleate [cholesteryl-1,2–³H(N)] (Perkin Elmer) per mg carbamyl-LDL was dried under nitrogen flow. Dimethylsulphoxide (10% of final volume) was added and vortexed for 1 min, mixed with carbamyl-LDL and incubated for 2 hr at 40° in a water bath followed by dialysis against 150 mm NaCl at + 4°. J774A.1 mouse macrophages were plated in 96-well plates (1 × 10⁵ cells/well) in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 100 U/ml penicillin and 100 μg/ml streptomycin and grown overnight at + 37° with 5% CO₂. The cells were washed and incubated at + 37° for 2 hr with ³H-carbamyl-LDL (2·5 μg/ml) in the presence or absence of unlabelled carbamyl-LDL as a competitor (75 μg/ml) or Fab106 antibody (305 μg/ml) mixed with labelled carbamyl-LDL. The cells were washed and lysed with 0·1 m NaOH for 2 hr in a plate shaker (1000 rpm) at room temperature. Cell lysate (50 μl or 25 μl) and 175 μl OptiPhase supermix scintillation liquid (Perkin Elmer) per well were added to a Wallac 96-well sample plate, shaked for 30 min (1000 rpm) and incubated for 15 min at room temperature. Cellular radioactivity was measured using a Wallac MicroBeta TriLux 1450 LSC & Luminescence Counter (Perkin Elmer). The cell protein concentrations were determined and radioactivity in cells (counts per minute, CPM)/μg cell protein was calculated.

The MDA-LDL was labelled with IRDye 800CW (LI-COR Biosciences) according to the manufacturer's protocol. IRDye800-conjugated MDA-LDL (1 μg/ml) in serum-free Dulbecco's modified Eagle's medium was incubated with macrophages (2 × 10⁵ cells/well) in the presence (50 μg/ml) or absence of the anti-carbamyl Fab106 for 3 hr at + 37°. A 50-fold excess of unlabelled MDA-LDL in separate wells was used as a control. The cells were washed three times with PBS and analysed with an Odyssey infrared imager (LI-COR Biosciences). After analysis the cells were lysed with 0·1 m sodium hydroxide overnight at room temperature, the amount of cell protein was measured, and the results were calculated as an infrared signal normalized with the protein contents of each well.

Immunohistochemical staining of mouse atherosclerotic lesions

Formalin-fixed and paraffin-embedded heart cross-sections of aortic origin from LDLR^−/− mice fed with a high-fat diet were stained with Fab106. As a control, aortic origin heart cross-sections from normal C57BL/6 mice without atherosclerosis were stained similarly. Antigen retrieval was performed by a 10-minute treatment with 10 mm sodium citrate (pH 6·0) near its boiling point. The cross-sections were stained with the Goat-on-Rodent HRP-polymer kit (Biocare Medical, Concord, CA) according to the kit protocol using DAB+ (3,3′-diaminobenzidine) chromogen (Dako, Glostrup, Denmark). Briefly, the endogenous peroxidase activity was quenched with the Peroxidase Block. Fab106 (5 μg/ml) or PBS was applied onto separate sections for 1 hr, followed by secondary goat anti-human IgG (Fab-specific) antibody (2 μg/ml, Sigma). HRP-labelled polymer conjugated with anti-goat antibody and DAB+ chromogen were used for detection. The cross-sections were further counterstained with Mayer's haematoxylin. Images were acquired with a Leica DM 3000 microscope and Leica Application Suite (LAS) V4.1.0 (Leica microsystems).

Statistical analysis

Data analysis was performed with SPSS Statistics 19.0. All the results for continuous variables are presented as mean ± SD. The differences between the groups were analysed by Student's t-test or Mann–Whitney U-test, as appropriate. Associations were analysed by Spearman's correlation coefficient. P-value < 0·05 was regarded as statistically significant. P-values are marked with asterisks in the figures as follows: *P < 0·05; **P < 0·01; ***P < 0·001.

