Peptidylarginine Deiminase 2 Suppresses Inhibitory κB Kinase Activity in Lipopolysaccharide-stimulated RAW 264.7 Macrophages*

Peptidylarginine deiminases (PADs) are enzymes that convert arginine to citrulline in proteins. In this study, we examined PAD-mediated citrullination and its effect on pro-inflammatory activity in the macrophage cell line RAW 264.7. Citrullination of 45–65-kDa proteins was induced when cells were treated with lipopolysaccharide (LPS; 1 μg/ml). Protein citrullination was suppressed by the intracellular calcium chelator BAPTA/AM (30 μm). LPS treatment up-regulated COX-2 levels in cells. Interestingly, overexpressing PAD2 reduced LPS-mediated COX-2 up-regulation by 50%. PAD2 overexpression also reduced NF-κB activity, determined by NF-κB-driven luciferase activity. The effect of PAD2 on NF-κB activity was further examined by using HEK 293 cells transfected with NF-κB luciferase, IκB β/γ kinase (IKKβ/γ) subunits, and PAD2. IKKβ increased NF-κB activity, but this increase was markedly suppressed when PAD2 was present in cells. IKKβ-mediated NF-κB activation was further enhanced by IKKγ in the presence of calcium ionophore A23187. However, this stimulatory effect of IKKβ/γ was abolished by PAD2. Coimmunoprecipitation of cell lysates showed that IKKγ and PAD2 can coimmunoprecipitate in the presence of the Ca2+ ionophore. IKKγ coimmunoprecipitated truncation mutants, PAD2(1–385) and PAD2(355–672). The substitution of Gln-358 (a putative ligand for Ca2+ binding) with an Ala abolished coimmunoprecipitation. Conversely, PAD2 coimmunoprecipitated truncation mutants IKKγ(1–196) and IKKγ(197–419). In other experiments, treating RAW 264.7 cells with LPS induced citrullination in the immunoprecipitates of IKKγ. In vitro citrullination assay showed that incubation of purified PAD2 and IKKγ proteins in the presence of Ca2+ citrullinated IKKγ. These results demonstrate that PAD2 interacts with IKKγ and suppresses NF-κB activity.

Protein citrullination can be induced by inflammatory stimuli such as LPS and TNF (10). Citrullination of proteins such as histones regulates gene transcription (11,12). Smoking increases PAD expression and citrullination in human lung (13). Among five different PAD isotypes identified in mammals (2), PAD2 and PAD4 are responsible for citrullination in macrophages (14,15). PAD2 protein is observed in macrophages, although the mRNA is expressed in monocytes and monocytes-derived macrophages (14). PAD4 is expressed mainly in the granulocytes, lymphocytes, endothelial cells, monocytes, and macrophages (16 -18). PAD4 is predominantly found in nuclei in these cells. However, the precise role of the citrullination in diseases and inflammation remains unclear. Moreover, the physiological role of PADs in macrophages has yet to be fully determined.
Macrophages are effector cells that regulate inflammatory responses in a variety of tissues including lung. The lung harbors abundant alveolar macrophages and is constantly challenged by inflammatory stimuli. Macrophages are the major responders to LPS (endotoxin) (19,20), and LPS-activated macrophages produce pro-inflammatory proteins including cyclooxygenase-2 (COX-2) (21). COX-2 is responsible for formation of prostaglandins (21)(22)(23), which promote inflammation through a variety of mechanisms. The synthesis of COX-2 is mediated by nuclear translocation and activation of NF-B that normally resides in the cytoplasm and forms a protein complex with inhibitory B (IB). The dissociation of NF-B from IB is mediated by IB kinase (IKK) signalosome, which consists of two catalytic subunits, IKK␣ and IKK␤, and a regulator, IKK␥. IKK␥ (also known as NF-B essential modulator; NEMO) is an important regulatory component of the IKK complex. When cells are stimulated by LPS, IKK signalosome is activated to cause NF-B dissociation from IB␣, and NF-B is translocated to nucleus, where it activates target genes (24,25).
In this study, we tested whether PAD2-mediated citrullination affects NF-B activity in the mouse leukemic monocyte macrophage cell line RAW 264.7. We examined LPS-induced citrullination, NF-B luciferase activity, and in vitro interaction between PAD2 and IKK␥. Our data demonstrate that PAD2 suppresses NF-B activity probably by interacting with IKK␥.

