CXCR3-dependent accumulation and activation of perivascular macrophages is necessary for homeostatic arterial remodeling to hemodynamic stresses

To gain mechanistic insight into our previous observation that the expression of IP-10, Mig, IL-1β, IL-6, TNF, and iNOS is induced within conduit arteries in response to outflow reduction (Tang et al., 2008), we investigated if any of these proinflammatory factors are necessary for flow-mediated vascular remodeling. To test the role of IP-10 and Mig, we compared adaptive vascular remodeling in WT versus CXCR3-deficient (CXCR3^−/−) mice by measuring common carotid artery size and wall thickness at 2 wk after left external carotid artery ligation (Fig. 1 a). In WT mice, the left common carotid arteries displayed a 10–15% reduction in external perimeter, whereas medial thickness remained unchanged compared with the right common carotid arteries from the unligated side (Fig. 1, b–d). Strikingly, left common carotid arteries in CXCR3^−/− mice did not inwardly remodel but responded by medial thickening instead. In contrast, mice with a deficiency of other proinflammatory factors induced by decreased blood flow (or with a deficiency of their receptors), such as IL-1R, IL-6, TNFR1/R2, and iNOS, showed similar vascular remodeling responses to that in WT mice, except for modestly less vessel shrinkage in IL-6^−/− mice (Fig. 1 d). These results suggest a nonredundant role for CXCR3 ligands, such as IP-10 and Mig, in adaptive vascular remodeling that is not a general feature of other proinflammatory factors induced by perturbations in blood flow.

Although CXCR3 is known to be expressed by both mouse T cells and macrophages (Luster and Leder, 1993; Taub et al., 1993; Luster et al., 1995), we have previously shown that vessel shrinkage in response to decreased blood flow is dependent on macrophages and not T cells (Tang et al., 2008). Therefore, we tested if monocytes from WT mice were sufficient to restore flow-mediated vascular remodeling in CXCR3^−/− recipients. We were unable to achieve detectable adoptive transfer of blood monocytes, likely because of limited yield (Fig. S1). Larger numbers of bone marrow myeloid cells could be obtained after density centrifugation and immunomagnetic bead depletion of lymphoid cells and erythrocytes. The enriched cell population from WT mice was 95% positive for the myeloid cell marker CD11b and 60% positive for CXCR3, albeit at relatively low levels (Fig. 2 a). CXCR3 expression by the selected bone marrow cells was limited to Ly-6C⁺ monocytes (Fleming et al., 1993) and was not detectable on small numbers of contaminating Ly-6G⁺ granulocytes or CD42d⁺ platelets (Fig. S2). Adoptively transferred bone marrow myeloid cells persisted in the circulation of WT and CXCR3^−/− hosts for at least 3 d (Fig. S1), which was the peak period of perivascular macrophage recruitment in this model (Tang et al., 2008). We also used polymorphic determinants to the CD45 leukocyte common marker to confirm that donor cells tracked to the remodeling vessel wall in WT and CXCR3^−/− hosts (Fig. S1). i.v. administration of bone marrow myeloid cells from CXCR3^−/− donors to CXCR3^−/− recipients did not alter the abnormal response of absent inward remodeling and aberrant medial thickening in common carotid arteries after external carotid artery ligation (Fig. 2, b and c), which was similar to that seen in CXCR3^−/− mice without adoptive transfer of cells. In contrast, adoptive transfer of bone marrow myeloid cells from WT donors to CXCR3^−/− recipients restored the phenotype of inward remodeling without medial thickening in response to outflow reduction, as in WT mice. CXCR3 was detected in the ligated, but not unligated, carotid arteries of CXCR3^−/− recipients of WT myeloid cells (Fig. 2 d), confirming selective homing of the adoptively transferred cells to the remodeling vessel. These experiments implicate monocytes/macrophages as relevant cell types involved in CXCR3-dependent flow-mediated vascular remodeling.

Because CXCR3 is necessary for vascular remodeling, we examined for its expression within the vessel wall of WT animals. Quantitative PCR analysis showed a time-dependent increase in CXCR3 transcripts within common carotid arteries after external carotid artery ligation, with peak expression at 3 d (Fig. 3 a). Immunohistochemical analysis verified the presence of CXCR3 and one of its ligands, IP-10, within the adventitia of remodeling arteries (Fig. 3, b and c). These results mirrored our previous finding that macrophages constitute the major population of artery-associated leukocytes in this model, with peak accumulation between 3 and 7 d after outflow reduction (Tang et al., 2008), and the expression of CXCR3 largely coincided with that of the macrophage marker F4/80 by immunofluorescence imaging (Fig. 3 d). However, not all F4/80⁺ cells expressed CXCR3 and certain CXCR3⁺ cells did not express F4/80. We quantified the expression of these markers by flow cytometric analysis of myeloid cells selected by CD11b antibody-coated magnetic beads from various sites (initial selection of relevant cells with immunomagnetic beads was necessary to avoid confounding signals from excessive debris in enzyme-digested artery preparations). The majority of adventitial macrophages, defined as CD11b-selected F4/80⁺ cells, expressed CXCR3 as did circulating monocytes with the same surface markers (Fig. 3, e and f). The spleen and bone marrow contained fewer such cells, although these differences did not achieve statistical significance. Thus, CXCR3 expression is widely detected on diverse populations of F4/80⁺ monocytes/macrophages.