Results

Human plasma contains antibodies binding to carbamyl-epitope that associate with antibodies to MDA-/MAA-adducts

Human plasma IgG and IgM binding to carbamyl-LDL, CuOxLDL, MDA-LDL and MAA-LDL and to carbamyl-BSA, MDA-BSA and MAA-BSA were measured from 42 healthy adults aged between 24 and 59 years (mean 36 ± 9 years). Statistically significant, strong positive associations were observed between plasma antibody levels, both IgG and IgM, to carbamylated and oxidation-specific epitopes in LDL (Fig. 1a) and BSA (Fig. 1b). Competitive immunoassays were then performed to study the specificity of human plasma antibodies bound to carbamyl-epitope. Carbamyl-LDL and MDA-LDL competed for binding of plasma IgG and IgM antibodies to carbamyl-LDL (Fig. 2a) and also to MDA-LDL (Fig. 2b). Carbamyl-BSA, MDA-BSA and MAA-BSA also competed for plasma IgG binding to carbamyl-BSA (Fig. 2c) and MDA-BSA (Fig. 2d). MDA- and MAA-adducts were more efficient soluble competitors than carbamylated proteins when measuring polyclonal plasma IgG and IgM binding to immobilized carbamylated antigens (Fig. 2a,c). Plasma IgM antibody levels to carbamyl-BSA and MDA-BSA were very low, and therefore competition assay was not performed.

Figure 1.

Association of human plasma anti-carbamyl antibodies with antibodies to oxidation-specific epitopes. (a) Plasma samples of healthy adult humans (n = 42) were analysed for IgG and IgM binding to carbamyl-low-density lipoprotein (LDL), copper oxidized LDL (CuOxLDL), malondialdehyde (MDA) -LDL and malondialdehyde acetaldehyde (MAA) -LDL. (b) Assays were also performed using carbamylated and MDA- or MAA-modified BSA. Associations were determined using Spearman rank correlation coefficient. RLU = relative light unit.

Figure 2.

Specificity of human plasma antibodies to carbamyl epitopes. Specific binding of human plasma IgM and IgG to immobilized (a) carbamyl-low-density lipoprotein (LDL) and (b) malondialdehyde (MDA) -LDL was analysed using carbamyl-LDL and MDA-LDL as competitors in a liquid-phase competition immunoassay. Specificity of human plasma IgG to (c) carbamyl-BSA and (d) MDA-BSA was analysed using carbamyl-BSA, MDA- and malondialdehyde acetaldehyde (MAA) -BSA as soluble competitors. Mean (SD) of four samples is shown.

Human monoclonal anti-carbamyl Fab antibody (Fab106) cross-reacts with oxidation-specific epitopes

The origin and specificity of the human antibodies to carbamylated proteins were further investigated by constructing a phage display Fab-antibody library from peripheral blood lymphocytes of four healthy humans. After five rounds of panning against carbamyl-LDL, several phage clones were randomly selected (n = 19) from both V_H/V_κ and V_H/V_λ libraries, all demonstrating binding to carbamyl-LDL. The characteristics of 10 human monoclonal anti-carbamyl Fab antibody clones including V_H/D_H/J_H and V_L/J_L gene use are presented in Table 1. Nine out of the ten clones originated from the large V_H3 family (six from V_H3-33, two from V_H3-23 and one from V_H3-66 germline gene), and only one clone originated from the V_H4 family (V_H4-39 germline gene). The D_H motifs were very variable, but all clones demonstrated an almost exclusive use of J_H4. In addition, the V_H sequences demonstrated a surprisingly high degree of identity to the corresponding germline genes (between 86% and 100%).

Table 1.

Characteristics of 10 human monoclonal anti-carbamyl-Fab clones

Clone	VDJ-usage	Identity of V-region to germline (%)
1	VH3-33*01/DH6-25*01/JH4*02	100·0
	Vκ3-20*01/Jκ1*01	97·3
21	VH3-33*01/DH6-25*01/JH4*02	100·0
	Vκ3-20*01/Jκ1*01	91·8
3	VH3-33*01/DH3-10*01/JH4*02	86·1
	Vκ1-39*01/Jκ1*01	90·7
4	VH3-23*01/DH3-3*01/JH4*02	100·0
	Vκ1-39*01/Jκ3*01	89·9
5	VH4-39*01/DH1-26*01/JH4*03	91·1
	Vκ1-5*03/Jκ1*01	90·6
6	VH3-33*01/DH3-3*01/JH4*02	99·2
	Vλ3-1*01/Jλ3*02	57·2
7	VH3-33*01/DH2-02*01/JH4*02	93·2
	Vλ2-14*03/Jλ2*01	73·6
8	VH3-23*01/DH5-18*01/JH4*02	97·0
	Vλ3-1*01/Jλ2*01	93·2
9	VH3-33*01/DH2-8*01/JH4*02	93·5
	Vλ10-54*01/Jλ3*02	90·7
10	VH3-66*02/DH3-3*01/JH4*03	93·5
	Vλ2-28*01/Jλ2*01	100·0