EXPERIMENTAL PROCEDURES
Reagents-Ultrapure TLR4-specific LPS from Escherichia coli (Alexis Biochemicals, San Diego, CA) was directly added to culture medium. The intracellular calcium chelator BAPTA/AM (Calbiochem) was dissolved in dimethyl sulfoxide (DMSO), and the calcium ionophore A23187 (Sigma-Aldrich) was dissolved in ethanol.
RT-PCR and Full-length cDNA for PAD2-Total RNA was extracted from untreated RAW 264.7 cells using the RNeasy mini kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. cDNA was synthesized from total RNA by using the SuperScript TM first-strand synthesis system for RT-PCR (Invitrogen). To confirm the presence of PAD2 and PAD4 in RAW 264.7 cells, we performed PCR. For PCR, two sets of primers were designed based on the published sequence of mouse PAD2 (GenBank TM accession number: NM_008812.1). The forward primer was 5Ј-CTTCAAGATG-GATGAAAATCACCAGG-3Ј, and the reverse primer was 5Ј-CACCATGTGCCACCACTTGAAGGC-3Ј, which amplify a 277-bp PAD2 product. The forward primer of the other sets was 5Ј-GTTATGTTCAAGGGCCTGGGAGGCATG-3Ј, and the reverse primer was 5Ј-TAGCACGATCATGTTCAC-CATGTTAGG-3Ј, which amplify a 246-bp PAD2 product. PCR was performed for 30 cycles with Pfu DNA polymerase (Promega, Madison, WI) at 94°C for 40 s, 60°C for 30 s, and 72°C for 40 s. The final extension time at 72°C was 7 min. The products were sequenced in the DNA Sequencing Facility at Vanderbilt University.
The full-length PAD2 cDNA was cloned by PCR with primers corresponding to the 5Јand 3Ј ends of mouse PAD2. The sequence of the full-length PAD cDNA was identical to the published PAD2 of mouse (the amino acid sequence of cloned PAD2 is available in the supplemental material).
Cell Transfection-The murine macrophage cell line RAW 264.7 and human embryonic kidney (HEK) 293 cells were purchased from the American Type Culture Collection (Manassas, VA). Cells were cultured in DMEM (Invitrogen) supplemented with 10% endotoxin-free, heat-inactivated fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 g/ml) in a 5% CO 2 atmosphere at 37°C in a humidified incubator. Transfection was achieved with Geneporter TM 2 transfection reagent (Gene Therapy Systems, San Diego, CA) for RAW 264.7 cells and Lipofectamine 2000 (Invitrogen) for HEK 293 cells according to the manufacturer's instructions. Cells at 24 h after transfection were used for experiments.
Detection of Citrullinated Proteins-Protein citrullination was analyzed by immunoblot with an antibody specific to modified peptidylcitrulline residues. Cells were lysed with radioimmunoprecipitation assay cell lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% sodium orthovanadate, 1% Triton X-100, 0.5% deoxycholate, 0.1% sodium dodecyl sulfate) supplemented with EDTA-free protease inhibitor cocktails (Roche Diagnostics). Cell debris was removed by microcentrifugation for 10 min, and protein concentration was determined by the Bradford assay reagent (Bio-Rad). Equal amounts of proteins were separated by SDSpolyacrylamide gel electrophoresis, transferred to a nitrocellulose membrane, and then cross-linked by formaldehyde incubation to improve protein retention. Citrullinated proteins were detected using the anti-citrulline (modified) detection kit (Millipore, Billerica, MA) with modified anti-citrulline rabbit polyclonal antibody (1:1000) and a goat anti-Rabbit IgG antibody conjugated with horseradish peroxidase (GAR-HRP) (1:5000; Millipore). Citrullinated protein bands were detected using the enhanced chemiluminescence ECL Plus (GE Healthcare). As an internal control, the membrane was stripped and reprobed with ␤-actin antibody (Santa Cruz Biotechnology, Santa Cruz, CA).
