ApoE^–/– (B6.129P2-Apoetm1Unc/J) mice were purchased from the Jackson Laboratory. The generation of Pad4^–/– mice was previously reported by our group (26). ApoE^–/–/Pad4^–/– mice were generated by crossing ApoE^–/– with Pad4^–/– mice without the Cre recombinase gene. All mice were on a C57/Bl6J genetic background.

Adult 8–12 weeks old male mice (22.86 ± 1.94 g) were used in this study. Mice were maintained in the local animal facility at a 12-h light/dark cycle with food and water ad libidum.

8–12 weeks old ApoE^–/– mice were subjected to 9-Gy whole body γ-irradiation to eliminate bone marrow cells. Bone marrow cells were extracted from male ApoE^–/–/Pad4^–/– donor mice by flushing the femurs and tibias. Irradiated mice were injected intravenously with 5 × 10⁶ bone marrow cells in 100 μl PBS. Mice reconstituted with bone marrow cells from ApoE^–/–/Pad4^+/+ (= ApoE^–/–) mice served as control.

Mice were kept on 0.01 mg/ml polymyxin B sulfate and 0.1 mg/ml neomycin, administered through drinking water, for 5 weeks. During the first 4 weeks after transplantation, mice were fed a standard chow diet to allow for bone marrow reconstitution. Afterward, mice were switched to a high-fat diet containing 0.2% cholesterol, 21% butter fat (TD.88137, Ssniff Spezialdiaeten, Soest, Germany) for 4 and 10 weeks, respectively. At the end of experiment, mice were euthanized by terminal anesthesia and blood samples were obtained from the right ventricle. In some groups, animals were fasted for 6 h. The aortic trees including the hearts were immediately dissected and prepared for further analyses. Successful bone marrow reconstitution was verified and confirmed by PCR using gene-specific primers and genomic DNA isolated from blood samples collected at the end of the experiment.

Aortas were dissected, minced using scissors and enzymatically digested with 200 U/mL Liberase (Roche) and 40 U/mL DNase I (Sigma Aldrich) in HBSS plus 5% FCS for 1 h at 37°C. Cells were filtered through a 40 μm nylon strainer, washed with HBSS plus 5% FCS, collected by centrifugation at 400 g for 5 min at 4°C and then suspended in FACS buffer (PBS plus 0.2% FCS plus 1 mM EDTA). Murine Fc receptors were blocked using anti-CD16/32 antibodies (BioLegend) for 10 min on ice. Violet 510 Viability Dye (Cell Signaling Technology) was used to discriminate between live and dead cells. The cells were stained with the following antibodies for 30 min at 4°C: PerCP-Cy5.5-conjugated anti-CD45, APC-Cy7-conjugated anti-CD11b, FITC-conjugated anti-Ly6G, PE-Cy7-conjugated anti-F4/80, and APC-conjugated anti-Ly6C (all purchased from BioLegend). Flow cytometric analysis was performed using FACSCanto II flow cytometer (BD Biosciences) and DIVA Software (BD Biosciences).

Serial sections (7 μm) of aortic root were collected and stored at -80°C under further processing. Frozen sections were air-dried, fixed with 4% paraformaldehyde for 10 min, stained with Oil Red-O (ORO) for 15 min and counterstained with mayers hematoxylin (two slides per animal). The lesion areas were quantified using Image J software and expressed as mm².

Masson trichrome staining for collagen analysis was performed using Trichome Stains (Masson) (Sigma Aldrich) according to the manufacturer’s protocol. Collagen quantification was determined by the method described by Chen et al. (27). Histological evaluation of necrotic core size was performed on hematoxylin and eosin (H&E-stained sections, two slides per animal).

For immunofluorescence, frozen sections (12–16 sections per animal) were fixed with 4% paraformaldehyde, permeabilized, blocked with 5% serum and stained with the following primary antibodies: rabbit anti-α-SMA (ab5694, Abcam) and anti-CD68 (ab53444, Abcam). Following labeling with the fluorochrome-coupled secondary antibodies Alexa Fluor 488-conjugated goat anti-rabbit IgG (Cell Signaling Technology) and Alexa Fluor 568-conjugated donkey anti-rat IgG (Thermo Fisher), sections were counterstained with DAPI and mounted with fluorescent mounting medium (Dako). All images were captured using an inverted Eclipse Ti-U 100 microscope and NIS Element software package (Nikon). Image J software was used to define the percentage of positive cells.

