The infarcted myocardium solicits GM-CSF for the detrimental oversupply of inflammatory leukocytes

During myocardial infarction (MI), interruption of coronary artery blood flow causes sterile cardiac injury. Although advances in medical and interventional management have improved outcomes after MI, many patients still develop progressive left ventricular dysfunction as a result of myocyte loss, suboptimal healing, and adverse left ventricular remodeling. Thus, beyond current standard revascularization and neurohumoral blockade measures, we urgently need new approaches to minimize acute cardiac ischemic damage and optimize the healing response to maximize the avoidance of chronic ischemic cardiomyopathy.

Wound healing after MI involves a variety of molecular and cellular events that replace necrotic myocardial debris with granulation tissue and a collagen-rich scar (Prabhu and Frangogiannis, 2016). Ischemic injury triggers massive recruitment of BM and spleen-derived neutrophils and monocytes (Nahrendorf et al., 2007; Swirski et al., 2009; Dutta et al., 2012; Leuschner et al., 2012; Swirski and Nahrendorf, 2013; Heidt et al., 2014; Hilgendorf et al., 2014a), which, upon accumulation, contribute to inflammation and repair by producing cytokines, proteases, reactive oxygen species, angiogenic substances, and growth factors. An optimal response should remove the irreversibly damaged myocytes while limiting further ischemic damage and avoiding complications such as heart failure or left ventricular rupture. Understanding the endogenous immune inflammatory mechanisms that operate during MI may identify strategies for avoiding ischemic cardiomyopathy, thus complementing current interventional and neurohumoral blockade approaches.

The multifunctional growth factor GM-CSF controls leukocyte production, proliferation, differentiation, and survival (Bradley and Metcalf, 1966; Hamilton, 2008; Becher et al., 2016). Although hematopoiesis under normal physiological conditions does not require GM-CSF, its production rises substantially after injury, which is one reason that current clinical trials target GM-CSF for diverse inflammatory conditions such as rheumatoid arthritis and multiple sclerosis (Hamilton et al., 2016; Wicks and Roberts, 2016). With the exception of a few inconclusive and contradictory findings, the role of GM-CSF in MI remains largely unresolved. For example, serum GM-CSF levels rise in patients with MI, correlating with severity of acute decompensated heart failure and subsequent cardiac remodeling (Parissis et al., 2000, 2004), whereas exogenous GM-CSF administration exacerbates heart failure (Maekawa et al., 2004). In some patients, however, GM-CSF may be beneficial after MI by improving neoangiogenesis (Seiler et al., 2001; Bruno et al., 2006). Here, we explored GM-CSF’s function in MI using human tissue and mice deficient in GM-CSF or its receptor (GM-CSFR). We demonstrate that GM-CSF impairs healing after MI by acting locally in the heart and at a distance in the BM to produce and recruit inflammatory leukocytes.

To explore GM-CSF’s function after MI, we first evaluated postinfarction survival. We found that WT mice died of cardiac rupture in considerably higher numbers than Csf2^−/− mice (Fig. 1 A). Among survivors, Csf2^−/− mice had improved left ventricular systolic function 21 d after MI, as assessed by cardiac magnetic resonance imaging (MRI), despite similar initial (1 d post-MI) infarct size or left ventricular volume in the groups (Fig. 1, B and C; and Fig. S1 A). On histological examination, 7-d-old infarcts of Csf2^−/− mice had reduced CD11b⁺ myeloid cell numbers, reduced collagen accumulation, and smaller α smooth muscle actin areas, with no differences in CD31⁺ endothelial cell number (Fig. 1 D), whereas flow cytometry revealed attenuated neutrophil, Ly-6C^high monocyte, and macrophage accumulation in Csf2^−/− hearts 3 and 7 d after MI (Fig. 1, E and F). At day 3, compared with WT, infarcts in Csf2^−/− mice had much lower levels of mRNAs that encode inflammatory cytokines such as IL-1β, IL-6, and matrix metalloproteinase (MMP)-9 (Fig. 1 G). Though GM-CSF may promote macrophage apoptosis through IL-23 (Subramanian et al., 2015), we found no differences in IL-23 mRNA levels in the infarct (Fig. 1 G). Mice lacking the common β subunit of the GM-CSF receptor (CD131; Csf2rb) yielded similar results (Fig. S1 B). Administration of recombinant mouse GM-CSF to WT mice decreased survival while increasing leukocyte accumulation (Fig. 1, H and I), whereas administration of anti-GM-CSF neutralizing antibody protected WT mice from cardiac rupture and improved survival during the first 4 wk after permanent coronary ligation (Fig. 1 J). These data suggest that endogenous GM-CSF causes impaired wound healing resulting from aggravated inflammation and enhanced leukocyte accumulation after MI.

