Chemokine-coupled β₂ integrin–induced macrophage Rac2–Myosin IIA interaction regulates VEGF-A mRNA stability and arteriogenesis

Nuclear-to-cytosolic translocation of HuR is a marker of, and required for, its mRNA stabilizing effect. In our previous studies, either ligand-mimetic anti–LFA-1 antibodies or adhesion to ICAM-1 in the presence of supraphysiologic divalent cation concentrations were used to promote integrin activation and signaling. Although this does lead to integrin affinity modulation and productive ICAM-1 adhesion, stimulation with the monocyte chemokine CCL2 represents a physiological mechanism that during inflammation can promote and couple with β₂ integrin engagement to drive signaling and HuR-dependent transcript stabilization. Freshly isolated human peripheral blood monocytes adhered to and spread on recombinant ICAM-1. However, rapid and robust HuR translocation occurred only when a CCL2 stimulus was coupled with β₂ integrin engagement on ICAM-1 (Fig. 1, A and B). Peripheral monocytes are heterogeneous and the maximal stimulus (CCL2 plus ICAM-1) resulted in robust HuR translocation in <50% of cells, suggesting that only a subset of monocytes may be capable of responding. Thus, to address whether inflammatory (CD14^hi) monocytes (recruited to sites of chemokine release and endothelial activation) are more likely to activate this posttranscriptional expression program upon ICAM-1 adhesion than “atypical” or “patrolling” (CD16^hi) monocytes, freshly isolated cells were sorted based on these two markers. Although the overall responsiveness was less after the duration of sorting, the HuR translocation response was almost exclusively found within the CD14^hiCD16^lo inflammatory monocyte population (Fig. 1, C and D). In part, this may be a consequence of greater CCR2 expression levels on the inflammatory monocytes (Fig. 1 E and Fig. S1; Geissmann et al., 2003; Swirski et al., 2009). This is the first report that chemokine signaling can couple with integrin engagement to activate HuR in primary human monocytes and supports the likely scenario that, in response to chemokine release and endothelial activation, inflammatory monocytes are recruited, adhere, and transmigrate through integrin engagement, after which there is rapid, enhanced production of multiple factors.

Although β₂ integrin engagement does trigger Rac2 activation (GTP-loading), as seen in a p21-activated kinase (PAK) binding domain (PBD) affinity pulldown assay (Fig. 2 A, lane 4), it is apparent that this stimulus alone is insufficient to promote HuR translocation. BM-derived macrophages (BMDMs) isolated from Rac2 gene-deleted mice maintain WT levels of CCR2 (Fig. 2 B). There is also no significant difference in thioglycollate-elicited peritoneal macrophage (CD11b^hiF4/80^hi) surface expression of CCR2 between Rac2^+/+ and Rac2^−/− mice (Fig. 2 C). However, BMDMs from Rac2^−/− mice were not able to translocate HuR in the setting of CCL2-coupled, ICAM-1 adhesion despite adequate adhesion and cell spreading (Fig. 2, D and E). Consequential to the effects on HuR translocation, the threefold increase in VEGF-A mRNA half-life induced by the combination of CCL2 stimulation and β₂ integrin engagement observed in BMDMs from Rac2^+/+ mice was lost in BMDMs from Rac2^−/− mice (Fig. 2 F). The HuR translocation and VEGF-A mRNA decay analyses support that integrin signal transduction leading to HuR activation depends on Rac2 as an intermediary. Thus, in efforts to define novel Rac2 interactors that trigger this posttranscriptional switch, the screening activation stimulus included both Rac2 activation (through either integrin engagement or, as a surrogate, Rac2 loading with nonhydrolyzable GTPγS) and CCR2 engagement. In the human monocytic cell line THP-1, upon GTPγS loading, Rac2, Rac1, and Cdc42 were detected in the PBD pulldown material (Fig. 3 A). With the additional CCL2 stimulus, a 226 kD protein was detected in the PBD pulldown material by Coomassie staining (Fig. 3 A). This stimulus-driven protein association was Rac2-specific, as it was undetectable in pulldown material from CCL2-activated, GTPγS-loaded Rac2^−/− BMDM lysates (Fig. 3, B and C). By liquid chromatography mass spectrometry analysis of in-gel trypsin-digested material, this 226-kD Rac2-associated protein was identified as Myh9, the heavy chain of NMMIIA. 61.5% of the Myh9 protein sequence was mapped by peptides identified from this tryptic digestion (Fig. S2). This proteomic identification was validated in Myh9 immunoblots of PBD pulldown material, where Myh9 was detected in lysates from GTPγS-loaded CCL2-treated Rac2^+/+, but not from Rac2^−/− BMDMs or from GTPγS-loaded non–CCL2-treated Rac2^+/+ BMDMs (Fig. 3 C). This again demonstrates Rac2 specificity, as Rac1-GTPγS and Cdc42-GTPγS were both present in pulldowns of Rac2^−/− BMDMs but without associated Myh9. Additional validation was obtained by the presence of Myh9 in PBD pulldown material obtained from WT BMDMs that were CCL2-stimulated and adhered to ICAM-1, but not from cells singly CCL2-stimulated or ICAM-1–bound, despite the latter leading to Rac2 activation (Fig. 2 A).

