Genetic inhibition of CARD9 accelerates the development of atherosclerosis in mice through CD36 dependent-defective autophagy

Caspase recruitment-domain containing protein 9 (CARD9) is a key signaling pathway in macrophages but its role in atherosclerosis is still poorly understood. Global deletion of Card9 in Apoe-/- mice as well as hematopoietic deletion in Ldlr-/- mice increases atherosclerosis. The acceleration of atherosclerosis is also observed in Apoe-/-Rag2-/-Card9-/- mice, ruling out a role for the adaptive immune system in the vascular phenotype of Card9 deficient mice. Card9 deficiency alters macrophage phenotype through CD36 overexpression with increased IL-1β production, increased lipid uptake, higher cell death susceptibility and defective autophagy. Rapamycin or metformin, two autophagy inducers, abolish intracellular lipid overload, restore macrophage survival and autophagy flux in vitro and finally abolish the pro-atherogenic effects of Card9 deficiency in vivo. Transcriptomic analysis of human CARD9-deficient monocytes confirms the pathogenic signature identified in murine models. In summary, CARD9 is a key protective pathway in atherosclerosis, modulating macrophage CD36-dependent inflammatory responses, lipid uptake and autophagy.

such as Toll-like receptors (TLRs) and Dectin receptors [3][4][5] .A broad range of molecules and particles bearing danger-associated molecular patterns, including oxidized LDL (oxLDL), can be taken up by macrophages, leading ultimately to the formation of pro-inflammatory foam cells 6 .CD36 is an archetypal pattern recognition receptor that binds polyanionic ligands of both pathogen-and self-origin 7 including oxLDL 8 .CD36 forms a complex with the TLR TLR4-TLR6 heterodimer, which recognizes oxLDL and stimulates pro-inflammatory pathways, including the NLRP3 inflammasome activation responsible for proIL-1β.cleavage and IL-1β secretion 9 .A large body of evidence suggests that cumulative metabolic/inflammatory signals and impaired efferocytosis foster apoptosis and secondary necrosis of foam cells, which contributes to the growth of the necrotic core and progression of atherosclerosis 10 .However, the critical downstream pathways that drive both macrophage activation and conversion into foam cells are still poorly understood.
Here, we investigated the role of Caspase recruitment-domain containing protein 9 (Card9), a key regulator of inflammation, in atherosclerosis.Card9 is an adapter protein that integrates pattern recognition receptor downstream signals in macrophages and dendritic cells 11 .Card9 is particularly involved in response to fungi via C-type lectin sensing, but also in response to bacteria by mediating nucleotide-binding oligomerization domain 2 (NOD2)-dependent p38/ JNK signaling and TLR signaling 12 .Card9 is required to mount appropriate immune responses through the production of interleukin (IL)−6, IL-17A, IFN-γ, and IL-22, which can affect gut microbiota 13 , with potential impact on atherosclerosis 14 .The role of Card9 in inflammatory diseases is ambiguous, being pathogenic in post-ischemic cardiac remodeling 15 , but protective in experimental colitis 13 .In two recent studies, the role of CARD9 in atherosclerosis has been explored, but results reported contradictory findings.One study showed increased lesion size in chimeric Ldlr -/-Card9 -/-mice 16 , whereas the other found that deletion of haematopoietic Card9 did not affect atherosclerosis in chimeric Ldlr -/-mice under hyperglycaemic conditions 17 .Moreover, the underlying mechanisms linking Card9 engagement to atherosclerosis development remain largely unknown.
Here, using several complementary approaches and state-of-theart models, we show that Card9 signaling pathway in macrophages regulates cytokine production, lipid upload, and cell survival.Global, as well as hematopoietic deletion of Card9, markedly accelerates the development of atherosclerosis, independently of the adaptive immune system.Mechanisms of the pro-atherogenic effects of Card9 deficiency mainly involve CD36-dependent defective autophagy.

Genetic invalidation of Card9 accelerates atherosclerosis in Apoe -/-mice
To gain insight into the immune cells expressing Card9, single-cell analysis of total cells from mouse atherosclerotic aortas (see methods) was performed (Fig. 1A, B and Supplementary Fig. 1).Card9 transcript was detected in myeloid cells, and macrophages in particular, including atherosclerosis-associated inflammatory and Trem2 hi /Foamy macrophages (Fig. 1A, B and Supplementary Fig. 1a, b).Analysis of a second, independent scRNA-seq dataset of Ldlr -/-mouse atherosclerotic aortas 18 corroborated preferential detection of Card9 transcripts in myeloid cells, including macrophages (Supplementary Fig. 2).Immunofluorescence staining confirmed that Card9 was expressed in atherosclerotic lesions of Apoe -/-mice at both early (Supplementary Fig. 3a) and advanced stages of atherosclerosis (Fig. 1C), and mainly colocalized with MOMA + macrophages (Fig. 1C).In vitro, Card9 expression in macrophages was induced by oxLDL (Supplementary Fig. 3b).
Card9 deletion in Apoe -/-mice induced a switch toward a more inflammatory plaque phenotype with a significant increase in macrophage accumulation (Fig. 1F) and necrotic core size (Fig. 1G).Collagen content was increased in plaques of Apoe -/-Card9 -/-mice (Supplementary Fig. 4d), but T cell accumulation was similar (Supplementary Fig. 4e).