Immunoblot for COX-2 in RAW 264.7 Cells-Equal amounts of proteins were separated by SDS-polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane. The membrane was preincubated in phosphate-buffered saline (PBS) containing 0.05% Tween 20 and 5% nonfat dry milk and then incubated with polyclonal anti-COX-2 (Cayman Chemical) at room temperature for 2 h. The blot was washed with PBS containing 0.05% Tween 20 for 20 min (4 ϫ 5 min) and then incubated with GAR-HRP (1:5000) for 1 h. Tubulin (Santa Cruz Biotechnology) was used as an internal control.
Immunoprecipitation-HEK 293 cells were co-transfected with (i) HA-tagged IKK␥ (WT) and FLAG-tagged PAD2 (WT) or truncation mutants or (ii) FLAG-tagged PAD2 (WT) and HA-tagged IKK␥ (WT) or truncation mutants for determining domains responsible for interaction. At 24 h after transfection, cells were treated with either 1 M A23187 or no A23187 in serum-free medium for 1 h. Cell lysates were pre-pared in immunoprecipitation buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1.5 mM EDTA, 3% glycerol, 0.5% Nonidet P-40, 1 mM NaF, and 1 mM Na 3 VO 4 ) supplemented with 1 mM dithiothreitol and EDTA-free protease inhibitor cocktails (Roche Diagnostics). Lysates were incubated with the monoclonal anti-HA antibody (1-2 g) (Cell Signaling, Danvers, MA) at 4°C overnight and then with protein A-Sepharose beads (Zymed Laboratories Inc., San Francisco, CA) at 4°C for 30 min. The beads were washed with immunoprecipitation buffer containing 250 mM NaCl more than five times, and the immunoprecipitates were dissociated from the beads by adding SDS-sample buffer and boiling. The immunoprecipitates were subjected to immunoblotting with monoclonal anti-FLAG M2 antibody (Sigma-Aldrich). Cell lysates for input were immunoblotted with monoclonal anti-FLAG M2 antibody or monoclonal anti-HA antibody and with a goat horseradish peroxidase-conjugated anti-mouse IgG antibody (1: 5000; Millipore). For immunoprecipitation of endogenous IKK␥ from RAW 264.7 cells, cells were treated with BAPTA/AM (30 M) for 1 h and then with LPS (1 g/ml) for 4 h. Cells were then lysed with immunoprecipitation buffer and subjected to immunoprecipitation with anti-IKK-␥ antibody (BD Pharmingen). After washes, protein citrullination was determined. To separate immunoprecipitated IKK␥ (48 kDa) from IgG heavy chains (50 kDa), the SDS sample buffer without ␤-mercaptoethanol was used for SDS-PAGE.
NF-B Reporter Gene Assay-The NF-B reporter assay was performed by the Dual-Luciferase assay kit (Promega) as described previously (22). Briefly, RAW 264.7 cells were transfected with NF-B-Luc reporter plasmid, and 24 h after transfection, cells were treated with LPS (1 g/ml) for 4 h. For HEK 293 cells, cells were transfected in combination with NF-B-Luc, FLAG-tagged PAD2, FLAG-tagged-IKK␤, and HA-tagged IKK␥. Cells at 24 h after transfection were then treated with A23187 (1 M) for 1 h to induce PAD2 activation. After washes with PBS, cell extracts were prepared with lysis buffer, and 20-l aliquots of the supernatant were mixed with 100 l of luciferase assay solution provided by the company. Luciferase activity was measured using the Synergy 2 SL luminescence microplate reader (BioTek, Winooski, VT).
In Vitro Citrullination of IKK␥-To purify IKK␥ protein, HEK 293 cells were transfected with T7-IKK␥ plasmid and lysed 24 h later. T7-IKK␥ proteins were purified using the T7⅐Tag affinity purification kit (EMD Chemicals, Gibbstown, NJ). After washing away unbound proteins, T7-IKK␥ was eluted from beads at pH 2.2. The protein fractions were analyzed by SDS-PAGE followed by silver staining. For in vitro citrullination assay, 0.1 mg/ml purified IKK␥ was mixed with 1 unit of rabbit muscle PAD2 (Sigma-Aldrich) in reaction buffer (0.1 M Tris-HCl, pH 8.0, 10 mM CaCl 2 , 2 mM DTT) at 55°C for 2 h. Reaction was stopped by adding 0.5 M EDTA. Citrullination was determined by the immunoblot described above. Silver staining of the proteins loaded on the gel was performed using the Silver Stain Plus kit (Bio-Rad).