The proximity ligation assay (PLA) was used to identify the native, intracellular Rac2–Myh9 interaction, and to quantify whether there is a relationship between this interaction and the noted downstream posttranscriptional switch. Fig. 3 D displays an augmented PLA signal in CCL2-stimulated ICAM-1–bound human peripheral blood monocytes above either stimulus alone. When performed in THP-1 cells stably expressing basal, nuclear HuR-YFP, we were able to quantify and correlate the PLA signal with induced HuR translocation. Fig. 3 E displays a strong, direct correlation (r = 0.90, P < 0.0001) between the induced PLA signal and HuR translocation, with the strongest PLA signal and HuR translocation in the combined CCL2-stimulated ICAM-1–bound cells. Although PLA does not definitively determine direct protein–protein interactions, Rac2 and Myh9 must be within 40 nm to generate a PLA signal, consistent with either a direct interaction or a close association within a larger adhesion-based signaling complex (Fredriksson et al., 2002, 2007; Söderberg et al., 2006). The correlation between PLA signal and degree of HuR translocation supports that the Rac2 and Myh9 interaction is required for the signal transduction events leading to the activation of the HuR-dependent posttranscriptional switch.

Rac2 gene deletion leads to decreased flow recovery after femoral artery ligation in a manner independent of collateralization potential and collateral maturation (De et al., 2009; Distasi et al., 2009). Rac2 expression is gene dose dependent, yet Rac2^+/− heterozygote animals have not been demonstrated to have an in vivo phenotype to date (Roberts et al., 1999; Li et al., 2002). Rac2^+/− neutrophils are known to have minimal defects in neutrophil superoxide production; however, heterozygote neutrophils have moderate reductions of FMLP-stimulated F-actin formation, chemotaxis, and NADPH oxidase activity. We hypothesized that with a 50% reduction in Rac2 (Fig. 2 B), macrophages would demonstrate a decreased β₂ integrin–induced Rac2 activation and consequent Myh9 interaction, as the reduced availability of Rac2 would fall below threshold levels required to promote such signaling. Indeed, BMDMs from Rac2^+/− mice demonstrated a deficiency in CCL2-coupled, ICAM adhesion-driven Rac2 activation and consequent Myh9 association (Fig. 3 F). If β₂ integrin–induced Rac2 activation were required for macrophage VEGF-A mRNA stabilization and subsequent vasculogenic responses, we would expect a defect in the femoral artery ligation model of hind limb ischemia. Laser Doppler-imaged flow recovery to hind limb ischemia was reduced at day 3 after ligation and reached a 50% reduction by 3 wk (Fig. 3, G and H). This was not a consequence of impaired monocyte/macrophage recruitment and localization, as numbers of CD11b^hiF4/80^hi cells recovered from day 3 collagenase-treated ischemic muscle tissue were not reduced (Fig. 3 I). This is consistent with intact β₂ integrin engagement (required for localization) but impaired integrin-mediated Rac2 activation and Myh9 association (required for maximal VEGF-A production) as seen in vitro, allowing us to segregate the membrane receptor-dependent adhesion from this important integrin-induced signaling cascade in vivo.