Dampened systemic pro-inflammatory cytokine signature in Apoe -/-Card9 -/-mice Because Card9 is known to modulate cytokine production and T cell polarization 13 , we next investigated the immuno-inflammatory response in male Apoe -/-Card9 +/+ and Apoe -/-Card9 -/-mice.Leukocyte populations were analyzed by flow cytometry in both blood and spleen at sacrifice.We did not observe any significant difference in leukocyte percentages in blood between groups.We only found a slight increase in neutrophil and classical monocyte counts in the blood of Apoe -/-Card9 -/-mice, compared to control Apoe -/-Card9 +/+ mice (Supplementary Fig. 5).Splenocyte number was significantly higher in Card9-deficient mice but the proportion of myeloid and lymphoid populations was not different between groups (Supplementary Fig. 6).Splenocytes from Apoe -/-Card9 -/-mice stimulated with IFN-γ and LPS produced less TNF-α (Fig. 1H) than those from control mice, but the production of IL-10 and IL-1β was not different.We then purified splenic CD4 + T cells from control Apoe -/-Card9 +/+ and Apoe -/-Card9 -/-mice and performed functional tests.In vitro, the proliferation of CD4 + T cells from Apoe -/-Card9 -/-mice was significantly increased compared with control cells (Data non shown), and their production of IFN-γ and IL-17A was increased.There were no differences in IL-10 and IL-22 production (Supplementary Fig. 4f).

Gut microbiota unlikely contributed to the acceleration of atherosclerosis induced by Card9 deficiency
Previous studies have shown that Card9 plays a critical role in gut microbiota homeostasis.Card9 -/-mice have been shown to display dysbiosis 19 , which has been implicated in atherosclerosis development 20 .To evaluate a potential impact of Card9-induced dysbiosis in our experimental conditions, we analyzed the bacterial microbiota was performed using 16S rRNA based sequencing.While there were no significant differences in alpha diversity between Apoe -/-Card9 +/+ and Apoe -/-Card9 -/-mice (Supplementary Fig. 9a, b), beta diversity analysis showed a significant difference between the 2 groups, as demonstrated by the PCoA plot of Bray-Curtis distance (Supplementary Fig. 9c).To determine which taxonomic groups accounted for these differences, we performed linear discriminant analysis with effect size (Lefse) 21 .Compared to Apoe -/-Card9 +/+ mice, Apoe -/-Card9 -/-mice displayed an increase in the pathobiont Helicobacter with a concomitant decrease in beneficial members of the Firmicutes phylum, including the order Clostridiales, as well as in Candidatus arthromitus, segmented filamentous bacteria.These bacteria are essential for Th17 maturation in the murine gut 22 , and in the genus Akkermansia, a genus associated with a lean body type and favorable metabolic outcomes 23 (Supplementary Fig. 9d).
Given the marked dysbiosis in Card9-deficient Apoe -/-mice, we aimed to determine whether this dysbiosis was also observed in Ldlr -/- mice transplanted with Card9 -/-bone marrow cells.As found in Apoe -/- mice, 16S rRNA based sequencing showed no significant difference in alpha diversity among the 2 groups, but the beta diversity analysis showed significant differences in mice transplanted with Card9 -/-BM cells (Supplementary Fig. 9e, f).Of note, because of transient aplasia and increased risk of sepsis chimeric Ldlr -/-mice received antibiotics during 14 days following lethal irradiation and BM cell transplantation.The administration of antibiotics caused marked changes in the microbiota in the 2 groups.In particular, at genus level, Parasutterella was increased in Ldlr -/-Card9 -/-mice and the sulfate-reducing bacteria Desulfovibrio was enhanced in Ldlr -/-Card9 +/+ mice.Only one bacterial family, Clostridiaceae_1 that was increased in Ldlr -/-Card9 +/+ mice, was concordantly altered in the two sets of experiments (Supplementary Fig. 9g-h).Taken together, these data confirm an effect of the Card9 on the gut microbiota composition.However, the highly divergent microbiota composition between Apoe -/-and Ldlr -/ genetic backgrounds despite similar Card9 effects on atherosclerosis suggests that the gut microbiota was unlikely involved in the vascular phenotype induced by Card9 deficiency.

Card9 deficiency upregulates CD36 expression and increases foam cell formation
Next, we speculated that the marked increase of acellular area in atherosclerotic plaques of Card9-deficient mice might be related to increased foam cell formation.To explore this hypothesis, we performed in vitro experiments investigating the uptake of oxidized LDL (ox-LDL) by BM-derived macrophages and their ability to accumulate intracellular lipids.Interestingly, foam cell formation was significantly increased in Card9-deficient macrophages, compared with control macrophages, after 6 and 24 h of incubation with ox-LDL (Fig. 4A, B).Total cholesterol and cholesterol ester content, after ox-LDL exposure, were significantly increased in macrophages from Apoe -/-Card9 -/-mice, confirming the intracellular cholesterol overload (Fig. 4C, D).Next, we investigated the mechanisms that could drive lipid overload in the absence of Card9.We measured a significant increase in Abca1, Abcg1 and Scarb1 mRNA levels in macrophages from Apoe -/-Card9 -/-mice exposed to ox-LDL, compared to those from Apoe -/-Card9 +/+ mice (Fig. 4E).Cholesterol transfer to HDL and to ApoA1 was also enhanced in Apoe -/-Card9 -/-macrophages (Fig. 4F).This finding highly suggests that increased foam cell formation in Card9-deficient macrophages was not due to impaired cholesterol efflux.Next, we investigated the expression of scavenger receptors involved in lipid uptake.We found no difference in Msr1 mRNA content between groups but Cd36 mRNA levels were markedly increased in macrophages from Apoe -/-Card9 -/-mice exposed to ox-LDL (Fig. 4G), which was confirmed at the protein level by immunofluorescent staining (Fig. 4H, I).In agreement with our in vitro experiments, Cd36 mRNA levels were higher in aortas from Apoe -/-Card9 -/-mice, compared with control mice (Supplementary Fig. 10a), as well as CD36+ macrophage numbers (Supplementary Fig. 10b).

