Aortic intimal resident macrophages are essential for maintenance of the non-thrombogenic intravascular state

Leukocytes and endothelial cells frequently cooperate to resolve inflammatory events. In most cases, these interactions are transient in nature and triggered by immunological insults. Here, we report that, in areas of disturbed blood flow, aortic endothelial cells permanently and intimately associate with a population of specialized macrophages. These macrophages are recruited at birth from the closing ductus arteriosus and share the luminal surface with the endothelium, becoming interwoven in the tunica intima. Anatomical changes that affect hemodynamics, such as in patent ductus arteriosus, alter macrophage seeding to coincide with regions of disturbed flow. Aortic resident macrophages expand in situ via direct cell renewal. Induced depletion of intimal macrophages leads to thrombin-mediated endothelial cell contraction, progressive fibrin accumulation and formation of microthrombi that, once dislodged, cause blockade of vessels in several organs. Together the findings reveal that intravascular resident macrophages are essential to regulate thrombin activity and clear fibrin deposits in regions of disturbed blood flow. Hernandez et al. show that aortic intima resident macrophages (MacAIR) seed the mouse aorta at birth, self-replicate and line the aortic lumen together with the endothelium, protecting the aorta from clot formation in regions of disturbed flow by clearing fibrin deposits and blunting thrombin activity.

s a gatekeeper of cellular traffic between blood and tissues, the endothelium is well equipped to interact with hematopoietic cells. Endothelial cells enable the well-coordinated process of diapedesis that includes capture, rolling and finally transmigration of leukocytes across the vascular barrier [1][2][3][4] . These events rely on complex and sequential molecular interactions by which these cell types cooperate to mount and resolve inflammatory responses at the level of venules and capillaries 2,3 . In large arteries, interactions of the endothelium with inflammatory cells are mostly known for their association with atherosclerosis, a chronic inflammation of the vascular wall 5,6 . In this pathology, leukocytes occupy the subendothelial layer, forming a neointima that progressively expands by the constant influx and local proliferation of inflammatory cells 5,[7][8][9] .
Other types of interactions between arterial endothelium and inflammatory cells have been shown, particularly a population of non-classical monocytes (Ly6C lo CCR2 lo CX3CR1 hi ) better referred to as patrolling monocytes that are thought to promote endothelial integrity and vascular health [10][11][12] . In addition to patrolling monocytes, myeloid cells with a highly dendritic appearance were also described in the luminal aspect of the aorta, especially at sites prone to develop atherosclerosis, such as the aortic arch 13,14 . Despite their conspicuous location, the contribution of these myeloid cells to atherosclerosis was proved to be minimal, as their elimination only slightly delayed onset of the disease with no impact on duration or burden 15 . Thus, the mechanisms behind their peculiar distribution, specific seeding time and, more importantly, their function have remained puzzling.
Here we show that the emergence of aortic myeloid cells is not pathologically induced; instead, it is developmentally triggered as part of natural hemodynamic changes at birth that result in localized disturbed flow dynamics. Genetic ablation of this aortic myeloid resident population promotes fibrin deposition and microthrombus formation, clarifying its function as a critical regulator of hemostasis.

Results
Hemodynamics at birth promote seeding of myeloid cells in the tunica intima of the aorta. Fetal circulation includes two parallel circuits with equal left and right ventricular pressures. At birth, this balance changes drastically due to multiple concurrent events that include interruption of placental circulation, inflation of the lungs and a shift in pulmonary blood pressure. These changes result in high left ventricular pressure and closure of the ductus arteriosus, a fetal vessel that connects the pulmonary arteries to the aorta, an event that further magnifies oscillatory flow in the lower curvature of the aortic arch (Fig. 1a). These alterations in hemodynamics are quickly sensed by endothelial cells, which transition from an elongated to a polygonal shape in the lesser curvature of the aortic arch (Fig. 1b,c). Furthermore, evaluation of mouse embryos and neonates uncovered a burst of inflammatory cells exiting from the constricted ductus arteriosus that seeds the aorta, in tandem with the changes in hemodynamics (Fig. 1d-f). Interestingly, from the onset of birth, this population of CD45 + cells continued to reside in areas experiencing oscillatory and disturbed flow, including the lesser curvature of the aortic arch and branch openings (Fig.1g-i and Extended Data Fig. 1a). Initial characterization indicated that they also expressed CD11c (Fig. 1h,j). Curiously, CD11c + cells were previously detected in the aortic arch of adult healthy mice 13,14 , raising the possibility that they might be the same population. We also found that CD11c + cells progressively accumulated with age in the absence of pathologies or hypercholesterolemia (Extended Data Fig. 1a-c). However, they were not found in large veins, such as the vena cava (Extended Data Fig. 1d,e), indicating that an arterial niche, including flow patterns, may be required for their accumulation. Intimal immune cells were also absent from the carotid arteries of healthy, adult mice (Extended Data Fig. 1f) and from the descending young aorta, except for branches (Extended Data Fig. 1g). We also observed that the distribution of these cells progressively broadened with age. In fact, the descending aorta of 52and 78-week-old mice revealed ongoing accumulation of intimal immune cells even in areas of laminar flow (Extended Data Fig. 1g-i), suggesting that vascular aging might also be a supportive niche for the seeding of intimal myeloid cells.
A definitive link between onset of oscillatory flow and recruitment of CD11c + cells was established using mouse models of patent ductus arteriosus (PDA). Failure in PDA closure significantly alters cardiovascular hemodynamics 16 . While viable and fertile 17 , Vimentin −/− (Vim −/− ) mice exhibit PDA in about 88% of adults, making them an ideal model to study myeloid cell distribution in adult aortae 18 . At 10 weeks of age, the ductus arteriosus in wild-type (WT) littermates becomes a solid fibrous structure that persists as the ligamentum arteriosum (Fig. 1k). By contrast, Vim −/− mice exhibit a viable ductus arteriosus, which impacts patterns of disturbed flow and the location of myeloid cells (Fig. 1l). The lesser curvature of the aortic arch in Vim −/− mice showed no intimal myeloid accumulation; instead, myeloid cells were noted, surrounding the openings of the ductus arteriosus, the subclavian artery and onset of the descending aorta (Fig. 1l).
The thin nature of the endothelial lining makes it difficult to ascertain the precise topology of myeloid cells in relation to the endothelium. Using Cdh5 CreERT2 ;R26 tdTomato reporter mice to label the endothelial monolayer (tdTomato combined with CD45 and vascular endothelial cadherin (VE-Cad) staining), we generated a three-dimensional (3D) rendering of the tunica intima. Myeloid cells were neither above nor below the endothelium but were instead interwoven within the endothelial monolayer, with cell processes projecting into and others below the lumen (Fig. 2a). En face scanning electron microscopy (SEM) of the aortic arch stained with anti-CD45 antibodies confirmed these findings (Fig. 2b). Additionally, 3D surface rendering using a reporter model that labels the tunica media using the Tagln (Sm22) promoter (Sm22 Cre ;R26 tdTomato ) combined with VE-Cad immunohistochemistry further validated that these CD45 + cells were located in the tunica intima (Fig. 2c). Finally, myeloid cells were detected by intravascular injection of anti-CD45 antibodies in vivo (Fig. 2d,e).
During diapedesis, leukocytes adhere to endothelial cells through platelet endothelial cell adhesion molecule (PECAM1), preserving junctional integrity as they cross the endothelial barrier [2][3][4] . We also detected PECAM1 expression in these myeloid cells, indicating that they bind to the endothelium via homophylic, heterotypic interactions and prevent barrier disruption (Fig. 2f). En face images after injection of non-blocking PECAM1-specific antibodies revealed lumen-exposed regions in inflammatory cells (Fig. 2g). When combined, these experiments revealed that, in regions of disturbed flow, the constituency of the endothelial layer is enriched by a population of myeloid cells that intimately coexists with the endothelium in the absence of pathology and without breach of permeability.

scRNA-seq reveals the transcriptional identity of aortic intimal immune cells.