Myh9, in its rapidly induced association with Rac2, appears to be a key component of a chemokine-coupled integrin-induced pathway, resulting in labile mRNA stabilization in macrophages. To address this requirement for Myh9 in vitro and in vivo, myeloid-specific Myh9 gene-deleted mice were generated, using Myh9 gene-targeted and Lysozyme M–Cre (LysM^+/Cre) mice. NMMIIA heavy chain levels were 20 and 80% of WT (Myh9^+/+) levels in homozygous (LysM^+/CreMyh9^fl/fl, herein Myh9^−/−) and heterozygous (LysM^+/CreMyh9^+/fl, herein Myh9^+/−) gene-deleted mice, respectively (Fig. 4 A). Neither Rac2 nor Myh9 gene deletion affects each other’s or CCR2’s expression in BMDMs (Figs. 2 B and 4 A). Myeloid Myh9^−/− animals were of similar size and body weight, and appeared to develop normally, compared with their heterozygous littermates. Myh9^−/− mice did not have a deficiency in peripheral, circulating inflammatory monocyte levels, as determined by CD11b^hiLy6C^hi flow cytometric analysis (Fig. 4 B; Geissmann et al., 2003; Swirski et al., 2009). Most importantly for our in vivo studies, similar numbers of macrophages were recruited to an inflammatory thioglycollate-induced peritonitis (Fig. 4 C). This is consistent with a recent report that Myh9-deficient cells can migrate efficiently despite diminished traction forces at the cell periphery (Jorrisch et al., 2013). Thus, a myeloid-specific Myh9^−/− mouse phenotype has the potential to reflect a deficient chemokine-coupled integrin-induced signaling cascade rather than just impaired myeloid mobility/migration.

As observed with Rac2^−/− BMDMs (Fig. 2, C and D), CCL2-coupled β₂ integrin engagement failed to induce HuR translocation in BMDMs isolated from myeloid-specific Myh9^−/− mice, despite adequate cell spreading (Fig. 4, D and E). This is further reflected in VEGF-A mRNA decay assays, in which the threefold increase in VEGF-A mRNA half-life induced by the combination of CCL2 stimulation and adhesion to ICAM-1in Myh9^+/− BMDMs is lost in Myh9^−/− BMDMs (Fig. 4 F). The in vitro adhesion data are consistent with a chemokine–integrin signal transduction axis that depends on Myh9, resulting in HuR nuclear-to-cytosolic translocation and labile VEGF mRNA stabilization.

If Myh9 were required for inflammation-based (chemokine plus integrin signaling) VEGF-A mRNA stabilization, we would expect a defect in macrophage-dependent vascular responses that require leukocyte recruitment and tissue localization, and so we returned to the femoral artery ligation model. 3 d after femoral artery ligation, there is a significant reduction in laser Doppler-imaged flow recovery in the myeloid-specific Myh9^−/− mice, compared with Myh9^+/− littermate controls, and this early difference persists as a 50% decrease in perfusion recovery out to 21 d (Fig. 4, G and H). There was no difference in absolute perfusion at baseline nor in the non-ischemic limbs throughout the analysis, confirming that the relative difference is driven by a macrophage-dependent absolute flow recovery deficiency in the ischemic limbs. At day 3 after ligation, Myh9^−/− ischemic thigh muscle tissue VEGF-A mRNA levels were 40% that of control, with a consequent 80% reduction in VEGF-A protein (Fig. 4 I; Fig. 5, A and B). Of note, CCL2 levels were comparable in ischemic tissue homogenates from both Myh9^−/− and Myh9^+/− mice (Fig. 5, A and C). This difference is not a consequence of reduced macrophage recruitment and localization to the ischemic limb, as quantity of F4/80⁺ cells by immunohistochemistry and relative CD68 mRNA levels in ischemic tissue homogenates are not decreased in Myh9^−/− mice (Fig. 5, D–F). Moreover, CD11b^hiF4/80^hi cells isolated from collagenase-treated ischemic muscle tissue at day 3 reflected no difference in early recruitment of macrophages between Myh9^−/− and Myh9^+/− mice. However, there was a 70% reduction in Myh9^−/− macrophage VEGF-A mRNA expression, demonstrating the global tissue reduction in VEGF-A protein to be a consequence of the macrophage VEGF-A expression defect (Fig. 5, G and H).

