TP53-mediated clonal hematopoiesis confers increased risk for incident atherosclerotic disease

Study Cohorts and Risk of Hematologic Malignancy

After excluding individuals with a known history of hematologic malignancy at enrollment, we identified 37,657 unrelated individuals from the UKB and 12,465 individuals from MGBB with whole exome sequencing data available for downstream analysis. Using a previously validated somatic variant detection algorithm¹¹, we identified 2,194 (5.8%) and 657 (5.4%) CHIP carriers in the UKB and MGBB, respectively (Supplementary Table 1). Demographic and clinical characteristics of these individuals, stratified by CHIP status, are depicted in Supplementary Table 2. CHIP carriers tended to be older, male, current or previous smokers, and have a history of coronary artery disease, hypertension, and hyperlipidemia (two-tailed chi-squared and Wilcoxon-rank sum P< 0.05).

We first replicated known CHIP associations¹¹ with white blood cell (Beta 0.09 SD; 95% CI 0.05-0.13; P=1.6x10⁻⁵), monocyte (Beta 0.05 SD; 95% CI 0.01-0.09; P=0.009), neutrophil (Beta 0.10 SD; 95% CI 0.06-0.14; P=2.1x10⁻⁶), and platelet counts (Beta 0.07 SD; 95% CI 0.03-0.11; P=0.0005) in UKB, with larger CHIP clone size as measured by variant allele fraction (VAF) having stronger effects on blood counts (Extended Data Figure 1). Consistent with the existing literature^6,11, CHIP also associated with incident hematologic malignancy (HR 2.20; 95% CI 1.70-2.85; P=1.8x10⁻⁹) - specifically for acute myeloid leukemia (HR 8.08; 95% CI 4.36-14.97; P=3.2x10⁻¹¹), myeloproliferative neoplasms (HR 5.89; 95% CI 3.69-9.89; P=9.7x10⁻¹⁴), and polycythemia vera (HR 12.37; 95% CI 4.85-31.54; P=1.4x10⁻⁷). This risk increased with larger VAF (Extended Data Figure 2).

CHIP and Incident PAD Risk

We next tested the association of CHIP status with incident PAD. Using available electronic health record (EHR) data and a previously validated PAD definition¹², we identified 338 and 419 incident PAD cases in UKB and MGBB, respectively. CHIP associated with a 58% increased risk of incident PAD in the UKB (HR_UKB = 1.58, 95% CI: 1.11-2.25; P=0.01, Figure 1), results that were replicated in MGBB (Overall HR = 1.66, 95% CI: 1.31-2.11; P=2.4x10⁻⁵). We then sought to evaluate whether those with larger CHIP clone sizes (i.e., higher VAF) had greater risk for PAD, as larger CHIP clones associate more strongly with adverse clinical outcomes⁷. We observed a graded relationship between CHIP VAF and PAD, as those with a VAF > 10% had even greater risk for an incident PAD event (Overall HR = 1.97, 95% CI: 1.44-2.71; P=2.3x10⁻⁵, Figure 1). Additional sensitivity analyses, including propensity score adjustment and a marginal structural Cox proportional hazards model estimated through stabilized inverse-probability-treatment-weight revealed similar results in the UKB (Extended Data Figure 3). Subsequent analyses showed no significant interaction between CHIP status and either age, sex, or smoking status on incident PAD risk. In sensitivity analyses accounting for a concomitant CAD diagnosis among those with incident PAD, the CHIP-PAD association remained significant (Supplementary Tables 3, 4).

CHIP and Incident Atherosclerosis Across Multiple Vascular Beds

We next assessed whether CHIP was associated with 8 other incident atherosclerotic diseases across multiple vascular beds. Using EHR-based disease definitions¹³, we tested the association of CHIP with atherosclerotic disease across the mesenteric (acute and chronic), coronary, and cerebral vascular beds, as well as with aneurysmal disease (aortic and any other aneurysm). We observed significant associations for coronary artery disease (HR 1.40, 95% CI: 1.20 to 1.63; P=1.9x10⁻⁵), any aortic aneurysm (HR 1.74; 95% CI: 1.21 to 2.51; P=0.0028), other aneurysms (HR 1.70; 95% CI: 1.23 to 2.34; P=0.0013), and mesenteric ischemia (HR 3.22; 95% CI: 2.01 to 5.17; P=1.3x10⁻⁶) across both cohorts, with directionally consistent effect estimates observed for all the tested phenotypes (Figure 2a). These associations were consistently stronger for large CHIP clones (Extended Data Figure 4). We then created a composite, incident atherosclerosis outcome combining all nine atherosclerotic phenotypes (“pan-arterial atherosclerosis”, Supplementary Table 5). CHIP associated with this combined incident pan-arterial atherosclerosis endpoint (HR 1.31, 95% CI: 1.14 to 1.49, P=9.7x10⁻⁵), again with stronger effects conferred by large CHIP clones (HR 1.45; 95% CI: 1.20 to 1.75; P=0.00013) (Figure 2b,c).