To recover cellular identities independently of defined labeling strategies, we turned to single-cell RNA sequencing (scRNA-seq) using specimens that were not sorted by flow cytometry 19,20 (Extended Data Fig. 2a). Three independent scRNA-seq libraries were generated, capitalizing on regions with abundant numbers of intimal immune cells: the aortic arch of 8-week-old C57BL/6 mice ('young arch') and the descending aorta (thoracic and abdominal) of 78-week-old C57BL/6 mice ('aged descending A,B') (Extended Data Fig. 2b-e and Supplementary Table 1). Using dimensionality reduction by t-distributed stochastic neighbor embedding (t-SNE) analysis, we identified ten distinct cell types and assigned cellular identities (Extended Data Fig. 2f and Supplementary Tables 2,3) based on canonical lineage markers (Extended Data Fig. 2g,h). Within the ten cell types, two distinct macrophage populations were identified based on Fcgr1 (CD64) and Adgre1 (F4/80) expression (Extended Data Fig. 2h). One of the macrophage clusters expressed Lyve1 (encoding lymphatic endothelium hyaluronan receptor 1 (LYVE1)), F13a1 and Mrc1 (CD206), well-known markers of adventitia macrophages 15,21-24 (Extended Data Fig. 2i,j), suggesting that this macrophage population came from the adventitia. While our isolation method enriches for cells in the tunica intima, few cells from the other aortic layers (adventitia and media) were also captured 19,20 (Extended Data Fig. 2f,g).  18.5, connection between pulmonary arteries and aorta through the ductus arteriosus. Postnatal day (P)1, the ductus arteriosus constricts, contributing to disturbed flow. Eight weeks, independent aorta and pulmonary artery. Red, high oxygen levels; blue, low oxygen levels; purple, intermediate oxygen levels. b, Schema illustrating aorta dissection. GC, greater curvature; LC, lesser curvature; desc., descending aorta. c, Confocal images of the lesser curvature and descending aorta from mice at E18.5, P7 and 8 weeks. Elongation factor in the lesser curvature is on the adjacent graph (E18.5, n = 3; P7, n = 4; 8 weeks, n = 5 mice), Mann-Whitney t-test, mean ± s.d., two tailed, P < 0.0001 (exact), ****P ≤ 0.0001; scale bar, 10 μm). d, Bright-field images of ductus arteriosus (DA) remodeling (scale bars, 150 μm, E18.5 (n = 8 embryos); 200 μm, P1 (n = 10 mice); 500 μm, P3 (n = 12 mice); and 1 mm, 3 weeks (n = 12 mice)). e, Exit of immune cells (green) from the ductus arteriosus. VE-Cad or the transcription factor ERG are indicated in red. Scale bars, 300 μm and 70 μm (e(i-iv)); n = 12 mice per time point. f, Lumen of the ductus arteriosus (dotted white lines) at P0; immune cells are shown in green. Scale bar, 50 μm; n = 5 mice. g, Illustration indicating regions of disturbed flow where intimal immune cells (green) accumulate. Branch openings (dashed boxes) are shown in h,i. h, Large branch openings with accumulation of intimal myeloid cells (CD11c, red; CD45, green) (white arrows). ERG staining (white) identifies endothelial nuclei. BA, brachiocephalic trunk; LCCA, left common carotid artery; LSA, left subclavian artery. Scale bars, 200 μm (left) and 20 μm (right); n = 15 mice. i, Time course (1 week, 8 weeks, 78 weeks) of intimal CD45 + cell (green) deposition in intercostal arteries (yellow arrows). VE-Cad, red; intimal CD45 + cells, green. Scale bar, 50 μm; n = 12 (1 week), n = 498 (8 weeks), n = 6 (78 weeks) mice. j, In total, 99.4% of intimal CD45 + cells in the lesser curvature are also labeled as CD11c + (n = 3 mice, mean ± s.d.). k, Bright-field images of control and Vim −/− aortae. Insets show the remnant (control) and PDA (Vim −/− ) with blood (n = 8 mice per group; scale bar, 1,300 μm). l, Aortic arch of 10-week-old control and Vim −/− mice; immune cells in control (green arrows) and Vim −/− (dotted pink ovals) mice are shown. Right, high magnification of squares (l(i,ii)). Scale bars, 500 μm and 40 μm (l(i,ii)); n = 3 mice per group. Therefore, we predicted that one of the two macrophage populations identified was from the adventitia. To test this prediction, we performed single-cell sequencing of dissected aortic adventitia (Extended Data Fig. 3a-g) and compared the two distinct myeloid populations identified. For this, we selected CD14-positive cells from the purified adventitia and intimal-enriched libraries (Extended Data Fig. 3c-e). This approach definitively confirmed the identity of the macrophage clusters from the aorta libraries. The Lyve1 + Fcgr1 + group, also present in libraries from the adventitia, represented typical adventitial macrophages, while the second cluster with distinct expression of Mmp12 and Mmp13 was unique to intima-enriched aortic libraries and absent from the adventitia (Extended Data Fig. 3f,g)  Atlas 25 (Extended Data Fig. 4a-h) and resident macrophages characterized in more recent publications 22 (Extended Data Fig. 4i). From this analysis, it became clear that the aortic population of intimal macrophages was especially distinct from resident macrophages in other organs, with the exception of sympathetic nerve-associated macrophages found in the lung 22 . surface rendering of adult aortae to visualize the spatial location of intimal CD45 + cells (green) relative to the endothelium (tdTomato (tdTom)). VE-Cad is shown in white. Scale bars, 8 μm; n = 3 mice. Max-int. proj., maximum-intensity projection. b, Confocal and scanning electron microscopy (SEM) images of the same aorta were overlaid to determine the location of intimal CD45 + cells (green in the confocal view) in relation to the endothelium. ERG was used to visualize endothelial nuclei in confocal images. High-vac., high-vaccume. Scale bar, 10 μm; n = 3 mice. c, Sm22 Cre ;R26 tdTomato construct of the double transgenic mice used for 3D rendering. Side view of a 3D surface image of adult aortae to visualize the spatial location of intimal CD45 + cells (green) relative to vascular smooth muscle cells (tdTom). Endothelial cells were visualized with anti-VE-Cad antibodies (white). Note that the green CD45 + cell is embedded in an area with white staining (VE-Cad). Scale bars, 5 μm; n = 3 mice. d, Schema of the experimental design related to e,g. Inj., injection; i.v., intravenous. e, Tail vein injection of rat anti-CD45 antibody (detected in green) was followed by euthanasia and fixation of the aorta 30 min later. Subsequent permeabilization and staining with an additional, distinct anti-CD45 antibody (detected using far red, labeled in white). The illustration was created using https://biorender.com/. e, En face images of the experiment described. Yellow arrowheads indicate the portions of immune cells exposed to the lumen (green). Mouse anti-CD45 biotinylated antibody labeled with streptavidin-A647 used in immunostaining after permeabilization is shown in white. PECAM1, red; 4,6-diamidino-2-phenylindole (DAPI), blue. Scale bar, 5 μm; n = 3 mice. f, En face images of the lesser curvature, staining for PECAM1 (green). Yellow arrowheads show faint PECAM1 positivity in CD45 + cells residing in the intima. Endothelial cells were detected in green with PECAM1. Scale bar, 5 μm; n = 6 mice. g, C57BL/6 mice were injected intravenously with a non-blocking PECAM1-specific (390) antibody to further examine exposure of cell bodies to the lumen. Mice were killed and harvested 15-30 min post-injection. 15 min after injection. CD45 (in green) was used to identify immune cells, and ERG (red) was used to identify endothelial cells. Yellow arrowheads indicate PECAM1 + regions of intimal immune cells. Scale bar, 4 μm; n = 3 mice.
Subsequently, we compared the Mmp12 + Mmp13 + macrophage population to a recently identified group of macrophages isolated from whole aortas and referred to as Mac AIR cells 15 . Comparisons between the intimal Mmp12 + Mmp13 + population presented here to the pre-hypercholesterolemic aortic Mac AIR population 15 revealed that they were transcriptionally identical (Supplementary Table 4). Interestingly, we also found that the young arch and old thoracic endothelial-associated macrophage populations were identical (Extended Data Fig. 4j and Supplementary Table 5), suggesting that the endothelial niche is responsible for the underlying tissue-specific imprinting of these macrophages. This realization prompted two immediate questions: were these populations progeny of CD45 + cells exiting the ductus arteriosus? And, more importantly, what was their biological relevance?