We further assessed the relative expression of CXCR3 by various populations of monocytes and macrophages defined by the Gr1 marker using flow cytometric analysis. Antibodies to the Gr1 determinant react with both Ly-6C (expressed by monocytes and granulocytes) and Ly-6G (expressed by granulocytes but not monocytes), and low versus high expression of Gr1 characterizes distinct subpopulations of mouse blood monocytes and plaque-infiltrating macrophages (Sunderkötter et al., 2004; Tacke et al., 2007). Two separate groups of CD11b-selected F4/80⁺ blood monocytes were identified by low versus high Gr1 expression with modest surface expression of CXCR3 (Fig. 4 a, R1 and R2). In contrast, CD11b-selected F4/80⁻ Gr1^high cells had little CXCR3 expression (Fig. 4 a, R3). Both the Gr1^high cell populations (R2 and R3) expressed Ly-6C, but not Ly-6G, identifying them as monocytes (Fig. S3). The F4/80⁻ population (R3) likely represents monocytes recently emigrated from bone marrow or spleen. Artery-infiltrating macrophages had increased F4/80 expression, far fewer Gr1^high cells, more frequent Gr1^low cells, and 10-fold greater CXCR3 expression as compared with blood monocytes (Fig. 4 b). CD11b-selected F4/80⁻ Gr1⁺ cells were not detected within the vessel wall. Forward versus side scatter profiles revealed two populations of small and large artery-associated macrophages with overlapping expression patterns and less than twofold differences for F4/80, Gr1, and CXCR3 (Fig. S4). The flow cytometry data indicate that a unique subtype of macrophage accumulates around the vessel wall in response to decreased blood flow, and a distinguishing feature of the remodeling-associated macrophage is the higher expression of CXCR3.

Having shown that macrophages associated with remodeling arteries express CXCR3, we next investigated if CXCR3 signaling contributes to monocyte trafficking to the vessel wall. The early induction of IP-10 and Mig transcripts in vascular cells at 6 h after external carotid artery ligation did not differ between WT and CXCR3^−/− mice (Fig. 5 a). However, the accumulation of F4/80 transcripts was less in common carotid arteries from CXCR3^−/− mice compared with WT mice at 3 d (Fig. 5 b) and 7 d (not depicted). We verified decreased recruitment and/or retention of macrophages in CXCR3^−/− animals by counting F4/80⁺ cells in artery sections (Fig. 5, c and d). To determine if CXCR3 signaling was sufficient for monocyte infiltration of the vessel wall independent of lymphocyte trafficking, we applied IP-10 to the tissues surrounding the common carotid artery of RAG1^−/− mice deficient in T and B cells. This resulted in infiltration of F4/80⁺ cells to the adventitia of IP-10–treated, but not contralateral, vehicle-treated vessels at 3 d (Fig. 5 e). Macrophages were not detected within the media or intima of the arteries similar to the limited distribution noted after blood flow reduction. These results demonstrate that CXCR3 was necessary and sufficient for perivascular macrophage accumulation.

We further investigated if CXCR3 signaling was required for the production of macrophage mediators of vascular inflammation and remodeling. iNOS, but not TNF, transcripts were diminished in common carotid arteries from CXCR3^−/− mice compared with WT mice at 3 d after outflow reduction (Fig. 6 a), which was previously documented as the time of peak expression for these inflammatory mediators in this experimental model (Tang et al., 2008). Because our results showed that neither iNOS nor TNF played nonredundant roles in inward vascular remodeling (Fig. 1 d), we also assessed the expression of other effector molecules that have been associated with extracellular matrix turnover and flow-mediated vascular remodeling, such as the proteinases, MMP2 and MMP9, and the transglutaminase FXIIIA (Bassiouny et al., 1998; Galis et al., 2002; Bakker et al., 2006). In common carotid arteries of WT mice, MMP9 transcripts that were undetectable preoperatively were induced early between 6 h and 3 d after external carotid artery ligation, whereas the basal expression of MMP2 and FXIIIA transcripts was up-regulated later between 3 and 7 d after operation (Fig. 6 b). By immunofluorescence imaging, the induction of MMP9 expression was only detected in the adventitia, whereas the up-regulation of MMP2 expression was restricted to the media at 3 d after outflow reduction (Fig. 6 c). In contrast, FXIIIA expression was expressed in both the media at baseline and was also associated with perivascular infiltrating cells (Fig. 6 c). We confirmed increased gelatinase activity as a result of MMP9 and MMP2 expression in remodeling arteries by gel and in situ zymography and increased transglutaminase activity within the adventitia, in addition to basal medial activity, after decreased blood flow (Fig. 6, d and e). The early accumulation of MMP9 transcripts and the later accumulation of MMP2 and FXIIIA transcripts were diminished in common carotid arteries from the ligated side of CXCR3^−/− mice compared with WT mice (Fig. 6 f). These results do not exclude the possibility that the diminished expression of factors produced by activated macrophages was entirely the result of decreased monocyte recruitment and/or macrophage retention.