On day 14 after ligation, we performed micro-computed tomography (micro-CT) arteriography on a subset of mice to quantify arteriogenesis as a consequence of early macrophage VEGF-A expression in the ischemic tissue. Fig. 5 I displays micro-CT angiograms at the thigh (adductor) and calf (soleus, gastrocnemius) levels. Quantitative segmentation analysis demonstrates that there was a trend of decreased arteries overall with significantly fewer small arteries (<48 µm) at both the thigh and calf levels in Myh9^−/− mice (Fig. 5 J). This defines the macrophage, and its expression of Myh9, as a critical cell responder and molecular effector, respectively, in ischemia-induced arteriogenesis. These results confirm that Myosin IIA, a molecule involved in macrophage integrin-triggered signal transduction leading to VEGF-A mRNA stabilization and enhanced protein production, is critical to tissue ischemia–driven arteriogenesis.

Because the impairment in flow recovery was strongly evident at an early time point (day 3) and there was a trend toward decreased arteries overall in the ischemic limb, we addressed whether macrophage Myh9 deletion confers differences in developmental vasculogenesis by studying baseline angiogenesis, arteriogenesis, and arteriolar collateral formation. Quantitative measures of postnatal retinal angiogenesis, renal arteriogenesis, or pial arterial collateral networks have been established (Jones et al., 2008; del Toro et al., 2010; Pitulescu et al., 2010; Tirziu et al., 2012). In the developmental model of retinal capillary angiogenesis, analyzed at postnatal day 4 (P4), there was no difference in relative vessel area, vessel branch points, and vessel length between Myh9^−/− and Myh9^+/− pups (Fig. 6, A–D). Micro-CT arteriograms at P7 failed to display baseline differences in development of the kidney arterial system (Fig. 6, E and F). Pial collateral artery quantification using smooth muscle actin–based imaging at P4 demonstrated an equal number of collateral arteries per cerebral hemisphere in Myh9^−/− and Myh9^+/− pups (Fig. 6, G and H). Thus, there were no detectable deficiencies in developmental capillary angiogenesis, arteriogenesis, and arterial collateral network formation in myeloid Myh9-deleted mice. Although this strengthens our inflammation-based, ischemia-induced arteriogenesis data, we were surprised by these results, as resident-tissue macrophages have been shown to play a role in developmental angiogenesis (Fantin et al., 2010). However, early resident macrophages are derived from the yolk sac, predating liver hematopoiesis at embryonic day 10.5 (E10.5), when monocyte-derived macrophages first appear and Lysozyme M expression begins (Clausen et al., 1999; Lichanska et al., 1999; Fantin et al., 2010). Also, macrophages are involved in capillary development, downstream of VEGF-A–mediated tip cell induction, facilitating tip cell fusion to capillary networks in a VEGF-A–independent manner, leading to increased vessel intersection and capillary network complexity (Fantin et al., 2010). It is likely that both the early timing of the tissue macrophage role (relative to Lysozyme M promoter activity for Cre expression) and VEGF-A–independent effects lead to the absence of an apparent developmental phenotype in these Myh9^−/− models.

In summary, our results add a new mechanistic dimension to the cellular contributors and soluble factors involved in arteriogenesis, defining an inflammatory macrophage Myh9-dependent signaling cascade that drives VEGF-A production in tissue with consequent arteriogenesis. In the setting of ischemia (or other perturbations of tissue homeostasis, such as disturbed shear stress or wounding), chemokine release (CCL2) and endothelial activation (ICAM-1 augmentation) result in recruitment of inflammatory monocytes, which undergo phenotypic modulation of secreted factors through a Rac2–Myh9 interaction-dependent posttranscriptional switch. This switch is certainly required for stabilization of VEGF-A mRNA with consequent enhanced translation and secretion of VEGF-A, a critical mediator of arteriogenesis. It is highly likely that transcripts encoding other molecular mediators of arteriogenesis are also affected. That is, VEGF-A is absolutely required, but we believe this to be a powerful switch, affecting a larger gene expression program. Although our data support a role for this mechanism in adult ischemic (inflammatory) arteriogenic responses but not in developmental vasculogenesis, it remains possible that earlier embryonic deletion of macrophage Myh9 may result in a developmental phenotype. This integrin–Rac2–Myh9 axis highlights the interdependence of signal transduction and cytoskeletal network dynamics, as the major phenotype was reflected by the loss of a signaling cascade that modulates gene expression rather than the loss of macrophage adhesion and cell spreading or migration and recruitment to sites of inflammation. We believe this molecular pathway has the potential to be therapeutically targeted, either for the enhancement (during ischemia) or inhibition (during tumor metastasis) of macrophage-induced neovascular responses.