Aortic intima resident macrophages (Mac AIR ) are distinct macrophages that seed the aorta at birth. To clarify the function of Mac AIR cells and delve into their developmental origin, we first performed differential expression analysis to seek as many unique markers as possible. As shown by direct comparisons with adventitial macrophages and the Tabula Muris Atlas, we found that Mac AIR cells expressed significantly higher transcriptional levels of matrix metalloproteinases (MMPs) (Mmp12 and Mmp13; Extended Data To match the expression profile to their presumed progenitors, immunohistochemistry was performed on aortae of mice at postnatal day (P)7 and adult mice for targets unique to Mac AIR cells (MMP13, CXCL16 and CD11c). Intimal CD45 + cells colocalized with Mac AIR markers (Extended Data Fig. 5a,b); whereas, in the adventitia, no CD45 + cells showed expression of Mac AIR markers (Extended Data Fig. 5c). We also found that Mac AIR cells expressed Cx3cr1 transcripts (Extended Data Fig. 5d) and CX3CR1 (fractalkine receptor) protein (Extended Data Fig. 5e).
Moreover, using an inducible Cx3cr1 reporter model (Cx3cr1 CreERT2 ; R26 tdTomato ), 98% of all intimal CD45 + cells in adult aortae were labeled with tdTomato (Extended Data Fig. 5f), which also colocalized with CXCL16 (Extended Data Fig. 5g), thus demonstrating the activity of the Cx3cr1 promoter, which was later used for lineage tracing. Additionally, using this transgene, intimal CD45 + cells in aortae of mice at P5 were also labeled by the reporter after tamoxifen treatment (P1 and P3), further indicating that Mac AIR cells seed the aorta immediately after birth (Extended Data Fig. 5h). Overall, these findings support the conclusion that immune cells accumulating and residing in the tunica intima after birth were Mac AIR cells.
To more definitively confirm the origin and the macrophage identity of Mac AIR cells (versus dendritic cells (DCs)), we assessed recombination labeling using the Csf1r MerCreMer ;R26 tdTomato macrophage fate-mapping model. In this transgenic model, we found all intimal immune cells (CD45 + ) to also be positive for tdTomato (Extended Data Fig. 5i). Moreover, all intimal immune cells expressed CD68 (Extended Data Fig. 5j). Additionally, Mac AIR cells did not express the DC master regulator transcription factor encoded by Zbtb46 (ref. 26 ) nor classical DC markers Cd8a, Ccr7 (Extended Data Fig. 5k), Cd103, Dcir2, or Mycl (not detected). Furthermore, Mac AIR cells were shown to phagocytose dying (Annexin + ) endothelial cells in vivo (Extended Data Fig. 5l), providing functional evidence of their macrophage identity. Thus, Mac AIR cells are a transcriptionally unique macrophage population that takes residency in the tunica intima of the aorta shortly after birth in regions of disturbed flow.
whether monocytes participated in replenishment of Mac AIR cells. Although the labeling frequency of Mac AIR cells decreased by 10% 10 weeks after tamoxifen injection, we noted that reporter expression remained constant for an additional 10 weeks (20 weeks in total) (Extended Data Fig. 6e,h). This finding supports the conclusion of local self-renewal with negligible contribution from monocytes.  To further confirm these findings, we performed parabiosis experiments using adult GFP + and WT mice in which both mice shared chimeric circulation for 5 weeks (Extended Data Fig. 6i). The experiment allowed all circulating cells to access the aortic regions of interest (lesser curvature). The presence of GFP + intimal immune cells incorporated in WT mice would infer monocyte contribution and vice versa (Extended Data Fig. 6i). Only mice with excellent level of parabiosis were used (Extended Data Fig. 6j). Evaluation of the lesser curvature in WT mice revealed no GFP + intimal immune cells (Extended Data Fig. 6k), proving further strength to the conclusion that Mac AIR cell maintenance was independent of monocytes. This was also supported by clonal analysis of Cx3cr1 CreERT2 ; R26 Rainbow mice in which inducible Cre recombinase in cells expressing Cx3cr1 would randomly recombine, generating three fluorescently labeled populations (Extended Data Fig. 6l). The data strongly support the notion that Mac AIR cells are self-maintained.
Mac AIR cells originate from definitive hematopoietic cells and expand through self-renewal. Mac AIR cells colonize the aorta as they exit from the ductus arteriosus and migrate to areas of disturbed flow. After this initial wave of migration, Mac AIR cells expand in number by local proliferation (Fig. 3a). Nonetheless, lineage tracing was required to fully ascertain adult progeny. Thus, we performed lineage analysis using the inducible Cx3cr1 CreERT2 ;R26 tdTomato reporter model to label Mac AIR cells immediately after birth and follow descendants over time (Fig. 3b,c). After administration of tamoxifen at birth, 63% of intimal CD45 + cells were positive for tdTomato by P7. At 4 weeks, 91% of intimal CD45 + cells in aortae retained tdTomato labeling ( Fig. 3d-f). We interpret the increase from P7 (63%) to 4 weeks (91%) to indicate that only those inflammatory cells expressing Cx3cr1 were successfully retained in the endothelium. Importantly, at 4 weeks, less than 0.5% of peripheral blood cells were positive for tdTomato (Fig. 3f), supporting the absence of contributions from circulating cells. Therefore, these data indicate that the Mac AIR cells that enter the aorta postnatally expand via self-renewal and colonize the lesser curvature with minimal (if any) input from circulating monocytes.
To assess clonal expansion of Mac AIR cells at early time points, we again used the inducible Cx3cr1 CreERT2 ;R26 Rainbow model (Fig. 3g). Transgenic pups were treated with tamoxifen immediately after birth to induce stochastic recombination, yielding permanent expression of one of three mutually exclusive fluorescent proteins: Cerulean, mOrange and mCherry ( Fig. 3g). At 8 weeks of age, clones of labeled Mac AIR cells were observed (Fig. 3h), confirming that Mac AIR cells self-expand in situ after seeding the aorta immediately after birth. We also examined Mac AIR cells in chemokine receptor 2 (CCR2)-deficient (Ccr2 −/− ) mice, in which Ly6C hi (classical) monocyte emigration from the bone marrow is defective 29 . Compared to control mice, we found no difference in the proportion and number of intimal CD11c + CD45 + (Mac AIR ) cells in the aortae of either young (P7) or adult (8-week-old) mice (Fig. 3i,j). This further solidifies the notion that postnatal aortic intimal macrophages self-expand and maintain their population independently of monocytes at steady state (Fig. 3k).
Assessment of whether Mac AIR cells were derived from primitive or definitive hematopoietic lineages was determined through experiments with a defined lineage-tracing model (Flt3 Cre ;R26 mTmG ) that labels cells arising from definitive hematopoietic stem cells as GFP + (refs. 27,30,31 ) (Extended Data Fig. 6m). Aortae from Flt3 Cre ;R26 mTmG mice at P5 showed that a large proportion of intimal CD45 + cells were GFP + relative to circulating CD11b hi Gr1 lo monocytes (Extended Data Fig. 6n,o). This labeling, combined with the observation that Mac AIRs were detected post-birth, strongly indicates that these cells were derived from a definitive hematopoietic lineage and not from the yolk sac, as is known to occur for some resident macrophage populations 27,28 .

Mac AIR cells blunt thrombin activity in areas of oscillatory flow.