To determine if CXCR3 signaling may contribute to macrophage activation independent of chemoattractant effects, we first assessed the products of different macrophage populations defined by CXCR3 expression from remodeling common carotid arteries of WT mice. The abundance of FXIIIA transcripts was significantly higher in CXCR3-expressing CD11b-selected F4/80⁺ cells, and a similar trend for MMP9 transcripts did not achieve statistical significance (Fig. 7 a). In contrast, TNF was detected in both CXCR3⁺ and CXCR3⁻ macrophages and MMP2 transcripts were not detected in either macrophage population. To examine the production of FXIIIA and MMP9 at the single cell level, we analyzed CD11b-selected CD3⁻/B220⁻/CD11c⁻ perivascular macrophages by immunofluorescence microscopy. There was a heterogeneous expression pattern in macrophages activated with PMA/ionomycin ex vivo with minor populations expressing predominantly FXIIIA, predominantly MMP9, or neither molecule, although the majority of cells displayed dual FXIIIA and MMP9 immunoreactivity (Fig. 7 b). These diverse macrophage subgroups contained both smaller cells with more rounded nuclei and larger cells with kidney- and horseshoe-shaped nuclei (Fig. 7 c). Similar findings with lesser intensity of signal were seen in untreated adventitial macrophages (Fig. 7 d). These data suggested that many CXCR3-expressing remodeling-associated macrophages displayed both proteinase and transglutaminase activities.

We next examined for direct evidence of IP-10 effects on macrophages in vitro. Bone marrow monocytes were differentiated in culture to a macrophage-type phenotype in the presence of macrophage (M) CSF–conditioned medium for 7–10 d, which increased CXCR3 expression (unpublished data), before treatment with various proinflammatory agents. IP-10 selectively induced the expression of FXIIIA transcripts, whereas a prototypical alternative activator of macrophages, IL-4, and a prototypical classical activator of macrophages, LPS, induced the expression of MMP9 and TNF transcripts, respectively (Fig. 8 a). We verified the IP-10–mediated up-regulation of FXIIIA transcripts with additional dose-response experiments, as well as induction of FXIIIA protein by Western blotting which was associated with increased transglutaminase activity (Fig. 8 b). Similarly, FXIIIA was induced by IP-10 in blood monocytes independent of the need for M-CSF–mediated differentiation (Fig. S5). In contrast, IL-4 and LPS diminished the basal expression of FXIIIA by bone marrow–differentiated macrophages and inhibited the effects of IP-10 but not vice versa (Fig. 8 c). Further analyses were performed for other markers of alternative and classical macrophage activation. As expected, IL-4 up-regulated arginase I, Ym1, and Fizz1 transcripts (Nair et al., 2003), whereas LPS induced iNOS transcripts. IP-10 at 300 ng/ml for 72 h did not significantly affect the expression of any of these molecules (Table S1). Unlike IL-4 or LPS, IP-10 did not modulate the production of the immunoregulatory cytokine IL-10. Additionally, the expression of TGM2, an alternative source of transglutaminase activity in remodeling vessels, was not modulated by IP-10 but was inhibited by IL-4 and LPS. We verified that the IP-10–mediated induction of FXIIIA was CXCR3 dependent, which is in contrast to the qualitatively similar effects of IL-4 and LPS on WT cells in the presence of CXCR3 blocking antibodies or on CXCR3^−/− cells (Fig. 8 d). Together, these in vitro data demonstrate that CXCR3 signaling contributes to selective and unique features of macrophage activation in the absence of other cell types.

To determine if our findings in mouse models may be of relevance in human biology, we examined the phenotype of perivascular macrophages in clinical specimens. CD11b⁺ cells were selected by magnetic beads from blood and aortic adventitia and analyzed by flow cytometry. CXCR3 was expressed by a distinct population of CD11b-selected CD14⁺ adventitial macrophages but not by circulating monocytes (Fig. 9 a). CD11b⁺/CD14⁺/CXCR3⁺ perivascular cells did not express lymphocyte or neutrophil markers and were CD3⁻/CD15⁻/CD19⁻/CD56⁻/CD177⁻ (Fig. S6). The frequency of CD11b⁺/CD14⁺/CXCR3⁺ cells was higher in the adventitia of aneurysmal aortas than of nonaneurysmal aortas (Fig. 9 a). Further analysis of CD11b-selected adventitial cells revealed that CXCR3 was more frequently expressed by macrophages characterized by a CD14⁺/CD16⁺ resident phenotype than by a CD14⁺/CD16⁻ inflammatory phenotype (Fig. 9 b). The expression of the scavenger receptor CD163 was also greater in CD14⁺/CD16⁺ than in CD14⁺/CD16⁻ perivascular macrophages (Fig. 9 b). Finally, there were more numerous CD68⁺ macrophages resident in adventitia at sites of disturbed blood flow, such as the common carotid artery bifurcation (carotid sinus), than at sites of laminar flow, for example, the midpoint of the common carotid artery (26.5 ± 3.2 vs. 14.5 ± 2.0 mm⁻³, respectively; P < 0.05) with a higher expression of CXCR3 (Fig. 9 c). These data confirm that CXCR3 is expressed by a subset of human perivascular macrophages and that accumulation of this cell population spatially correlates with sites of disturbed hemodynamic stresses on the vessel wall.