Statistical Analysis-Data were reported as means Ϯ S.E. Levels of significance were assessed using the unpaired, twotailed Student's t test for comparison between two samples and one-way analysis of variance with Bonferroni post test for comparison among more than two samples. The p value of less than 0.05 was considered significant. Data were analyzed using Origin 8.1 software (OriginLab Corp., Northampton, MA).

PAD Activity Is Induced by LPS Treatment in RAW 264.7
Cells-To examine PAD activity in RAW 264.7 cells, we treated cells with LPS (1 g/ml in DMEM containing 2 mM Ca 2ϩ ) for 0.25-30 h and determined protein citrullination by immunoblot with the antibody to modified citrulline. Fig. 1A shows the time course of citrullination after LPS treatment (n ϭ 4). Proteins with the molecular mass of 45-65 kDa were prominently citrullinated, comparable with a previous report (26). Most of time, two or three citrullinated major bands appeared on the Western blot. The reason for this is unclear. Vertical stripes or dots at 0.25, 0.5, 16, and 30 h are immunoblot artifacts. Citrullination was gradually observed over time. When compared with control (0 h), substantial levels of citrullination were found at 1 h determined by densitometric measurements of citrullinated proteins normalized to ␤-actin (p Ͻ 0.05; one-way analysis of variance with Bonferroni post test) (Fig. 1B). Activation of PAD enzymes requires Ca 2ϩ (1,2,4), and thus, to further characterize citrullination, we applied LPS to cells for 4 h after pretreating with the Ca 2ϩ chelator BAPTA/AM (30 M) for 1 h. Fig. 1C shows a representative immunoblot result. Control cells without LPS had negligible citrullination regardless of BAPTA/AM in the medium (Fig.  1C, lanes 1 and 2). A slight citrullination in the presence of BAPTA/AM is often observed, the reason for which is unclear. Cells with LPS had typical citrullination in the absence of BAPTA/AM (lane 3). However, the citrullination was suppressed by BAPTA/AM (lane 4). These results demonstrate that LPS induces protein citrullination in RAW 264.7 cells and that citrullination is calcium-dependent.
PAD2 Reduces COX-2 Expression in LPS-treated Cells-Among five PAD isotypes characterized in mammals (2), PAD2 and PAD4 are closely associated with rheumatoid arthritis (7). PAD2 is expressed in cytosol, whereas PAD4 is predominantly found in nucleus (27), and thus, we focused on PAD2 in this study to determine the effect of cytosolic citrullination on inflammatory gene expression in macrophage cells. RT-PCR with primers specific to mouse PAD2 detected PAD2 mRNA in RAW 264.7 cells (Fig. 2A). PCR without reverse transcriptase detected none. We also performed RT-PCR with primers for PAD4 but could not detect its mRNA (data not shown). We then cloned the full-length PAD2 cDNA by PCR with primers corresponding to mouse PAD2 (amino acid sequence comparison of the cloned mouse PAD2 with human PAD4 is provided in the supplemental material).
RAW 264.7 cells were transfected with PAD2 and treated with LPS for 4 h, and an immunoblot was performed to determine COX-2 protein expression. Fig. 2B shows that COX-2 expression was induced by LPS, consistent with previous reports (21)(22)(23). The induction occurred in both groups of cells transfected with PAD2 or vector only (Mock in the figure), but the COX-2 expression level was substantially lower in cells transfected with PAD2. This decrease corresponded to 50% of the controls (p Ͻ 0.05; n ϭ 3; Fig. 2C). The internal control tubulin had negligible change.
This finding suggests that PAD2 can suppress COX-2 expression induced by LPS. To further characterize this effect, we compared NF-B activity between control cells and PAD2overexpressing cells. Cells were transiently transfected with PAD2 and the NF-B luciferase construct, and 24 h later, luciferase activity in cells was measured after 4 h of treatment with LPS (Fig. 2D). Controls were cells without PAD2 (Mock in the figure). The application of LPS increased luciferase activity by 3.2-Ϯ 0.3-fold (n ϭ 4) in control cells. However, in PAD2-expressing cells, LPS increased luciferase activity by 1.7-Ϯ 0.2-fold (n ϭ 4). The decrease corresponded to 47% of the control (p Ͻ 0.05).