¹Clone IFP-02_106 (Fab106).

One clone (IFP-02_106) was chosen for the production of soluble Fab antibody (Fab106). The binding of Fab106 to native and modified LDL and BSA was tested with chemiluminescence immunoassay (Fig. 3a). Fab106 showed a dose-dependent increase in binding to carbamyl-LDL, but also to MDA-LDL, MAA-LDL and CuOxLDL (Fig. 3a). Similar binding was observed to carbamylated BSA and MDA-/MAA-modified BSA. Fab106 did not, however, bind to PC conjugated to BSA (PC-BSA), which suggests that binding to CuOxLDL was a result of the small amount of MDA-epitopes generated during preparation of CuOxLDL¹ (Fig. 3a). The binding specificity of Fab106 to carbamyl-modification was determined in liquid-phase competition immunoassay (Fig. 3b). Anti-carbamyl-Fab binding to carbamyl-LDL was competed out by adding an increasing amount of carbamyl-BSA and carbamyl-LDL, but not by native LDL or native BSA (Fig. 3b). This suggests that the antibody was specific for the carbamyl-epitope, and not the carrier protein. Fab106 binding to carbamyl-LDL was also competed with MDA-LDL and MAA-LDL (Fig. 3b) demonstrating cross-reactive epitopes in carbamylated and MDA-/MAA-modified proteins. In addition, CuOxLDL was shown to compete for Fab106 binding to carbamyl-LDL but no competition for PC-BSA was observed, verifying that the MDA-epitopes in CuOxLDL were recognized by Fab106 antibody (Fig. 3b). Native BSA, carbamyl-BSA, MDA-BSA and MAA-BSA were analysed on reducing SDS–PAGE (Fig. 3c) and Western blotted with Fab106 antibody (Fig. 3d). Fab106 bound to carbamyl-BSA and also to MDA-BSA and MAA-BSA, but not to native BSA in Western blot analysis (Fig. 3d).

Figure 3.

Characterization of binding properties of Fab106. (a) Direct binding assay for anti-carbamyl-Fab (Fab106). Increasing amount of Fab106 (0–20 μg/ml) bound to native low-density lipoprotein (LDL), carbamyl-LDL, malondialdehyde (MDA) -LDL, malondialdehyde acetaldehyde (MAA) -LDL and copper oxidized LDL (CuOxLDL) (upper panel) or to native BSA, carbamyl-BSA, MDA-BSA, MAA-BSA and phosphocholine (PC) -BSA (lower panel). (b) Competitive liquid phase immunoassay to demonstrate Fab106 binding specificity. Fab106 binding to immobilized carbamyl-LDL was tested when native LDL, carbamyl-LDL, MDA-LDL, MAA-LDL and CuOxLDL and also native BSA, carbamyl-BSA, MDA-BSA, MAA-BSA and PC-BSA were used as competitors. RLU = relative light unit. (c) SDS–PAGE of modified BSA and (d) Western blot with Fab106 antibody. Molecular weight marker (lane 1), native BSA (lane 2), carbamyl-BSA (lane 3), MDA-BSA (lane 4) and MAA-BSA (lane 5).