Gene-specific analyses of CHIP with incident atherosclerotic diseases

Next, we sought to understand whether the clonal hematopoiesis putative driver gene differentially affected the risk of acquiring atherosclerosis. Previous work has focused primarily on the epigenetic regulators DNMT3A and TET2^4,14, and whether DDR CHIP confers an increased risk of atherosclerosis is unknown. We stratified the CHIP-PAD, CHIP-CAD, and CHIP pan-arterial atherosclerosis analyses by putative driver genes and specific mutations - focusing on DNMT3A, TET2, ASXL1, JAK2, the DDR genes PPM1D and TP53, and mutations that specifically disrupt splicing factor genes (LUC7L2, PRPF8, SF3B1, SRSF2, U2AF1, and ZRSR2)¹⁵. We observed an association of CHIP with PAD across three of the four common CHIP genes (TET2, ASXL1, and JAK2), with significant heterogeneity of incident PAD effect sizes across the CHIP genes (P_{heterogeneity} = 0.03) (Figure 3a). This heterogeneity persisted in sensitivity analysis after excluding JAK2 carriers (P_{heterogeneity} = 0.046). A similar set of associations was observed for CHIP and CAD (Supplementary Table 6). These data also revealed the finding that DDR TP53 and PPM1D CHIP associates with incident PAD (HR 2.72; 95% CI: 1.20 to 1.75; P=0.00013) and incident CAD (HR 2.51; 95% CI: 1.52-4.13; P=0.00032), with a numerically stronger effect on PAD conferred by TP53 mutations (HR 4.98; 95% CI: 1.23-20.09; P=0.024, Figure 3a-c). Similar findings were observed for the incident pan-arterial atherosclerosis outcome when stratifying by putative driver gene (Extended Data Figure 5). Further sensitivity analysis for DDR-CHIP and incident PAD when excluding solid organ malignancy did not significantly change the associations (P_{heterogeneity} > 0.05).

Atherosclerosis development in p53−/− CHIP mice

Based on the novel finding of an association between somatic TP53 mutations and atherosclerotic disease, we next characterized further the effects of TP53-mutant CHIP on atherosclerosis through experimental studies in mouse models. The most frequent somatic TP53 variants in CHIP are mainly missense mutations that act in a dominant negative manner and have an effect comparable to that of full p53 deficiency in mouse studies¹⁶. Therefore, p53-deficient mice were used as a murine surrogate of human TP53 mutations. To mimic CHIP and test whether the expansion of p53-deficient hematopoietic cells contributes to atherosclerosis, a competitive bone marrow transplantation (BMT) strategy was used to generate atherosclerosis-prone Ldlr−/− chimeric mice carrying 20% Trp53−/− hematopoietic cells (20% KO-BMT mice) (Figure 4a and Supplementary Figure 1). This chimerism is consistent with the VAF of TP53 mutations identified in our human study population (mean VAF=0.13 i.e. ~26% mutant cells if monoallelic mutations; mean VAF=0.23 in those defined as carrying large CHIP clones, Supplementary Table 7). Transplanted mice then consumed a high fat/high cholesterol (HF/HC) diet for 9 weeks to induce atherosclerosis development. To distinguish donor Trp53−/− and Trp53+/+ cells in this experimental setting, Trp53+/+ cells were obtained from mice carrying the CD45.1 variant of the CD45 cell surface marker, whereas Trp53−/− cells were obtained from mice carrying the CD45.2 variant of this protein. Control mice (20% WT-BMT) were transplanted with 20% CD45.2+ Trp53+/+ cells and 80% CD45.1+ Trp53+/+ cells. Flow cytometry analysis of CD45.2+ blood cells established that this BMT strategy led to a modest, but significant expansion of donor Trp53−/− BM-derived cells compared to Trp53+/+ cells in both BM hematopoietic stem/progenitor cells (LSK cells, defined as Lineage- c-Kit+ Sca1+) and circulating white blood cells (Figure 4b,c), consistent with previous studies^17-19. Transplanted Trp53−/− BM cells expanded into all blood cell lineages to a similar extent, but this relative expansion did not affect absolute blood cell counts within the timeframe of the experiment (Extended Data Figure 6a,b). Having validated this mouse model of p53-deficient CHIP based on a competitive BMT strategy, we next assessed whether this phenomenon affects the development of atherosclerosis or related metabolic abnormalities. The presence and expansion of Trp53−/− cells led to a significant ~40% increase in plaque size in the aortic root of male Ldlr−/− mice (Figure 4d), without affecting body weight, spleen weight or serum cholesterol levels (Extended Data Figure 6c-e). Similar results were obtained in female mice (Extended Data Figure 6f-k). In contrast to previous findings on DNMT3A/TET2 CHIP^14,20-23, the expansion of p53-deficient cells did not affect the expression of the pro-inflammatory cytokines IL-6 and IL-1β or that of the NLRP3 inflammasome in the atherosclerotic aortic wall (Extended Data Figure 7a).