The striking improvements in survival and cardiac function, along with the attenuated inflammation, in mice lacking GM-CSF, prompted us to delve deeper into the orchestrating mechanisms. The infarct, and not the uninfarcted myocardium, spleen, BM, and lymph nodes, augmented Csf2 mRNA, which peaked on day 1 and subsided thereafter (Fig. 2 A), suggesting a direct link between point of injury and GM-CSF induction. Expression of Il3 and Il5, the growth factors that also signal to Csf2rb, were undetectable and unchanged, respectively, indicating an exclusive and dynamic GM-CSF/GM-CSFR signaling dyad after MI (Fig. 2 B). Infarcted tissue also produced abundant Csf2 protein in both WT and Csf2rb^−/− mice (Fig. 2 C), reflecting robust local translation independent of GM-CSF signaling. That Csf2^−/− mice did not produce any GM-CSF confirmed the integrity of the mouse strain and the assays used (Fig. 2 C).

The presence of the GM-CSF transcript in the heart argued in favor of the idea of a cardiac-resident cell population as the growth factor’s source. Using FACS, we identified four cell types in the heart (Fig. 2 D): leukocytes, which express CD45; endothelial cells, which do not express CD45 but are abundant for CD31; fibroblasts, which express the fibroblast-specific marker MEFSK4 (Pinto et al., 2016), along with other classic fibroblast-associated markers (Fig. 2, D and E); and so-called other stromal cells, which lack the aforementioned markers and thus comprise a mixture of cells, some of which are cardiomyocytes (Fig. 2, D and F). Strikingly, fibroblasts were the most abundant, and nearly exclusive, Csf2 producers 1 d after coronary artery ligation (Fig. 2 G), with minor production by other cells and no production otherwise. The increase depended on a mixture of danger-associated molecular patterns given reduced GM-CSF production in Myd88^−/−, Tlr9^−/−, and Ticam1^−/− mice (Fig. 2 H). To test whether cardiac fibroblasts can directly sense danger signals via TLR3 and TLR9, we sorted fibroblasts, placed them in culture, stimulated with the two TLR ligands alone or in combination, and measured GM-CSF in supernatants. Our in vitro data confirm that fibroblasts can produce GM-CSF through Ticam1-dependent (Tlr3) and Myd88-dependent (Tlr9) mechanisms, suggesting the importance of multiple signals for GM-CSF production (Fig. 2 I).

The data obtained in the mouse led us to consider whether infarcted human hearts likewise produce GM-CSF after MI. We obtained heart tissue specimens from deceased patients with infarcts that, upon pathological examination, were classified as acute (1–2 d because of C4d staining) or late (5–7 d, given signs of leukocyte accumulation and granulation tissue formation). We also examined hearts from patients who died from causes unrelated to MI. Consistent with data obtained in the mouse, we found that the acutely infarcted myocardium, and not the late infarct or the uninfarcted myocardium, contained GM-CSF⁺ cells. Of these, the majority were spindle-shaped cells that expressed vimentin, a morphology and phenotype consistent with cardiac fibroblasts (Fig. 2, J–L). These data indicate that cardiac fibroblasts predominate as the source of GM-CSF in the acutely infarcted myocardium in both mice and humans.

Having established GM-CSF’s source, we next sought to determine whether GM-CSF influences the response to ischemic injury via local effects in the heart. Because of the correlation between GM-CSF expression and leukocyte accumulation, we reasoned that the growth factor may contribute to leukocyte accrual, which depends on influx, survival, and proliferation. We eliminated the likelihood that GM-CSF promotes leukocyte survival (Fig. 3, A and B) and proliferation (Fig. 3 C), and thus focused on influx. We found that 3-d-old infarcts in Csf2^−/− mice expressed lower levels of Cxcl2, Ccl2, Ccl3, Ccl5, and Ccl7 compared with WT mice (Fig. 3 D). Among those chemokines that are GM-CSF dependent, Cxcl2 attracts neutrophils whereas Ccl2, Ccl3, Ccl5, and Ccl7 attract Ly-6C^high monocytes. Because we found no differences in expression of either neutrophil Cxcr2 (ligand for Cxcl2) or monocyte Ccr2 (ligand for Ccl2) between WT and Csf2rb^−/− mice (Fig. 3 E), we concluded that GM-CSF–dependent effects on recruitment, if any, likely rely on the attracting chemokines produced in the heart.

To explore neutrophil recruitment, we profiled Cxcl1 and Cxcl2. We found that cardiac fibroblasts produced abundant levels of GM-CSF–independent Cxcl1, whereas leukocytes were the largest source of GM-CSF–dependent Cxcl2 (Fig. 4 A). Among the leukocytes, neutrophils generated considerable Cxcl2, and Csf2rb^−/− neutrophils expressed significantly less Cxcl2 compared with WT cells (Fig. 4 B). These data collectively indicate that enhanced neutrophil accumulation is a function of increased recruitment rather than survival. This recruitment is driven by GM-CSF, which stimulates neutrophil-attracting Cxcl2 production.

