**4. Discussion**

The results of this study show that ERK isoforms have a di fferential e ffect on arteriogenesis. While a global *Erk1* knockout impaired blood flow recovery due to ine fficient arteriogenesis, it took a combination of endothelial- and macrophage-specific knockout of this isoform to match the global deletion phenotype. The principal driver of response appeared to be a large increase in tissue macrophage levels that likely resulted in abnormally high VEGF levels and exuberant, albeit ine fficient, arteriogenesis. In contrast, endothelial-specific deletion of the *Erk2* isoform resulted in reduced blood flow recovery even though the anatomical extent of arteriogenesis appeared normal. The culprit in this case was a dramatic reduction in endothelial eNOS expression that led to vasoconstriction. Neither ERK isoform deletion by itself in smooth muscle cells a ffected either blood flow recovery or arteriogenesis per se, while deletion of both genes resulted in a transient decrease of blood flow recovery.

Arteriogenesis is a process leading to the formation of arteries and arterioles. It can proceed either by remodeling of pre-existing collateral arteries or by expansion and arterialization of the capillary bed [2,5,29]. Arteriogenesis is distinct from angiogenesis, a process defined as sprouting and proliferation of the existing capillary bed [2]. Importantly, the two processes are regulated by distinctly di fferent sets of factors. While hypoxia is the primary driver of angiogenesis, arteriogenesis is induced by a combination of shear stress and other mechanical factors [2,5,30]. At the same time, VEGF, and its subsequent induction of endothelial ERK activation, are crucial to both angiogenesis and arteriogenesis [31–35]. One important distinction, however, is the source of VEGF: while, in angiogenesis, VEGF is produced locally due to tissue ischemia, in arteriogenic settings macrophages are the key source of the growth factor [24].

While the importance of the VEGF/VEGFR2/ERK signaling cascade in both angiogenesis and arteriogenesis has been clearly recognized, how this signaling cascade promotes the two distinct means of vascular growth has been unclear. ERK activation is thought to be involved in the proliferation and migration of endothelial cells. Interestingly, our data indicate that endothelial proliferation is largely controlled by ERK2 while migration is the additive function of both isoforms. These results are in agreemen<sup>t</sup> with the study of Lefloch et al. who found that ERK2 controls cell proliferation in NIH 3T3 cells [36]. Other isoforms-specific e ffects of ERK signaling are regulation of macrophage accumulation by ERK1 and regulation of eNOS expression by ERK2.

Both global and a combination of macrophage- and/endothelial-specific *Erk1* knockouts led to markedly increased macrophage accumulation at the site of arteriogenesis after the common femoral artery ligation that was coupled with exuberant but ine ffective arteriogenesis. The critical role macrophages play in arteriogenesis is well established. While the M2 subset (macrophages involved in wounds healing and vascular growth) have been described as a principle source of VEGF [37,38], other cell population, including blood-derived inflammatory cells, and mechanical factors can also contribute [13,39]. The observed increase in tissue macrophage levels in these mutant strains is likely derived from circulating monocytes [11] although a proliferation of resident M2 macrophages cannot be ruled out [39,40]. Endothelial cells play a crucial role in monocyte recruitment by increasing expression of the Notch ligand Dll1 [11], as has been observed in this study. Activation of Notch signaling in recruited monocytes polarizes them to an arteriogenic M2 phenotype [11]. Similarly, haploinsu fficiency of *Phd2*, encoding the PHD2 oxygen sensor, leads to an expansion of tissue-resident M2-like macrophages, an increased release in arteriogenic factors, and an improved vascular reperfusion in the hindlimb ischemia model [37]. Here, we found that ERK1 controls endothelial-macrophage crosstalk, and that exacerbated macrophage infiltration increases arteriogenesis extent but decreases arteriogenesis functionality. However, a combination of the macrophage- and endothelial-specific *Erk1* knockout phenotype is not as severe as in the *Erk1* global knockout mice, suggesting that other cell types are also may be involved in the phenotype found in the *Erk1* global knockout mice.

In addition to the Dll1 signaling, endothelial cells also regulate macrophage infiltration via MAPK pathways. There are four distinct MAPK pathways: ERK1/2, ERK5, p38, and JNK. While we show that ERK1 controls macrophage recruitment, p38 MAPK pathway has also been shown to be involved in this process [41]. Indeed, a p38 downstream e ffector MAP-kinase-activated protein kinase 2 (MK2) induces MCP-1 expression in endothelial cells, which promotes monocyte chemoattraction. Thus, at least two di fferent MAPK pathways are involved in the promotion of macrophage infiltration.