Expansion of p53-deficient macrophages in the murine atherosclerotic aorta

Increased atherogenesis in mice carrying Trp53−/− cells was paralleled by a substantial increase in plaque macrophage content, as assessed by immunohistological staining of Mac2 (Figure 5a), suggesting a contribution of increased arterial macrophage burden to accelerated atherosclerosis in conditions of p53-deficient CHIP. No significant changes were observed in other plaque components, such as collagen content, vascular smooth muscle cell content, necrotic core extension or lipid content. (Figure 5a and Extended Data Figure 7b). Flow cytometry analysis of matched samples from blood and digested atherosclerotic aortae from 20% KO-BMT mice revealed substantially higher chimerism in aortic macrophages (~76% Trp53−/−) than in blood classical monocytes (~48%), the main source of plaque macrophages (Figure 5b). Immunofluorescent staining of aortic root sections confirmed that plaque macrophages are predominantly Trp53−/− in 20% KO-BMT mice (Figure 5c). These findings suggest that p53-deficient macrophages have a selective advantage to expand within the atherosclerotic plaque.

Increased proliferation of murine p53-deficient macrophages

Atherosclerotic plaques in mice carrying p53-deficient hematopoietic cells exhibited a significant increase in the proportion of cells positive for Ki-67, a marker of proliferating cells (Extended Data Figure 8a). Flow cytometry analysis indicated that this is at least in part driven by increased proliferation of p53-deficient macrophages, as there was a 2-fold increase in the frequency of Ki-67+ cells within the CD45.2+ Trp53−/− aortic macrophage population compared to CD45.2+ Trp53+/+ macrophages (Figure 6a). Consistent with this in vivo observation, cultured Trp53−/− macrophages exhibited accelerated mitotic cell cycle progression, with a >2-fold increase in the percentage of DNA-replicating (S-phase) cells upon stimulation with macrophage colony stimulating factor (MCSF), a major determinant of plaque macrophage proliferation²⁴ (Figure 6b). Furthermore, Trp53−/− macrophages showed and increased percentage of S-phase cells when proliferating asynchronously and higher rates of BrdU incorporation when stimulated with MCSF (Extended Data Figure 8b,c). p53 was expressed in quiescent macrophages at the transcript and protein level and further induced after MCSF stimulation (Figure 6c,d), suggesting a central role of p53 in the normal regulation of macrophage cell cycle progression. Consistent with this possibility, transcriptomic profiling by RNA-seq revealed a widespread alteration of gene expression in MCSF-stimulated Trp53−/− macrophages compared to WT controls (Extended Data Figure 9). Functional annotation of differentially expressed genes showed that categories related to quantity of myeloid cells, cell cycle progression and immune cell proliferation were among the most upregulated in Trp53−/− macrophages (Extended Data Figure 9). qPCR analysis at different timepoints of MCSF stimulation further confirmed that p53-deficient-macrophages exhibit major changes in the expression of pivotal regulators of cell cycle entry and progression, such as Cdkn1a/p21^Cip1 and Cyclin B1 (Figure 6e). RNA-seq of cultured p53-deficient macrophages also revealed a significantly downregulated expression of several genes related to cell death and apoptosis (Extended Data Figure 9). However, apoptosis rates were comparable in p53-deficient and WT macrophages within atherosclerotic plaques (Supplementary Figure 2), suggesting that differences in apoptosis are not a relevant contributor to increased macrophage burden and atherosclerotic plaque size in mice exhibiting p53-deficient CHIP. Accordingly, we found that p53 deficiency does not affect oxysterol-induced apoptosis in cultured macrophages (Supplementary Figure 2). Similarly, neither the uptake of modified LDL, the uptake of apoptotic cells, or the expression of central mediators of these cellular processes were affected by p53 deficiency (Supplementary Figure 3). p53-deficient neutrophils isolated from 20%-KO-BMT mice exhibited an expression of phenotypic markers and a production of reactive oxygen species comparable to those of WT neutrophils (Supplementary Figure 4). Neutrophil extracellular traps were not detected in murine atherosclerotic plaques in the experimental conditions used in this study (Supplementary Figure 5).

Cohorts and exclusion criteria

The UKB is a population-based cohort of approximately 500,000 participants recruited from 2006-2010 with existing genomic and longitudinal phenotypic data and median 10-year follow-up³⁷. Baseline assessments were conducted at 22 assessment centres across the UK with sample collections including blood-derived DNA. Of ~49,960 individuals with WES data available, we analysed 37,657 participants consenting to genetic analyses after our exclusion criteria. Use of the data was approved by the Massachusetts General Hospital Institutional Review Board (protocol 2013P001840) and facilitated through UK Biobank Application 7089.

The MGBB contains genotypic and clinical data from >105,000 patients who consented to broad-based research across 7 regional hospitals and median 3-year follow-up³⁸. Baseline phenotypes were ascertained from the electronic medical record and surveys. We analysed 12,465 individuals consenting to genetic analysis after our exclusion criteria. Use of the data was approved by the Massachusetts General Hospital Institutional Review Board (protocol 2020P000904).