PAD2 Inhibits IKK Complex Activity-We tested whether PAD2 inhibits IKK complex from activating NF-B. For this analysis, we used HEK 293 cells to ensure robust expression of wild type and truncated mutant DNA plasmids. HEK 293 cells were transiently transfected with IKK␤ and IKK␥ subunits, PAD2, and NF-B-luciferase construct. PAD2 activation was achieved by inducing Ca 2ϩ influx using 1 M calcium ionophore A23187 (18). Fig. 3A shows NF-B activity in cells transfected with or without PAD2 (no subunits of IKK complex). The values were presented relative to the value for cells without PAD2. Overall, NF-B activities were low in these control cells regardless of PAD2 expression and A23187 treatment (n ϭ 3 for each). Fig. 3B shows NF-B activity in cells transfected with IKK␤ with or without PAD2. IKK␤ increased NF-B activity by 2.1fold when compared with control cells without IKK␤ (p Ͻ  3). The intensity of COX-2 was normalized to the intensity of tubulin. D, NF-B activity in control cells and PAD2-expressing cells. Cells transfected with the NF-B reporter plasmid and PAD2 or none were incubated for 24 h and then treated with LPS for 4 h. NF-B activity was measured by luciferase assay (n ϭ 4) as described previously (22). The activity was presented relative to the value in mock cells without PAD2 and LPS (leftmost bar).

PAD2-mediated Regulation of IKK␥ in Macrophages
0.05). A23187 treatment had negligible effect. However, the increased NF-B activity by IKK␤ was significantly suppressed in cells expressing PAD2 (p Ͻ 0.05; (a) in the figure). The suppression also occurred in the presence of A23187 (p Ͻ 0.05; (b) in the figure), thus indicating that PAD2 reduces IKK␤ activity regardless of Ca 2ϩ in the bath. In the third set of cells transfected with IKK␤/␥ with or without PAD2 (Fig. 3C), IKK␤/␥ coexpression increased NF-B activity by 2.4-fold (p Ͻ 0.05). The activity was further stimulated in the presence of A23187 (p Ͻ 0.05; (c) in the figure), consistent with the report (28) that nuclear transport of IKK␥ requires Ca 2ϩ . However, this IKK␤/␥-mediated effect was suppressed by PAD2 (p Ͻ 0.05; (d) and (e) in the figure). The suppression was more substantial in the presence of A23187 as the NF-B activity was substantially reduced to the basal level for control cells.
PAD2 Interacts with IKK␥-The above finding suggests that PAD2 suppresses NF-B activity by affecting the IKK complex. Given the fact that IKK␥ interacts with many different proteins (29,30), we then hypothesized that PAD2 might also interact with IKK␥ to achieve suppression. To test this hypothesis, we performed coimmunoprecipitation of HEK 293 cells expressing PAD2 and IKK␥. Cells were transfected with FLAG-tagged PAD2 and HA-tagged IKK␥, and lysates were immunoprecipitated with the anti-HA antibody and immunoblotted with anti-FLAG antibody. Fig. 4A shows that PAD2 was detected in the immunoprecipitates from cells. The detection required Ca 2ϩ as it was absent under A23187-free conditions. This also indicates that the detection is not due to incomplete washing of protein A-Sepharose beads. PAD2 levels are similar in the two groups of cells treated with and without the ionophore (bottom panel; one-tenth of the lysates used for coimmunoprecipitation was loaded on the gel). Densitometric measurements of PAD2 signal intensity in the immunoprecipitates versus lysates (one-tenth) were estimated to be 4.3% (n ϭ 4) of PAD2 interacting with IKK␥ (Fig. 4B).