In this study, we report several novel roles for CXCR3 in mouse systems: CXCR3 is necessary for flow-mediated inward vascular remodeling; CXCR3 is expressed by a subset of circulating monocytes and perivascular macrophages with resident phenotypic markers; CXCR3 is required for perivascular macrophage accumulation and activation after decreased blood flow; and CXCR3 signaling selectively induces FXIIIA expression by monocytes and macrophages. We also find that CXCR3-expressing human macrophages of resident phenotype accumulate at sites of disturbed blood flow. These findings provide new insights into the mechanisms of macrophage-mediated vascular remodeling in response to hemodynamic forces and suggest that a unique macrophage phenotype is involved in homeostatic responses triggered by tissue stress.

In our previous study, we found two distinct roles for MyD88 in flow-mediated inward vascular remodeling (Tang et al., 2008). It was required both for ROS-initiated vascular cell inflammatory responses and consequent macrophage accumulation and for activation of macrophage effectors of vascular remodeling. We did not define inflammatory or remodeling mediators with nonredundant roles in this initial work. In our present study, we find that CXCR3 is necessary for perivascular macrophage recruitment in response to blood flow reduction and that IP-10 induces FXIIIA expression. The two studies are related in both phases of inward vascular remodeling. First, MyD88-dependent vascular cell inflammatory responses are required for the production of CXCR3 ligands. Second, both MyD88-dependent and CXCR3 signaling are required for production of macrophage mediators of vascular remodeling. The MyD88-dependent effectors of vascular remodeling are as yet undefined. FXIIIA is identified as a CXCR3-dependent macrophage product that is not induced by ligands that signal through MyD88, although it has not been proven to play an essential role in our model.

Although we have previously described the production of IP-10 and Mig during flow-mediated inward vascular remodeling, it was surprising to find a nonredundant role for their receptor CXCR3, as we had also found that this process is largely independent of T cells and only minimally reduced in RAG1^−/− mice (Tang et al., 2008). CXCR3 is known to play an essential role in intimal remodeling in hypercholesterolemic mice (van Wanrooij et al., 2005, 2008; Veillard et al., 2005) and after wire-mediated injury of the mouse femoral artery (Schwarz et al., 2009). However, the function of CXCR3 in these models of neointima formation is as a chemoattractant for T cells. Specifically, monocyte recruitment to atherosclerotic plaques is not reduced in hypercholesterolemic animals with genetic deficiency or pharmacologic inhibition of CXCR3 (van Wanrooij et al., 2005, 2008; Veillard et al., 2005). This discrepancy with our findings may reflect different homing requirements or retention signals for adventitial versus intimal compartments under physiological and pathological settings, respectively. Leukocyte–endothelial interactions are dependent on hemodynamic factors and the endothelial phenotype, which differ between adventitial microvessels and the lumen of conduit arteries (Langheinrich et al., 2007). Temporal and spatial variables may also play a role. Notably, the perivascular inflammation in response to altered shear stress in our model is transient and resolves without neointima formation after the adaptive changes in vessel size have occurred within 1–2 wk (Tang et al., 2008). The interpretation of our adoptive transfer experiments in which WT monocytes rescued the inward vascular remodeling phenotype in CXCR3^−/− mice is limited by the caveats that negatively selected bone marrow myeloid cells may contain rare nonmonocyte precursor cells with sufficient proliferation potential to influence the results, small numbers of contaminating granulocytes that do not express CXCR3 may have nonredundant effects, and the initial perivascular accumulation of relatively few WT macrophages may secondarily recruit CXCR3^−/− cells thus playing a critical but not sufficient role.

We document variable expression of CXCR3 by mouse CD11b⁺ F4/80⁺ cells from different anatomical sites. Robust expression of CXCR3 was previously reported for mouse thioglycollate-elicited peritoneal macrophages (Janatpour et al., 2001). In contrast, CXCR3 was undetectable in many analyses of human monocytes and macrophages (Janatpour et al., 2001; Katschke et al., 2001; Geissmann et al., 2003; Cho et al., 2007). We confirm that human circulating monocytes do not express CXCR3, but we demonstrate that perivascular macrophages, particularly of cells characterized by a CD14⁺/CD16⁺ resident phenotype (Geissmann et al., 2003), selectively express CXCR3 in addition to the scavenger receptor CD163 (Kim et al., 2006). Tissue macrophages that coexpress CD163 and FXIIIA have been described in human skin (Zaba et al., 2007). Because unstimulated human monocytes, unlike mouse myeloid cells, do not express detectable CXCR3, signaling by this receptor may recruit rare circulating precursors or contribute to the retention and activation of differentiated human tissue macrophages at sites of disturbed hemodynamic stresses. IP-10 elicits mouse monocyte recruitment in vivo (Luster and Leder, 1993; Taub et al., 1993, 1996), and deficient CXCR3 signaling is associated with diminished accumulation of macrophages in various mouse models of inflammation (Hancock et al., 2000; Janatpour et al., 2001; Waeckel et al., 2005; Heller et al., 2006; van Wanrooij et al., 2008). However, the interpretation of these findings is limited without confirmation of CXCR3 expression by the recruited macrophages and by not excluding indirect effects on monocyte trafficking by CXCR3-expressing T cells. More direct evidence of CXCR3-mediated effects includes chemotaxis of purified populations of human monocytes in response to IP-10 in vitro (Taub et al., 1993, 1996; Kohrgruber et al., 2004). We extend these observations by demonstrating a physiological role for CXCR3-mediated macrophage accumulation in adaptive vascular remodeling and show that IP-10 recruits monocytes to the arterial adventitia independent of lymphocytes in RAG1^−/− mice.