Across both cohorts, we excluded individuals with prevalent hematologic cancer, individuals without genotypic-phenotypic sex concordance, and one of each pair of 1^st or 2^nd degree relatives at random. For the UKB, samples were further restricted to individuals with Townsend deprivation index, a marker of socioeconomic status, available for analysis. Follow-up time was defined as time from enrolment to disease diagnosis for cases, or to censorship or death for controls.

Whole exome sequencing and CHIP calling

UKB WES were generated from whole blood-derived DNA at the Regeneron Sequencing Center³⁹. MGBB WES were generated using whole blood-derived DNA using Illumina sequencing (mean coverage 55x). Somatic CHIP variants were detected with GATK MuTect2 (v1) software with parameters as previously described^7,11. Briefly, aligned short-read exome sequences were analyzed using the GATK MuTect2 software to detect putative somatic genetic variants, which were then filtered to exclude common sequencing artifacts (including minimum variant read counts of 3 reads, required evidence of a variant on both forward and reverse reads, exclusion of frameshift mutations within homo-polymer repeat regions). Common germline variants and sequencing artifacts were excluded as before. Samples were annotated with the presence of any CHIP if MuTect2 identified one or more of a pre-specified list of pathogenic somatic variants reported in the literature and/or the Catalog of Somatic Mutations in Cancer (COSMIC, http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/ ), following manual curation, as previously described^21,40. Additionally, samples were annotated with the presence of Large CHIP (variant allele frequency >10%), as larger CHIP clones have previously been more strongly associated with adverse clinical outcomes⁷.

Phenotype definitions

Across both UKB and MGBB, PAD was defined by grouping together ICD-10 and ICD-9 billing codes for aortic atherosclerosis (I70), peripheral vascular disease (I73.8, I73.9), and operative procedures including amputation of leg (X09.3-5), bypass of artery of leg (L21.6, L51.3, L51.6, L51.8, L59.1-8), endarterectomy or angioplasty of leg artery (L52.1-2, L54.1,4,8, L60.1-2, L63.1,5), and other transluminal operations on leg arteries or peripheral stent placement (L63.9,L66.7), as previously described¹². Additionally, in the UKB self-reported peripheral vascular disease, leg claudication/intermittent claudication, arterial embolism, femoral-popliteal leg artery bypass, leg artery angioplasty +/− stent, or amputation of leg were also included (Data fields 20002, 20004) as performed previously¹². Coronary artery disease and cerebral atherosclerosis phenotype definitions for each cohort are detailed in Supplementary Tables 8 and 9. Other atherosclerotic conditions were defined using the Phecode Map 1.2⁴¹ ICD-9 (https://phewascatalog.org/phecodes) and ICD-10 (https://phewascatalog.org/phecodes_icd10) phenotype groupings for “abdominal aortic aneurysm”, “aortic aneurysm”, “other aneurysm”, “chronic vascular insufficiency of intestine”/“acute vascular insufficiency of intestine” (combined into a mesenteric ischemia phenotype), “atherosclerosis of renal artery”. The composite atherosclerosis phenotype was created by combining all analyzed atherosclerosis phenotypes (including PAD, coronary artery disease, cerebral atherosclerosis, abdominal aortic aneurysm, aortic aneurysm, other aneurysm, chronic vascular insufficiency of intestine, acute vascular insufficiency of intestine, and atherosclerosis of renal artery) into one phenotype, whereby the first instance across all of these phenotypes was used to determine time of first diagnosed atherosclerotic disease for survival analysis. Other phenotypic covariates (never/prior/current smoking status, hypertension, hyperlipidemia, principal components of ancestry, etc.) were used as previously defined⁴².

Association Analysis

In the UKB and MGBB a traditional cox-proportional hazards model was utilized using the Survival package in R-3.5 adjusting for age, age², sex, smoking status, normalized Townsend deprivation index as a marker of socioeconomic status (only available in UKB), and the first ten principal components of genetic ancestry. Demographic and clinical characteristics found to differ between individuals with and without CHIP were tested using chi-squared (categorical) and Wilcoxon-rank sum (continuous) tests with a two-tailed P< 0.05 determining significance. Sensitivity analysis including other covariates (normalized BMI, prevalent hypertension, prevalent Type 2 diabetes, and prevalent hyperlipidemia) was not found to significantly change associations with PAD in the UKB (Extended Data Figure 10). Additional sensitivity analyses were utilized in the UKB including propensity score adjustment, as well as a marginal structural cox proportional hazards model estimated through stabilized inverse-probability-treatment-weight (IPTW)⁴³ to estimate the total causal effect of CHIP on PAD. Further details of propensity score methods and stabilized IPTW analysis are described below. For our primary outcome PAD, results were combined across the UKB and MGBB using an inverse-variance weighted fixed effects meta-analysis, a two tailed association P < 0.05 determined statistical significance. In our secondary analysis of CHIP with 10 additional, incident atherosclerotic diseases, a two-tailed Bonferroni p-value threshold of P<0.05/10=0.005 was used to declare statistical significance.