To explore monocyte recruitment, we profiled Ccl2. To determine Ccl2’s cellular source after MI, we took advantage of Ccl2 reporter mice, in which all Ccl2⁺ cells express the fluorescent protein mCherry. mCherry⁺ cells appeared in 2-d-old infarcts, with minimal to no expression in naive hearts, blood cells, or cells in the BM (Fig. 4 C). Among the cardiac Ccl2 producers, myeloid cells, such as Ly-6C^high monocytes and macrophages, predominated, with additional contribution from neutrophils and cardiac fibroblasts (Fig. 4, C and D). In Ccr2^−/− mice, which have profoundly reduced numbers of circulating monocytes, Ccl2 production in the heart was diminished, along with diminished monocyte and macrophage accumulation, further substantiating the idea that monocytes are Ccl2 sources (Fig. 4 E). Ccl2 production depended on signaling via GM-CSFR (Fig. 4 F), thus recapitulating observations made in Csf2^−/− mice (Fig. 3 D). In vitro, GM-CSF strongly stimulated Ccl2 production from sorted Ly-6C^high monocytes and macrophages. We also profiled cells from Csf2rb^−/− mice as controls and found no Ccl2 production in response to GM-CSF, indicating a direct, specific stimulation (Fig. 4 G). However, neither neutrophils nor fibroblasts produced additional Ccl2 upon GM-CSF stimulation, indicating that Ccl2 production in these cells involves GM-CSF–independent pathways. To dig deeper, we found phosphorylated STAT5 in Ly-6C^high monocytes and macrophages, indicating GM-CSF signaling (Fig. 4 H). Finally, we adoptively transferred Ly-6C^high monocytes from GFP⁺ mice into WT and Csf2^−/− animals subjected to coronary ligation 3 d earlier. Even though the same number of GFP⁺ cells were injected, we retrieved far fewer GFP⁺ cells from Csf2^−/− infarcts (Fig. 4, I and J). Our data show that cardiac GM-CSF attracts monocytes to the heart by stimulating Ccl2 production.

Our data thus far indicate that GM-CSF promotes leukocyte influx to the infarcted myocardium via neutrophil- and monocyte-attracting chemokine production. We next tested whether GM-CSF influences leukocyte function. We elected to profile inflammatory and proteolytic activity, as these can disrupt wound repair (Nahrendorf et al., 2007; Hilgendorf et al., 2014a) and contribute to left ventricular heart rupture (Heymans et al., 1999; Matsumura et al., 2005), respectively. Using a probe that reports on proteolytic activity (Nahrendorf et al., 2007), we found considerably reduced proteolysis in Csf2^−/− mice (Fig. 5, A and B). Among the cells, Ly6C^high monocytes and neutrophils contributed comparably to proteolysis, but the highest proteolytic activity tracked with macrophages (Fig. 5, C and D). Importantly, all three cell types had reduced proteolytic activity in Csf2^−/− mice. In keeping with these observations, we likewise detected reduced expression of Il1β, Il6, and Mmp9 in sorted neutrophils, Ly-6C^high monocytes, and macrophages of Csf2rb^−/− mice, compared with WT mice (Fig. 5 E). Together, these data indicate that GM-CSF not only induces monocyte and neutrophil influx but is also important to leukocyte function, including proteolysis and inflammation.

Having shown the local action of GM-CSF, we next considered whether the growth factor also acts systemically. In support of this idea, GM-CSF protein levels increased in serum and BM 1 and 2 d after coronary artery ligation (Fig. 6, A and B). To determine whether heart-derived GM-CSF can influence the BM via the circulation, we joined WT mice with either WT or Csf2rb^−/− partners by parabiosis, waited 14 d to establish equilibrium (Theurl et al., 2016), subjected the WT mouse to coronary ligation, and performed a 2-h BrdU pulse-chase experiment 2 d later (Fig. 6 C). Compared with non-MI controls, we found higher Ly-6C^high monocyte production rates in WT partner BM (i.e., WT partner of a WT mouse that underwent coronary ligation), indicating that MI can stimulate hematopoiesis via the blood. Csf2rb^−/− partners, in contrast, showed blunted production (Fig. 6 C), indicating that infarct-derived GM-CSF travels through the circulation to stimulate BM hematopoiesis.

To elucidate GM-CSF’s specific target, we examined CD131 (i.e., Csf2rb) expression among hematopoietic stem progenitor cells and found a distinct CD131⁺ population within Lin^–ckit⁺Sca1⁺ LSK that was flk2^–CD48⁺CD150^–, and thus akin to the recently described myeloid-biased MPP3s (Pietras et al., 2015; Fig. 7 A and Fig. S2). In vitro experiments using different LSK subpopulations stimulated with GM-CSF supported this observation and further revealed that MPP3s responded most to GM-CSF (Fig. 7 B), giving rise to GMP and MDP (Fig. 7 C) before becoming CD11b⁺ mature myeloid cells, comprised of monocytes/macrophages (CD11b⁺ CD115⁺ Ly6G^–) and neutrophils (CD11b⁺ CD115^–Ly6G⁺; Fig. 7 D). In vivo, CD131⁺ MPP3s augmented proliferation in WT mice 2 d after MI, whereas CD131^– MPP3s and CD131⁺ (and CD131^–) MPP3s from Csf2^−/− mice proliferated at baseline rates (Fig. 7 E). In agreement with these data, Csf2rb^−/− mice had a lower percentage of BrdU⁺ BM MPP3s than WT mice 2 d after coronary artery ligation (Fig. 7 F). Moreover, experiments involving competitive adoptive transfer of CD131⁺ and CD131^– MPP3s into WT mice with MI revealed a higher proportion of myeloid cells accumulating in the myocardium that had differentiated from the CD131⁺ population, suggesting that CD131⁺ MPP3s have an enhanced capacity for recruitment and differentiation into mature myeloid cells in response to myocardial ischemia (Fig. 7 G). Consequently, the BM of WT mice had more neutrophils and Ly-6C^high monocytes during MI than the BM of Csf2^−/− and Csf2rb^−/− mice (Fig. 7 H and Fig. S2 B).