Acceleration of atherosclerosis in Card9 deficiency is due to impaired autophagy
The phenotype of Card9-deficient mice, in terms of size and composition of atherosclerotic plaques, as well as macrophage apoptosis susceptibility, resembled that of Atg5 fl/fl LysM Cre+/− Ldlr −/− mice that are characterized by impaired autophagy flux 24 .Interestingly, CD36, which was up-regulated in Card9-deficient mice, has been reported to participate in the regulation of autophagy in hepatocytes through AMPK downstream pathway 25 .Therefore, we next focused on the potential role of autophagy, a compensatory survival mechanism involved in atherosclerosis 26 , in the context of Card9 deficiency both in vitro and in vivo.Macrophages isolated from Apoe -/-Card9 +/+ and Apoe -/-Card9 -/-mice were cultured and stressed to activate autophagy.Interestingly, we found a significant decrease in AMPK phosphorylation in Card9-deficient macrophages (Fig. 5A) but no difference in CHOP levels as well as Beclin-1 and LKB-1 phosphorylation (Supplementary Fig. 11).In addition, we found higher p62 protein content in macrophages from Apoe -/-Card9 -/-mice compared to Apoe -/-Card9 +/+ mice (Fig. 5A).Confocal analysis of immunofluorescence staining of the p62 protein confirmed it was increased in the cytoplasm of unstimulated Card9-deficient macrophages, and much more in the presence of oxLDL (Fig. 5B).The p62 protein was co-localized with large inclusion bodies.LC3B dot size was significantly larger in the cytoplasm of Apoe -/-Card9 -/-macrophages, with specific aberrant colocalization with these large p62+ inclusion bodies (Fig. 5B).P62+ inclusions bodies have been previously described in Atg5-null macrophages 27 , which strongly supports impaired autophagy in the absence of Card9.The accumulation of p62 in Card9-deficient macrophages was confirmed in vivo, as revealed by the higher number of p62+ MOMA+ macrophages in atherosclerotic plaques of Apoe -/-Card9 -/-mice in reference to plaques of Apoe -/-Card9 +/+ mice (Fig. 5C).Next, we evaluated a pharmacological approach to restore autophagy treated with Metformin, a well-known activator of autophagy through AMPK stimulation 28 .In vitro, metformin treatment abolished the accumulation of p62+ protein in the Card9-deficient macrophages and the formation of inclusion bodies (Fig. 5D).In vivo, Apoe -/-Card9 +/+ and Apoe -/-Card9 -/-mice, under a high-fat diet, were treated or not with metformin during 6 weeks 28 .At sacrifice, there was no significant difference in body weight and in plasma cholesterol levels between Apoe -/-Card9 +/+ and Apoe -/-Card9 -/-groups treated with metformin (Supplementary Fig. 12a, b).Metformin treatment abolished the acceleration of atherosclerosis observed in Card9 deficiency, with no difference in plaque size (Fig. 5E), plaque composition (Fig. 5F,  G) and P62 accumulation (Fig. 5H) between the 2 treated groups.
Next, we used an addition complementary pharmacological approach to activate autophagy with rapamycin, an inhibitor of both mTORC1 and mTORC2 29 .In vitro, rapamycin treatment restored autophagy flux, as shown by strong reduction of p62 protein accumulation in the cytoplasm of Card9-deficient macrophages and an almost disparition of p62+ inclusion bodies (Fig. 6A, B).In addition, rapamycin treatment abolished intracellular lipid overload (Fig. 6C),
Finally, we examined CARD9 expression in atherosclerotic plaques from human carotid arteries.CARD9 was not detected in normal aorta (Supplementary Fig. 14a).However, CARD9 was detected in fatty streak lesions (Fig. 8D) and in lipid-rich areas surrounding the necrotic core of advanced atherosclerotic plaques (Fig. 8E).CARD9 expression was higher in atheromatous plaques than in fibrous lesions (Supplementary Fig. 14b).Fluorescent staining confirmed that CARD9 expression was mostly confined to CD68 + intimal macrophages (Fig. 8D, E), and analysis of previously published single-cell RNA-seq 33 revealed CARD9 mRNA expression specifically in macrophages from atherosclerotic human coronary arteries (Fig. 8F and Supplementary Fig. 15a-d).In an integrated single-cell analysis of human mononuclear phagocytes in atherosclerosis 34 , CARD9 was detected in all the major plaque macrophage subsets (Supplementary Fig. 15e-g).
in the absence of Card9 might be counterintuitive since Card9 is a downstream adapter of fungal-and bacteria-induced activation of TLRs, as well as activation of ITAM-containing non-TLRs and Dectin-1 35,36 .Card9 engagement and Card9-Bcl-10-MALT1 complex formation lead to NF-κB transcription and subsequent secretion of proinflammatory cytokines 37 .Decreased spleen production of TNF-α is consistent in all murine models and in line with previous studies 35 .The effect of Card9 deficiency on IL-1β production is more complex.We found decreased IL-1β production by stimulated mixed immune cells in the spleen, but higher production by murine Card9 -/-macrophages as well as higher IL-1β transcripts in monocytes from CARD9-deficient patients.In the context of Salmonella infection, it has also been reported that CARD9 negatively regulates IL-1β by fine-tuning pro-IL-1β expression, SYK-mediated NLRP3 activation and repressing inflammasome-associated caspase-8 activity 38 .Higher production of IL-1β might be involved, at least in part, in the acceleration of atherosclerosis in Apoe -/-Card9 -/-mice and might be due to increased CD36 expression.Liu et al. have shown that CD36 promoted the expression of NLRP3 and consecutive IL-1β production through ROS generation in ox-LDL-stimulated macrophages 39 .In our study, we found that Card9 deficiency had significant effects on the adaptive immune system and particularly on T cell polarization.We observed discrepancies in cytokine production by CD4 + T cells between male Apoe -/-and female chimeric Ldlr -/-, which could be due to gender or background difference, Apoe having by itself immune-modulatory functions 40 .In chimeric Ldlr -/-mice, Card9-deficient CD4 + T cells produced less IL-17A than control CD4 + T cells, which is consistent with previous studies in normocholesterolemic mice 13 .However, our findings of increased atherosclerosis in immune-deficient Apoe -/-Rag2 -/-Card9 -/-mice ruled out the possibility that the acceleration of atherosclerosis in the absence of Card9 was mediated by a modulation of the adaptive immune system.