For NETs staining, frozen sections of aortic root were fixed with 4% paraformaldehyde, followed by permeabilization and blocking with 5% donkey serum. Subsequently, the samples were incubated with goat anti-MPO (R&D Systems) and rabbit anti-citrullinated histone H3 (citH3, Abcam) antibodies, followed by incubation with Alexa Fluor 488-conjuagted anti-goat IgG and Alexa Fluor 555-conjugated anti-rabbit IgG (both from Abcam) for 2 h at room temperature. Tissue sections were counterstained with DAPI and mounted in Dako Fluorescent mounting medium (Dako). NETs were defined as colocalization of MPO and citH3 and visualized using an inverted microscope (Eclipse Ti-U 100, Nikon) and the NIS elements software (BR3.10, Nikon). The percentage of NETs-positive area and percentage of citH3-positive area were calculated using Image J software.

For quantification of atherosclerotic lesions, aorta from root and abdominal area was fixed with 4% paraformaldehyde followed by careful removal of connective tissues. Then, the aorta was opened longitudinally, pinned en face and stained with Oil Red-O (ORO) for 3–4 h. Images were taken with a digital camera and both, total surface area and ORO-positive lesion area were determined using Image J software. The extent of atherosclerotic lesion development was defined as the percentage of total ORO-positive lesion area over the total surface area.

Total RNA from dissected aortas and cultured cells was extracted using Trizol (Sigma Aldrich) or the RNeasy Mini Kit (Qiagen), respectively, according to manufacturers’ instructions. RNA was reverse transcribed using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time PCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems) and specific primers for IL-6, IL-12, IL-1β, MMP-2, MMP-9 (28), iNOS, Ym-1, SPHK1, LIGHT, FIZZ/RELM-α, MerTK (29), TNF-α (30), TGF-β (31) and CCL2 (32). All samples were run in triplicates. Expression of target genes was normalized to the GAPDH housekeeping gene and expressed as relative expression (2^{Δ CT} formula).

Bone marrow-derived macrophages from 9 to 12 weeks old ApoE^–/– and ApoE^–/–/Pad4^–/– mice were prepared by flushing femurs and tibias followed by cell culture in RPMI medium supplemented with 20% FCS and 20 ng/mL murine M-CSF (Peprotech) for 7 days (26). To induce a M1- or M2a-like phenotype, macrophages (M0) were further incubated in the presence of 20 ng/mL IFN-γ (Peprotech) and 100 ng/mL LPS (Sigma Aldrich) or in medium supplemented with 20 ng/mL IL-4 (Peprotech), respectively. Macrophages’ phenotype was characterized by Real-time PCR and flow cytometry (FACS Canto II, BD). For flow cytometry, M0, M1-like and M2a-like macrophages were stained for CD11b, CD86 (M1 marker) and CD206 (M2 marker). The following antibodies were used: APC-Cy7-conjugated anti-CD11b, APC-conjugated anti-CD86 and Brilliant Violet 421-conjugated anti-CD206 (BioLegend).

M1- or M2a-polarized bone marrow-derived macrophages were serum starved for 24 h and further incubated with 20 μg/mL Dil-labeled oxLDL (Themo Fisher) for 6 h. The cells were washed twice with PBS and analyzed by flow cytometry (FACS Canto II, BD).

Plasma cholesterol was measured by standard colorimetric assay kit from Sigma Aldrich (No. MAK043).

Plasma concentrations of IL-6, IL-12, TNF-α and CCL2 were determined by using a customized ProcartaPlex multiplex immunoassay for mouse (ProcartaPlex 4-plex, Thermo Fisher) according to the protocol provided by the manufacturer. Data were collected and analyzed using a Magpix instrument equipped with xPONENT software (Luminex Corporation, Austin, Texas, United States).

cfDNA levels in plasma samples were quantified by Pico green staining as previously described (33).

Experimental data were analyzed with GraphPad Prism software (GraphPad Software, San Diego, CA, USA) and are reported as mean ± standard deviation (SD). Normal distribution of variables was tested using the Kolmogorov-Smirnov test. Comparisons of two groups were analyzed using unpaired t-test or the non-parametric Mann-Whitney U-test. Differences between groups of two independent variables were determined using two-way ANOVA and Tukey’s post hoc test. A p-value of less than 0.05 was considered as statistically significant.