An inducible diphtheria toxin (DTx) model (Cx3cr1 CreERT2 ;Csf1r lslDTR ) in which dual tamoxifen and DTx injections result in elimination of cells expressing both Cx3cr1 and Csf1r was used to deplete Mac AIR cells and clarify their biological relevance (Fig. 4a). Efficient loss of Mac AIR cells was noted 24 h after DTx injection (Fig. 4b) and was associated with altered endothelial morphology and apparent reduction in cell size ( Fig. 4c and Extended Data Fig. 7a). To clarify whether cell size changes were due to cell contraction, we evaluated the expression of phosphorylated myosin light chain 2 (pMLC2). In control mice, Mac AIR cells showed high expression of pMLC2 ( Fig. 4d and Extended Data Fig. 7b). Upon Mac AIR depletion, endothelial cells became positive for pMCL2, indicating a highly contractile functional state, likely the reason behind the drastic change in cell size (Fig. 4c,d and Extended Data Fig. 7b). Together, these findings suggested that Mac AIR cells prevented a contractile endothelial phenotype that would otherwise manifest in areas of disturbed flow through an unclear mechanism.

weeks Cx3cr1
CreERT2 ;Csf1r lslDTR contraction, as shown by pMLC2 expression and cell size quantification (Fig. 4h,i), demonstrating that endothelial contraction was thrombin dependent but mitigated by Mac AIR cells. Thrombin was still present in areas of oscillatory flow after macrophage depletion (Extended Data Fig. 7e). These findings indicate that Mac AIR cells prevent endothelial contraction driven by thrombin in regions of disturbed flow. Importantly, we tested other aspects of clotting in the presence and absence of macrophages, including tail bleed times (Extended Data Fig. 7f) as well as other mediators or regulators of clotting such as nitric oxide and prostacyclin (Extended Data Fig. 7g,h). None of these were altered by depletion of Mac AIR cells.
The next critical question was to sort out the molecular mechanism by which Mac AIR cells impaired thrombin action and, in particular, pMLC2. It is well established that thrombin mediates signals on endothelial cells through PAR1, PAR2 and PAR4) 32,33 , all expressed by aortic endothelial cells but not by Mac AIR cells (Extended Data Fig. 7d). Interestingly, PAR receptors can be cleaved by thrombin, mediating signaling, but also by other proteases such as MMP12 and MMP13 blunting these signals by cleaving PAR1 on the carboxyl-terminal side of the thrombin site [34][35][36][37] . Considering the high levels of MMP12 and MMP13 expressed by Mac AIR cells, it was only logical to predict that MMPs secreted by Mac AIR cells impaired thrombin-mediated contraction. While testing this hypothesis in mouse aortae was impossible, we evaluated thrombin signaling, as shown by pMLC2 levels on cultured endothelial cells in the presence and absence of MMP13 (Fig. 4j). The findings showed that thrombin activates pMLC2 in the endothelium and that this effect is impaired by co-incubation with MMP13.

Mac AIR residency coincides with deposition of fibrin(ogen).
The presence of thrombin in areas of disturbed flow led us to inquire whether fibrinogen, a thrombin substrate, was also found in these regions. Indeed, en face immunostaining revealed buildup of fibrin(ogen) in the lesser curvature (Fig. 5a). Higher magnification images (Fig. 5b) and 3D surface rendering (Fig. 5c) showed fibrin(ogen) decorating the surface of Mac AIR cells. To confirm that fibrin(ogen) was deposited in regions of disturbed flow, we depleted endogenous fibrinogen in vivo and mice with fluorescently tagged fibrinogen to assess binding and accumulation (Fig. 5d). Knockdown of Fbg (fibrinogen gamma chain) was accomplished by delivery of small interfering (si)RNA targeting hepatic fibrinogen mRNA (siFibrinogen), which was encapsulated in lipid nanoparticles containing an ionizable cationic lipid 38 . Seven days after treatment, circulating fibrinogen was ~90% depleted (Fig. 5e), and, at this time, fibrinogen-Alexa Fluor (A)488 (fbg-a488) avidly bound to Mac AIR cells and accumulated throughout the lesser curvature 3 h after injection (Fig. 5f). By contrast, fbg-a488 did not accumulate in the greater curvature (Extended Data Fig. 8a), confirming the predilection of fibrinogen for regions of disturbed flow.
We validated concurrent associations of Mac AIR cells and fibrin(ogen) in intercostal arterial openings (Fig. 5g). The correlation was also noted in descending aortae of 78-week-old mice (Extended Data Fig. 8b). Moreover, a time course evaluation showed progressive fibrin(ogen) accumulation in the aortic arch of mice at P7, 3 weeks and 8 weeks, coinciding with the age-dependent expansion of Mac AIR cells (Extended Data Fig. 8c). As fibrinogen is a substrate for macrophages, we predicted that fibrinogen deposits might be required for seeding Mac AIR cells in the regions of disturbed flow. Thus, we analyzed Mac AIR accumulation in fibrinogen-deficient mice (Fbg −/− ) 39 and in mice expressing a mutant form of fibrinogen that could not be converted to fibrin polymer (Fbg AEK ) 40 . Our prediction was incorrect, as no difference in the abundance or distribution of Mac AIR cells was found in either mouse model (Fig. 5h). These findings revealed that, while overlapping in location, fibrin(ogen) was dispensable for seeding or anchoring Mac AIR cells to the tunica intima.
Mac AIR cells are necessary to clear fibrin(ogen) deposits in regions of disturbed flow. The concurrent presence of both thrombin and fibrin(ogen) in areas of disturbed flow implied that Mac AIR cells might be involved in preventing fibrin formation; therefore, we explored fibrin(ogen) accumulation in the Cx3cr1 CreERT2 ;Csf1r lslDTR model over time (Fig. 6a). Evaluation of fibrin(ogen) levels at 7 and 14 d after continued depletion showed progressive and significant accumulation in the lesser curvature (Fig. 6b-d and Extended Data Fig. 8d). Additionally, fibrin fibrils, which were not detected in any control mice in our experiments, were clearly visible in the aortae of mice depleted of macrophages for 14 d (Fig. 6c). The data imply that Mac AIR cells promote clearance of fibrin(ogen) and/or prevention of fibrin formation. Further, en face SEM images of aortae ( Fig. 6e and Extended Data Fig. 8e) revealed microthrombi and polymerized fibrin decorating the lesser curvature of Cx3cr1 CreERT2 ;Csf1Rr lslDTR mice but not present in littermate controls. Microthrombi were also noted by confocal microscopy along with rupture of the endothelial lining (Fig. 6f,g). As a direct readout of disseminated microthrombi, we evaluated d-dimer levels and found markedly elevated levels in macrophage-depleted mice compared to undetectable levels in control mice (Fig. 6h), further supporting hemostatic imbalance. Histological examination of tissues from Mac AIR -depleted mice revealed hemorrhagic foci in multiple tissues, including the kidney, liver and lung (Fig. 6i). Images of kidneys from Cx3cr1 CreERT2 ;Csf1r l slDTR mice that had to be euthanized due to health decline exhibited abundant fibrin(ogen) throughout the tissue (Fig. 6j), consistent with vascular rupture. Based on the data presented, it is likely that dislodged microthrombi traveling through the circulatory system were responsible for occluding smaller-diameter vessels, leading to hemorrhagic foci and organ damage. Additional support for the requirement of Mac AIR cells in clearing fibrin(ogen) deposits also emerged from evaluation of Cd11c −/− mice. The absence of CD11c resulted in a significant reduction in Mac AIR numbers, indicating that that CD11c was necessary, albeit not fully sufficient, for anchorage of Mac AIR cells to the intimal niche, as we could still detect some macrophages (approximately 45% in comparison to controls) in the tunica intima of this mouse. Importantly, this mouse model also exhibited an impressive accumulation of fibrin(ogen) (Extended Data Fig. 8f).