Given the marked increase in aortic atherosclerosis in the absence of hematopoietic Card9 and the colocalization between Card9+ macrophages and lipid-rich areas in both mouse and human plaques, we then focused on the role of Card9 in macrophage foam cell formation.We found a marked increase in ox-LDL uptake and lipid accumulation in Card9 -/-macrophages.Among the receptors that govern foam cell formation in macrophages, Card9 deletion selectively increased both Cd36 gene expression and cell surface protein levels, in vitro, as well as in vivo in plaque macrophages.The effect of Card9 deficiency on the upregulation of CD36 expression and lipid uptake might account for the accelerated atherosclerosis in Card9-deficient mice.Several studies have previously reported a pro-atherogenic role of CD36 in Apoe -/-41 and Ldlr -/-mice 42 .
Apoptosis susceptibility of Card9-deficient macrophages might be due to intracellular lipid overload.In the context of atherosclerosis, other mechanisms, including autophagy, could account for enhanced apoptosis susceptibility of macrophages in the absence of Card9.Notably, we found that autophagy was impaired in cultured Card9deficient macrophages, as well as in atherosclerotic plaques of Apoe -/-Card9 -/-mice, as shown by the accumulation of p62 accumulation.Confocal analysis revealed that the p62 protein accumulated in large inclusion bodies with LC3B.These inclusion bodies that are p62enriched polyubiquitinated protein aggregates, have been previously described in the context of impaired autophagy 27 .These observations highly suggested that Card9 deficiency impaired autophagy flux, but the exact molecular mechanisms requires further investigation.Ohman et al. previously suggested a link between the Card9 signaling pathway and autophagy, and found that Curdlan, an activator of the Dectin-1 receptor, induced LC3I conversion into LC3II in cultured macrophage 43 .More recently, Rubicon, a Beclin-1-binding partner, was identified as a physiological feedback inhibitor of Card9-BCL10-MALT1-mediated PRR signaling 44 .Impaired autophagy has also been shown to promote atherosclerosis.Autophagy blockade in LysMCre + Atg5l ox/lox Ldlr -/-mice led to increased lesion size and larger necrotic core 24 , which phenocopies the genetic loss of Card9 in Apoe -/- mice.Our study showed that impaired autophagy in Card9-deficient macrophages was mediated, at least in part, by AMPK blockade.We focused on AMPK because CD36 is known to inhibit AMPK phosphorylation 25,45 and also because AMPK is an upstream regulator of autophagy through several mechanisms, including Ulk1 activation 46 and mTORC1 inhibition 47 .In Card9-deficient macrophages, oxLDLinduced AMPK phosphorylation was abolished and P62 protein accumulated in cell cytoplasm.Interestingly, AMPK phosphorylation and autophagy were restored in Cd36 -/-Card9 -/-cells, supporting a critical role of CD36 in the regulation of autophagy and subsequently atherosclerotic plaque development.Rapamycin and metformin, two pharmacological activator of autophagy, respectively through mTOR blockade and AMPK activation 29 , restored in vitro autophagy flux in Card9-deficient macrophages, with a strong reduction of p62+ inclusion bodies formation.In vivo, both treatment abolished the proatherogenic effect of Card9 deficiency and reduced p62 accumulation in atherosclerotic lesions.As rapamycin and metformin have several molecular targets, it cannot be ruled out that some of their antiatherogenic effects were independent of their effect on autophagy.
Experiments performed in Cd36 -/-animals confirmed that CD36 upregulation was involved in the acceleration of atherosclerosis in Card9 -/-animals.However, how the two functions of CD36 (lipid uptake versus autophagy) contribute to the pro-atherogenic impact of Card9 deficiency remains unknown.Addressing this issue is very challenging because autophagy, by itself, can regulate Cd36 levels 48 .Finally, in our study, cholesterol efflux was not impaired in Card9-deficient macrophages suggesting that autophagy-independent mechanisms such as neutral lipolysis 49,50 were activated in Card9 -/-macrophages to limit lipid overload.
Finally, given the marked differences in the profiles of gut microbiota between Apoe -/-and Ldlr -/-models, but the very similar proatherogenic effects of Card9 deficiency in these two murine models of atherosclerosis, it is unlikely that these vascular effects were due to Card9-associated dysbiosis.However, in the absence of microbiota transfer experiments, a contribution of gut dysbiosis in the vascular phenotype cannot be definitively ruled out.Altogether, our studies identify CARD9 as a major protective pathway in the development and complications of atherosclerosis.Pro-apoptotic and pro-atherogenic effects of Card9 deficiency are mediated by CD36-dependent defective autophagy that can be reversed by rapamycin and metformin.

Animals
Experiments were conducted according to the guidelines formulated by the European Community for experimental animal use (L358-86/ 609EEC) and were approved by the Ethical Committee of INSERM and the French Ministry of Agriculture (agreement A75-15-32).Mouse breeding occurred in our animal facility in accordance with local recommendations.Animals were provided with food and water ad libitum.All the animals were maintained under identical standard conditions (housing, regular care and high-fat diet).Mice were maintained in isolated ventilated cages under specific pathogenfree conditions.Before euthanasia by cervical dislocation, animals were anesthetized with isoflurane (3% in oxygen).

Human carotid plaques
Formalin-fixed and paraffin-embedded arterial tissue sections were used after antigen retrieval by heating in microwave oven in Tris/EDTA buffer pH9 (Dako).buffer.For single labeling, CARD9 rabbit polyclonal antibody (Abcam, Cambridge, UK) was used at 1:500 dilution, incubated for 1 h and revealed using ABC-peroxidase technique (Vector Laboratories, Burlingame, CA, USA).For double labeling, the sections were first incubated with mouse monoclonal anti-CD68 antibodies diluted at 1:50 (to detect macrophages) or anti-α smooth muscle actin antibodies diluted at 1:100 (to detect smooth muscle cells), both from Dako-Agilent (Trappes, France).Sections were then incubated with species-specific secondary antibodies (1:500 dilution, 45 min at room temperature) conjugated to AlexaFluor 488 or 594 (Fischer Scientific) and mounted on microscope slides using the Prolong Antifade Diamond kit (Thermofisher).Image acquisition was performed on a laser scanning confocal microscope (Leica TCS SP8, Leica Microsystems).