To further characterize PAD2/IKK␥ interaction, we made truncation mutants: PAD2(1-385), which has the N-terminal 385 amino acids of the protein, and PAD2(355-672), which has the C-terminal 318 amino acids. In addition, a point mutation replacing Gln-358 (a putative ligand for Ca 2ϩ binding; see supplemental material) with an Ala was introduced on PAD2(355-672) to evaluate the effect of Ca 2ϩ on PAD2/ IKK␥ interaction. Fig. 4C shows both PAD2(1-385) and PAD2(355-672) being coimmunoprecipitated with IKK␥ (the 50-kDa band in all lanes is IgG heavy chain). The signal intensity of PAD2(1-385) was relatively weak when compared with those for wild type PAD2 and PAD2(355-672), but PAD2(1-385) protein expression in lysates also appeared to be low. Thus, the weak PAD2(1-385) signal in the immunoprecipitates does not represent weak interaction between this mutant and IKK␥. In contrast, the point mutant PAD2(355-672)/Q358A was not detected. There was no immunoreactive band corresponding to the truncation protein that serves as the template protein. Thus, the mutation at Gln-358 completely abolished PAD2/IKK␥ interaction despite the presence of A23187 in the medium.
In other experiments, we also constructed truncation mutants of IKK␥: IKK␥ , which contains the N-terminal 196 amino acids of the protein, and IKK␥(197-419), which contains the remaining amino acids. Coimmunoprecipitation was performed with cells transfected with PAD2 and these mutants as well as wild type IKK␥. PAD2 was coimmunoprecipitated with these truncation mutants (Fig. 5A). The amount of PAD2 being coimmunoprecipitated with IKK␥(1-196) appeared to be lower, but the amount of IKK␥  bound to protein A-Sepharose beads was also low. Thus, densitometric measurements of PAD2 relative to IKK␥ (Fig. 5B) exhibited similar values among different immunoprecipitates (p Ͼ 0.05; n ϭ 4). These results indicate that the interaction of PAD2 and IKK␥ involves multiple bindings, at least one in the region of 1-196 residues of IKK␥ and another in the region of 197-419 residues.
PAD2 Can Citrullinate IKK␥-The finding of the ability of PAD2 to suppress IKK␤/␥ activity and interact with IKK␥ led us to postulate that PAD2 might citrullinate IKK␥ and/or its associated proteins. To test this hypothesis, we performed two sets of experiments. First, RAW 264.7 cells were treated with LPS after pretreatment with BAPTA/AM (30 M), and immunoprecipitation was performed with an anti-IKK␥ antibody. The immunoprecipitates were then immunoblotted with anti-citrulline antibody. Fig. 6A shows representative  DECEMBER 17, 2010 • VOLUME 285 • NUMBER 51 results. LPS-induced citrullination was observed in IKK␥ immunoprecipitates (lane 2) but abolished by BAPTA/AM (lane 4), consistent with the Ca 2ϩ -dependent PAD activity. The detection of citrullinated proteins indicates that either IKK␥ or its binding partners, or both, is citrullinated. The anti-citrulline antibody recognized double bands, whereas the anti-IKK␥ antibody recognized a single IKK␥ band in a sister immunoblot membrane. Although this could be due to a relatively small amount of IKK␥ being citrullinated, we do not exclude the alternative possibility that other proteins interacting with IKK␥ might be citrullinated.

PAD2-mediated Regulation of IKK␥ in Macrophages
Thus, in the second set of experiments, we performed the in vitro citrullination assay in which purified IKK␥ and PAD2 proteins are incubated in the buffer containing Ca 2ϩ (10 mM). Probed with the anti-citrulline antibody, the immunoblot showed citrullination of IKK␥ (Fig. 6B). Citrullination did not occur without PAD2 or with IKK␥ and PAD2 in Ca 2ϩ -free buffer.

DISCUSSION
Overview-The major findings from the present study are the following. 1) LPS induces Ca 2ϩ -dependent citrullination in the macrophage cells RAW 264.7; 2) PAD2 overexpression in macrophage cells reduces COX-2 production; 3) activated PAD2 suppresses IKK␤/␥ activity; 4) PAD2 interacts with IKK␥ in the presence of Ca 2ϩ ; and 5) PAD2 can citrullinate IKK␥. These findings provide novel evidence that PAD2 can reduce inflammatory responses by suppressing IKK␤/␥ and interacting with, and probably citrullinating, the regulatory protein IKK␥ in macrophage cells.