An alternative explanation for fibrin(ogen) accumulation and thrombosis upon Mac AIR depletion could relate to breach of CD45 Fibrin(ogen)

VE-Cad Merge
Intercostal artery (8 weeks) barrier integrity in areas of oscillatory flow, implicating Mac AIR cells in the maintenance of endothelial junctional integrity in these areas. We tested this possibility by examining fluorescent microsphere (40 nm) deposition in the tunica intima of littermate control and Cx3cr1 CreERT2 ;Csf1r lslDTR mice that both received tamoxifen and DTx injections (1 d of Mac AIR depletion) (Extended Data Fig. 8g). While the positive control (buffered EDTA intracardiac injection for 5 min) resulted in robust deposition of fluorescent beads in between cells, no accumulation of fluorescent beads was found in any of the other groups (Extended Data Fig. 8g). Thus, Mac AIR cells do not play a role in maintaining barrier integrity; instead, they appear to be necessary to clear fibrin(ogen) deposits and mitigate PAR1-thrombin signaling in regions of disturbed flow. The drastic phenotype observed from depletion of Mac AIR cells, a relatively small population, prompted the question of whether the approach might affect a broader group of macrophages. Thus, we evaluated alterations in the macrophage populations of multiple organs by flow cytometry analysis (Fig. 6k). The findings revealed that, while bone marrow and peripheral blood were affected by the dose of tamoxifen and DTx used, none of the other CX3CR1 + CSF1R + macrophage populations in the evaluated organs were altered. The implication is that areas of oscillatory and disturbed flow in the arterial tree are sites of fibrin accumulation that absolutely depend on Mac AIR cells for clearance with critical consequences. A few last important pieces of evidence to solidify this conclusion are still pending, including how Mac AIR cells degrade fibrin(ogen).
Macrophages are known to express plasminogen receptors to degrade fibrin(ogen) extracellularly 41 . Importantly, Mac AIR cells expressed plasminogen receptors (Fig. 6l) and bound to fluorescently conjugated plasminogen when injected intravenously (Fig. 6m). Thus, in addition to MMP12 and MMP13, Mac AIR cells are capable of generating cell surface-associated plasmin, particularly with the aid of endothelial cells, which express high levels of tissue-type plasminogen activator (Extended Data Fig. 8h,i). We also confirmed that neither endothelial cells nor Mac AIR cells expressed plasminogen (Extended Data Fig. 8j). These findings indicate that Mac AIR cells are required to clear fibrin(ogen) deposits, prevent fibrin formation and maintain an anti-thrombotic state in areas of disturbed flow.
Additional support for the conclusion that Mac AIR cells are responsible for clearing fibrin(ogen) deposits in regions of disturbed flow came from experiments in which Mac AIR cells were allowed to return after a 2-week depletion. Much like resident macrophages in other organs 27,28,42-44 , we found that, upon removal of Mac AIR cells and elimination of depletion pressure (tamoxifen and DTx), monocytes seeded areas of disturbed flow and reconstituted the Mac AIR population (Fig. 7a-d). A gradual increase in macrophage number was noted 1 and 2 weeks after depletion (Fig. 7d). Importantly, the increase in Mac AIR numbers was associated with a reduction of accumulated fibrin(ogen) (Fig. 7c,e). Moreover, circulating d-dimer levels (Fig. 7f) and endothelial cell size in the lesser curvature of the aorta (Fig. 7g) also returned to control levels, demonstrating full rescue of the phenotype. Overall, these findings further support the conclusion that Mac AIR cells are required to clear fibrin(ogen) deposits in regions of disturbed flow.

Discussion
Vascular endothelial and hematopoietic cells are well known to coordinate inflammatory responses. Here, we have expanded these functions to also include regulation of intravascular hemostasis. Indeed, our findings indicate that, while endothelial cells provide a non-thrombogenic surface, facilitating blood circulation in areas of disturbed flow, this function is challenged by the accumulation of fibrinogen and thrombin (aortic arch and branches with rapid flow). In these regions, the presence of a population of intima resident macrophages (Mac AIR ), summoned to areas of disturbed flow from birth, is critical to effectively clear fibrinogen and prevent intravascular clotting.
Macrophage association with the endothelium is not necessarily surprising, and, when seen in aortic tissue sections, the assumption is that this heterotypic interaction might be part of an inflammatory or pre-atherosclerotic lesion. However, the unusual feature that captured our attention was that seeding of this macrophage population occurred after birth; furthermore, these cells displayed a unique topology in relation to the endothelium. These elements indicated that the presence of these macrophages in the luminal aspect of the aorta was not part of an inflammatory response; instead, the process was a normal developmental program by which Mac AIR cells become a constitutive component of the tunica intima in regions of disturbed flow. While here we addressed multiple points related to origin, lineage and self-renewal, questions such as why these cells are attracted to areas of disturbed and oscillatory flow and how and why do they migrate from the closing ductus arteriosus to these regions remain unanswered. A logical assumption is that disturbed flow alters the endothelium, creating unique niche conditions that attract monocytes to these sites and promote Mac AIR differentiation into a stable population capable of self-renewal.
healing has been established long ago 45 . However, the repertoire of proteases is distinct in different macrophage subtypes; some proteases are expressed only upon induction and in situations of wound healing. We found that, unlike adventitial and other macrophage populations, Mac AIR cells constitutively express MMP12 and MMP13, known to cleave fibrinogen [46][47][48] . In addition, Mac AIR cells also bind to plasminogen, and this anchorage enables endothelial tissue plasminogen activator (tPA) to generate plasmin, which also degrades fibrinogen and fibrin. Together, these proteases enable intravascular macrophages to efficiently remove fibrinogen deposits and antagonize fibrin accumulation driven by procoagulant pathways and the presence of thrombin in areas of oscillatory and disturbed flow.
Supporting the biological role attributed to Mac AIR cells, it is pertinent to remember that, while compatible with development and reproduction, plasminogen deficiency results in a severe thrombotic phenotype in both normal and inflamed tissues of adult animals 39  thrombotic lesions in multiple organs and a median survival of 176 d, with about 40% of mice succumbing to death. Additionally, these death phenotypes are effectively reversed by the simultaneous imposition of fibrinogen deficiency 39 . These findings underlie a critical constitutive function of the plasminogen-activation system for fibrin surveillance and clearance in non-pathological settings and resonate extremely well with the findings described here.
Given the association of Mac AIR cells with areas of disturbed flow, also known to be pro-atherogenic sites, their potential contribution to atherosclerotic lesions is an important question. Specifically, one   15 . The authors found that, while Mac AIR cells can take up lipids, elimination of these cells only slightly delays but does not alter the burden of atherosclerotic lesions, nor does it change the accumulation of foam cells in lesions. Altogether, the conclusions of that study indicate that, while participating in the process, it is the influx of circulating monocytes into the neointima that is the main source of foam cells. In context, these data together with the fact that Mac AIR cells embed themselves in the tunica intima immediately after birth concur with the notion that the presence of these macrophages is not a response to a pathological insult. The present report adds to the long list of studies that have recently identified self-renewing tissue-resident macrophages in multiple organs 21,22,28,[51][52][53][54] , now including the aortic endothelium in this list. These tissue-resident macrophages were found to seed multiple organs either during embryonic development or shortly thereafter and are derived from either yolk sac erythroid-myeloid or circulating myeloid progenitors 27,28 .
The findings presented here shift several paradigms. First, they challenge the concept that associations between the endothelium and leukocytes are always transient and triggered by acute inflammatory events. In fact, our results highlight a long-term partnership between endothelial cells and macrophages that is not dependent on immune responses; instead, it is triggered by the drastic hemodynamic changes associated with birth. Second, they change the view that a homotypic endothelial layer forms the luminal side of vessels, which now needs to be amended to include macrophages in areas of disturbed flow and aged arteries. Lastly, our findings indicate that intravascular clotting in arteries is constantly antagonized in regions of disturbed flow and the 'aged' endothelium. While the non-thrombogenic function of vessels was attributed exclusively to the endothelium, the data presented here provide clear evidence that, in some regions of the vascular tree, this can only be accomplished with the aid of macrophages.

Methods
Mice. All animal procedures were approved and performed in accordance with the University of California, Los Angeles (UCLA) and Northwestern University (NU) Institutional Animal Care and Use Committee. All other mouse information can be found in Supplementary Table 9. Mouse strains were maintained on a C57BL/6J background, with the exception of Csf1r merCremer mice, which had a mixed background (FVB:C57BL/6). Mice were genotyped by Transnetyx. Male and female mice were used in approximately equal numbers for all experiments except for scRNA-seq experiments in which only male C57BL/6 mice were used to minimize potential sex and strain differences at the transcriptional level for the arch (8 weeks) and 'aged descending (78 weeks) A,B' datasets. However, validation of transcripts was performed in male and female mice. Unless specified, all adult mice used were 8-10 weeks of age.