We have previously demonstrated that the spleen contributes monocytes to the infarcted myocardium 1 d after MI (Swirski et al., 2009). In this study, we wondered whether the spleen may be contributing to accumulation on day 3, a time point at which we detected GM-CSF–dependent differences. Whereas splenectomy at the time of infarction recapitulated the contribution of the reservoir in the early monocyte response on day 1, splenectomy on day 2 showed no contribution of the spleen to leukocyte accumulation on day 3, indicating that GM-CSF–dependent leukocyte accumulation was spleen independent (Fig. S3, A and B). In agreement with these observations, neutrophils and Ly-6C^high monocytes in the spleen were not altered in the absence of GM-CSF signaling 3 d after MI (Fig. S3 C). Together, these data reveal that GM-CSF produced in the heart can promote hematopoiesis in the BM by targeting a specific resident progenitor subset.

Our data thus far indicate that GM-CSF alerts the hematopoietic compartment for increased inflammatory cell production and recruitment after MI. We next sought to determine whether hematopoietic cells were indeed responsible for the difference in survival between WT and Csf2^−/− mice, as depicted in Fig. 1. To this end, we generated three types of BM chimeras: lethally irradiated WT mice reconstituted with WT or Csf2rb^−/− BM cells (hereafter termed WT^WT and WT^{Csf2rb−/−}, respectively) and lethally irradiated Csf2rb^−/− mice reconstituted with WT BM cells (Csf2rb^−/−WT). After coronary artery ligation, we observed increased Csf2 in WT^WT mice (Fig. 8 A), along with abundant leukocyte, neutrophil, and Ly-6C^high monocyte accumulation in both WT^WT and Csf2rb^−/−WT mice (Fig. 8 B). In contrast, fewer leukocytes accumulated in WT^{Csf2rb−/−} mice, suggesting that GM-CSF–dependent increases in leukocyte numbers were linked directly to hematopoiesis. To test this conjecture, we enumerated cells in the BM, and observed that BM cells lacking the GM-CSF receptor failed to generate progenitors and mature myeloid cells in addition to those generated in the basal state (Fig. 8 C). Moreover, infarcts of WT^{Csf2rb−/−} mice contained lower mRNAs encoding the inflammatory mediators Il-1β, Il-6, Mmp9, and Ccl2 than those in the other groups (Fig. 8 D). Finally, when followed for 4 wk after coronary ligation, WT^{Csf2rb−/−} mice survived longer compared with the other two groups (Fig. 8 E). Collectively, these data indicate that GM-CSF reduces survival by acting on hematopoietic cells.

Many studies have implicated leukocytosis, and specifically the rise of neutrophils and monocytes, as independent risk factors in cardiovascular disease (Kyne et al., 2000; Brown et al., 2001; Hajj-Ali et al., 2001; Maekawa et al., 2002; Madjid et al., 2004; Tsujioka et al., 2009; Nahrendorf et al., 2010). In experimental MI, excessive myeloid cell accumulation likely contributes to adverse cardiac remodeling and death (Panizzi et al., 2010; Kempf et al., 2011; Anzai et al., 2012), whereas hearts with reduced leukocyte infiltration heal and function better than controls (Romson et al., 1983; Kaikita et al., 2004; Leuschner et al., 2010). Apoe^−/− mice, which develop monocytosis (Swirski et al., 2007) by augmenting responsiveness to GM-CSF (Murphy et al., 2011), have detrimental wound healing and accelerated heart failure after MI (Panizzi et al., 2010). The BM, and to a lesser extent the spleen, manufactures neutrophils and monocytes through myelopoiesis, and although billions of cells are made every day in the steady state, output rises in response to injury or infection. Given leukocytes’ central role in post-MI inflammation and repair, identifying signals that contribute to increased cell production has high importance.

Our study’s findings, that GM-CSF controls neutrophil and monocyte production by acting on a specific BM progenitor, have several implications that extend beyond MI. Unlike M-CSF, the related growth factor that controls myeloid cell production in the steady state, GM-CSF contributes little to homeostatic hematopoiesis; mice lacking GM-CSF or its receptor have normal leukocyte levels and, aside from developing alveolar proteinosis as a result of defective surfactant clearance, have no demonstrable abnormalities. That said, accumulating evidence shows that the growth factor propels the inflammatory cascade. Our data contribute to this concept by demonstrating that GM-CSF–dependent leukocytosis relies not so much on a general increase in cell production as on a specific GM-CSFR–expressing subset within the MPP3 progenitor (Pietras et al., 2015). The BM, it seems, remains poised in a state of alert for GM-CSF should a need for increased cell production arise.