Human ethics.Our research complies with all relevant ethical regulations for all the human samples used.Immunostaining studies were performed on arteries obtained after surgery (Ethical Committee CPP Ile de France 2013-13-19) and written informed consent was obtained.
Protein of plaques were from the Athero-Express study, a longitudinal vascular biobank study in which participants provided written informed consent, and the study was approved by the Medical Ethics Committee of the University Medical Center Utrecht (NL45885.041.13,METC 13/597, Medical Ethical Committee of University).Blood monocytes were obtained from a longitudinal biobank study in which participants provided written informed consent, and the study was approved by the ethical Committee for the Protection of Human Subjects in Biomedical Research (Inserm N°C10-14).
Pharmacological in vivo treatment.Eight-week old male Apoe -/-Card9 +/+ and Apoe -/-Card9 -/-mice were treated with daily intraperitoneal injections of rapamycin (4 mg/kg body weight) for 6 weeks and were put on a high-fat diet.Eight-week old male Apoe -/-Card9 +/+ and Apoe -/-Card9 -/-mice were treated with metformin (300 mg/kg body weight, drinking water) for 6 weeks and were put on a high-fat diet.

Extent and composition of atherosclerotic lesions
Plasma cholesterol was measured using a commercial cholesterol kit (DiaSys® Cholestérol FS*).Quantification of lesion size was standardized 51 .Briefly, the basal half of the ventricles and the ascending aorta were perfusion-fixed in situ with 4% paraformaldehyde, then transferred to a PBS-30% sucrose solution, embedded in frozen OCT and stored at −80 °C.Serial 10-μm sections of the aortic sinus with valves (80 per mouse) were cut on a cryostat 52 .One section out of 5 was used for plaque size quantification after Oil red O staining.In total, 16 sections spanning over 800 μm of the aortic root were used to determine the mean lesion area for each mouse.After PBS flushing, the aorta from the root to the iliac bifurcation was removed and fixed with 10% neutral-buffered formalin.After thorough PBS washing, the adventitial tissue was removed and the aorta was longitudinally opened to expose the luminal surface for en-face visualization of atherosclerotic lesions after Oil Red O staining.Quantification of Oil Red O positive surface area was performed by a blinded operator.Aortic collagen content was detected using Sirius red staining.Necrotic core surface was quantified after Masson's Trichrome staining.At least 4 sections per mouse were examined for each immunostaining, and appropriate negative controls were used.Morphometric studies were performed using Histolab software (Microvisions) 53 .For immunostaining on mouse atherosclerotic plaques, we used antibodies raised against Card9 (AA 274-530), MOMA-2 (macrophage detection, MAB1852, Merck Milllipore®) and CD3 (T cell detection, A0452, Dako®) 53 .TUNEL (Terminal dUTP nick end-labeling) staining was performed using In Situ Cell Death Detection Kit (histochemistry staining) and TMR Red kit (Fluorescent staining) (Roche).The investigators were blinded to group allocation during data collection and analysis.
Macrophage experiments.Primary macrophages were derived from mouse BM cells (BMDMs).Tibias and femurs of C57Bl6/J male mice were dissected and their marrow flushed out.Cells were grown in RPMI 1640 medium, 10% FCS, and 15% Macrophage-Colony-Stimulating Factor (M-CSF)-rich L929-conditioned medium for 7 days at 37 °C.To analyze oxidized LDL uptake, BMDMs were exposed to human oxidized LDL (25μg/ml) for 24 h (see oxidation method below).Cells were then washed, fixed and stained using Bodipy (493/503, Thermofischer Scientific D3922).Foam cells were quantified blindly on 6-8 fields and the mean was recorded.To analyze apoptosis susceptibility, macrophages were incubated with OxLDL (200 μg/ml) for 6, 12, and 24 h.Apoptosis was determined by independent experiments using Annexin V-(FITC) apoptosis detection kit with 7-AAD (PerCP) (BD Biosciences) according to the manufacturer's instructions.Intracellular cholesterol (total and ester) quantification was done using Amplex® Red Cholesterol Assay Kit (Invitrogen A12216).For cytokine measurements, BM-derived macrophages were exposed to human oxLDL (25 μg/ml) and stimulated with LPS (1 μg/ml) for 24 h.Cytokine production in the supernatants were measured using specific ELISA immunoassay kits (BD Biosciences).
Immunofluorescence on macrophages.BM-Derived macrophages were fixed in ice-cold methanol, then washed twice in PBS.Cells were permeabilized with TBS + 0,1% triton for 10 min, then non-specific epitopes were blocked in TBS + 0,1% tween + 3% BSA for 20 min.Primary antibodies diluted in blocking buffer were incubated overnight at 4 °C.The following antibodies were used: Guinea-pig anti-P62 (Progen, GP62-C) and Rabbit anti-LC3B (Cell signaling technology, #43566).After washes in TBS + 0,1% tween, cells were incubated with alexa-568 donkey anti-guinea-pig and alexa-488 donkey anti-rabbit secondary antibodies (Invitrogen) for 2 h at room temperature.After incubation, cells were washed, nuclei were stained with Hoechst and slides were mounted with ibidi fluorescent mounting medium.Images were acquired on a Leica SP8 confocal with lightning super resolution module.