Aorta en face collection. Adult mice were injected intraperitoneally (i.p.) with 10 mg methacholine to promote smooth muscle cell relaxation and facilitate en face imaging. Immediately after injection, mice were killed and perfused with 10 ml 2% paraformaldehyde (PFA) through the left ventricle (for embryos and neonates, 0.5-3 ml 2% PFA was used). Following perfusion, aortae were removed, and the adventitia was dissected under a microscope. Aortae were opened longitudinally, transferred to a 35-mm silicon-coated dish filled with 2% PFA and pinned to lay flat, exposing the endothelium. Fixation proceeded for 1 additional hour at 4 °C followed by washes with PBS.
Aortic en face immunostaining. For immunostaining, tissue was washed three times with 1× Hank's balanced salt solution (HBSS) and incubated in blocking-permeabilization buffer (0.3% Triton X-100, 0.5% Tween-20, 3% normal donkey serum) for 1 h. The primary antibody cocktail was prepared in blockingpermeabilization buffer and incubated overnight at 4 °C (the endothelial marker ERG, VE-Cad or PECAM1 were always used in conjunction with other markers to label the endothelium). Aortae were washed three times with 1× HBSS and incubated with secondary antibodies for 1 h, washed with 1× HBSS and mounted on glass slides with ProLong Gold (Thermo, P36930). Antibodies and dilutions used are listed in Supplementary Table 9. Aortae were imaged using either an LSM880 confocal microscope (Zeiss) or an A1R HD25 confocal microscope (Nikon). Z-stack and tile scan features were used to image the large, wavy surfaces of the aortae. Resulting tiles were then stitched into a single large image (ZEN 2.0 Black software, Zeiss or NIS-Elements, Nikon), which enabled visualization of the large aortic arch at high resolution. Imaris software (Imaris 9.5.1 and 9.7.0, Bitplane) was used to visualize images in 3D. For a list of software used for analysis, see Supplementary Table 9. Additionally, Denoise.AI (Nikon) was employed to remove Poisson shot noise. Images were acquired using either ×20, ×63 or ×100 objectives.
In vivo labeling. Mice were injected via the tail vein with either rat anti-mouse CD45 antibody or non-blocking PECAM1-specific (390)-daylight 650 antibody diluted in sterile PBS to label the lumen-facing surfaces of Mac AIR cells. Mice were killed 15 min after injection, and aortae were fixed, collected and stained with additional antibodies. For in vivo labeling of CD45, the following antibodies were additionally used: mouse anti-mouse CD45-biotin and streptavidin-A647, donkey anti-rat-A488 and anti-PECAM1 (2H8)-A568. For in vivo labeling of mice injected with PECAM1-specific (390) antibody, the following antibodies were used: anti-ERG and donkey anti-rabbit-A568, and anti-rat CD45 and donkey anti-rat-A488 (Supplementary Table 9).
Immunostaining and imaging of sections. PFA-fixed, paraffin-embedded specimens from kidneys were sectioned at 4 µm. Antigen retrieval was performed using 1× citrate buffer, and then samples were incubated in blockingpermeabilization buffer for 1 h. Sections were then incubated with primary antibodies (Supplementary Table 9) overnight at 4 °C. The following day, samples were incubated with species-specific secondary antibodies for 1 h before mounting in ProLong Gold. Samples were evaluated using an A1R HD25 confocal microscope (Nikon).
Single-cell RNA sequencing. Isolation of intima cells was performed as previously described 20 . In summary, mice were anesthetized and perfused with 10 ml Versene buffer through the left ventricle. Under a dissecting microscope, the adventitia was removed, and the aorta was cut open in Versene buffer, exposing the endothelium. After Versene washes, aortae were bathed in 1× trypsin and incubated twice for 5 min at 37 °C. The endothelium was then gently removed using a microscalpel (EMS, 72046-30) and repeat pipetting, with sc-HBSS (containing 0.04% BSA and 2% FBS to inactivate trypsin and actinomycin D at 1 µg ml −1 to block transcription) applied to obtain a single-cell suspension (Extended Data Fig. 2a). Cells were pelleted at 300g and then treated with 1× red blood cell (RBC) lysis buffer (eBioscience, 00-4333-57) for 1 min and washed twice with 0.04% BSA. To obtain enough cells, six male (78-week-old; 'aged descending') or eight male (8-week-old; arch) C57BL/6 mice were used per library.
To isolate adventitia cells into a single-cell suspension, mice were anesthetized and perfused with 10 ml DMEM. The adventitia was dissected from aortae and dissociated using the Miltenyi Adipose Tissue Dissociation kit (130-105-808). Cells were pelleted at 300g. Following the Miltenyi protocol, the single-cell suspension was additionally treated with 1× RBC lysis buffer (eBioscience, 00-4333-57) and 1 U DNase. The final cell suspension was washed multiple times and resuspended in 0.04% BSA.
scRNA-seq libraries were generated using the 10x Genomics Chromium Single Cell 3′ Library & Gel Bead kit version 2. Cells were loaded accordingly following the 10x Genomics protocol with an estimated targeted cell recovery of 5,000 cells. Sequencing was performed on the Illumina HiSeq 4000 (paired end, 100 bp per read, 8-week-old mice, arch and 'aged descending A,B'). The digital expression matrix was generated by demultiplexing, processing barcodes and counting gene unique molecular indices using the Cellranger count pipeline (version 4.0.0, 10x Genomics). Multiple samples were merged using the Cellranger aggr pipeline. To identify different cell types and find signature genes for each cell type, the R package Seurat (version 3.1.2) was used. Cells that expressed <100 genes or <500 transcripts were filtered out. We ran the DoubletFinder algorithm and set the doublet rate at 2% as recommended by the vendor (10x Genomics). The algorithm predicted that only 0.01% of doublets were captured. Variable genes were selected using the FindVariableGenes function for further analysis. Data were normalized using the NormalizeData function with a scale factor of 10,000. Genes were then scaled and centered using the ScaleData function. Principal-component analysis and t-SNE were used to reduce the dimensionality of the data. Cluster marker genes were found using the FindAllMarkers function. Cell types were annotated based on cluster marker genes. Heatmaps, violin plots and gene expression plots were generated by DoHeatmap, VlnPlot and featurePlot functions, respectively.
Flow cytometry of peripheral blood. To measure reporter labeling (lineage tracing) and donor chimerism (parabiotic mice) of circulating cells, blood was collected by retro-orbital bleeding into tubes containing FACS Buffer at 4 °C. Blood cells were pelleted at 300g and treated with 1× RBC lysis buffer. Additionally, cells were stained on ice with anti-CD45-APC-Cy7 (BD, 557659) and then analyzed on a BD Fortessa machine.
Lineage tracing. Homozygous lox-stop-lox-tdTomato reporter mice were crossed with homozygous Cx3cr1 CreERT2 mice. Tamoxifen induction of Cre activity in the resulting F 1 compound heterozygotes was initiated by dissolving tamoxifen (free base, MP Biomedicals, 0215673891) in sunflower seed oil (Sigma, S5007) and administering 0.01 mg tamoxifen via oral gavage to neonates at P1, P3 and P5. The penetration of intimal tdTomato + CD45 + cells in the aorta at P7 ranged from 40% to 62%, determined by en face confocal imaging (Fig. 3f); thus, analysis of 4-week-old adult mice needed to be normalized to data from littermates at P7. For baseline controls, half of the litter was killed, and blood as well as aortae were collected to determine reporter labeling efficiency. At 4 weeks, the remaining littermates were killed, and blood as well as aortae were collected to determine reporter expression.
Reporter expression in adult Cx3cr1 CreERT2 ;R26 lox-stop-lox-tdTomato mice was induced after i.p. injection of 1 mg tamoxifen three times, every other day. The penetration of tdTomato-positive intimal CD45 + cells in the aorta ranged from 97% to 100%. Aortae and blood were collected at 1 week after tamoxifen treatment as a baseline control for normalization purposes. Littermates were killed 10 weeks after injection to assess retention of reporter expression in adulthood.