To study the role of PAD4 in bone marrow-derived cells in atherosclerosis, ApoE^–/– mice were subjected to lethal irradiation and reconstituted with bone marrow from ApoE^–/–/Pad4^–/– or ApoE^–/– mice as illustrated in Figure 1A. Challenging of ApoE^–/–/Pad4^–/– bone marrow-reconstituted mice with HFD for 10 weeks resulted in a significant increase of body weight gain when compared to mice fed a HFD for 4 weeks and no differences in body weight could be observed between the experimental groups (Figure 1B). Further analysis revealed widely comparable plasma cholesterol levels, indicating that PAD4 does not affect plasma cholesterol (Figure 1C).

Impaired NETs formation in PAD4-deficient mice was already well documented by several groups (19, 22, 23). However, to confirm that bone marrow-restricted PAD4 deficiency impedes NETs formation, we performed immunofluorescent staining to detect colocalization of citrullinated histone 3 (citH3) and myeloperoxidase (MPO) (= NETs) in aortic root lesions. Consistent with previous reports (19, 23), NETs formation was clearly evident in control mice, but not in aortic root lesions of ApoE^–/–/Pad4^–/– bone marrow-transplanted mice after 10 weeks of HFD feeding (Figure 1D). Moreover, reduced histone H3 citrullination, confirming impaired PAD activity, could be observed.

Plasma cfDNA levels have been proposed to represent biomarkers of NETs release and tissue damage in cardiovascular disease (34). Quantification of plasma cfDNA levels did not reveal any differences between the experimental groups, suggesting that circulating cfDNA does not reflect NETs formation (Figure 1E). Nevertheless, at 10 weeks, cfDNA levels were slightly reduced in mice with bone marrow-restricted PAD4 deficiency.

Analyses of en face-stained aortas showed significant increase of lesion area in ApoE^–/– control mice after 10 weeks of HFD when compared with the 4 weeks cohort (Figure 2A). However, plaque burden did not differ between mice with bone marrow-restricted PAD4 deficiency and fed a HFD for 4 and 10 weeks respectively, suggesting that PAD4 deficiency impedes plaque progression in atherosclerosis.

Likewise, at 10 weeks, lesion size in aortic root sections of ApoE^–/–/Pad4^–/– bone marrow-reconstituted mice was significantly reduced compared to the ApoE^–/– control group, although lesions detected after 10 weeks of HFD were generally larger than after 4 weeks (Figure 2B). Notably, reduced lesion size in PAD4-deficient mice was associated with impaired lipid accumulation after 10 weeks of HFD, as demonstrated by reduced ORO-positive lesion area (Figure 2C). This reduction was not related to a decreased number of recruited macrophages, as no differences in CD68-positive cells were detected between both groups (Figure 2D).

The phenotype of macrophages was previously reported to influence the ability of lipid accumulation in atheorosclerosis (35). We therefore next questioned if PAD4 deficiency affects macrophage lipid burden in dependence on their polarization state. For this, bone marrow-derived in vitro polarized M1- and M2a-like ApoE^–/– and ApoE^–/–/Pad4^–/– macrophages were incubated with Dil-labeled oxLDL for 6 h and lipid accumulation was quantified by flow cytometry. As shown in Figure 3A, M2a-like macrophages were found to internalize higher amounts of oxLDL compared to M1-like cells independent on macrophages’ genotype. However, the percentage of ApoE^–/–/Pad4^–/– M2a-positive cells was somehow lower than of ApoE^–/– M2a macrophages and did not significantly increase. Enhanced oxLDL uptake by M2a-like cells was found to correlate with increased expression of the oxLDL receptor CD36 (Figure 3B) and no differences between both genotypes could be observed. These results suggest that the ability to accumulate modified lipids is markedly higher in M2 like-macrophages compared to M1-like cells, while no substantial influence by PAD4 deficiency became evident.

Plaque vulnerability and the risk of rupture highly depend on necrotic core size, collagen content as well as inflammation. Necrotic core size increased significantly in both experimental groups after 10 weeks of HFD compared to the size determined after 4 weeks (Figure 4A). However, after 10 weeks, necrotic core areas in ApoE^–/–/Pad4^–/–-reconstituted mice were found to be significantly decreased compared to ApoE^–/– control mice, indicating that PAD4 deficiency is associated with reduced cell death.