Anti-Rac2 mouse monoclonal antibody was purchased from Proteintech. Anti–Myosin IIA was purchased from Abcam (mouse monoclonal) and Cell Signaling Technology (rabbit polyclonal). Anti-HuR (clone 3A2, mouse monoclonal) and Anti-Actin (rabbit and goat polyclonals) were purchased from Santa Cruz Biotechnology, Inc. Goat anti–human IgG, Fcγ fragment-specific, was purchased from Jackson ImmunoResearch Laboratories, Inc. Recombinant ICAM-1 chimera with Fcγ fragment of human IgG₁, Fcγ of human IgG₁ fragment alone, and human and murine CCL2 were purchased from R&D Systems. Where appropriate, murine ICAM-1 and murine CCL2 were used with mouse cells, and human ICAM-1 and human CCL2 were used with human cells. PAK1 PBD agarose beads were purchased from Cell Biolabs. Alexa Fluor–conjugated secondary antibodies and phalloidin, DAPI (4,6 diamidino-2-phenylindole), and RPMI 1640 culture medium were purchased from Invitrogen. Actinomycin d, poly-l-lysine, Ficoll-Histopaque, GTPγS, and endotoxin-free BSA were purchased from Sigma-Aldrich. Duolink II Proximity Ligation kit was purchased from Olink. THP-1 cell line was purchased from American Type Culture Collection. Sequences of primers used in quantitative real-time PCR were all synthesized by the Keck Biotechnology Resource Facility at Yale University School of Medicine as follows: VEGF, 5′-ACTGGACCCTGGCTTTACTGCT-3′ (sense) and 5′-TGATCCGCATGATCTGCATGGTG-3′ (antisense); CD68, 5′-CTTCCCACAGGCAGCACAG-3′ (sense) and 5′-AATGATGAGAGGCAGCAAGAGG-3′ (antisense); and GAPDH, 5′-AAGGCCGGGGCCCACTTGAA-3′ (sense) and 5′-GGACTGTGGTCATGAGCCCTTCCA-3′ (antisense).

C57BL/6 mice were obtained from The Jackson Laboratory. Rac2^−/− mice were a gift from D. Williams (Harvard Medical School, Boston, MA; Roberts et al., 1999). LysM^+/cre were obtained as described and backcrossed onto C57BL/6 for 11 generations by our laboratory (Clausen et al., 1999; Zhang et al., 2012). Myh9^fl/fl mice were a gift from R.S. Adelstein (National Institutes of Health, Bethesda, MD; Jacobelli et al., 2010). For BMDMs, animals 6–8 wk of age were used. For hind limb ischemia or thioglycollate-elicited macrophage experiments, animals 12–14 wk of age were used. Animals were excluded from the study if BMDMs demonstrated a <50% reduction in Myh9 levels normalized to Actin, as measured by Western blotting. Animal studies did not involve randomization; however, the investigators were blinded to the genotypes of individual animals in each litter until after the study outcomes were assessed. The Yale University School of Medicine Institutional Animal Care and Use Committee approved all housing protocols and experimental animal procedures.

Human peripheral blood monocytes were isolated by density centrifugation and, where indicated, fluorescence-activated cell sorting (Geissmann et al., 2003; Repnik et al., 2003; Swirski et al., 2009). Monocytes were isolated from anonymous healthy donors. Informed consent was obtained before drawing blood from all donors. The Yale University School of Medicine Human Investigation Committee approved all protocols and experiments involving human leukocytes.

BMDMs were obtained by in vitro differentiation of primary femur and tibia BM cells using an established protocol (Davies and Gordon, 2005). In brief, femurs and tibiae from Rac2^+/+, Rac2^−/−, myeloid-specific Myh9^+/−, or myeloid-specific Myh9^−/− mice were dissected, cleaned, disinfected in 70% ethanol, and washed with fully supplemented RPMI 1640 medium. Red blood cells were removed by ammonium-chloride-potassium lysis buffer and subsequent centrifugation. Remaining BM cells were then cultured at a density of 4 × 10⁶ cells/10 cm Petri dish in fully supplemented RPMI with 30% (vol/vol) L929 cell-conditioned medium. Cells were matured to phenotypic macrophages over 6–8 d. Nonadherent cells were washed away with PBS. Adherent cells were recovered with gentle pipetting in PBS with 1 mM EDTA. Flow cytometric analysis confirmed the recovered cells to be >97% CD11b^hiF4/80^hi.