Fab106 binds with high affinity to carbamyl- and MDA-/MAA-adducts

The steady-state binding affinity of monoclonal Fab106 antibody to carbamyl- and MDA-/MAA-modified antigens (BSA, Fig. 4a and LDL, Fig. 4b) was tested by surface plasmon resonance analysis. Fab106 bound most strongly to MDA-LDL and MAA-LDL with steady-state affinity (K_D) values 5·24 × 10⁻⁶ m and 4·07 × 10⁻⁶ m, respectively. Binding to carbamyl-LDL was also high compared with native LDL, with the K_D value 2·13 × 10⁻⁵ m. Fab106 also bound modified BSA antigens, but with a slightly weaker affinity than modified LDL. K_D values for MAA-BSA, MDA-BSA and carbamyl-BSA were 3·28 × 10⁻⁵ m, 5·09 × 10⁻⁴ m and 1·58 × 10⁻⁴ m, respectively. Binding affinity to native LDL and native BSA was very weak, with K_D values 8·08 × 10⁻¹ m for LDL and 9·70 × 10⁻² m for BSA. Steady-state affinity K_D (m) values are shown in Fig. 4(c).

Figure 4.

Biacore interaction analysis. Steady-state affinity curves for immobilized Fab106 bound to soluble antigens (a) native BSA, carbamyl-BSA, malondialdehyde (MDA) -BSA and malondialdehyde acetaldehyde (MAA) -BSA and also to (b) native low-density lipoprotein (LDL), carbamyl-LDL, MDA-LDL and MAA-LDL. (c) Steady-state affinity K_D (m) values.

Fab106 binds to apoptotic Jurkat cells and binding can be inhibited with carbamyl-, MDA- and MAA-modified proteins

To study further the cross-reactivity between oxidation-specific epitopes and carbamyl-epitopes, the binding of Fab106 to apoptotic cells was investigated. Previously, apoptotic cells have been shown to contain oxidation-specific epitopes.⁸ Fab106 demonstrated binding to apoptotic cells in immunofluorescence microscopy (Fig. 5a, b). Fab106 binding to Jurkat T cells with UV-induced apoptosis was further investigated in a flow cytometry assay (Fig. 5c–i). Fab106 did not bind to non-apoptotic, PI-negative cells (Gate 1, Fig. 5d–f), but bound to PI-positive late apoptotic cells (Gate 2, Fig. 5g–i). Fab106 bound approximately 30% of the apoptotic cell population, as shown in Fig. 5(h). The binding of Fab106 to apoptotic cells was reduced to 18% when carbamyl-LDL was added or to 8% when carbamyl-BSA was added as a competitor (Fig. 6). Fab106 bound < 4% of the apoptotic cells when MDA-LDL, MDA-BSA or MAA-BSA was used as competitor (Fig. 6). This provides further evidence for the Fab106 cross-reaction between carbamylated and MDA-/MAA-epitopes. MAA-LDL was not used as a competitor in this flow cytometry assay, because of its tendency to form aggregates when added to apoptotic cells. Native LDL and native BSA did not compete for Fab106 binding to apoptotic Jurkat cells (Fig. 6). The binding of the negative control Fab antibody (clone IFUp-08_109) to Jurkat cells was also tested as described in the Supplementary material (Data S1 and Fig. S1).

Figure 5.

Fab106 binding to apoptotic Jurkat T cells. Immunofluorescence microscopy image of Fab106 bound to apoptotic Jurkat cells (a) and secondary antibody control (b). Fab antibody bound to apoptotic cells is shown in green (FITC) and nuclear DNA staining in blue (DAPI). (c–i) Flow cytometry assay for non-apoptotic, propidium iodide (PI) -negative (Gate 1, d-f) and apoptotic, PI-positive (Gate 2, g-i) Jurkat cells after UV irradiation of 51 mJ/cm². Density plots demonstrate secondary antibody (d, g) and Fab106 antibody (e, h) bound to Jurkat cells and histograms (f, i) illustrate apoptotic cells only (black), secondary antibody binding (red) and Fab106 antibody binding (blue) to Jurkat cells.

Figure 6.

Flow cytometry competition assay. Fab106 binding to apoptotic Jurkat cells was analysed with flow cytometry using soluble competitors carbamyl-low-density lipoprotein (LDL), malondialdehyde (MDA) -LDL, native LDL, carbamyl-BSA, MDA-BSA, malondialdehyde acetaldehyde (MAA) -BSA and native BSA. Propidium iodide (PI) -positive apoptotic Jurkat cells were gated and cells in the upper left (UL) quadrant of forward scatter versus FL-1 density plots were analysed (see Fig.  5g and 5h). Samples were analysed in triplicates.