Circulating monocytes and tissue macrophages, including those present in blood vessels, are characterized as long-lived resident cells or short-lived inflammatory cells with distinct migratory patterns (Geissmann et al., 2003). Resident macrophages have been described in the aortic adventitia of normocholesterolemic mice (Galkina et al., 2006), and macrophage-derived foam cells in developing atherosclerotic lesions do not derive from resident adventitial macrophages (Lessner et al., 2002). However, Gr1^low resident monocytes that patrol the resting vasculature can rapidly invade the vessel wall in response to inflammatory signals (Auffray et al., 2007). We have not determined the origin of macrophages accumulating in remodeling arteries, a subject of future studies, although a 65-fold increase in cell number within 3 d of altered hemodynamic stresses (Tang et al., 2008) makes it unlikely that they are all derived from preexisting resident macrophages and suggests that the influx of cells mostly accrue from the circulation. Inflammatory macrophages, including those within atherosclerotic arteries, are described as Gr1^high (Geissmann et al., 2003; Tacke et al., 2007), although it is known that the expression of Ly-6C decreases as monocytes remain in the circulation (Sunderkötter et al., 2004) and that adoptively transferred monocytes down-regulate Ly-6C expression after entering atherosclerotic aortic plaques (Swirski et al., 2007). Thus, Gr1^low macrophages in the adventitia of remodeling arteries may be derived from circulating Gr1^high cells that subsequently decreased the expression of this marker or, alternatively, may be selectively derived from circulating Gr1^low cells. Our findings of smaller and larger perivascular macrophages are similar to previous descriptions of blood and spleen monocytes in which size and nuclei shape defined different cell differentiation states (Swirski et al., 2009).

In addition to its chemotactic effects, CXCR3 signaling is also known to play a role in the proliferation, effector cytokine production, and reactive oxygen species generation of T cells (Whiting et al., 2004; Schwarz et al., 2009). However, CXCR3 has not been described to contribute to macrophage activation. The differences in expression of macrophage mediators of inflammation and remodeling between WT and CXCR3^−/− mice in our study may be entirely the result of differences in cell numbers, although the differential gene expression between CXCR3⁺ and CXCR3⁻ perivascular macrophages in WT mice and the effects of IP-10 on isolated monocytes and macrophages argue against this interpretation. We show induction of FXIIIA expression by IP-10 treatment in addition to the up-regulation of this molecule by M-CSF–dependent differentiation of mouse bone marrow–derived macrophages. LPS and IL-4 inhibit these responses, and the cross-regulation of effector molecules by different macrophage activators is well established (Mosser and Edwards, 2008). IP-10 is strongly induced in macrophages by IFN-γ so that FXIIIA induction could arise from autocrine or paracrine signaling in M1 type activation. Alternatively, the characteristics of CXCR3-mediated macrophage activation may be considered to be partial M2 polarization even though there are striking differences with the phenotype induced by the prototypical alternative activator IL-4 in mouse macrophages, as it recapitulates IL-4–mediated induction of FXIIIA in human macrophages (Gratchev et al., 2005; Töröcsik et al., 2005; Martinez et al., 2008). The limited but unique pattern of macrophage activation by CXCR3 ligands, which is not seen with IL-4 or LPS, suggests a role for a macrophage population in the arterial wall that is neither M1 nor M2 and is responsible for adaptive vascular remodeling. M1- and M2-independent effects of another chemokine, PF4/CXCL4 (which shares binding sites with IP-10), have recently been described as M4 activation of blood-derived macrophages (Gleissner et al., 2010). Although the IP-10–dependent induction of FXIIIA in cultured macrophages was only several-fold, it was similar in magnitude to that in remodeling arteries. In a previous microarray study of human blood-derived macrophages, modest differences in gene expression were found in response to relatively high doses of MCP-1/CCL2 and GRO-α/CXCL1 without obvious polarization signatures (Cho et al., 2007). In vitro studies require comparatively high concentrations of IP-10 or other chemokines as a result of lower affinities for their cognate receptors than those of many proinflammatory cytokines. High local concentrations of IP-10 may be reached in vivo by binding to low-affinity nonspecific binding sites on surface proteoglycans, such as on endothelial cells (Ranjbaran et al., 2006). The activation of monocytes and macrophages by chemokines may also be dependent on their state of differentiation, and M-CSF has a significant transcriptional program (Martinez et al., 2006, Cho et al., 2007). The ability of CXCR3 ligands to both recruit and activate macrophages (this study) and to both recruit and activate lymphocytes (Whiting et al., 2004; Schwarz et al., 2009) qualifies these inflammatory mediators as alarmins or immune amplifiers (Oppenheim and Yang, 2005).