Once generated and released into the bloodstream, leukocytes follow chemotactic cues to find their destination tissue. Among the signals, the ischemic heart produces the chemokines Ccl2 (Hayashidani et al., 2003; Dewald et al., 2005), which attracts Ccr2⁺ monocytes, and Cxcl2, which attracts neutrophils (Soehnlein and Lindbom, 2010). We discovered that GM-CSF stimulates substantial Ccl2 and Cxcl2 in the heart after ischemic injury. Though many cells produce Ccl2 and Cxcl2, Ly-6C^high monocytes and macrophages are among the most abundant contributors of Ccl2, whereas neutrophils themselves produce Cxcl2. This provides for a parallel, feed-forward response that ensures massive monocyte and neutrophil accumulation over a short period of time (Fig. 9). Although they are required for wound healing, an overabundance of these cells results in excessive inflammation and proteolysis, which eventually leads to heart failure and death.

Our previous studies in sepsis, pneumonia, and atherosclerosis identified a B cell subset, which we called innate response activator, as a major GM-CSF source (Rauch et al., 2012; Hilgendorf et al., 2014b; Weber et al., 2014). This study identified no GM-CSF–producing B cells in the heart or spleen. However, the observation that cardiac fibroblasts make GM-CSF is intriguing, as fibroblasts also produce GM-CSF in other heart-related inflammatory diseases, such as autoimmune myocarditis and Kawasaki disease (Wu et al., 2014; Stock et al., 2016). As nonhematopoietic cells, fibroblasts are strategically located to sense the cardiac microenvironment. During MI, when dying cardiomyocytes release danger-associated molecular patterns (Mann, 2015; Prabhu and Frangogiannis, 2016), cardiac fibroblasts can easily respond as they bear TLRs and other scavenger receptors. Our data accordingly indicate that multiple signals converging on the adapter molecules MyD88 or TRIF trigger GM-CSF production in the infarcted heart, which helps to generate and recruit hematopoietic cells that fuel inflammation. Experiments involving chimeras lacking the GM-CSF receptor underscore the importance of GM-CSF’s action on the hematopoietic rather than the stromal compartments because left ventricular rupture ultimately depends on this dyad. This result should not be surprising given that inflammatory leukocytes are major sources of oxidases, proteases, and other factors that can antagonize the formation of a durable scar.

Drugs targeting GM-CSF or its receptor are available for human use, and several clinical trials are underway testing the efficacy of anti–GM-CSF therapy in various inflammatory diseases (Hamilton et al., 2016; Wicks and Roberts, 2016). As a mediator that governs production, recruitment, and function of myeloid cells after MI, GM-CSF should be considered a potential therapeutic target to foster salutary healing of the ischemic myocardium and prevent ischemic cardiomyopathy.

C57BL/6 (WT), B6.SJL-Ptprc^aPepc^b/BoyJ, C57BL/6-Tg(UBC-GFP)30Scha/J (GFP⁺), B6.129P2(SJL)-Myd88^tm1.1Defr/J, B6.129-Tlr2^tm1Kir/J, B6(Cg)-Tlr4^tm1.2Karp/J, B6.129S1-Tlr7^tm1Flv/J, C57BL/6J-Tlr9^M7Btlr/Mmjax, B6.129S7-Il1r1^tm1/mx/J, C57BL/6J-Ticam1^Lps2/J, B6.Cg-Ccl2^tm1.1Pame/J, and B6.129S4-Ccr2^tm1lfc/J (The Jackson Laboratory) were used in this study. Csf2^−/− and Csf2rb^−/− mice on C57BL/6 background (10 generations) were bred in-house (Rauch et al., 2012; Hilgendorf et al., 2014b; Weber et al., 2014). All experiments were conducted with male mice aged 8–16 wk at the time of death. All protocols were approved by the Animal Review Committee at Massachusetts General Hospital.

MI was induced by permanent ligation of the left anterior descending coronary artery (LAD) as previously described (Nahrendorf et al., 2007; Hilgendorf et al., 2014a). In brief, mice were anesthetized with isoflurane (2%/2 liters O₂), intubated, and ventilated with an Inspira Advanced Safety Single Animal Pressure/Volume Controlled Ventilator (Harvard Apparatus). Left thoracotomy was performed in the fourth intercostal space after shaving the chest wall. The left ventricle was visualized, and the LAD was ligated with monofilament 8-0 suture (Ethicon). The chest and skin were closed with a 7-0 nylon suture followed by removal of air from the thorax via a pleural catheter. The procedure was performed by the same surgeon blinded to genotypes.

Naive C57BL/6 or Csf2rb^−/− mice were lethally irradiated (950 cGy) and reconstituted with either WT or Csf2rb^−/− BM cells to generate three groups: (a) WT^WT (irradiated WT mice reconstituted with WT BM), (b) WT^{Csf2rb−/−} (irradiated WT mice reconstituted with Csf2rb^−/− BM), and (c) Csf2rb^−/−WT (irradiated Csf2rb^−/− mice reconstituted with WT BM). After 7–8 wk to allow reconstitution, the mice were subjected to MI surgery.