Fab106 blocks the uptake of carbamyl-LDL and MDA-LDL by macrophages and binds to epitopes in mouse atherosclerotic lesion

Uptake of modified LDL into macrophages is a key step in the pathogenesis of atherosclerosis. The functional role of antibodies to modified lipid and protein epitopes in vivo is not fully understood. The role of Fab106 antibody was investigated in macrophage uptake assay (Fig. 7). Fab106 was able to significantly inhibit the uptake of ³H-carbamyl-LDL by J774A.1 mouse macrophages, a 63% decrease (P < 0·001) was observed when Fab106 antibody was added together with labelled carbamyl-LDL (Fig. 7a). A similar assay was performed using IRDye800-labelled MDA-LDL. Fab106 was able to inhibit 33% (P < 0·001) of the IRDye800-MDA-LDL uptake by J774A.1 macrophages (Fig. 7b). Fab106 was also used for staining mouse heart aortic origin cross-sections. The Fab106 antibody bound to epitopes found in advanced atherosclerotic lesions that develop within intima in LDLR^−/− mice (Fig. 8). No similar staining was observed in normal aortic origin cross-sections from C57BL/6 mice without atherosclerotic lesions (see Supplementary material, Fig. S2).

Figure 7.

Carbamyl-low-density lipoprotein (LDL) and malondialdehyde (MDA) -LDL uptake by mouse J774A.1 macrophages. Uptake of (a) ³H-labelled carbamyl-LDL and (b) IRDye800-labelled MDA-LDL was tested in the presence and absence of Fab106. 30× or 50× excess of unlabelled carbamyl- or MDA-LDL was used as an assay control.

Figure 8.

Immunohistochemical staining of atherosclerotic lesions in mouse heart aortic origin cross-section with the anti-carbamyl Fab106 antibody. Secondary antibody control (a) and Fab106 staining (b, c). Magnification 5× (a, b) and 10× (c), scale bar = 400 μm.

Discussion

Both carbamyl-epitopes and oxidation-specific MDA-LDL epitopes are found in humans^23,24,35 and have been shown to contribute to the development of atherosclerosis.^19,20,25,26 Furthermore, both epitopes are immunogenic and they induce antibody production.^22,36–38 Previously, we showed that carbamyl-LDL immunization induces a specific IgG immune response that cross-reacted with MDA-adducts in mice.²² This prompted us to study further if cross-reactive antibodies to carbamyl-adducts and MDA-adducts exist in humans.

Carbamyl-adducts are generated in vivo in reaction with isocyanate formed by dissociation from urea or by myeloperoxidase catalysed oxidation of thiocyanate,²¹ and the chemistry has been well described.^39,40 Carbamylation of LDL is suggested to contribute to a higher prevalence of atherosclerosis in uraemic patients^11–13 and protein carbamylation has been linked with mortality in patients with end-stage renal disease in two separate studies.^19,20 Protein-bound homocitrulline¹⁹ and carbamylated serum albumin²⁰ were detected in patients with end-stage renal disease, and the amount of carbamylated protein was associated with mortality in these patients during the follow-up period. Carbamyl-epitopes are found in atherosclerotic plaque and have been shown to co-localize with myeloperoxidase.²¹ Other epitopes found in atherosclerotic lesions include oxidation-specific lipid epitopes such as MDA-/MAA-adducts.