Quantifications were done with ImageJ (NIH) using semiautomatic macros.Briefly, cells were manually circled to create ROI, then for each staining a threshold was applied and quantification of the number of particles, mean area of particles and area of ROI was measured using "Analyze particle".For P62, inclusion bodies were defined as particles with area >100px.Macros are available upon request.Experiments were done in 4 replicates.For LC3+ dots per cell, 25 to 47 cells per condition were quantified and for P62+ inclusion bodies, 20 to 53 cells per condition were quantified.
Cholesterol efflux assays.BMDMs were obtained by differentiation of BM cells in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum, 2 mM glutamine, 20% L929 cellconditioned media (as a source of M-CSF), and penicillin-streptomycin for 5 days.BMDMs were loaded with 50 µg/ml [3H]cholesterol-labeled acetylated LDL (acLDL, 1 µCi/mL) for 48 h in serum-free DMEM supplemented with 50 mM glucose, 2 mM glutamine, 0.2% BSA (RGGB), and 100 µg/ml penicillin / streptomycin.The labeling medium was then removed and cells were washed twice in PBS and then equilibrated in RGGB for additional 16-24 h.To measure cholesterol efflux, cells were incubated 4 h at 37 °C in the presence of 60 µg/ml lipid-free apoAI (Sigma), or 30 µg/ml HDL-PL (density = 1.063-1.21g/ml), isolated from normolipidemic plasma by preparative ultracentrifugation, as cellular cholesterol acceptor.Finally, culture media was harvested and cleared of cellular debris by brief centrifugation.Fractional cholesterol efflux (expressed as a percentage) was calculated as the amount of radiolabel detected in the supernatants divided by total radio-label in each well (radioactivity in the supernatant plus radioactivity in the cells) obtained after lipid extraction from cells in a mixture of 3:2 hexane:isopropanol (3:2 vol/vol).The background cholesterol efflux obtained in the absence of any acceptor was subtracted from the efflux obtained with samples.
Forward scatter (FSC) and side scatter (SSC) were used to gate live cells excluding red blood cells, debris, and cell aggregates in total blood cells and splenocytes preparations.Cells were acquired using a BD LSRII Fortessa flow cytometer (BD Biosciences) and analyzed with FlowJo™ (TreeStar, Inc.).
Quantitative real-time PCR.RNA extraction was done either with Trizol or with Qiagen columns (RNeasy MiniSpin Columns) using a polytron (T25 basic, IKA, Labortechnik).The phase containing RNAs was then recuperated and washed with molecular biology water.RNA quality control and concentration were performed using Nanodrop 2000 (Thermofisher scientific).Reverse transcription was done following manufacturer instruction [kit QuantiTect Reverse Transcription (Qiagen)].Real-time fluorescence monitoring was performed with the Applied Biosystems, Step One Plus Real-Time PCR System with Power SYBR Green PCR Master Mix (Eurogentec).qPCR was performed in triplicate for each sample.GAPDH cycle threshold was used to normalize gene expression: (F: 50 -CGTCCCGTAGACAAAATGGTGAA-30; R: 50 -GCC GTGAGTGGAGTCATACTGGAACA-30).Relative expression was calculated using the 2-delta-delta CT method followed by geometric average, as recommended.
Transcriptomic analysis on human monocytes.RNA sequencing libraries were prepared from 100 to 200 ng of total RNA using the Illumina® Stranded Total RNA Prep, Ligation with Ribo-Zero Plus library preparation kit, which allows performing a strand specific sequencing.This protocol includes a first step of enzymatic depletion of abundant transcripts from multiple species (including human cytoplasmic & mitochondria rRNA, mouse rRNA, rat rRNA, bacteria Gram +/-rRNA, human beta globin transcripts) using specific probes.cDNA synthesis was then performed and resulting fragments were used for dA-tailing followed by ligation of RNA Index Anchors.PCR amplification with indexed primers (IDT for Illumina RNA UD Indexes) was finally achieved, with 13 cycles, to generate the final cDNA libraries.Individual library quantification and quality assessment were performed using Qubit fluorometric assay (Invitrogen) with dsDNA HS (High Sensitivity) Assay Kit and LabChip GX Touch using a High Sensitivity DNA chip (Perkin Elmer).Libraries were then equimolarly pooled and quantified by qPCR using the KAPA library quantification kit (Roche).Sequencing was carried out on the NovaSeq 6000 instrument from Illumina using paired-end 2 × 100 bp, to obtain around 100 million clusters (200 million raw paired-end reads) per sample.Raw and normalized counts are provided in Supplementary data.
We performed the gene set enrichment analysis using cluster-Profiler v4.0.5 54 with selected pathways from Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases, and Benjamini-Hochberg correction was applied.Used keywords were: "apopto*", "atheroscleros*", "NF-kappa B", "TNF".All pathways with adjusted p-value below 0.05 were considered as significantly enriched.For each category, median expression of all gene included in core enrichment were calculated for each patient before plotting the heatmap.
Single-cell analysis of Card9/CARD9 expression patterns.Singlecell RNA sequencing (scRNA-seq) datasets of immune cells from mouse aortic atherosclerotic plaques, reported in various publications [55][56][57][58][59][60][61] , (see ref. 34 for details), were pooled and integrated using canonical correlation analysis (CCA) in Seurat v4.3.0 62,63 .All datasets used for analysis were pre-processed as follows: cells containing >200 detected genes, and genes detected in at least 3 cells were included in the analysis using the 'CreateSeuratObject' function with 'min.features= 200' and 'min.cells=3'.Quality control filtering was further performed to remove dead/damaged cells with a high proportion of mitochondrial transcripts, and outlier cells with high UMI numbers (probable doublets/mutliplets).For mitochondrial transcripts, a < 5% cutoff was applied for all datasets, except for the data from Williams et al. 60 and from Gil-Pulido et al. 61 , where <7.5% and <10% cutoffs were applied, respectively.Pre-processing code for aortic leukocyte datasets can be found as supplemental files of Zernecke et al. 34 .All data were log normalized using the 'NormalizeData' function in Seurat with default parameters.Data integration was performed using a canonical CCA workflow in Seurat with default parameters.After CCA integration, data were scaled using 'ScaleData', and principal component analysis was performed using 'RunPCA'.Dimensional reduction was performed using 'RunUMAP' with 30 principal components.Clustering was performed using 'FindNeighbors' with 30 principal components, and 'FindClusters' with a resolution of 0.4.Immune cell annotations are based on Zernecke's works 34,64 .