Petri dishes or glass coverslips were coated with 10 µg/ml goat anti–human IgG, Fcγ fragment-specific, in 50 mM Tris-HCl, pH 9.5, for 1 h at room temperature. Blocking was performed for 1 h at room temperature with PBS containing 2% (wt/vol) BSA. Dishes or coverslips were then incubated with 100 ng/ml human or mouse rICAM-1-Fcγ fragment chimeric protein or, as a negative control, 100 ng/ml Fcγ alone or 20 mg/ml poly-l-lysine, overnight at 4°C. Where appropriate, cells were resuspended in fully supplemented RPMI 1640 with or without 50 ng/ml human or mouse recombinant CCL2 at the time of adhesion.

Plated cells were incubated for 30 min at 37°C, washed with PBS, and then fixed with 4% PFA for 15 min at room temperature. Cells were then washed with PBS and permeabilized with 0.1% Triton X-100 in PBS, followed by blocking with 5% goat serum for 60 min at room temperature. Immunostaining with 1/250 dilution of anti-HuR antibody was performed overnight at 4°C. Cells were incubated with Alexa Fluor–conjugated secondary antibodies for 2 h in the dark at room temperature. Counterstaining was performed with phalloidin and 0.005% DAPI as per manufacturer protocols. Confocal microscopy was performed on a five-laser inverted microscope (Ti-E Eclipse; Nikon) equipped with a Perfect Focus, motorized XY stage on a Nano Focusing Piezo Stage. Oil immersion objective lenses (60 and 100×; CFI Plan Apochromat VC Series; Nikon) with a numerical aperture of 1.4 were used. Samples were imaged at 21°C. Flourochromes included Alexa Fluor 488, 568, and 647 (Invitrogen). Cells were mounted in Fluorogel with Tris Buffer (Electron Microscopy Sciences). Images were taken with the UltraVIEW VoX confocal imaging system (Perkin Elmer) using Volocity 6.2.1 (Perkin Elmer) acquisition software. We have defined an algorithm (Matlab-based) to quantify the relative translocation of HuR in each cell. Background subtracted cell images were identified and segmented using the watershed algorithm and then tracked across the z-plane by using the centroid of the nucleus. From this point, the relative translocation of HuR was calculated by multiplying the relative external (to nucleus) intensity with the relative external (to nucleus) volume. Threshold for the quantification of cells with a robust translocation in a 30-min period are based on a minimum intensity-volume product of 0.25.

To demonstrate a complex between native Rac2 and Myh9 in adhered primary human monocytes, a PLA kit was used as per manufacturer protocols (Fredriksson et al., 2002, 2007; Söderberg et al., 2006). Primary antibodies to Rac2 and Myosin IIA were used, and secondary antibodies were conjugated to oligonucleotides for ligation and subsequent rolling circle amplification. To correlate Rac2–Myh9 complex formation with HuR translocation, HuR was cloned into a modified pEGFP-C3 vector. The monocytic cell line, THP-1, was transfected with the construct and stable clones expressing YFP-HuR fusion protein were selected by a combination of antibiotic and fluorescence-activated cell sorting. Clones expressing YFP-HuR were used to track HuR translocation without use of an antibody. We have defined an algorithm (Matlab-based) to quantify both the proximity ligation signal and relative HuR translocation simultaneously in the same cell for correlative analysis.

After cells were adhered for 30 min, transcriptional arrest was imparted by 10 µg/ml actinomycin d (time 0 in assay). RNA was harvested at 20-min intervals for 60 min mRNA expression was analyzed by RT-qPCR as previously described (Wang et al., 2006). VEGF-A mRNA levels were normalized to GAPDH mRNA.