In this study we showed a specific cross-reactivity between carbamyl- and MDA-/MAA-adducts with the monoclonal antibody Fab106. This is an interesting finding because there seems to be no similar chemistry in the modifications. During atherogenesis LDL is accumulated in the arterial wall and exposed to oxidation, leading to formation of OxLDL. Oxidation-specific epitopes are generated when highly reactive lipid peroxidation end-products modify proteins and lipids.¹ One of the abundant dialdehydes generated during lipid peroxidation is MDA, which reacts widely with primary amines to create MDA-adducts.⁴¹ Acetaldehyde, another oxidation product, forms MAA-adducts in the presence of MDA.^42,43 CuOxLDL is commonly used in laboratory experiments and contains a wide range of different epitopes, including MDA-adducts and, as the major epitope, PC-head groups of oxidized phospholipids, e.g. 1-palmitoyl-2-(5′-oxovaleroyl)-sn-glycero-3-phosphocholine.⁴⁴ These epitopes are also found on apoptotic cells⁸ and atherosclerotic lesions,⁴⁴ and have been suggested to participate in the immune modulation of atherogenesis. The wide diversity of epitopes on OxLDL and apoptotic cells is recognized by antibodies that may have various effects on atherogenesis. The human Fab106 monoclonal antibody was cloned against carbamyl-LDL and cross-reacted with oxidation-specific MDA-/MAA-epitopes, but not with PC-epitopes. Also, Fab106 bound to epitopes found on apoptotic cells and in the atherosclerotic lesion. Whether Fab106 recognizes oxidation-specific epitopes or carbamyl-epitopes on apoptotic cells remains to be studied.

Published data concerning the antibodies to carbamylated LDL or carbamylated proteins are sparse. Here we demonstrated that healthy humans have antibodies to carbamylated LDL and carbamylated proteins, and they are associated with antibodies to oxidation-specific epitopes. We showed an association on both IgM and IgG isotypes, suggesting that the roles of these antibodies could be both natural and adaptive in vivo. We and others have also shown that IgG antibodies to carbamylated proteins are associated with end-stage renal disease²² and rheumatoid arthritis.^45,46 In both of these chronic diseases, the patients have a higher incidence of atherosclerotic cardiovascular disease that cannot be explained only by traditional risk factors.^47–50 In our previous study we showed IgG and IgM antibodies to carbamylated LDL and carbamylated proteins in patients with end-stage renal disease as well as in healthy subjects.²² The levels of IgG antibodies were significantly higher in uraemic patients than in healthy controls, and also higher in smokers than in non-smokers within the study groups.²² Human IgG and IgA antibodies recognizing carbamylated proteins in the sera of patients with rheumatoid arthritis are shown to predict a more severe disease course in patients that tested negative for antibodies against citrullinated protein antigens, which are used in the prognosis and diagnosis of rheumatoid arthritis.⁴⁵ Anti-carbamyl antibodies have also been found in patients with arthralgia and are shown to predict the development of rheumatoid arthritis.⁴⁶ Citrulline and homocitrulline (carbamyl-lysine) have similar structures; they differ only by one methylene (-CH₂) group, homocitrulline being longer.⁵¹ A recent study revealed that also citrulline-epitopes are present in atherosclerotic plaque and could be a target for anti-citrullinated protein antibodies in patients with rheumatoid arthritis, and the resulting immune complexes could promote atherogenesis in these patients.⁵² Antibodies to both citrullinated and carbamylated proteins exist in patients with rheumatoid arthritis.^45,46 An interesting question is whether these two chronic diseases, namely rheumatoid arthritis and chronic kidney disease, share some common factors that have an impact on atherogenesis. Both diseases demonstrate antibodies to carbamylated proteins and increased risk for atherosclerotic cardiovascular disease. Our present data suggest that human antibody cross-reactivity between carbamylated and oxidation-specific epitopes may be involved in enhanced atherogenesis in these patients.

The role of antibodies in the development of atherosclerosis remains inconclusive. The plasma antibodies and the relation to atherosclerosis have been investigated in several studies that have used different detection methods and different target antigens in the assays, and the results are difficult to compare. Technical issues related to unstandardized methods of measuring these antigen-specific antibodies might lead to divergent results and interpretation. Some early studies have shown that IgG antibodies to OxLDL predict atherosclerotic disease.¹⁰ Antibody levels to OxLDL, when measured as a ratio of anti-OxLDL and anti-native LDL, have been shown to be elevated in patients with chronic kidney disease and uraemia.^53–56 The anti-OxLDL antibodies could be markers of oxidative stress during atherogenesis. Also, an inverse association of IgG antibodies to OxLDL and coronary artery disease has been shown, and the study suggested that the IgG immunocomplexes formed affect the measurement of free IgG antibodies.⁵⁷ The IgM antibodies, however, have been shown to be associated with diminished atherosclerosis in several studies.^2,3,5 Natural IgM antibodies to oxidation-specific epitopes have been cloned from mice and shown to have atheroprotective properties, e.g. to block the uptake of OxLDL by macrophages.^9,58