BM Ly-6C^high monocytes were collected from C57BL/6-Tg(UBC-GFP)30Scha/J mice by flow-assisted cell sorting after enrichment of cells negative for CD3, CD19, B220, NK1.1, and Ly6G using magnetic cell separation. At 3 d after MI, 3 × 10⁶ GFP⁺ Ly-6C^high monocytes were injected intravenously into WT or Csf2^−/− mice. For BM MPP3 transfer, CD131⁺ MPP3s were sorted from C57BL/6-Tg(UBC-GFP)30Scha/J mice (CD45.2⁺ GFP⁺), and CD131^– MPP3s were sorted from C57BL/6 mice (CD45.2⁺ GFP⁻). Those cells were intravenously injected at a 1:1 ratio (10⁴ cells each) into B6.SJL-Ptprc^aPepc^b/BoyJ (CD45.1⁺) mice subjected to MI 1 d earlier.

Parabiosis surgery was performed as previously described (Robbins et al., 2013). In brief, after shaving the corresponding lateral aspects of each mouse, matching skin incisions were made from behind the ear to the tail of each mouse, and the subcutaneous fascia was bluntly dissected to create 0.5 cm of free skin. The scapulae were sutured using mononylon 5.0 (Ethicon), and the dorsal and ventral skins were approximated by continuous suture. Mice were joined for 2 wk and then subjected to MI.

To assess cell proliferation, 1 mg BrdU was injected intraperitoneally 2 or 24 h before death and subsequent organ harvest. BrdU flow kit (BD Biosciences) was used to stain BrdU⁺ cells.

Splenectomy was performed as previously described (Swirski et al., 2009). In brief, during isoflurane anesthesia, the abdominal cavity of mice was opened, and the spleen vessels were cauterized. The spleen was carefully removed. For control experiments, the abdomen was opened, but the spleen was not excised.

After coronary ligation, mice were randomly allocated to either anti–GM-CSF treatment or control IgG treatment groups. Neutralizing antibody against GM-CSF (MP1-22E9, eBioscience; 300 µg) or control IgG2a (eBR2a, eBioscience; 300 µg) was intravenously injected once a day for 3 d.

Recombinant mouse GM-CSF (BioLegend; 20 µg/kg) was intravenously injected into mice after the MI surgery twice a day for 3 d. PBS was used as a control.

To assess cardiac function and morphology, MRI was performed on days 1 and 21 after MI as described previously (Sager et al., 2015). Cine images of the left ventricular short axis were obtained by using a 7-Tesla horizontal bore Pharmascan (Bruker) and a custom-built mouse cardiac coil (Rapid Biomedical). Left ventricular volume, ejection fraction, and infarct size were acquired and calculated as previously described.

2 nmol of an MMP sensor (MMPSense-680; PerkinElmer) was intravenously injected into mice subjected to MI 2 d earlier. 24 h later, hearts were excised, and ex vivo fluorescence imaging was performed to interrogate magnitude and quality of MMP activity.

Peripheral blood was collected by retroorbital bleeding, and erythrocytes were lysed in RBC lysis buffer (BioLegend). Spleens, femurs, tibias, and hearts were excised after vascular perfusion with cold PBS. Minced spleen and flushed BM were strained through 40-µm nylon mesh (BD Biosciences). Spleen cell suspensions were further subjected to RBC lysis. The naive hearts or infarcted areas with the border zone were minced and digested with 450 U/ml collagenase I, 125 U/ml collagenase XI, 60 U/ml DNase I, and 60 U/ml hyaluronidase (Sigma-Aldrich) in PBS for 1 h at 37°C with shaking. Total viable cell numbers were counted using Trypan blue (Cellgro, Mediatech).

Single-cell suspensions of heart tissue and BM from indicated animals were made and stained to identify indicated cell populations. Cells were sorted on a FACS Aria II (BD Biosciences) directly into either RLT buffer for subsequent RNA isolation or complete medium for cell culture experiments.

Single-cell suspensions were stained in PBS supplemented with sterile 2% FBS and 0.5% BSA. The following monoclonal antibodies were used for flow cytometric analysis: anti-CD45 (30-F11), anti-CD45.1 (clone A20), anti-CD45.2 (clone 104), anti-CD3e (clone 145-2C11), anti-CD90.2 (clone 53–2.1), anti-CD19 (clone 6D5), anti-B220 (clone RA3-6B2), anti-NK1.1 (clone PK136), anti-Ly6G (clone 1A8), anti–Ly-6C (AL-21), anti-MHCII (clone AF6-120.1), anti-F4/80 (clone BM8), anti-CD11b (clone M1/70,), anti-CD11c (clone HL3), anti-CD115 (clone AFS98), anti-Ter119 (clone Ter-119), anti-CD34 (clone RAM34), anti-flk2 (clone A2F10), anti-CD150 (clone TC15-12F12.2), anti-Sca1 (clone D7), anti–c-kit (clone 2B8), anti-CD48 (clone HM48-1), anti-CD16/32 (clone 93), anti-Gr1 (clone RB6-8C5), anti-CD127 (clone A7R34), anti-CD131 (clone REA193), anti-CD31 (clone 390), anti-MEFSK4 (clone mEF-SK4), anti-CCR2 (clone 475301), anti-CXCR2 (clone TG11/CXCR2), anti–activated caspase 3 (clone C92-605), pSTAT5 (clone pY694), and isotype controls. Antibodies were purchased from BioLegend, BD Biosciences, eBioscience, or Miltenyi Biotec. Viable cells were identified as unstained with Zombie Aqua (BioLegend).