Based on these findings, we propose a model for the effect of protein citrullination on cellular function in immune effector macrophages. LPS causes intracellular Ca 2ϩ to rise (31), which then induces the conversion of inactive PAD2 to active PAD2 to mediate protein citrullination (supplemental material). Activated PAD2 then suppresses the IKK complex activity by binding to IKK␥. The binding of PAD2/IKK␥ may lead to citrullination of IKK␥ and/or IKK␥-interacting proteins. The essential part of the model is that PAD2 acts to downregulate pro-inflammatory gene expression induced by LPS. Activation of PAD2 in inflammation might serve to self-limit inflammatory processes by modifying key LPS-sensing signaling molecules.
LPS Treatment Induces PAD Activity in RAW 264.7 Cells-Many reports show that citrullination occurs in a variety of proteins, including keratin (32), histones (10), vimentin (26), and myelin basic proteins (33). Citrullination also occurs in chemokines (34 -36) such as CXCL8 -12. Given the fact that the conversion of positively charged arginine into uncharged citrulline causes changes in protein folding, citrullination has been known to alter protein structure and function in cells (1,2,4). Citrullination has attained strong pathological and clinical interests because of its close association with autoimmune diseases such as rheumatoid arthritis (7). The goal of our study was to understand whether PAD2-mediated protein citrullination in macrophages is associated with inflammatory responses.
We found that protein citrullination is induced by LPS in RAW 264.7 cells (Fig. 1). Citrullination starts to occur within 1 h after LPS treatment and continues over time. LPS increases intracellular free Ca 2ϩ levels by ϳ150-fold within 4 min in RAW 264.7 cells (31). This increase is mediated by L-type Ca 2ϩ channels (31). PAD is a Ca 2ϩ -dependent enzyme that requires Ca 2ϩ for substrate binding. Thus, Ca 2ϩ influx by LPS is sufficient to bring out PAD activity. Protein citrullination at early time points after LPS treatment is probably due to the conversion of inactive PAD to active PAD by transient increase in intracellular Ca 2ϩ . As for citrullination at later time points, PAD activity may be induced from pro-inflammatory proteins produced by LPS. Prostaglandin D 2 produced by COX-2 acts to cells in autocrine and paracrine fashions (37). Prostaglandin D 2 can induce PAD activity in RAW 264.7 cells (38). Prostaglandin D 2 receptor (D 2 receptor) activation also results in intracellular Ca 2ϩ increase by Ca 2ϩ influx (39).
PAD2 Reduces NF-B Activity-The novel finding in our study is that PAD2 can reduce COX-2 production mediated  . The immunoprecipitates were then immunoblotted (IB) with anticitrulline antibody (a-citrulline). Protein cit , citrullinated proteins. B, in vitro citrullination assay. Purified IKK␥ and PAD2 proteins were incubated in reaction buffer containing 10 mM Ca 2ϩ for 2 h. The reaction mixture was then immunoblotted with the anti-citrulline antibody. Silver staining was done with sister mixture loaded on the gel (bottom panel). IKK␥ cit , citrullinated IKK␥.

PAD2-mediated Regulation of IKK␥ in Macrophages
by LPS in macrophage cells (Fig. 2, B and C). This reduction is due to the capability of PAD2 to suppress NF-B activity (Fig.  2D). PAD2-mediated suppression of NF-B is also observed in HEK 293 cells expressing PAD2 and IKK␤ and IKK␥ subunits (Fig. 3). IKK␤ is a catalytic subunit that can phosphorylate IB␣, which binds to NF-B and inhibits its function (25). Phosphorylation of IB␣ results in ubiquitination. In contrast, IKK␥ (also known as NEMO) is the regulatory subunit. IKK␥ interacts preferentially with IKK␤ and is required for the activation of the IKK complex (40). Our study shows that PAD2 markedly suppresses IKK␤ alone as well as IKK␤/␥. The effect of PAD2 on IKK␤ occurs regardless of Ca 2ϩ in the medium, whereas the effect on IKK␤/␥ appears to be dependent upon Ca 2ϩ . Thus, PAD2 appears to be activated to suppress IKK␤/␥ activity. Liu et al. (41) reported that PAD4 overexpression in hematopoietic cells induces apoptosis. One might thus think that the reduced expression of COX-2 in our study would be due to apoptotic events caused by PAD2 overexpression in cells. However, in the previous report, the apoptotic event occurs Ͼ48 h after PAD4 overexpression, and we found that there is no difference in cell viability 24 -32 h after PAD2 transfection (data not shown).