In this study, we cloned from healthy humans a monoclonal Fab antibody with high germline homology and cross-reactivity between carbamylated and oxidation-specific epitopes, and apoptotic cells. These data suggest a natural origin of the Fab106 antibody. Natural antibodies are usually an IgM isotype and share high identity with germline genes. The isotype origin of Fab106 remains unknown because of the technique used in constructing the phage display Fab library. Only Fab variable regions were amplified from the cDNA, whereas antibody isotype-determining constant regions were amplified from the template cloned into the pComb3X vector. Natural IgM antibodies have low affinity, but they are polyreactive and can successfully bind previously unencountered antigens.^59,60 IgM antibodies also prime the adaptive IgG immune response; the primary response after antigen encounter is low-affinity IgM followed by the secondary response of high-affinity IgG.⁵⁹ Additionally, one important property of natural IgM antibodies is to maintain cellular homeostasis and participate in the clearance of apoptotic cells.⁶¹ It has been shown that IgM, but not IgG, antibodies bind to late apoptotic PI-positive human Jurkat T cells via the antibody's Fab domain after trypsin digestion.⁶²

The Fab antibody cloned in this study was originally selected for binding to carbamyl-LDL but was observed to also bind carbamyl-adducts in other proteins and to cross-react with oxidation-specific epitopes, especially with MDA- and MAA-modified proteins. Previously published human monoclonal Fab antibodies binding to MDA-LDL⁶³ and OxLDL⁶⁴ have been cloned from phage display libraries constructed from atherosclerotic patients. Sequence analysis of these human Fabs against OxLDL shows high homology with germline genes, ability to block the uptake of OxLDL by macrophages and binding to epitopes found in atherosclerotic lesions. The Fab106 antibody to carbamyl-LDL, although cloned from healthy humans, showed similar functional properties to the previously published monoclonal Fab antibodies to OxLDL. The Fab106 shared homology with germline genes, bound to epitopes in atherosclerotic plaque and inhibited the uptake of modified LDL by macrophages, which is a crucial step in foam cell formation and atherogenesis. Nevertheless, Fab106 can be distinguished from the previously characterized human Fabs to OxLDL in its ability to cross-react with carbamylated and MDA-/MAA-modified oxidation-specific epitopes but not with PC-epitopes. The polyclonal repertoire of human plasma antibodies also recognized the cross-reactivity between carbamylated proteins and MDA-/MAA-adducts supporting the presence of natural antibodies that possess features equivalent with the Fab described in this study.

Conclusions

We demonstrated an association between antibodies binding to carbamyl- and oxidation-specific malondialdehyde-derived adducts in human plasma, and also cloned a human monoclonal Fab antibody with characteristics of a natural antibody and ability to bind both carbamyl- and malondialdehyde-derived epitopes. This cross-reactivity between antibodies binding to carbamylated and oxidation-specific epitopes, and their possible role in explaining the link between atherogenesis and kidney disease will require additional studies.

Glossary

BHT

butylated hydroxytoluene

CPM

counts per minute

CuOxLDL

copper oxidized LDL

DAPI

4′,6-diamidino-2-phenylindole

K_D

steady state affinity

LDL

low-density lipoprotein

MAA

malondialdehyde acetaldehyde

MDA

malondialdehyde

OxLDL

oxidized low-density lipoprotein

PC

phosphocholine

PI

propidium iodide

RLU

relative light unit

RU

resonance unit

Author contributions

OK, SPT, PP, JL, MV, CW and SH designed the study, OK, SPT, PP, JL, MV and CW performed the experiments, OK, SPT, PP, JL, MV, CW and SH wrote the paper.

Disclosures

The authors declare no conflicting interests.

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Data S1

Methods.

Figure S1

Fab binding to apoptotic Jurkat T cells.

Figure S2

Immunohistochemical staining of normal C57BL/6 mouse heart aortic origin cross-section with the anti-carbamyl Fab106 antibody.