Mouse scRNA-seq data given in Supplementary Fig. 2 were obtained from Pan et al. 18 and downloaded from Gene Expression Omnibus GSE155513, pre-processed in cellranger-6.1.2,and further analyzed in Seurat v4.3.0 63 .We used data from Ldlr -/-mice fed normal chow or a western diet for 8, 16 or 26 weeks (i.e. the following data from Gene Expression Omnibus GSE155513: GSM4705592, GSM4705593, GSM4705594, GSM4705595, GSM4705596, GSM4705597, GSM4705598, GSM4705599).Individual datasets were pre-processed with quality control filtering in Seurat: cells containing >200 detected genes, and genes detected in at least 3 cells were included in the analysis using the 'CreateSeuratObject' function with 'min.features= 200' and 'min.cells= 3'.Quality control filtering was further performed to remove dead/damaged cells with a high proportion of mitochondrial transcripts (>10%), and outlier cells with high UMI numbers.All data were log normalized using the 'NormalizeData' function in Seurat with default parameters.Data were pooled and batch corrected using Harmony 65 within Seurat.2000 highly variable genes were identified using 'FindVariableFeatures' (with selection.method= "vst").Data were scaled using 'ScaleData' with default parameters, and principal component analysis performed using 'RunPCA' with default parameters, and batch corrected using 'RunHarmony' with default parameters.Dimensional reduction was performed using 'RunUMAP(reduction = "harmony", dims = 1:20)', and clustering was performed at a 0.4 resolution using 'FindNeighbors(reduction = " harmony", dims = 1:20)' followed by 'FindClusters(resolution = 0.2)'.Positive marker genes for each cluster were identified using 'FindAllMarkers'.
Human scRNA-seq data given in Fig. 7 were obtained from total cells of human atherosclerotic coronary arteries 33 and analyzed in Seurat v3 62 starting from the author provided cell-count matrix (downloaded from Gene Expression Omnibus GSE131778).Cells containing <200 detected genes were excluded, and genes detected in at least 3 cells were included in the analysis using the 'Create-SeuratObject' functions with 'min.features= 200' and 'min.cells= 3'.Further quality control filtering was performed and cells with >5% mitochondrial transcripts were excluded, as well as cells with outlier number of UMIs (nCount_RNA > 15,000).A total of 10,934 cells were analyzed.As a pre-analysis indicated a substantial patient-driven batch effect, we performed batch correction using Harmony 65 within Seurat, considering each patient as an independent sample.Data were normalized using the 'NormalizeData' function in Seurat with default parameters.In all, 2000 highly variable genes were identified using 'FindVariableFeatures' (with selection.method= "vst").Data were scaled using 'ScaleData' with default parameters, and principal component analysis performed using 'RunPCA' with default parameters, and batch corrected using 'RunHarmony' with default parameters.Dimensional reduction was performed using 'RunUMAP(reduction = "harmony", dims = 1:20)', and clustering was performed at a 0.4 resolution using 'FindNeighbors(reduction = "harmony", dims = 1:20)' followed by 'FindClusters(resolution = 0.4)'.Positive marker genes for each cluster were identified using 'FindAllMarkers'.Cell type annotation was performed based on expression of known cell lineage markers, and on cluster annotations in Wirka et al. 33 .In Supplementary Fig. 15e-g, the expression of CARD9 was examined in scRNA-seq data of mononuclear phagocytes from human coronary and carotid arteries in an integrated dataset described in details in Zernecke et al. 34 .
Membranes were then blocked with Tris Buffered Saline supplemented with 0.1% Tween-20 (TBST) and 2% BSA (2 h at room temperature), then incubated with primary antibodies (overnight, 4 °C) diluted following the manufacturer's recommendations.After three washes in TBST, membranes were incubated with species-specific horseradish peroxidase-conjugated secondary antibodies (1:8000 dilution, 45 min at room temperature).After three washes in TBST, the peroxidase activity was detected using Clarity Western ECL Substrate (Biorad) using Cytiva's ImageQuant Fluor 800.The migration position of transferred proteins was compared to the PageRuler Prestain Protein Ladder (10 to 170 kDa, Thermo Fisher Scientific).Densitometric analysis was performed using ImageJ software (NIH).Phosphorylated protein signals were normalized on total protein levels, whereas non-phosphorylated proteins were normalized on β-actin protein levels.Primary and HRP-coupled secondary antibodies used for immunoblotting experiments are listed in supplementary data 3. Uncropped Western blots are available as supplementary material.