Additionally, at the same time point, ApoE^–/–/Pad4^–/–-reconstituted mice showed reduced abundance of α-SMA⁺ smooth muscle cells (SMCs) (Figure 4B), which are considered to have a beneficial role by contributing to the stabilization of atherosclerotic lesions (36). Accordingly, reduced collagen content in lesions of ApoE^–/–/Pad4^–/– bone marrow-transplanted mice was observed, as determined by Masson’s Trichrome staining (Figure 4C). We further found the number of α-SMA⁺ cells as well as collagen-positive area to strongly increase in ApoE^–/–-reconstituted control mice after 10 weeks of HFD when compared to the earlier time point (4 weeks). However, this increase was abrogated in ApoE^–/–/Pad4^–/–-transplanted mice. Altogether, these data indicate that PAD4 deficiency diminishes the abundance of collagen-producing SMCs.

Given that atherosclerotic burden is mainly driven by inflammatory processes (37), we next determined the impact of PAD4 deficiency on the recruitment of immune cells to the lesions by performing flow cytometric analysis. Gating strategy is depicted in Figure 5A. The percentage of recruited CD45⁺ leukocytes in bone marrow-reconstituted mice was comparable (Figure 5B). After 10 weeks of HFD, mice transplanted with ApoE^–/–/Pad4^–/– bone marrow showed significantly increased number of lymphocytes and reduced content of myeloid cells vs. mice fed a HFD for 4 weeks. Among myeloid cells, no alterations in neutrophil and macrophage contents could be observed. However, PAD4 was found to influence the phenotype of lesional macrophages. Newly recruited inflammatory monocytes (Ly6C^high) differentiate into Ly6C^high macrophages which in turn may convert into anti-inflammatory Ly6C^low macrophages through phenotypic switching (5). Based on the expression of F4/80 and Ly6C, three different macrophage subsets were identified, characterized by high, low or absent Ly6C expression. At both time points (4 and 10 weeks HFD), ApoE^–/–/Pad4^–/–-reconstituted mice displayed significantly higher content of F4/80⁺/Ly6C^high macrophages (M1-like) and strongly reduced number of F4/80⁺/Ly6C^neg cells, reflecting resident macrophages (M2-like) when compared to ApoE^–/– control mice. The number of F4/80⁺/Ly6C^low macrophages did not differ between groups. As the content of Ly6C⁺ inflammatory monocytes/macrophages visibly increased in lesions with PAD4 deficiency, we assume that more recruited Ly6C⁺ monocytes become pro-inflammatory, M1-like macrophages. Moreover, conversion into Ly6C^low/neg macrophages seems to be impaired.

As these data demonstrate altered distribution of macrophage subsets in PAD4-deficient lesions, we further aimed to assess the consequences of PAD4 knockout on the expression of macrophage phenotype marker genes using in vitro-polarized macrophages. Of note, M1-like ApoE^–/–/Pad4^–/– macrophages showed significantly higher expression of IL-6 and iNOS as well as a moderate, not significant, upregulation of TNF-α expression compared to ApoE^–/– M1-like macrophages. No differences regarding the expression of the M2-specific genes Ym-1, Fizz-1/RELM-α1, LIGHT, SPHK1 and MerTK were found between M2a-polarized macrophages (Supplementary Figure 1A). Moreover, increased expression of inflammatory genes in M1-polarized ApoE^–/–/Pad4^–/– macrophages was accompanied by elevated expression of the costimulatory molecule CD86 (Supplementary Figure 1B), suggesting that PAD4 deletion increases the activity of M1 macrophages.

Having shown that atherosclerotic lesions of mice with PAD4 deficiency are enriched with M1-like macrophages, we next analyzed the expression of inflammatory mediators in mice aortas after 10 weeks of HFD by Real-time PCR (Figure 6). PAD4 deficiency resulted in significantly elevated mRNA expression of the monocytes-attracting chemokine CCL2 and the pro-inflammatory gene iNOS (Figure 6A). Besides, a slight not significant increase of TNF-α, IL-6, and IL-1β could be observed. Conversely, the expression of the collagen-degrading matrix metalloproteinases MMP2 and MMP9 and the pro-fibrotic gene TGF-β did not differ. Moreover, plasma levels of CCL2, TNF-α, IL-6, and IL-12 were quantified by multiplex assay and were found to be comparable between both groups (Figure 6B). Thus, bone marrow-restricted PAD4 deficiency does not influence systemic inflammation. Taken together, these findings indicate that PAD4 deficiency strongly affects macrophages’ phenotype resulting in a pro-inflammatory state.