Rac1, Rac2, and CDC42 activity was measured by an affinity precipitation assay with Pak1 PBD agarose beads (del Pozo et al., 2000; Knaus et al., 2007). Cells were stimulated with CCL2 and adhered as described above, after which they were lysed in 500 µl lysis buffer (20 mM Hepes, pH 7.4, 150 mM NaCl, 1% Triton X-100, 4 mM EDTA, 4 mM EGTA, and 10% glycerol, supplemented with protease inhibitor cocktail [Roche]). 50 µl was set aside for whole lysate Rac analysis. To the remainder of the sample, 10 µg Pak1 PBD beads were added, followed by a 45-min incubation at 4°C. Where GTPγS-loading was required, cell lysates were incubated with 0.1 mM GTPγS for 10 min at 37°C before addition of PBD beads. Beads were washed with lysis buffer and subsequently brought up in Laemmli sample buffer. Activated (GTP- or GTPγS-bound) Rac1, Rac2, or CDC42 was detected by Western blot analysis with whole cell lysates as loading controls.

Human peripheral blood monocytes were enriched by density gradient centrifugation in Histopaque (Sigma-Aldrich), followed by Percoll (GE Healthcare) and stained with the following mouse monoclonal antibodies to gate out contaminating NK cells and to separate monocyte subpopulations: PE-conjugated anti-CD56, allophycocyanin (APC)-conjugated anti-CD14, Alexa Fluor 700–conjugated anti-CD16, and PerCP-Cy5.5–conjugated anti-CCR2. Mouse peripheral blood mononuclear cells were obtained from whole blood by density gradient centrifugation and stained with the following rat monoclonal antibodies to identify monocytes: PE-conjugated anti-CD11b and FITC-conjugated anti-Ly6C. Mouse thioglycollate-elicited peritoneal macrophages were stained with rat monoclonal PE-conjugated anti-CD11b and APC-conjugated anti-F4/80 antibodies. Staining of mouse cells was always performed in the presence of FcBlock to eliminate nonspecific binding by Fc receptors. The antibodies were purchased from BD, eBioscience, and BioLegend. LSR II flow cytometer and FACS Diva software (BD) were used to acquire fluorescence signal for live single cells based on gating for forward and side scatter; at least 2 × 10⁵ cells were acquired for each sample. Subsequent analyses of cell populations were performed using FlowJo software (v. 7.6.4; Tree Star). Sorting of monocytes was performed using iCyt Reflection high-speed sorter (Sony) at Yale University School of Medicine Cell Sorter Facility.

Hind limb ischemia surgery was performed by the Microsurgery Core Facility at the Yale Cardiovascular Research Center. In brief, mice were anesthetized by intraperitoneal injection of ketamine/xylazine (100/10 mg/kg) solution. The femoral artery was ligated at two positions spaced 5 mm apart, one just below the inguinal ligament and the second distal to superficial epigastric artery. All branches between the two ligatures were ligated and the femoral artery segment was excised. Flow images of the foot were acquired using a Moor Laser Doppler Imager (Moor Instruments) at 37 ± 0.50°C under ketamine/xylazine anesthesia. The data were analyzed with moorLDI image processing software (V5.3; Moor instruments) and reported as the ratio of flow in the ischemic/non-ischemic contralateral control hind limbs (Tirziu et al., 2005, 2012; Lanahan et al., 2010). Tissue perfusion was assessed before (day −1), after (day 0), and 3, 7, 14, and 21 d after surgery. Early inflammatory assessment was performed on day 3 after ligation, where thigh and calf muscles were dissected, harvested, and segmented for subsequent analysis. Portions of thigh and calf were homogenized for subsequent TRIzol-based fractionation and isolation of DNA, RNA, and protein (Chomczynski, 1993). RNA was treated with a DNase kit (Life Technologies) to remove genomic contaminant and then mRNA was reverse transcribed and analyzed by RT-qPCR for relative VEGF-A and CD68 expression, using GAPDH for normalization. VEGF-A protein was quantified by densitometry of immunoblot analysis with β-actin normalization. In some animals, ischemic muscle tissue was dissected on day 3 after femoral artery ligation and then treated with collagenase digestion to generate a single cell suspension (Swirski et al., 2009). After density gradient centrifugation with Percoll (GE Healthcare), cells were stained for CD11b and F4/80 and fixed in 1% paraformaldehyde for 5 min. Macrophages (CD11b^hi F4/80^hi cells) were isolated using FACS. Cross-linking was reversed and macrophage mRNA purified for subsequent RT-qPCR using the miRNeasy FFPE kit (QIAGEN). Other segmented portions of thigh and calf were fixed, dehydrated, embedded, and sectioned for histology. Quantitative high-resolution micro-CT was performed on ischemic limbs at day 14 after surgery.