Epidemiologic causal inference methods used to assess the CHIP-PAD association

The propensity score is the probability of treatment assignment (i.e. clonal hematopoiesis of indeterminate potential [CHIP] or large CHIP) conditional on the minimally sufficient set of confounders. Using a propensity score as a covariate allows observational studies to mimic some of the characteristics of a randomized controlled trial in the sense that conditional on the propensity score, the distribution of covariates will be similar between the CHIP and the non-CHIP carriers. In our scenario, the propensity score will help assess the ‘dimensionality problem’ in that it will reduce the number of covariates from 15 covariates in the minimally sufficient set to 1. Here, the propensity score is calculated using multivariable logistic regression where the outcome was CHIP and covariates consisted of the 15 covariates in the minimally sufficient set (age, age2 – included to fully account for age given significant associations in multivariate models, sex, smoking status, and the first ten principal components of genetic ancestry). The propensity score was then predicted for all samples in the UK Biobank for CHIP and, separately, for large CHIP.

Stabilized Inverse probability of treatment weighting (IPTW) uses weights based on the propensity score to create a pseudo-population in which the distribution of measured covariates is independent of CHIP status, i.e. removing the arrow from the confounders to the exposure. The conditional exchangeability assumption made in IPTW is that if the set of covariates are sufficient to block all backdoor paths from CHIP to peripheral artery disease (PAD), then all confounding is eliminated in the pseudo-population and thus the association between CHIP and PAD in the synthetic population is an estimate of the causal effect. To estimate the IPTW stabilized weight (SW) for each individual in the UKB, we utilized the following equation for CHIP carriers and controls, where PS reflects the propensity score for that individual:

Mice

C57Bl/6J Trp53−/− mice were obtained from the Jackson Laboratory. Ldlr−/− mice carrying the CD45.1 isoform of the CD45 hematopoietic antigen were generated by crossing Ldlr−/− mice from the Jackson Laboratory and B6.SJL-PtprcaPepcb/BoyCrl mice from Charles River Laboratories. Animal experiments followed protocols approved by the Institutional Ethics Committee at the Centro Nacional de Investigaciones Cardiovasculares and conformed to EU Directive 86/609/EEC and Recommendation 2007/526/EC regarding the protection of animals used for experimental and other scientific purposes, enforced in Spanish law under Real Decreto 1201/2005. All mice were maintained on a 12-h light/dark schedule in a specific pathogen-free animal facility in individually ventilated cages, and were given food and water ad libitum. Ambient temperature in the animal facility was 20-24°C, relative humidity was 45-65%.

Competitive bone marrow transplantation and atherosclerosis induction in mice

8-12 weeks-old CD45.1+ Ldlr−/− recipients were transplanted with suspensions of BM cells containing 20% CD45.2+ Trp53−/− cells and 80% CD45.1+ Trp53+/+ cells (20% KO-BMT mice) or 20 % CD45.2+ Trp53+/+ cells and 80% CD45.1+ Trp53+/+ cells (20% WT-BMT mice) similar to previous studies¹⁴ (Supplementary Figure 1). BM cells were isolated from femurs and tibias of 6-12 weeks-old donor mice after euthanasia. Recipient Ldlr−/− mice were exposed to two doses of 450 rad three hours apart. After the second irradiation, each recipient mouse was injected with 10⁷ BM cells i.v. Water was supplemented with antibiotics for 7 days before transplant and for 14 days post-transplant. Mice that did not recover full pre-irradiation body weight 28 days after transplant were excluded from further analysis. Starting four weeks after BMT, an interval sufficient for BM reconstitution in our experimental conditions, mice were fed a HF/HC Western diet (Harlan-Teklad, TD.88137, Adjusted Calories Diet; 42% from fat, 0.2% cholesterol) to promote hypercholesterolemia and the development of atherosclerosis. Mice were maintained on HF/HC diet for 9 weeks unless otherwise stated.