For clonal tracing, Cx3cr1 CreERT2 ;R26 Rainbow (ref. 55 ) neonates were administered 0.01 mg tamoxifen via oral gavage at P1, P3 and P5 to yield permanent expression of three mutually exclusive fluorescent proteins: Cerulean, mOrange and mCherry. At 8 weeks, mice were killed, and reporter labeling was assessed. Adult Cx3cr1 CreERT 2 ;R26 Rainbow mice were injected with one dose of 0.02 mg tamoxifen, and aortae were assessed 9 months later (at 11 months of age).
In vivo macrophage depletion. To deplete macrophages, adult (8-10-week-old) Cx3cr1 CreERT2 ;Csf1r lslDTR mice were first injected with tamoxifen i.p. to promote Cre-dependent induction of the DTR. The following day, mice were injected with 200 ng DTx (Sigma, D0564) in sterile 1× PBS to induce apoptosis of DTR-expressing cells. This resulted in cell death of this population in as quickly as 24 h. However, as intimal, aortic myeloid cells are replenished quickly, evaluation of longer time points required continuous injection with tamoxifen followed by DTx every 3 d until the end of the experiment.
EdU-incorporation assay. Mice were injected i.p. with EdU (Fisher Scientific, A10044 at 10 mM). Two hours after injection, mice were terminally anesthetized and perfused with 2% PFA. The aorta was removed, and the adventitia was dissected. Aortae were longitudinally cut to exposed the endothelium. Following fixation, EdU was revealed using A647 following the manufacture's protocol (Invitrogen, C10640). Additionally, aortae were stained for the nuclear endothelial marker ERG and the pan-hematopoietic marker CD45 and with DAPI before being imaged (Supplementary Table 9).
Scanning election microscopy. Mice were anesthetized and perfused with 10 ml 2% PFA through the left ventricle. Aortae were first immunostained en face following protocols listed above. High-and low-resolution images were then obtained by confocal microscopy. Aortae were then washed with 1× HBSS, dehydrated with increasing concentrations of ethanol and subjected to critical point drying followed by gold and/or palladium coating using a sputter coater. High-and low-resolution SEM images were also taken to use landmarks and find macrophages identified with the confocal images. Confocal and SEM images were then overlaid together using Adobe Photoshop (22.5.0).
To image microclots in aortae from macrophage-depleted (Cx3cr1 CreERT2 ; Csf1r lslDTR ) and control mice, mice were perfused with 4% glutaraldehyde through the left ventricle. Aortae were dissected and treated for an additional hour with 4% glutaraldehyde at room temperature and then washed several times with 1× PBS. Aortae were incubated in 1% osmium tetroxide for 1 h, dehydrated with a series of ethanol, dried to the critical point, mounted on pins and coated with 10-nm gold particles for SEM. Aortae were then imaged using a JEOL NeoScope microscope at 10 kV or 15 kV.
Dabigatran treatment. Mice were injected twice daily (morning and night) i.p. with 300 µg dabigatran (BIBR 953, Selleck, S2196) per injection for 5 d. Dosage was determined by the clotting test. After 5 d of treatment, mice were sacrificed, and aortae were analyzed.
In vivo siFibrinogen knockdown. Mice were injected with either siFibrinogen or an siRNA targeting luciferase (siLuciferase, control) at 1 mg siRNA per kg body weight via tail vein injection. siFibrinogen and siLuciferase were each encapsulated in lipid nanoparticles, composed of an ionizable lipid (DLin-MC3-DMA), phosphatidylcholine, cholesterol and a polyethylene glycol lipid, using methods previously described 56 . To determine knockdown efficiency, we quantified fibrinogen protein levels in plasma 7 d after injection. For this, we collected peripheral blood in EDTA-coated tubes (BD, 365974) and isolated plasma by centrifugation. Plasma was diluted (1:50), and fibrinogen levels were measured using a Mouse Fibrinogen ELISA kit (Abcam, ab213478).
d-dimer measurements. Mice were anesthetized, and blood was collected via right ventricle puncture with a 25G syringe. For d-dimer measurements, blood was collected in citrate buffer with a final concentration of 3.2% citrate (for d-dimer). Blood was centrifuged, and citrate plasma was collected and shipped to IDEXX on dry ice for quantification.

Comparison to the Tabula Muris Atlas and the Chakarov et al. study.
Mac AIR cells were compared to monocytes or macrophages from the Tabula Muris Atlas 25 . Specifically, expression values for all monocytes and macrophages were extracted from the Tabula Muris Atlas and merged with expression values of Mac AIR cells from our dataset. The Seurat pipeline described above was applied to cluster cells. As the datasets were generated by different laboratories, the RunHarmony function from the R package Harmony 57 was applied to remove potential batch effects among different tissues. After clustering, Mac AIR marker genes were used to calculate a module score, which was used to identify the cell population from the Tabula Muris Atlas that was similar to Mac AIR cells. The module score was calculated based on the average expression of genes in the list, subtracted by the aggregated expression of randomly chosen control genes.
Our monocyte and macrophage scRNA-seq datasets were also overlapped with data from the two macrophage populations obtained by Chakarov et al. 23 . For this marker, genes extracted from the comparison between monocytes, adventitial macrophages and Mac AIR cells in our study were compared to data from Lyve1 lo MHCII hi and Lyve1 hi MHCII lo cells (MHCII complex contains multiple genes: M2-Aa, M2-Ab1, M2-Eb2, M2-Eb1) from the Chakarov et al. study. The Jaccard index between each pair for cell types from the two datasets was calculated and plotted in the heatmap.
Parabiosis. Mice of similar weight and sex were housed together for 2 weeks before surgery to assess compatibility. Surgeries were performed as described previously 58 . In short, mice were anesthetized, and matching skin incisions were made from the olecranon to the knee joint of each mouse, and the subcutaneous fascia was bluntly dissected to create ~0.5 cm of free skin. The right olecranon of one animal was attached to the left olecranon of the other by a single 3-0 nylon suture and tie. The partners' knee joints were similarly connected. Dorsal and ventral skins were closed, approximated by staples, and the animals were warmed with heating pads and monitored until recovery. Parabiotic pairs were housed one pair per cage and given acidified water (pH 2.5). After 4 weeks of anastomosis, blood samples from each animal in a parabiont pair were analyzed using flow cytometry. Animal pairs with <30% blood chimerism were excluded from our studies.
Annexin staining in aortae. Adult mice were killed and perfused with 3 ml 5% Annexin-A488 conjugate (Invitrogen, A13201) through the left ventricle. Five minutes later, the aorta was dissected and transferred to a 35-mm silicon-coated dish filled with 20% Annexin-A488. Annexin-A488 (20%) was also used to flush the Annexin inside the intact vessel and then incubated at 37 °C for 10 min. Following incubation, tissue was fixed in 2% PFA for 1 additional hour at room temperature, followed by multiple washes with PBS. Aortae were then opened and pinned to lay flat, exposing the endothelium and immunostained as described previously with anti-ERG and anti-CD45 antibodies and DAPI (Supplementary Table 9).
For thrombin and shear stress experiments, HAECs were grown on gelatin-coated, glass-bottom, six-well plates (Cellvis, P06-1.5H-N) to confluency in complete MCDB-131 medium supplemented with 10% FBS. Once confluent, cells were washed with PBS to remove serum, and medium containing MCDB-131 and 4% dextran (Sigma-Aldrich, 31392) was added to confluent HAECs. Vehicle (PBS) or thrombin (Sigma, 10602400001) at a final concentration of 0.625 U ml −1 was then applied. Monolayers were subjected to unidirectional constant laminar flow for 48 h at 130 r.p.m. in a horizontal circular orbit (Benchmark, BT302). Static monolayers used the same dextran-containing medium and were cultured alongside flow-treated monolayers. After flow treatment (48 h), cells were fixed in 2% PFA for 15 min at room temperature and then washed with PBS.
For immunostaining, cells were incubated in blocking-permeabilization buffer (0.3% Triton X-100, 0.5% Tween-20, 3% normal donkey serum) for 1 h at room temperature. The primary antibody cocktail was prepared in blockingpermeabilization buffer and incubated with samples overnight at 4 °C. The following day, aortae were washed three times with 1× PBS and incubated with secondary antibodies for 1 h at room temperature.