PAD2 Interacts with IKK␥-A number of proteins have been identified to interact with IKK␥. The essential function of these IKK␥-binding proteins is to modulate IKK activity. At least 16 IKK␥-binding proteins (30) are reported to enhance IKK complex activity, and at least eight IKK␥-interacting proteins are reported to suppress the complex (29). The mechanisms of their interactions with IKK␥ vary depending upon the molecular and cellular nature of individual proteins. Our study shows that PAD2 is another member of the proteins interacting with IKK␥ (Fig. 4). The significance of this interaction is its dependence on Ca 2ϩ . The interaction does not occur under normal conditions, in which intracellular Ca 2ϩ is low. In response to LPS, however, intracellular Ca 2ϩ levels increase to mediate the conversion of inactive PAD to active PAD2, and activated PAD2 then interacts with IKK␥. Therefore, the interaction is an induced event that occurs in response to inflammatory stimuli.
What would be the consequence of PAD2/IKK␥ interaction? IKK␥ plays important roles in regulating IKK complex activation. Proteins interacting with IKK␥ activate or downregulate NF-B activity (29,30). Obviously, PAD2/IKK␥ interaction appears to down-regulate IKK complex activity. The interaction may serve as an inhibitory pathway to reduce the LPS-mediated signaling cascade. PAD2 decreases COX-2 expression by 50% (Fig. 2C) and NF-B activity by 47% (Fig. 2D), both of which are substantial changes. These lead us to reevaluate the pathological role of protein citrullination in inflammation. Citrullination causes changes in protein folding, which lead to abnormal protein structure and function (3)(4)(5). Therefore, citrullination induced by LPS is clearly disadvantageous to cells. On the other hand, the activation of PAD2 occurring in inflammation may be a process that limits inflammation. PAD2 may play an anti-inflammatory role by reducing IKK␥ activity and/or modifying key molecules responsible for LPS-induced inflammation (Fig. 5). In this sense, we note that Prostaglandin D 2 , which can induce citrullina-tion (38), functions as an anti-inflammatory molecule in lung inflammation (42).
Mechanism of PAD2/IKK␥ Interaction-Given the finding that the mutation of a putative Ca 2ϩ coordination site Gln-358 completely abolishes the interaction (Fig. 4C), we conclude that PAD2/IKK␥ interaction requires Ca 2ϩ . This result provides valuable information on the molecular mechanism of the interaction. Structural analysis of PAD4 (4) suggests that the binding of Ca 2ϩ to the acidic concave surface in the C-terminal domain is critical for substrate specificity and binding. The binding of Ca 2ϩ to PAD4 leads to conformational changes around the substrate-binding site, which then allows PAD4 to interact with arginine residues in other proteins. PAD4 has many residues for Ca 2ϩ coordination, and one of these residues is Gln-349 (corresponding to Gln-358 in PAD2). Replacing Gln-349 with an Ala abolishes citrullination activity of PAD4. Thus, in our study, Q358A probably inhibits PAD2 from binding Ca 2ϩ and subsequently fails to induce conformational change required for IKK␥ binding. Nonetheless, this interpretation makes it difficult to explain the interaction between IKK␥ and PAD2(1-385), which does not contain the site for substrate binding. The N-terminal domain of PAD4 is proposed to be responsible for a Ca 2ϩ -dependent regulation or protein-protein interaction (4). Thus, the interaction between IKK␥ and PAD2 not only requires the C-terminal substrate-enzyme interaction, but it may also involve N-terminal protein-protein interaction.
Summary-Our study shows that protein citrullination is induced by LPS in macrophages and that citrullination can suppress up-regulation of pro-inflammatory gene expression. Activated PAD2 can interact with IKK␥ and may citrullinate IKK subunits. PAD2-mediated suppression of IKK complex may be a novel mechanism that regulates inflammatory response. The exact mechanism underlying PAD2-mediated suppression needs further investigation.