Analysis of aortic atherosclerosis burden in mice

Mice were euthanized and plasma was collected in EDTA-coated tubes for cholesterol level quantification by an enzymatic assay (Cholesterol E, WAKO Diagnostics). Aortas were removed after in situ perfusion with phosphate-buffered saline (PBS) injected through the left ventricle of the heart. Tissue fixation was achieved by immersion in 4% paraformaldehyde in PBS overnight at 4°C. Aortic tissue was then embedded in paraffin to obtain histological sections from the aortic root as determined by the location of the aortic valve leaflets. Alternatively, unfixed aortic root specimens were embedded in Tissue-Tek O.C.T. Compound. An operator who was blinded to genotype quantified plaque size in aortic root sections by computer-assisted morphometric analysis of microscopy images. For each mouse, atherosclerotic plaque size in aortic root cross-sections was calculated as the average of 5 independent sections separated by ~16μm. Atherosclerotic plaque composition was examined by immunohistochemical techniques. Vascular smooth muscle cells were identified with an alkaline phosphatase-conjugated mouse anti-smooth muscle α-actin (SMA) monoclonal antibody (clone 1A4, Sigma) and Vector Red Alkaline Phosphatase Substrate (Vector Laboratories). Macrophages were detected with a rabbit anti-Mac2 monoclonal antibody (Cedarlane Labs), a biotin-conjugated goat anti-rat secondary antibody, streptavidin-HRP, and DAB substrate (all from Vector Laboratories), with hematoxylin counterstaining. Ki-67 was detected with a rabbit anti-Ki-67 antibody (clone SP6, Abcam), a biotin-conjugated horse anti-rabbit secondary antibody and streptavidin HRP. Collagen and lipid content were determined by a modified Masson’s trichrome staining and Oil Red O (ORO) staining, respectively. Microscopy images were analyzed using Fiji ImageJ software using the Color Deconvolution plugin. Plaque cell apoptosis was quantified by TUNEL analysis (In situ Cell Death Detection Kit, Fluorescein, Roche) according to manufacturer instructions. Prior to apoptotic cell staining, sections where blocked with 3% of goat serum in PBS and co-stained with Alexa Fluor 647-conjugated anti-CD45.2 antibody (BioLegend), anti-CD68 primary antibody (BioRad) and Alexa Fluor 568-conjugated goat anti-rat IgG secondary antibody (ThermoFisher). Nuclei were counterstained with DAPI. Images were acquired on a LEICA SP8 Lightning-Navigator confocal microscope with LAS X 3.1.1. 15751 software, using a z-stack of 5 μm-thickness frames. Results are shown as the percentage of TUNEL positive cells within the total plaque cell population, the plaque macrophage population or the CD45.2+ macrophage population in the aortic root. A list of antibodies and additional information, including antibody dilutions, can be found in Supplementary Table 10.

Flow cytometry analyses of blood and tissue samples.

Peripheral blood was obtained from the facial vein and collected into EDTA-coated tubes. BM cells were flushed out of two femurs and two tibias per mouse. Aortic arches were digested for 45 minutes at 37°C in RPMI containing 10% FBS and 0.25 mg/ml Liberase TM (Roche Life Science). Red blood cells were lysed in all samples by treatment with 1X Red Blood Cell Lysis Buffer (eBioscience ThermoFisher) for 5 minutes on ice. BM lineage-negative cells were defined as negative for CD11b, Gr-1, Ter119, B220, CD3e and CD127. BM LSK were defined as Lineage−, c-Kit+, Sca1+. Blood classical monocytes were identified as CD45+, CD115^High, CD43^Low, Ly6C^High; patrolling monocytes, as CD45+, CD115^High, CD43^High, Ly6C^Low; neutrophils, as CD45+, CD115^Int, Ly6G+; T lymphocytes, as CD45+, CD115−, B220−, CD3+; B lymphocytes, as CD45+, CD115−, CD3−, B220+; and aortic macrophages, as CD45+, CD3−, Ly6G−, CD11b+, F4/80^High. Gating strategies are summarized in Supplementary Figure 6. Samples were stained with combinations of biotinylated and/or fluorescently labeled antibodies in PBS with 1% FBS for 30 min on ice. The following fluorescent antibodies were used for staining and flow cytometry analysis (additional information, including antibody dilutions, can be found in Supplementary Table 10): eFluor450-conjugated anti-CD45.2, FITC-conjugated anti-CD45.2, Pe-Cy7-conjugated anti-CD45.1, FITC-conjugated anti-CD11b, eFluor450-conjugated anti-CD11b, PE-conjugated anti-CD115, PE-eFluor610-conjugated anti-CD3, FITC-conjugated anti-CD4 (from eBioscience); APC-Cy7-conjugated anti- B220, PerCP-Cy5.5-conjugated anti-Ly6G, BV510-conjugated anti-Ly6G, BV711-conjugated anti-CD43, PE-Cy7-conjugated anti-c-Kit, AlexaFluor647-conjugated anti-Sca-1 (from BD Biosciences); PE-conjugated anti-F4/80 (from R&D Systems); PerCP-Cy5.5-conjugated anti-CD45.1, BV510-conjugated anti-CD8a, PerCP-Cy5.5-conjugated anti-mouse Ki-67 (from BioLegend). Fixation/permeabilization for Ki-67 intracellular staining was achieved using commercially available kits following manufacturer’s instructions (Foxp3 Transcription Factor Staining Buffer Set, eBioscience ThermoFisher). Dead cells were excluded from analysis by DAPI staining in unfixed samples and by LIVE/DEAD Fixable Near-IR staining (ThermoFisher Scientific) in fixed samples. BD LSRFortessa and BD FACSymphony Cytometers were used for data acquisition with BD FACSDiva software (BD Biosciences). Data were analyzed with FlowJo Software (Tree Star).