The production of diverse macrophage-derived factors are found in remodeling arteries, including products of M1 polarized macrophages, such as iNOS and TNF, and products of M2 polarized macrophages, such as MMP9 and FXIIIA (Galis et al., 2002; Bakker et al., 2006, 2008; Tang et al., 2008; Nuki et al., 2009). Indeed, TNF, IL-4, and TGF-β are all expressed in remodeling arteries after reduction of blood flow (Bakker et al., 2008) and may trigger classical activation, alternative activation, and deactivation of macrophages, respectively. Smooth muscle cell death, in addition to extracellular matrix degradation and cross-linking, occurs in inward vascular remodeling (Rudic et al., 2000; Tang et al., 2008) and, thus, the activity of both cytotoxic (M1) and tissue repair (M2) macrophages may be expected to play a role. The lack of nonredundant roles for TNF and iNOS in flow-mediated inward vascular remodeling is surprising given their essential functions in arterial remodeling in models of hypertension and graft arteriosclerosis, respectively, although these inflammatory mediators were described as T cell products in these studies (Koh et al., 2004; Guzik et al., 2007). Given its role in vascular smooth muscle cell proliferation (Selzman et al., 1999), one may speculate that the persistence of TNF production in CXCR3^−/− animals in this study and in MyD88^−/− animals or in monocyte-depleted animals in our previous study (Tang et al., 2008) may be responsible for the paradoxical medial thickening observed in the absence of inward vascular remodeling. We have not investigated if the artery-associated macrophages in our in vivo model are polarized into mutually exclusive subsets or whether single cells retain plasticity to secrete both M1 and M2 products. Recent work on heterogeneous macrophage subsets involved in tissue healing and remodeling after myocardial infarction showed both MMP activity and TNF production by all Ly-6C^high macrophages using single event histograms, suggesting dual production of M1 and M2 products in vivo (Nahrendorf et al., 2007). The paradigm of macrophage polarization requires further confirmation using multiple event histogram analysis of intracellular products by flow cytometry as commonly performed in studies of T cell differentiation.

An additional point emphasized by the present study is that macrophages have important homeostatic roles in addition to those of host defense. The essential function of macrophages in conduit artery remodeling in response to altered hemodynamic stresses is an example of the former (Bakker et al., 2008; Tang et al., 2008; Nuki et al., 2009). The inflammatory component of flow-mediated vascular remodeling is remarkably well orchestrated, such that adaptive changes in vessel size occur without apparent pathological sequelae. Different initiators of inflammation, namely tissue stress, tissue injury, and infection, elicit markedly different responses of homeostasis, tissue repair, and host defense, respectively. This suggests that different types of inflammation use distinct combinations of sensors, mediators, and effectors. It has been proposed that inflammation triggered by tissue stress be referred to as parainflammation to distinguish the process as an intermediate between the basal homeostatic state and classical inflammation triggered by pathogen-associated or damage-associated molecular patterns (Medzhitov, 2008). Because all forms of inflammation may have pathological consequences if unproductive and persistent, it is important to understand their underlying mechanisms. Indeed, it has been hypothesized that dysregulated parainflammation caused by tissue malfunction may underlie many chronic illnesses (Medzhitov, 2008). However, relatively little is known about inflammatory responses to tissue stress. We find that, unlike Gr1^high monocyte homing to the intima of atherosclerotic arteries known to be CXCR3 independent, Gr1^low macrophage accumulation and function in the adventitia of remodeling arteries is CXCR3 dependent.

Blood, adventitia (stripped from the thoracic aorta), and/or common carotid arteries (midvessel and distal sinus) were obtained from five organ donors and adventitia was also obtained from three patients undergoing thoracic aortic aneurysm repair under protocols approved by the Yale University Human Investigation Committee, the West Haven VA Hospital, and the New England Organ Bank. Demographic and clinical information of the eight individuals included the following: mean age 42.1 ± 21.6 yr (range 5–67 yr); five males and three females; six Caucasians, one African-American, and one Hispanic; and four normal aortas, one atherosclerotic aorta, and three ascending aortic aneurysms.

C57BL/6, IL-1R^−/−, IL-6^−/−, iNOS^−/−, TNFR1^−/−, TNFR2^−/−, and RAG1^−/− mice were purchased from The Jackson Laboratory. CXCR3^−/− mice were obtained from C. Gerard (Harvard University, Cambridge, MA) and have been previously described (Hancock et al., 2000). All strains had been back bred to a C57BL/6 background for >10 generations. Procedures were performed in 8–12-wk-old male mice.

Animal studies were approved by the Yale University Institutional Animal Care and Use Committee. After anesthesia, a midline cervical incision was performed, the left external carotid artery was isolated and ligated with 6–0 silk, and the wound was closed with 5–0 prolene. After 6 h–10 d, RNA, protein, and cellular studies of the common carotid arteries were performed after saline perfusion of the vasculature. After 2 wk, morphometry studies of the common carotid arteries were performed after fixation-perfusion with 4% paraformaldehyde in PBS at a mean pressure of 100 mmHg, and then arteries were embedded in OCT medium. Alternatively, the tissues around the common carotid arteries of RAG1^−/− mice were infiltrated with 30 µl nonpyrogenic saline or mouse IP-10 at 300 ng/µl (PeproTech) and the vessels were procured after 3 d for immunohistochemical analysis.