For intracellular staining, cells were fixed and permeabilized with BD Cytofix/Cytoperm (BD Biosciences) according to the manufacturer’s instructions. For pSTAT5 staining, cells were stimulated with or without GM-CSF–containing medium (20 ng/ml) for 30 min at 37°C to induce phosphorylation of STAT5, followed by addition of 4% PFA solution (pH 7.4) to obtain a final concentration of 2%. After fixation for 20 min, cells were resuspended in 1 ml ice-cold methanol to permeabilize. After 30 min of incubation at 4°C, cells were washed twice and stained with anti-pSTAT5 antibody as well as antibodies for extracellular antigens for 45 min at room temperature before a final wash and acquisition. Specifically, cardiac cells were defined as (1) Ly-6C^high monocytes (CD45⁺Lin^–CD11b⁺F4/80^–Ly-6C^high), (2) neutrophils (CD45⁺CD11b⁺Lin⁺F4/80^–), (3) macrophages (CD45⁺Lin^–CD11b⁺F4/80⁺Ly-6C^low), (4) endothelial cells (CD45^–CD31⁺MEFSK4^–), (5) fibroblasts (CD45^–CD31^–MEFSK4⁺), and (6) other stromal cells (CD45^–CD31^–MEFSK4^–). BM cells were defined as (1) monocytes (CD45⁺Lin^–CD11b⁺CD115⁺F4/80^–), (2) neutrophils (CD45⁺Ly6G⁺CD11b⁺CD115^–F4/80^–MHCII^–), (3) ^LTHSC (Lin₂^–ckit⁺Sca1⁺flk2^–CD48^–CD150⁺), (4) ^STHSC (Lin₂^–ckit⁺Sca1⁺flk2^–CD48^–CD150^–), (5) MPP2 (Lin₂^–ckit⁺Sca1⁺flk2^–CD48⁺CD150⁺), (6) MPP3 (Lin₂^–ckit⁺Sca1⁺flk2^–CD48⁺CD150^–), (7) MPP4 (Lin₂^–ckit⁺Sca1⁺flk2⁺CD150^–), (8) GMP (Lin₂^–ckit⁺Sca1^–CD34⁺CD16/32^highCD115^–), and (9) MDP (Lin₂^–ckit⁺Sca1^–CD34⁺CD16/32^highCD115⁺). Lineages were defined as Lin: CD3, CD90.2, CD19, B220, NK1.1, Ter119, Ly6G; Lin₂: CD3, CD90.2, CD19, B220, NK1.1, Ter119, Gr1, CD127, CD11c. Data were acquired on a LSRII and analyzed with FlowJo (Tree Star).

For all cell culture experiments, cells were cultured in complete medium (RPMI-1640 supplemented with 10% FBS, 2 mM l-glutamine, 100 U/ml penicillin and streptomycin, 10 mM Hepes, 50 µM 2-mercaptoethanol, 1 mM sodium pyruvate, and 1× nonessential amino acids), and kept in a humidified 5% CO₂ incubator at 37°C. To determine the effect of GM-CSF on CCL2 production, neutrophils, Ly-6C^high monocytes, macrophages, and fibroblasts were sorted from WT and Csf2rb^−/− mice on the FACS Aria II, and 5 × 10³ cells were seeded into 96-well plates in 200 µl medium. After incubation with or without GM-CSF for 48 h, supernatants were collected, and CCL2 levels were measured by ELISA. To identify the cellular target of GM-CSF in BM LSK compartment, 5 × 10³ sorted ^LTHSC, ^STHSC, MPP2, MPP3, and MPP4 cells were cultured with and without GM-CSF for indicated periods, followed by counting and staining after retrieval from the 96-well plates. To determine the effect of danger signals on GM-CSF production in cardiac fibroblasts after MI, 3 × 10⁴ cardiac fibroblasts sorted from 1-d-old infarct were plated onto 96-well plates in 200 µl medium and incubated in the presence or absence of poly(I:C) or CpG (InvivoGen) for 48 h. Supernatants were collected, and GM-CSF levels were measured by ELISA.

Total RNA was isolated using the RNeasy Mini kit (Qiagen) according to the manufacturer’s instructions. cDNA was generated from 1 µg total RNA per sample using High Capacity cDNA Reverse Transcription kit (Applied Biosystems).

2 × 10⁴ total leukocytes, neutrophils, Ly-6C^high monocytes, macrophages, endothelial cells, fibroblasts, and other stromal cells were sorted from infarcted myocardium, and total RNA was extracted using the RNeasy Micro kit (Qiagen) followed by cDNA transcription. Quantitative real-time TaqMan PCR was performed using the following TaqMan primers (Applied Biosystems): Il-1β (Mm01336189_m1), Il-3 (Mm00439631_m1), Il-5 (Mm00439646_m1), Il-6 (Mm00446190_m1), Mmp9 (Mm00442991_m1), Col1a1 (Mm00801666_g1), Pdgfra (Mm00440701_m1), Postn (Mm01284919_m1), Csf2 (Mm01290062_m1), Cxcl1 (Mm04207460_m1), Cxcl2 (Mm00436450_m1), Ccl2 (Mm00441242_m1), Ccl3 (Mm00441259_g1), Ccl5 (Mm01302427_m1), Ccl7 (Mm00443113_m1), Cx3cl1 (Mm00436454_m1), Ccl2 (Mm00441242_m1), Myh6 (Mm00440359_m1), Il-23a (Mm00518984_m1), and housekeeping gene Gapdh (Mm99999915_g1). PCR was run on a 7500 thermal cycler (Applied Biosystems), and data were quantified with the 2^-ΔCt method.