Microsphere permeability. One day after DTx injection (Mac AIR depletion), littermate control and Cx3cr1 CreERT2 ;Csf1r lslDTR mice were injected in the left ventricle with 200 µl 40-nm microspheres (1:10, Thermo, F8795). Beads circulated for 5 min, and then mice were euthanized and perfused with 10 ml PBS, followed by 10 ml 2% PFA. Aortae were collected, and whole-mount staining was performed as described above. For the positive control, C57BL/6 mice were injected with 50 mM EDTA buffer solution in the left ventricle, which was allowed to circulate for 5 min to challenge endothelial junctions. Next, these positive control mice were injected with microspheres and harvested as described.
Tail bleeding. One day after DTx injection (Mac AIR depletion), littermate control and Cx3cr1 CreERT2 ;Csf1r lslDTR mice with same age and weight were anesthetized with ketamine-xylazine at a dose of 0.1 ml per 20 g. Using scalpel blade #11, tails of mice were resected exactly 3 mm distal from the tail end. Tails were quickly placed in tail-bleeding buffer (10 mM Tris-HCl, 2 mM CaCl 2 , warmed to 37 °C), and a stopwatch was started. The stopwatch was stopped after the blood stream halted.
Ex vivo nitric oxide measurements. A 5 mM stock solution of 4-amino-5-methylamino-2′,7′-difluorofluorescein (DAF-FM) diacetate (Invitrogen, D3844) was diluted to 10 µM in medium without phenol red (EBM, Lonza, CC-3129). One day after DTx injection (Mac AIR depletion), littermate control and Cx3cr1 CreERT2 ; Csf1r lslDTR mice were euthanized and perfused with 10 ml PBS followed by 3 ml 10 µM DAF-FM diacetate. Aortae were then quickly and gently collected in 10 µM DAF-FM. Intact aortae were then incubated in DAF-FM for 10 min at 37 °C. After 10 min, DAF-FM was replaced with fresh DAF-FM, and the sample was incubated for an additional 15 min. Aortae were then quickly filleted open and mounted with PBS. Confocal imaging occurred immediately after to evaluate nitric oxide levels.
Quantification and statistical analysis. Treatments were randomized; investigators were blinded to allocation for outcome assessment. No samples or animals were excluded from this study. Sampling size for animal experiments was determined by power analysis with a type 1 error rate of 5% and a minimum detectable effect of 20%. For cell surface experiments, we used sampling similar to previous published reports. For experiments for which the outcome was immunofluorescence, figures show representative images; however, the number of independent times that the experiment was reproduced using biological replicates is provided in the legend.
Quantification of intimal immune cells in aortae was performed using the spots function in Imaris 9.5.1 or 9.7.0 (Bitplane) on maximum-intensity Z projections. Only cells with clearly distinguishable bodies and nuclei (DAPI, not shown) were quantified. No samples or animals were excluded from this study.
Cell surface area and elongation factors of endothelial cells in aortae of control and macrophage-depleted (Cx3cr1 CreERT2 ;Csf1r lslDTR ) mice was determined in several regions of 44.1 mm 2 within the lesser curvature of each whole-mount, flat-mounted aorta. Measurements were performed with NIS-Elements, using a combination of manual and automated cell shape identification. VE-Cad or PECAM1 staining was used to define endothelial cell borders. For macrophage-depletion experiments, we used littermate controls that received both tamoxifen and DTx injections. For evaluation of fluorescence, mean fluoresce intensity was determined using the surface function in Imaris 9.5.1 or 9.7.0 on maximum-intensity Z projections. Subsequently, the mean for at least three independent samples and s.d. was determined. To assess whether two datasets were significantly different, we calculated P values with unpaired, nonparametric Student's t-test followed by the Mann-Whitney test; P < 0.05 was considered significant (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). Statistical analyses were performed with Prism 8 (GraphPad Software).
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data generated or analyzed are included in the main article and associated files. scRNA-seq data were deposited in the GEO database under accession numbers GSE161787, GSE125691 (ref. 22 ) and GSE154817 (ref. 15 ). The Tabula Muris Atlas dataset is deposited at https://www.czbiohub.org/tabula-muris/ 25 . Source data are provided with this paper. Fig. 1 | Intimal immune cell distribution in the aorta and vena cava of young and aged mice. a, Accumulation of intimal immune cells, as per CD45 (green arrows) in adult (n = 498 mice) and old 78wk (n = 6 mice) aortic arch. VE-Cadherin (red). Scale bars, 500μm & 15μm (a I-II ). b, High magnification of en face distribution of CD45 + (green) cells in the lesser curvature (LC) of adult and aged aortae. Scale bar, 20μm. c, Quantification of intimal CD45 + cells in LC of 8wk and 78wk old mice. (n = 4 mice per group, Mann-Whitney T-test, ± SD, p = 0.0286, two-tailed, *p ≤ 0.05). d, e Images of the vena cava of 8wk old mice. d, Low magnification of vena cava (VE-Cadherin in white). Scale bar, 500μm. e, No intimal CD11c + (red) CD45 + (green) cells were found in the vena cava, even in regions of disturbed blood flow (Scale bar, 10μm, n = 3 mice). f, Few intimal CD45 + cells (green) accumulate in carotid arteries (8wk), VE-Cadherin in red (Scale bar, 300μm, 50μm (f I-II ), n = 8 mice). High magnification inserts on the right showing carotid branches (disturbed flow). g, Whole mount scans of the descending aortae (thoracic and abdominal) of 8wk and 78wk old C57BL/6 mice. Except for branches, few intimal CD45 + cells (green) were detected in 8wk old descending aortae, (g I ) whereas 78wk old mice display large number of intimal CD45 + cells (g II ). Scale bar, 500μm (8wk thoracic), 700μm (8wk abdominal), 1000μm (78 wk thoracic and abdominal), and 50μm (g I-II ), n = 5 (8wk) and 6 (78wk) mice. h, Topographic map to guide the quantification of immune cells shown in 'i'. i, Intimal CD45 + cells per aortic region at 8wk, 52wk, and 78wk old mice (n = 3 mice per timepoint, two-tailed T-test, ± SD, p = 0.0022 (2: 8wk vs 52wk), p = 0.0031 (2: 8wk vs 78wk), p = 0.0175 (2: 52wk vs 78wk), p = 0.0396 (3: 8wk vs 78wk), p = 0.0052 (5: 8wk vs 52wk), (*p ≤ 0.05, **p ≤ 0.01). Fig. 4 | Mac AIR comparison to other published macrophage data sets. a, Uniform manifold approximation and projection (UMAP) plot of monocytes and macrophages extracted from the Tabula muris atlas and merged with the aortic intima-macrophage (Mac AIR ) data set ('Aorta') from this work. b, UMAP plot identifying distinct cell clusters based on transcriptional signatures. c, Mac AIR marker genes were used to calculate a module score, which was used to identify cells from the Tabula muris atlas that were similar to Mac AIR . The Mac AIR module score was applied and represented by heat map-style UMAP plot. d, Number (top graph) and percentage (bottom graph) of cells from each tissue that are in each cluster. e, UMAP heat map-style representation of MMP12 (Mac AIR marker), Cx3cr1, and Csf1r (top row) with violin plot representation of each given gene below. f, Heat map comparing the top 50 MacAIR markers to all the clusters found in b. g, UMAP plotting only cluster 2 and showing tissue origin. h, UMAP heat map-style representation of top Mac AIR markers: Mmp12, Mmp13, Cxcl16, Itgax (CD11c), i, Differential genes from the comparison among monocytes, adventitia macrophages, and Mac AIR (our data -this study) were overlapped with genes extracted from the comparison between the Lyve l°w MHCII high and Lyve high MHCII l°w macrophages from the Chakarov et al. dataset. The jaccard index between each pair for cell types from the two datasets were calculated and plotted in the heatmap. j, Comparison of Mac AIRs identified in the tunica intima-enriched young arch vs aged descending data sets.