Hematoxylin and eosin–stained sections of the mid-common carotid arteries were analyzed by microscopy. Internal and external elastic lamina perimeters, as well as medial thickness at four quadrants, were measured and averaged over 10 separate sections using an image software program (ImageJ; National Institutes of Health).

Primary antibodies included rat anti–mouse CXCR3 (R&D Systems), rat anti–mouse F4/80 (Abcam), and rabbit anti–mouse IP-10 (AbD Serotec). Isotype-matched nonbinding immunoglobulin was used as negative control. Binding of secondary antibodies (Jackson ImmunoResearch Laboratories) was detected with Vectastain ABC reagent and AEC peroxidase substrate kits (Vector Laboratories). Sections were counterstained with hematoxylin (Sigma-Aldrich). Nuclei surrounded by positive immunostaining were enumerated at high magnification.

Primary antibodies for tissue sections included FITC-conjugated rat anti–mouse F4/80 (eBioscience), rabbit anti–mouse CXCR3 (MLB International), rabbit anti–mouse MMP2 (Millipore), goat anti–mouse MMP9 (R&D Systems), and goat anti–human FXIIIA (sc-18013, cross-reactive with mouse FXIIIA; Santa Cruz Biotechnology, Inc.). For analysis of human arteries, rabbit anti–human CD68 (Santa Cruz Biotechnology, Inc.) and mouse anti–human CXCR3 (BD) were used. Primary antibodies for cytospin preparations included rabbit anti–mouse MMP9 (Abnova) and goat anti–human FXIIIA. Isotype-matched nonbinding immunoglobulin was used as negative control. Detection of nonconjugated primary antibodies was visualized with secondary antibodies conjugated with FITC, TRITC (Jackson ImmunoResearch Laboratories), PE (Invitrogen), Alexa Fluor 594, or Alexa Fluor 488 (Invitrogen). Sections were counter-stained with DAPI (Invitrogen).

Blood, single cell suspensions from minced spleens, and bone marrow cells from bilateral femurs and tibias were collected from 4–20 mice, red blood cells were lysed with ACK buffer (Bio Whittaker), and mononuclear cells were enriched by Lymphocyte Separation Medium (MP Biomedicals). For phenotyping experiments and adoptive transfer experiments with circulating cells, CD11b⁺ monocytes were positively isolated with antibody-coated magnetic beads and MACS LS cell separation columns (Miltenyi Biotec). Alternatively, for adoptive transfer experiments with bone marrow cells, CD11b⁺ myeloid cells were negatively selected by incubation with magnetic beads coated with antibodies to CD90 (T cell marker), B220 (B cell marker), DX5 (natural killer cell marker), and TER-119 (red blood cell marker), passage through MACS LS cell separation columns, and collection of the filtrate. 10⁶ blood monocytes or 2.7 × 10⁶ bone marrow myeloid cells were injected i.v. 12 h before left external carotid artery ligation.

For experiments with in vitro differentiated macrophages, unselected bone marrow cells were cultured in DME containing 10% FBS, 10% L929 cell–conditioned medium (enriched for M-CSF), 2 mM L-glutamine, 100 units/ml penicillin, and 100 µg/ml streptomycin (Invitrogen). Adherent cells were harvested after 7–10 d with a rubber scraper, resuspended in M-CSF–enriched and complete DME medium and treated with IP-10, LPS, and/or IL-4. FXIIIA-depleted FBS was used for analyses of FXIIIA protein expression and transglutaminase activity.

For isolation of artery-associated macrophages, common carotid arteries were pooled from 18–22 mice and digested with 125 U/ml collagenase XI, 60 U/ml hyaluronidase I, 60 U/ml DNase 1, and 450 U/ml collagenase I (Sigma-Aldrich) in phosphate-buffered saline containing 20 mM Hepes at 37°C for 1–2 h. The tissue pieces and supernatant were resuspended in RPMI-1640 medium at 4°C and sequentially passed through 0.5- and 0.1-mm sieves. Macrophages were isolated with CD11b antibody-coated magnetic beads and cell separation columns, and the selected cells were >95% CD11b⁺. In certain experiments, the cells were further selected by flow cytometric sorting as indicated. For isolation of human macrophages, similar techniques were used as for mouse cells, except that enzymatic digestion of the aortic adventitia was for 2–4 h.

Total RNA was isolated and DNase treated using the NanoPrep system (Agilent Technologies). Bulk reverse transcription was performed using random hexamer primers according to the MultiScribe RT system protocol (Applied Biosystems). Real-time RT-PCR reactions were performed in duplicate using Taqman PCR reagents and Taqman gene expression probes for GAPDH, IP-10, Mig, TNF, iNOS, F4/80, CXCR3, MMP2, MMP9, FXIIIA, Arginase I, Ym1, Fizz1, IL-10, and TGM2 (Applied Biosystems). DNase-treated RNA samples produced without RT enzyme served as negative controls. An iCycler and its system interface software (Bio-Rad Laboratories) were used to analyze the samples and data. Transcript expression levels were normalized to GAPDH.