Murine bone marrow-derived macrophage culture and cell cycle analysis

BM-derived macrophages (BMDM) were obtained by differentiating femoral BM cell for 7 days in RPMI Medium supplemented with antibiotics, 10% fetal bovine serum and 100 ng/ml MCSF. Cell-cycle dynamics were analyzed both in asynchronously proliferating macrophages and after synchronization in G0 phase by MCSF deprivation. MCSF concentration in cell culture medium was decreased to 5 ng/ml for 48h to induce progressive synchronization of macrophages in G0 phase (i.e. quiescence) and cell cycle re-entry was induced by treatment with 100 ng/ml MCSF. Macrophages were trypsinized and collected by centrifugation for 5 minutes at 300g. After fixation in 80% ethanol for at least 1 h at −20°C, cells were incubated for at least 30 minutes with 50 μg/mL propidium iodide containing 0.25 mg/mL RNAse A (both from Sigma). Labelled cells were analyzed in a BD FACSCANTO flow cytometer with FACSDiva software (BD Biosciences) and DNA histograms were fitted into cell cycle distributions using ModFit 3.0 software (Verity Software House). Macrophage proliferation was also assessed by evaluating 5-bromo-2’-deoxyuridine (BrdU) incorporation into BMDM after synchronization in G0 phase by MCSF deprivation and induction of cell-cycle re-entry by treatment with 100 ng/ml MCSF together with 30 μM BrdU (Sigma) for 16h. To detect BrdU incorporation, BMDM were fixed in ethanol at −20^oC for 20 min and subjected to acid treatment (10 mM citrate buffer pH 6.0). After neutralization, samples were then blocked with 5% goat serum in PBS-0.1% Tween, and incubated with anti-BrdU primary antibody (ThermoFisher), followed by Alexa Fluor 488-conjugated goat anti-mouse IgG secondary antibody (ThermoFisher). Additional information on these antibodies, including antibody dilutions, can be found in Supplementary Table 10. Nuclei were counterstained with DAPI. Images were captured on a Nikon Eclipse microscope with Nikon AR 4.30.02 (b925) 64bits software, and analyzed using Fiji ImageJ software and the StartDist plugin.

Gene expression analysis by quantitative real-time PCR (qPCR), mRNA sequencing and Western Blot

Total RNA from cultured macrophages or aortic arch tissue was isolated using Trizol reagent and RNeasy kits (QIAGEN). RNA was reverse transcribed with High-Capacity cDNA Reverse Transcription Kit (Applied Bosystems) and qPCR was performed with SYBR^® Green PCR Master Mix (Applied Biosystems) in a AB7900 Real time PCR system. Results were analyzed with the ΔΔCt method. The average of 36B4 and β-actin was used as reference for normalization. Primer sequences are listed in Supplementary Table 11.

Libraries for mRNA sequencing were prepared using the SMART-Seq v4 Ultra Low Input RNA Kit (Clontech-Takara). Full-length cDNAs were processed using a Nextera XT DNA Library Preparation Kit (Illumina). Quality control was performed using a High Sensitivity DNA ScreenTape Assay with a 4200 TapeStation System (Agilent). Libraries were then multiplexed in an equimolar pool and sequenced using a NextSeq 550 Platform (Illumina). Single-end unstranded libraries (75bp) were first subjected to a quality check FastQC v0.11.8 and aligned to the reference genome assembly mm10 with the following settings (STAR --outFilterMultimapNmax 20 --alignSJoverhangMin 8 –alignSJDBoverhangMin 1 --outFilterMismatchNmax 999 --outFilterMismatchNoverReadLmax 0.04 --alignIntronMin 20 -- alignIntronMax 1000000 --alignMatesGapMax 1000000 --outSAMattributes NH HI NM MD).

RNA-seq quality assessment was performed using RSeQC v3.0.1 tool. Differential expression analysis was assessed with edgeR v3.30.0 R Bioconductor package (PMID: 19910308) and only genes with a cut-off expression > 1 CPM in at least three samples were included in the analysis. TMM method was applied to normalize gene counts and glmQLFTest function was used to perform differential analysis, with a minimum cut-off of absolute fold change ≥ 1.5 and FDR ≤0.05. A complete list of differentially expressed genes and statistical information (pValue and FDR) for each condition is available at Gene Expression Omnibus (GEO: GSE184420). Functional enrichment analysis was performed using Ingenuity Pathway Analysis software (IPA v01-13; Qiagen). GOplot representations were used to combine and integrate the expression data with the results of the functional analysis⁴⁴.

For Western Blot analysis of protein levels, protein extracts from cultured macrophages were obtained using ice-cold lysis buffer (20mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM NA2EDTA, 1 mM EGTA, 1% Triton) supplemented with protease and phosphatase inhibitors (Roche Applied Science). Equal amounts of protein lysates were resolved by SDS-PAGE and the following antibodies were used for immunoblotting: anti-p53 (Cell Signaling) and HRP-conjugated anti-α tubulin (Abcam). Additional information on these antibodies, including antibody dilutions, can be found in Supplementary Table 10. An ImageQuant LAS 4000 biomolecular imaging system with Image Quant LAS4000 software (GE Healthcare) was used for image acquisition, and Fiji ImageJ software was used for band densitometric analysis. p53 protein levels were normalized to α-tubulin.