Microbiota analysis
Stool collection and DNA extraction.Fecal samples were homogenized and 0.2 g aliquots were stored at −80 °C for further analysis.DNA was extracted from fecal samples using a multi-step protocol 66 .Briefly, the feces samples were weighed and then resuspended for 10 min at room temperature in 250 μl of 4 M guanidine thiocyanate in 0.1 M Tris (pH 7.5) (Sigma-Aldrich) and 40 μl of 10% N-lauroyl sarcosine (Sigma-Aldrich).After the addition of 500 μl of 5% N-lauroyl sarcosine in 0.1 M phosphate buffer (pH 8.0), the 2-ml tubes were incubated at 70 °C for 1 h.One volume (750 ml) of a mixture of 0.1and 0.6-mm-diameter silica beads (Sigma-Aldrich) (sterilized by autoclaving) was added, and the tube was shaken at 6.5 m/s three times for 30 s each in a FastPrep (MP Biomedicals) apparatus.Polyvinylpolypyrrolidone (15 mg) was added to the tube, which was then vortexed and centrifuged for 5 min at 20,000 × g.After recovery of the supernatant, the pellets were washed with 500 μl of TENP (50 mM Tris (pH 8), 20 mM EDTA (pH 8), 100 mM NaCl, 1% polyvinylpolypyrrolidone) and centrifuged for 5 min at 20,000 × g, and the new supernatant was added to the first supernatant.The washing step was repeated two times.The pooled supernatant (about 2 ml) was briefly centrifuged to remove particles and then split into two 2-ml tubes.Nucleic acids were precipitated by the addition of 1 volume of isopropanol for 10 min at room temperature and centrifugation for 10 min at 20,000×g.Pellets were resuspended and pooled in 450 μl of 100 mM phosphate buffer, pH 8, and 50 ml of 5 M potassium acetate.The tube was placed on ice overnight and centrifuged at 20,000 × g for 30 min.The supernatant was then transferred to a new tube containing 20 μl of RNase (1 mg/ml) and incubated at 37 °C for 30 min.Nucleic acids were precipitated by the addition of 50 μl of 3 M sodium acetate and 1 ml of absolute ethanol.The tube was incubated for 10 min at room temperature, and the nucleic acids were recovered by centrifugation at 20,000 × g for 15 min.The DNA pellet was finally washed with 70% ethanol, dried, and resuspended in 100 μl of Tris-EDTA (TE) buffer.
Sequencing.Microbiota analysis was performed by amplicon sequencing of the V3-V4 region of the 16 S ribosomal RNA gene.This region was amplified using the following primers -16 S sense 5′-TACGGRAGGCAGCAG-3′ and anti-sense 5′-CTACCNGGGTATC-TAAT-3′according to an optimized and standardized 16 S amplicon library preparation protocol (Metabiote, GenoScreen, Lille, France).Briefly, PCR of the 16 S DNA was performed with 5 ng of genomic DNA according to the manufacturer's protocol (Metabiote), with bar-coded primers (Metabiote MiSeq Primers) to a final concentration of 0.2 μmol/l, with an annealing temperature of 50 °C for 30 cycles.Purification of the PCR products was performed with Agencourt AMPure XP-PCR purification system (Beckman Coulter, Brea, CA, USA), and quantified following the manufacturer's instructions.The samples were multiplexed at equal concentrations.Sequencing was performed on an Illumina MiSeq platform (Illumina, San Diego, CA, USA) using a 250 bp paired-end sequencing protocol at GenoScreen.Raw paired-end reads were subjected to the following processes: (1) quality filtering using the PRINSEQ-lite PERL script 67 , by truncating the bases from the 3′ end, that did not exhibit a quality <30, based on the Phred algorithm and (2) searching for and removing both forward and reverse primer sequences using CutAdapt, with no mismatches allowed in the primer sequences.Only sequences where perfect matching forward and reverse primers were detected were included.
16S sequence analysis.Sequences were quality filtered using the dada2 software package (version 1.12.1) 68in the R programming language (R version 3.6.1)to produce amplicon sequence variants (ASVs).Taxonomic classification was performed using the Silva reference database (version 132) 69 .Bacterial ASVs that could not be assigned to Phylum-level taxonomy were excluded.Alpha diversity was estimated using the number of observed species and the Shannon diversity index.Raw sequence data are accessible in the Sequence Read Archive (accession number pending).Beta diversity analysis was performed on proportion-normalized data using the Bray-Curtis index.Assessment for significant differences between clusters was performed using PERMANOVA with the adonis function in the vegan package (version 2.5-6) in R with 99999 permutations.
Differential abundance was tested using linear discriminant analysis with effect size (Lefse) using default settings 21 .
LDL isolation and oxidation.LDL from normal human pooled sera was prepared by ultracentrifugation and dialyzed against PBS containing 100 µM EDTA.The LDL pool was then diluted to 2 g/l with PBS into a final volume of 3 ml.LDLs were mildly oxidized by UV-C for 2 h in the presence of 5 µM CuSO4 70 .Oxidized LDL contained 4.2-7.4nmoles of TBARS (thiobarbituric acid-reactive substances)/µg apoB.Relative electrophoretic mobility (REM) and 2,4,6-trinitrobenzenesulfonic acid (TNBS) reactive amino groups were 1.2-1.3times and 85-92% of native LDL, respectively.

Statistical analysis
Graphs and statistical analyses were performed using Prism software (Graphpad).Values are expressed as mean ± s.e.m.Differences between values were examined using the nonparametric two-tailed Mann-Whitney, Kruskal-Wallis tests when appropriate and were considered significant at P < 0.05.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/.© The Author(s) 2023

Fig. 8 |
Fig. 8 | CARD9 related pathways in human.A protocol to obtain transcripts after isolation of monocytes from controls and CARD9-deficient patients.B Volcano-plot of the differentially expressed genes between monocytes from patients with CARD9 mutation and control patients.Red dots represent up-regulated genes and blue dots down-regulated genes (adjusted p-value < 0.05).C Heatmap of mean expression in each patient of the leading edge genes contributing to the enrichment of indicated pathways in the GSEA.Immunofluorescent micrograph of human healthy (D) and atherosclerotic (E) carotid artery sections stained for CARD9 (green), α-actin + smooth muscle (red) or CD68 (red) showing that CARD9 was strongly expressed by cells that engulf lipids and cholesterol crystals and of giant lipid-laden foam cells but not by smooth muscle cells.Magnitude X20 (D), X2.5 (E), X40 (E) (2 Pooled experiments, n = 6/staining).F CARD9 expression in 10,934 total human atherosclerotic coronary artery cells from 4 patients (data from Wirka et al. 33 ) with UMAP representation of single-cell RNA-seq gene expression data (left) and cellular lineage identification (right) where CARD9 expression in single cells projected onto the UMAP plot.(For clarity, expression cutoff have been applied and cells with detectable Card9/CARD9 transcripts were brought to the front of the plot).