Artery-associated macrophages were either immediately collected or treated with PMA at 500 ng/ml and ionomycin at 50 ng/ml (Sigma-Aldrich) in the presence of brefeldin A (BD) for 5 h. The cells were pelleted onto slides at 800 rpm for 10 min, air dried for 30 min, and fixed with 4% paraformaldehyde for 20 min. Before immunostaining, the cells were permeabilized with 0.1% Triton X-100 for 30 min and then blocked with 2% BSA for 1 h. The labeled slides were examined by fluorescence microscopy and the signal was quantified within the outline of individual cells as the percentage of area using image analysis software (ImageJ). Cells were considered positive for FXIIIA or MMP9 expression if FITC or TRITC signal exceeded 5% of cell area, respectively, and were classified as null, predominantly FXIIIA, predominantly MMP9, or dual expressors.

Cell lysates were prepared in RIPA buffer and boiled in SDS sample buffer for 10 min. An equal volume of samples was loaded, and proteins were separated by SDS-PAGE, transferred electrophoretically to a nitrocellulose membrane (Bio-Rad Laboratories), and blotted with goat anti–human FXIIIA (Santa Cruz Biotechnology, Inc.), followed by horseradish peroxidase-conjugated secondary antibodies. Bound antibody was detected with Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific).

CD11b immunomagnetic bead-selected mouse cells were labeled with PerCP-CD11b, APC-CD42d, PE-Cy7-CD45.1, APC-CD45.2, FITC-F4/80 (eBioscience), APC-Gr1, APC-Ly-6C, V450-Ly-6G (BD), and PE-CXCR3 (R&D Systems) antibodies or isotype-matched irrelevant IgG. Similarly, CD11b-selected human cells were labeled with PE-Cy7-CD3, FITC-CD19, APC-CD56 (eBioscience), Pacific blue–CD14, FITC-CD16, PE-CXCR3 (BD), PE-CD15 (Beckman Coulter), APC-CD163 (R&D Systems), and APC-CD177 (Abcam) antibodies or isotype-matched irrelevant IgG. The labeled cells were analyzed using an LSR II flow cytometer (BD) and data were analyzed using FlowJo 6.3 (Tree Star, Inc.). CD11b-selected cells were also labeled with FITC-F4/80 and PE-CXCR3 antibodies and sorted using a FACSAria flow cytometer (BD) into F4/80⁺/CXCR3⁺ and F4/80⁺/CXCR3⁻ cell populations to analyze further by quantitative PCR. Alternatively, CD11b-selected cells were labeled with FITC-CD3, FITC-B220, and FITC-CD11c antibodies and the population negative for these lymphoid and dendritic cell markers was collected for cytospin preparations and immunofluorescence imaging.

Protein extracts from common carotid arteries were purified with gelatin sepharose beads (GE Healthcare) and subjected to 8% SDS-PAGE containing 1 mg/ml gelatin. Standards included purified MMP2 and MMP-9 (Millipore). After electrophoresis, the gels were washed and incubated overnight at 37°C in 100 mM NaCl, 2.5% Triton X-100, and 10 mM CaCl₂ and subsequently stained with Coomassie blue (Bio-Rad Laboratories). Zones of proteolysis appeared as clear bands against a blue background. For in situ gelatin zymography, OCT-embedded common carotid artery sections were incubated in DQ-gelatin solution (Invitrogen) for 2 h at 37°C, washed in PBS, and DAPI mounting mixture (Invitrogen) was applied to stain nuclei. For transglutaminase activity in intact arteries, tissue sections were incubated with biotinylated cadaverine (Invitrogen) at 5 mM for 90 min, washed, incubated with a 1:50 dilution of stock fluorescein (DTAF)-conjugated streptavidin (Jackson ImmunoResearch Laboratories) for 30 min in the dark, washed, and then visualized after application of DAPI mounting mixture. For transglutaminase activity in tissue and cell lysates, the samples were preincubated with α-thrombin (Haematologic Technologies) at 5 U/ml for 30 min, and enzyme activity was then measured based on the catalysis of covalent bonds between plate-bound poly-l-lysine residues and γ-carboxamide groups of soluble biotin-TVQQEL-OH substrate, according to the manufacturer’s instructions (Sigma-Aldrich).

Paired Student’s t test was used to compare two groups within the same animals (i.e., left vs. right arteries). An unpaired Student’s t test was used to compare two groups between different animals (i.e., WT vs. CXCR3^−/−). One-way analysis of variance was used to compare between multiple groups. Statistical analyzes were performed using Prism 4.0 (GraphPad Software, Inc.). Differences with p-values <0.05 were considered to indicate statistical significance.

Fig. S1 shows successful adoptive transfer of WT bone marrow myeloid cells to WT and CXCR3^−/− hosts. Fig. S2 shows that negatively selected bone marrow myeloid cells for adoptive transfer experiments are highly enriched for monocytes. Fig. S3 shows the expression of monocyte, but not granulocyte markers, by the selected population of circulating cells. Fig. S4 shows the overlapping expression patterns of small and large artery-infiltrating macrophages. Fig. S5 shows that IP-10 induces FXIIIA expression in isolated blood monocytes. Fig. S6 shows that human perivascular macrophages do not express lymphocyte or granulocyte markers. Table S1 summarizes the production of macrophage mediators in response to different proinflammatory factors.

This work was supported by the National Institutes of Health (PO1 HL70295). P.C.Y. Tang was supported by a Vascular Research Postdoctoral Training Grant from the NIH (T32 HL07950).

The authors have no conflicting financial interests.