E ffective blood flow recovery requires not only expansion of the arterial bed thereby ensuring adequate blood supply, but also e ffective organization and function of this newly formed vasculature. Interestingly, these processes appear to be di fferentially regulated. We and others have previously reported a dissociation between the extent of anatomical arteriogenesis and e ffective blood flow. Thus, a mouse strain with an endothelial loss of NF-kB signaling due to deletion of *Rela* showed increased and disorganized arteriogenesis and decreased tissue perfusion after CFA ligation [42]. This was driven by decreased expression of Dll4 that is NF-kB dependent. Delta-like 4 (Dll4) promotes arterial di fferentiation and restricts vessel branching by direct endothelial cell–cell signaling. Indeed, a similar phenotype was observed in adult *Dll4*+/− mice. These animals also show reduced blood flow recovery after femoral artery occlusion despite exuberant arteriogenesis [43]. Finally, mice with an endothelial-specific deletion of HIF2 α also display increased arteriogenesis abut impaired blood flow recovery in the same hindlimb model [44]. Our description of excessive arteriogenesis ye<sup>t</sup> impaired perfusion in ERK1 null mice adds to this growing body of literature.

In contrast to *Erk1*, *Erk2* knockout in endothelial cells resulted in impaired blood flow recovery despite the normal anatomical extent of arteriogenesis. This is likely due to a decrease in endothelial proliferation combined with decreased expression of eNOS and a corresponding fall in NO production that is critical to the maintenance of arterial tone. Indeed, these observations match a similar decrease in blood flow recovery in endothelial eNOS knockout mice.

ERK1 and ERK2 isoforms share 84% of their amino-acid sequences. ERK1 is larger than ERK2 due to an extension of 17 amino-acids at its N-terminal and two amino-acids at its C-terminal. It has been long a matter of debate whether ERK1 and ERK2 have isoform-specific functions or are totally redundant [45]. ERK2 is expressed at higher levels than ERK1 in most mammalian tissues [46,47]. This di fference in expression level may account for the di fference in phenotype of the global knockout. Indeed, *Erk1*−/− mice are viable and fertile [20], whereas *Erk2*−/− mice die at an early stage in development [19]. Several studies support the functional redundancy of ERK1 and ERK2. Indeed, deletion of *Erk2* but not *Erk1* affected NIH 3T3 cell proliferation in vitro while overexpression of *Erk1* in *Erk2*-deficient NIH 3T3 cells rescued this proliferation defect [36]. ERK1 can also rescue the loss of ERK2 in vivo. Indeed, overexpression of *Erk1* in *Erk2*−/− mice, generating mice expressing only the ERK1 isoform, fully rescue the developmental defects associated with the loss of ERK2 [47]. We recently published that loss of endothelial *Erk2* in a global *Erk1*−/− background in adult mice is lethal, whereas loss of one of the two isoforms in the endothelium of adult mice has no vascular phenotype [22]. Interestingly, deletion of *Erk2* by a ubiquitously expressed Cre in adult *Erk1*−/− mice is lethal in less than three weeks due to multiple organ failure [48]. However, adult mice with only one allele of ERK regardless of the isoform survive [48]. These results sugges<sup>t</sup> redundant roles between ERK1 and ERK2. On the other hand, other studies sugges<sup>t</sup> isoform-specific functions (review in [45]). Notably, in a model of myocardial ischemia/reperfusion injury, myocardial infarction extent was found to be similar in *Erk1*−/− mice and WT mice [49]. However, mice lacking one *Erk2* allele (*Erk2*+/−), developed increased infarct areas compared with WT mice.

In summary, our data demonstrate specific roles of ERK isoforms in endothelial cells. While ERK1 controls macrophage infiltration following an ischemic event, ERK2 primarily controls endothelial cell proliferation and eNOS expression. Both isoforms are involved in regulation of migration.

**Author Contributions:** N.R. and M.S. designed experiments. J.Z. and Z.W.Z. carried out animal studies. N.R. and M.S. wrote the manuscript. M.S. supervised the study. All authors have read and agreed to the published version of the manuscript.

**Funding:** Funded in part by NIH gran<sup>t</sup> 2PO1 HL107205.

**Conflicts of Interest:** The authors declare no conflict of interest.