Analysis of modified lipoprotein uptake and efferocytosis in cultured macrophages

Thioglycolate-elicited peritoneal macrophages were cultured in RPMI medium supplement with antibiotics and 10% FBS. To asses modified LDL uptake, peritoneal macrophage were treated with 5 μg/ml fluorescent acetylated LDL (DiI acLDL, ThermoFisher) for 1h or 3h. Macrophage were then trypsinized, collected and analyzed by flow cytometry using a BD FACSymphony flow cytometer with BD FACSDiva software (BD Biosciences). Data were analyzed with FlowJo Software (Tree Star). To assess uptake of apoptotic cells (i.e. efferocytosis), thymuses from C57Bl/6J mice were collected, homogenized, filtered through a 70μm cell strainer and labeled fluorescently with 5 μM CellTrace Violet (ThermoFisher). Thymocyte apoptosis was induced by exposure to UV light for 10 minutes. 1 or 2 hours later, apoptotic thymocyte were added to macrophage cultures at a 5:1 ratio to allow apoptotic cell engulfment by macrophages. After 1h, macrophages were washed extensively, collected by trypsinization and stained with FITC-conjugated anti-CD11b (Invitrogen) and PE-conjugated anti-F4/80 antibodies (Tonbo Bioscience). Additional information on these antibodies, including antibody dilutions, can be found in Supplementary Table 10. Samples were analyzed by flow cytometry using a BD FACSCanto flow cytometer with BD FACSDiva software (BD Biosciences). Data were analyzed with FlowJo Software (Tree Star).

Analysis of neutrophil phenotype and function

Neutrophil phenotypes analyses and ROS quantification were performed in blood samples extracted from experimental mice and treated with RBC hypotonic lysis buffer (0.15M NH4Cl, 0.01M KHCO3 and 0.01M EDTA in water). For ROS quantification, cells were stimulated with 133 nM of phorbol 12-myristate 13-acetate (PMA) for 20 min or PBS as control, and then stained with 5 mM Dihydrorhodamine 123 (DHR123, Thermo Fisher) for 20 min, and with BUV737-conjugated anti-Ly6G antibody (BD Biosciences), PE-conjugated anti-CD11b antibody (Tonbo Biosciences), PE-Cy7-conjugated anti-CD45.1 antibody (BioLegend) and APC-conjugated anti-CD45.2 antibody (BioLegend) for 15 min. For surface marker analysis, samples were incubated with BUV737-conjugated anti-Ly6G antibody (BD Biosciences), PE-conjugated anti-CD62L antibody (eBioscience), PerCP-Cy5.5-conjugated anti-CXCR2 antibody (BioLegend), BV510-conjugated anti-CD11b antibody (BioLegend), PECy7-conjugated anti-CD101 antibody (ThermoFisher), AlexaFluor 647-conjugated anti-CD54 antibody (BioLegend), APC-Cy7-conjugated anti-CD45 antibody (BioLegend) and FITC-conjugated anti-CD45.2 antibody (BioLegend) for 15 min. Additional information on these antibodies, including antibody dilutions, can be found in Supplementary Table 10. Dead cells were excluded from analysis by DAPI staining. Samples were analyzed in a BD FACSymphony Cytometer with BD FACSDiva software (BD Biosciences) for data acquisition. Analysis was performed using FlowJo software (Tree Star).

NETosis was assessed in cryopreserved aortic root sections from hematopoietic chimeric mice, and cryopreserved lung sections from mice subjected to transfusion-related acute lung injury as positive control⁴⁵. After fixing in 4% PFA for 15 min at RT, samples were blocked with 5% BSA, 1% normal goat serum, 5% FBS and 0.1% triton-x 100 in PBS in a humid chamber for 1h followed by an incubation with antibodies against citrullinated histone 3 (Abcam), CD31 (Thermo Fisher) and biotinylated-MPO (R&D) overnight at 4°C. Sections were then incubated with goat anti-rabbit-Alexa 568 (Life Technologies), goat anti-hamster-Alexa-647 (Jackson Immunoresearch), Alexa-488 conjugated Streptavidin (BioLegend) and DAPI (Life Technologies), and mounted with Mowiol 4-88 (Mw 31,000, Sigma). Additional information on antibodies, including antibody dilutions, can be found in Supplementary Table 10. Imaging of NETs was performed using a Leica SP5 multi-line inverted confocal microscope with 63x magnification and LAS-AF 2.7.3. build 9723 software. Images were analyzed using Fiji ImageJ software (NIH). An event was considered positive for NETosis when it was triple stained (MPO, cit H3, and DAPI).

Statistical analysis of data in experimental studies

Data are shown as mean ± SEM unless otherwise stated. Statistical significance of differences in experiments with two groups and only one variable was assessed by unpaired Student’s t tests (with Welch correction for unequal variance when appropriate) or Mann-Whitney tests. Differences in experiments with more than one independent variable were evaluated by two-way analysis of variance (ANOVA) with post-hoc Sidak’s or Tukey’s multiple comparison tests. All statistical tests were performed using GraphPad Prism software (GraphPad Software Inc.).