Quantitative micro-CT was performed after bismuth-gelatin perfusion of small arteries by the Yale Translational Research Imaging Center. Ischemic hind limbs or P7 kidneys were staged in a high-resolution micro-CT scanner (GE eXplore Locus SP; GE Healthcare) with a cone beam filtered back projection algorithm, set to an 8-µm effective detector pixel size. Micro-CT was operated with 80-kVp x-ray tube voltage, 80-mA tube current, 3,000-millisecond per frame, 1 × 1 detector binning model, 360° angle, and 0.5° increments per view.

Retina labeling was done as previously described with some modification (del Toro et al., 2010; Pitulescu et al., 2010). In brief, the eyeballs were removed from neonates at P4 and prefixed win 4% paraformaldehyde for 15 min at room temperature. The dissected retinas were blocked overnight at 4°C in TNB (0.1 M Tris-HCl, pH 7.4, 150 mM NaCl, and 0.2% Blocking reagent [PerkinElmer]) supplemented with 0.5% Triton X-100. After blocking, the retinas were incubated with IsolectinB4, Alexa Fluor 488 Conjugate (Molecular Probes) in Pblec (1 mM MgCl₂, 1 mM CaCl₂, 0.1 mM MnCl₂, and 1% Triton X-100 in PBS) for 2 h at room temperature, washed 6× in PBS, fixed briefly in 4% PFA, and mounted in fluorescent mounting medium (DAKO). 20–25 images per group were acquired on the above-described 80i Ti-E Eclipse inverted microscope, and Biological CMM Analyzer software was used to quantify vessel branch points, vessel length, and number of segments per 10× magnification fields (Jones et al., 2008).

Whole brains were harvested and stained as described previously (Tirziu et al., 2012). In brief, at P4, mouse brains were extracted, fixed in 1:4 dimethyl sulfoxide/methanol overnight at 4°C, washed in methanol, and stored in methanol at −20°C. The samples were then rehydrated, washed in PBS, and permeabilized overnight at 4°C in TNBT (0.1 M Tris-HCl, pH 7.4, 150 mM NaCl, 0.2% Blocking reagent [PerkinElmer], and 0.5% Triton X-100). After incubation with smooth muscle α-actin antibody (1A4-Cy3, 1:200) overnight at 4°C, the brains were washed in PBS and imaged on a microscope (M205 FA; Leica) with Leica Application Suite (LAS) software. Pial arteriolar–arteriolar collateral anastomoses were quantified as mean number of collateral vessels between the anterior cerebral artery (ACA) and the middle cerebral artery (MCA) territories, per cerebral hemisphere.

Based on previous studies (Tirziu et al., 2005, 2012; Lanahan et al., 2010; Zhang et al., 2012), the pooled standard deviation of hind limb ischemia recovery over 21 d is assumed to be 0.13. Alpha is specified as 0.05, using a two-tailed test of statistical significance. The study had a 95% power to detect a between group difference of 25% with a sample size of 7. All data that demonstrated a normal distribution are presented as the mean ±SEM. Differences between multiple groups were assessed with one-way ANOVA, followed by Tukey’s post hoc multiple comparisons test. Comparisons between two independent groups were performed with a two-sample t test. All p-values were calculated with two-tailed statistical tests. Differences were considered significant when P < 0.05. Statistical data were analyzed with the use of Prism 6 (GraphPad Software).

Fig. S1 shows flow cytometric serial selection, identification, and subsequent separation (sorting) of human peripheral blood inflammatory versus patrolling monocytes. Fig. S2 shows MH⁺ ions detected in trypsin digests by LCMS.

We thank David Williams and Robert Adelstein for their generous gifts of the Rac2^−/− and Myh9^fl/fl mice, respectively. We thank Gautham Rao for his generation of the HuR-YFP construct.

This work was supported by the following National Institutes of Health grants: F32 HL097422 (A.R. Morrison), T32 HL007950 (B.D. Young), R01 HL043331 (J.R. Bender), R01 HL084619 (M. Simons), and R01 GM47214 (M.A. Schwartz). This work was also supported by a Raymond and Beverly Sackler Foundation Award (J.R. Bender) and an AIRC-IG (R. Pardi).

The authors declare no competing financial interests.