Hearts from WT and Csf2^−/− mice (7 d after MI) were embedded in Tissue-Tek OCT compound (Sakura Finetek), frozen in 2-methylbutane (Thermo Fisher Scientific) cooled with dry ice, and sectioned into 6-µm slices. Immunohistochemistry was performed using antibodies against CD11b (dilution 1:25, clone M1/70; BD Biosciences), collagen 1 (dilution 1:50, ab21286; Abcam), CD31 (dilution 1:5, clone MEC13.3; BD Biosciences), and α-smooth muscle actin (dilution 1:50, ab5694; Abcam). Biotinylated rabbit anti–rat IgG (dilution 1:100, BA4001) or biotinylated goat anti–rabbit IgG (dilution 1:100, BA1000) followed by Vectastain Elite ABC HRP kit (PK-6100; all from Vector Laboratories) were applied to the sections, and AEC substrate chromogen (K3464; Dako/Agilent) was used for color development. Human heart samples were obtained from autopsies performed in the Department of Pathology, Massachusetts General Hospital. The study was approved by the Human Subjects Institutional Review Board at Massachusetts General Hospital. Hearts from two patients who died within 2 d after MI (acute infarct), two patients who died 7 d after MI (late infarct), and two patients who died from other reasons outside the heart were used in this study. Samples were fixed in 10% formalin and embedded in paraffin for histological sectioning (4-µm-thick slices). After deparaffinization and heat-induced antigen retrieval, sections were blocked with 4% normal goat serum for 45 min and stained with rabbit anti–human C4d (dilution 1:50, 12-5000; American Research Products). For immunofluorescence staining, rabbit anti–human GM-CSF (dilution 1:100, bs-3790R, Bioss) and guinea pig anti–human vimentin (dilution 1:50, GP53; Progen) were incubated at 4°C overnight, followed by secondary antibodies, DyLight680 goat anti–rabbit IgG (dilution 1:100, 35568; Thermo Fisher Scientific) and biotinylated anti–guinea pig IgG (dilution 1:100, BA7000, Vector Laboratories). To detect vimentin, streptavidin Dylight594 (dilution 1:600, SA-5594; Vector Laboratories) was applied, and nuclei were counterstained by DAPI (D12490; Thermo Fisher Scientific). Images were captured by using a BX63 automated fluorescence microscope (Olympus).

GM-CSF levels were measured in heart homogenates, serum, and BM fluids with the Mouse GM-CSF ELISA kit (R&D Systems). CCL2/MCP-1 levels were measured in cell culture supernatants with the CCL2 ELISA kit (R&D Systems) according to the manufacturer’s instructions.

Results are shown as mean ± SEM. Statistical tests included unpaired, two-tailed nonparametric Mann–Whitney tests (when Gaussian distribution was not assumed) or unpaired, two-tailed parametric t tests with Welch’s correction (when Gaussian distribution was assumed). For multiple comparisons, nonparametric multiple comparisons test comparing mean rank of each group (when Gaussian distribution was not assumed) or one-way ANOVA followed by Tukey’s test were performed. Survival distributions were estimated by the Kaplan–Meier method and compared by the log-rank test. P values of 0.05 or less were considered to denote significance.

Fig. S1 shows additional readouts from MRI and cell quantification. Fig. S2 shows gating strategies for HSPCs. Fig. S3 shows experiments using splenectomy.

This work was supported in part by National Institutes of Health grants R35HL135752, R01 HL095612, and R01 HL128264; the American Heart Association’s Established Investigator Award; and the Patricia and Scott Eston MGH Research Scholar (to F.K. Swirski). A. Anzai was supported by grants from the Banyu Life Science Foundation International. J.E. Mindur was supported by a National Institutes of Health Ruth L. Kirschstein Institutional National Research Service Award training grant (T32 AI118692). M. Nairz was supported by an Austrian Science Fund (FWF) Erwin Schroedinger Fellowship (J3486-B13). P. Libby was supported by National Institutes of Health grant R01 HL080472 and the Robert R. McCormick Foundation Charitable Fund.

The authors declare no competing financial interests.

Author contributions: A. Anzai conceived the project, designed and performed experiments, analyzed and interpreted data, and wrote the manuscript; J. Choi, S. He, A.M. Fenn, M. Nairz, S. Rattik, C. McAlpine, J.E. Mindur, C. Chan, Y. Iwamoto, B. Tricot, and G.R. Wojtkiewicz performed experiments; R. Weissleder, P. Libby, and B. Becher provided intellectual input and edited the manuscript; M. Nahrendorf provided intellectual input, designed experiments, and edited the manuscript. J.R. Stone provided materials and edited the manuscript; and F.K. Swirski conceived the project, designed experiments, analyzed and interpreted data, and wrote the manuscript.